Put back the full ap

Bumps [codecov/codecov-action](https://github.com/codecov/codecov-action) from 4.6.0 to 5.0.2.
- [Release notes](https://github.com/codecov/codecov-action/releases)
- [Changelog](https://github.com/codecov/codecov-action/blob/main/CHANGELOG.md)
- [Commits](b9fd7d16f6...5c47607acb)

---
updated-dependencies:
- dependency-name: codecov/codecov-action
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>

.cargo/config.toml

@@ -1,2 +1,6 @@
 [alias]
 xtask = "run --manifest-path ./tasks/Cargo.toml --"
+
+# Accessed by wasm-bindgen when testing for the wasm target
+[target.wasm32-unknown-unknown]
+runner = 'wasm-bindgen-test-runner'

.github/workflows/aws_tfhe_backward_compat_tests.yml

@@ -26,7 +26,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -44,16 +44,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -75,7 +72,7 @@ jobs:
           echo "branch=${BRANCH}" >> "${GITHUB_OUTPUT}"
 
       - name: Clone test data
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           repository: zama-ai/tfhe-backward-compat-data
@@ -90,7 +87,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Backward compatibility tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -103,7 +100,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -114,7 +111,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (backward-compat-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/aws_tfhe_fast_tests.yml

@@ -26,6 +26,7 @@ jobs:
     outputs:
       csprng_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.csprng_any_changed }}
       zk_pok_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.zk_pok_any_changed }}
+      versionable_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.versionable_any_changed }}
       core_crypto_test: ${{ env.IS_PULL_REQUEST == 'false' ||
         steps.changed-files.outputs.core_crypto_any_changed ||
         steps.changed-files.outputs.dependencies_any_changed }}
@@ -50,24 +51,30 @@ jobs:
       any_file_changed: ${{ env.IS_PULL_REQUEST == 'false' || steps.aggregated-changes.outputs.any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             dependencies:
               - tfhe/Cargo.toml
               - concrete-csprng/**
+              - tfhe-fft/**
               - tfhe-zk-pok/**
+              - utils/tfhe-versionable/**
+              - utils/tfhe-versionable-derive/**
             csprng:
               - concrete-csprng/**
             zk_pok:
               - tfhe-zk-pok/**
+            versionable:
+              - utils/tfhe-versionable/**
+              - utils/tfhe-versionable-derive/**
             core_crypto:
               - tfhe/src/core_crypto/**
             boolean:
@@ -103,6 +110,7 @@ jobs:
         if: ( steps.changed-files.outputs.dependencies_any_changed == 'true' ||
           steps.changed-files.outputs.csprng_any_changed == 'true' ||
           steps.changed-files.outputs.zk_pok_any_changed == 'true' ||
+          steps.changed-files.outputs.versionable_any_changed == 'true' ||
           steps.changed-files.outputs.core_crypto_any_changed == 'true' ||
           steps.changed-files.outputs.boolean_any_changed == 'true' ||
           steps.changed-files.outputs.shortint_any_changed == 'true' ||
@@ -124,7 +132,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -144,16 +152,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -167,6 +172,11 @@ jobs:
         run: |
           make test_zk_pok
 
+      - name: Run tfhe-versionable tests
+        if: needs.should-run.outputs.versionable_test == 'true'
+        run: |
+          make test_versionable
+
       - name: Run core tests
         if: needs.should-run.outputs.core_crypto_test == 'true'
         run: |
@@ -182,10 +192,37 @@ jobs:
         run: |
           make test_user_doc
 
+      - name: Get Node version
+        run: |
+          echo "NODE_VERSION=$(make node_version)" >> "${GITHUB_ENV}"
+
+      - name: Node cache restoration
+        id: node-cache
+        uses: actions/cache/restore@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
+        with:
+          path: |
+            ~/.nvm
+            ~/.npm
+          key: node-${{ env.NODE_VERSION }}
+
+      - name: Install Node
+        if: steps.node-cache.outputs.cache-hit != 'true'
+        run: |
+          make install_node
+
+      - name: Node cache save
+        uses: actions/cache/save@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
+        if: steps.node-cache.outputs.cache-hit != 'true'
+        with:
+          path: |
+            ~/.nvm
+            ~/.npm
+          key: node-${{ env.NODE_VERSION }}
+
       - name: Run js on wasm API tests
         if: needs.should-run.outputs.wasm_test == 'true'
         run: |
-          make test_nodejs_wasm_api_in_docker
+          make test_nodejs_wasm_api_ci
 
       - name: Gen Keys if required
         if: needs.should-run.outputs.shortint_test == 'true' ||
@@ -208,14 +245,18 @@ jobs:
         run: |
           make test_high_level_api
 
-      - name: Run safe deserialization tests
+      - name: Run safe serialization tests
         run: |
-          make test_safe_deserialization
+          make test_safe_serialization
+
+      - name: Run zk tests
+        run: |
+          make test_zk
 
       - name: Slack Notification
-        if: ${{ always() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Fast AWS tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -228,7 +269,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -239,7 +280,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (fast-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/aws_tfhe_integer_tests.yml

@@ -19,28 +19,61 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   pull_request:
-    types: [ labeled ]
+    types: [labeled]
   push:
     branches:
       - main
-  schedule:
-    # Nightly tests @ 3AM after each work day
-    - cron: "0 3 * * MON-FRI"
 
 jobs:
+  should-run:
+    if:
+      (github.event_name == 'push' && github.repository == 'zama-ai/tfhe-rs') ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
+      (github.event_name == 'pull_request' && contains(github.event.label.name, 'approved')) ||
+      github.event_name == 'workflow_dispatch'
+    runs-on: ubuntu-latest
+    permissions:
+      pull-requests: write
+    outputs:
+      integer_test: ${{ github.event_name == 'workflow_dispatch' ||
+        steps.changed-files.outputs.integer_any_changed }}
+    steps:
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          persist-credentials: "false"
+
+      - name: Check for file changes
+        id: changed-files
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
+        with:
+          since_last_remote_commit: true
+          files_yaml: |
+            integer:
+              - tfhe/Cargo.toml
+              - concrete-csprng/**
+              - tfhe-fft/**
+              - tfhe-zk-pok/**
+              - tfhe/src/core_crypto/**
+              - tfhe/src/shortint/**
+              - tfhe/src/integer/**
+
   setup-instance:
     name: Setup instance (unsigned-integer-tests)
-    if: (github.event_name == 'push' && github.repository == 'zama-ai/tfhe-rs') ||
-        (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
-        (github.event_name == 'pull_request' && contains(github.event.label.name, 'approved')) ||
-        github.event_name == 'workflow_dispatch'
+    needs: should-run
+    if:
+      (github.event_name == 'push' && github.repository == 'zama-ai/tfhe-rs' && needs.should-run.outputs.integer_test == 'true') ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
+      (github.event_name == 'pull_request' && contains(github.event.label.name, 'approved')) ||
+      github.event_name == 'workflow_dispatch'
     runs-on: ubuntu-latest
     outputs:
       runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -58,16 +91,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
-          persist-credentials: 'false'
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+          persist-credentials: "false"
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -93,9 +123,9 @@ jobs:
           AVX512_SUPPORT=ON NO_BIG_PARAMS=${{ env.NO_BIG_PARAMS }} BIG_TESTS_INSTANCE=TRUE make test_unsigned_integer_ci
 
       - name: Slack Notification
-        if: ${{ always() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Unsigned Integer tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -103,12 +133,12 @@ jobs:
   teardown-instance:
     name: Teardown instance (unsigned-integer-tests)
     if: ${{ always() && needs.setup-instance.result != 'skipped' }}
-    needs: [ setup-instance, unsigned-integer-tests ]
+    needs: [setup-instance, unsigned-integer-tests]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -119,7 +149,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (unsigned-integer-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/aws_tfhe_signed_integer_tests.yml

@@ -19,28 +19,61 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   pull_request:
-    types: [ labeled ]
+    types: [labeled]
   push:
     branches:
       - main
-  schedule:
-    # Nightly tests @ 3AM after each work day
-    - cron: "0 3 * * MON-FRI"
 
 jobs:
+  should-run:
+    if:
+      (github.event_name == 'push' && github.repository == 'zama-ai/tfhe-rs') ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
+      (github.event_name == 'pull_request' && contains(github.event.label.name, 'approved')) ||
+      github.event_name == 'workflow_dispatch'
+    runs-on: ubuntu-latest
+    permissions:
+      pull-requests: write
+    outputs:
+      integer_test: ${{ github.event_name == 'workflow_dispatch' ||
+        steps.changed-files.outputs.integer_any_changed }}
+    steps:
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          persist-credentials: "false"
+
+      - name: Check for file changes
+        id: changed-files
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
+        with:
+          since_last_remote_commit: true
+          files_yaml: |
+            integer:
+              - tfhe/Cargo.toml
+              - concrete-csprng/**
+              - tfhe-fft/**
+              - tfhe-zk-pok/**
+              - tfhe/src/core_crypto/**
+              - tfhe/src/shortint/**
+              - tfhe/src/integer/**
+
   setup-instance:
-    name: Setup instance (signed-integer-tests)
-    if: (github.event_name == 'push' && github.repository == 'zama-ai/tfhe-rs') ||
-        (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
-        (github.event_name == 'pull_request' && contains(github.event.label.name, 'approved')) ||
-        github.event_name == 'workflow_dispatch'
+    name: Setup instance (unsigned-integer-tests)
+    needs: should-run
+    if:
+      (github.event_name == 'push' && github.repository == 'zama-ai/tfhe-rs' && needs.should-run.outputs.integer_test == 'true') ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
+      (github.event_name == 'pull_request' && contains(github.event.label.name, 'approved')) ||
+      github.event_name == 'workflow_dispatch'
     runs-on: ubuntu-latest
     outputs:
       runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -58,16 +91,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
-          persist-credentials: 'false'
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+          persist-credentials: "false"
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -97,9 +127,9 @@ jobs:
           AVX512_SUPPORT=ON NO_BIG_PARAMS=${{ env.NO_BIG_PARAMS }} BIG_TESTS_INSTANCE=TRUE make test_signed_integer_ci
 
       - name: Slack Notification
-        if: ${{ always() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Signed Integer tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -107,12 +137,12 @@ jobs:
   teardown-instance:
     name: Teardown instance (signed-integer-tests)
     if: ${{ always() && needs.setup-instance.result != 'skipped' }}
-    needs: [ setup-instance, signed-integer-tests ]
+    needs: [setup-instance, signed-integer-tests]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -123,7 +153,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (signed-integer-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/aws_tfhe_tests.yml

@@ -57,19 +57,20 @@ jobs:
       any_file_changed: ${{ env.IS_PULL_REQUEST == 'false' || steps.aggregated-changes.outputs.any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             dependencies:
               - tfhe/Cargo.toml
               - concrete-csprng/**
+              - tfhe-fft/**
               - tfhe-zk-pok/**
             csprng:
               - concrete-csprng/**
@@ -131,7 +132,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -151,16 +152,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -222,9 +220,9 @@ jobs:
           make test_kreyvium
 
       - name: Slack Notification
-        if: ${{ always() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "CPU tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -237,7 +235,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -248,7 +246,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cpu-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/aws_tfhe_wasm_tests.yml

@@ -27,7 +27,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -45,39 +45,68 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
+      - name: Get Node version
+        run: |
+          echo "NODE_VERSION=$(make node_version)" >> "${GITHUB_ENV}"
+
+      - name: Node cache restoration
+        id: node-cache
+        uses: actions/cache/restore@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
+        with:
+          path: |
+            ~/.nvm
+            ~/.npm
+          key: node-${{ env.NODE_VERSION }}
+
       - name: Install Node
+        if: steps.node-cache.outputs.cache-hit != 'true'
         run: |
           make install_node
 
+      - name: Node cache save
+        uses: actions/cache/save@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
+        if: steps.node-cache.outputs.cache-hit != 'true'
+        with:
+          path: |
+            ~/.nvm
+            ~/.npm
+          key: node-${{ env.NODE_VERSION }}
+
+      - name: Install web resources
+        run: |
+          make install_chrome_browser
+          make install_chrome_web_driver
+
       - name: Run fmt checks
         run: |
           make check_fmt_js
 
       - name: Run js on wasm API tests
         run: |
-          make test_nodejs_wasm_api_in_docker
+          make test_nodejs_wasm_api_ci
 
       - name: Run parallel wasm tests
         run: |
-          make test_web_js_api_parallel_ci
+          make test_web_js_api_parallel_chrome_ci
+
+      - name: Run x86_64/wasm zk compatibility tests
+        run: |
+          make test_zk_wasm_x86_compat_ci
 
       - name: Slack Notification
-        if: ${{ always() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "WASM tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -90,7 +119,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -101,7 +130,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (wasm-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_boolean.yml

@@ -29,7 +29,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -48,9 +48,10 @@ jobs:
     continue-on-error: true
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -60,13 +61,8 @@ jobs:
             echo "COMMIT_HASH=$(git describe --tags --dirty)";
           } >> "${GITHUB_ENV}"
 
-      - name: Set up home
-        # "Install rust" step require root user to have a HOME directory which is not set.
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
-
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -94,17 +90,17 @@ jobs:
       - name: Parse key sizes results
         run: |
           python3 ./ci/benchmark_parser.py tfhe/boolean_key_sizes.csv ${{ env.RESULTS_FILENAME }} \
-          --key-sizes \
+          --object-sizes \
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_boolean
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -113,21 +109,13 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Boolean benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -140,7 +128,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -151,7 +139,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (boolean-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_core_crypto.yml

@@ -26,7 +26,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -44,9 +44,10 @@ jobs:
       cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -56,13 +57,8 @@ jobs:
             echo "COMMIT_HASH=$(git describe --tags --dirty)";
           } >> "${GITHUB_ENV}"
 
-      - name: Set up home
-        # "Install rust" step require root user to have a HOME directory which is not set.
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
-
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -86,13 +82,13 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_core_crypto
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -101,21 +97,13 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on downloaded artifact"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "PBS benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -128,7 +116,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -139,7 +127,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (core-crypto-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_erc20.yml

@@ -0,0 +1,141 @@
+# Run all ERC20 benchmarks on an AWS instance and return parsed results to Slab CI bot.
+name: ERC20 benchmarks
+
+on:
+  workflow_dispatch:
+  schedule:
+    # Weekly benchmarks will be triggered each Saturday at 5a.m.
+    - cron: '0 5 * * 6'
+
+env:
+  CARGO_TERM_COLOR: always
+  RESULTS_FILENAME: parsed_benchmark_results_${{ github.sha }}.json
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
+jobs:
+  setup-instance:
+    name: Setup instance (erc20-benchmarks)
+    runs-on: ubuntu-latest
+    if: github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: aws
+          profile: bench
+
+  erc20-benchmarks:
+    name: Execute ERC20 benchmarks
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    concurrency:
+      group: ${{ github.workflow }}_${{ github.ref }}
+      cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+    continue-on-error: true
+    timeout-minutes: 720  # 12 hours
+    steps:
+      - name: Checkout tfhe-rs repo with tags
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Get benchmark details
+        run: |
+          {
+            echo "BENCH_DATE=$(date --iso-8601=seconds)";
+            echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
+            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+          } >> "${GITHUB_ENV}"
+
+      - name: Install rust
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: nightly
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Run benchmarks
+        run: |
+          make bench_hlapi_erc20
+
+      - name: Parse results
+        run: |
+          python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --hardware "hpc7a.96xlarge" \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512
+
+      - name: Parse PBS counts
+        run: |
+          python3 ./ci/benchmark_parser.py tfhe/erc20_pbs_count.csv ${{ env.RESULTS_FILENAME }} \
+          --object-sizes \
+          --append-results
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_erc20
+          path: ${{ env.RESULTS_FILENAME }}
+
+      - name: Send data to Slab
+        shell: bash
+        run: |
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "ERC20 benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (erc20-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, erc20-benchmarks ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (erc20-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_4090.yml

@@ -39,9 +39,10 @@ jobs:
 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -53,12 +54,12 @@ jobs:
           echo "FAST_BENCH=TRUE" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -82,7 +83,7 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_integer_multi_bit_gpu_default
           path: ${{ env.RESULTS_FILENAME }}
@@ -90,21 +91,13 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
       - name: Slack Notification
-        if: ${{ always() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Integer RTX 4090 full benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -121,7 +114,7 @@ jobs:
 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
@@ -134,12 +127,12 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -164,7 +157,7 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_core_crypto
           path: ${{ env.RESULTS_FILENAME }}
@@ -184,9 +177,9 @@ jobs:
           ${{ secrets.SLAB_URL }}
 
       - name: Slack Notification
-        if: ${{ !success() && !cancelled() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Core crypto RTX 4090 full benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_core_crypto.yml

@@ -27,7 +27,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -56,7 +56,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -64,10 +64,12 @@ jobs:
           make -j"$(nproc)"
           sudo make install
 
+
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -83,7 +85,7 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -128,13 +130,13 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_core_crypto
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -143,16 +145,8 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on downloaded artifact"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
   slack-notify:
     name: Slack Notification
@@ -162,7 +156,7 @@ jobs:
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-core-crypto-benchmarks.result }}
           SLACK_MESSAGE: "PBS GPU benchmarks finished with status: ${{ needs.cuda-core-crypto-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
@@ -175,7 +169,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -186,7 +180,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-core-crypto-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_erc20.yml

@@ -0,0 +1,195 @@
+# Run ERC20 benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: ERC20 GPU H100 benchmarks
+
+on:
+  workflow_dispatch:
+  schedule:
+    # Weekly benchmarks will be triggered each Saturday at 5a.m.
+    - cron: '0 5 * * 6'
+
+env:
+  CARGO_TERM_COLOR: always
+  RESULTS_FILENAME: parsed_benchmark_results_${{ github.sha }}.json
+  PARSE_INTEGER_BENCH_CSV_FILE: tfhe_rs_integer_benches_${{ github.sha }}.csv
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
+jobs:
+  setup-instance:
+    name: Setup instance (cuda-erc20-benchmarks)
+    runs-on: ubuntu-latest
+    if:  github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: hyperstack
+          profile: single-h100
+
+  cuda-erc20-benchmarks:
+    name: Execute GPU integer benchmarks
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    strategy:
+      fail-fast: false
+      # explicit include-based build matrix, of known valid options
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
+    steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+      - name: Checkout tfhe-rs repo with tags
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Get benchmark details
+        run: |
+          {
+            echo "BENCH_DATE=$(date --iso-8601=seconds)";
+            echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
+            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+          } >> "${GITHUB_ENV}"
+
+      - name: Set up home
+        # "Install rust" step require root user to have a HOME directory which is not set.
+        run: |
+          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+
+      - name: Install rust
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: nightly
+
+      - name: Export CUDA variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CUDA_PATH=$CUDA_PATH";
+            echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH";
+            echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc";
+          } >> "${GITHUB_ENV}"
+          echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
+
+      # Specify the correct host compilers
+      - name: Export gcc and g++ variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CC=/usr/bin/gcc-${{ matrix.gcc }}";
+            echo "CXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
+          } >> "${GITHUB_ENV}"
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run benchmarks
+        run: |
+          make bench_hlapi_erc20_gpu
+
+      - name: Parse results
+        run: |
+          python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --hardware "n3-H100x1" \
+          --backend gpu \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512
+
+      - name: Parse PBS counts
+        run: |
+          python3 ./ci/benchmark_parser.py tfhe/erc20_pbs_count.csv ${{ env.RESULTS_FILENAME }} \
+          --object-sizes \
+          --append-results
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_erc20
+          path: ${{ env.RESULTS_FILENAME }}
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Send data to Slab
+        shell: bash
+        run: |
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
+
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-erc20-benchmarks ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-erc20-benchmarks.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ needs.cuda-erc20-benchmarks.result }}
+          SLACK_MESSAGE: "Integer GPU benchmarks finished with status: ${{ needs.cuda-erc20-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-erc20-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-erc20-benchmarks, slack-notify ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (cuda-erc20-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_erc20_2h100.yml

@@ -0,0 +1,195 @@
+# Run ERC20 benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: ERC20 GPU 2xH100 benchmarks
+
+on:
+  workflow_dispatch:
+  schedule:
+    # Weekly benchmarks will be triggered each Saturday at 5a.m.
+    - cron: '0 5 * * 6'
+
+env:
+  CARGO_TERM_COLOR: always
+  RESULTS_FILENAME: parsed_benchmark_results_${{ github.sha }}.json
+  PARSE_INTEGER_BENCH_CSV_FILE: tfhe_rs_integer_benches_${{ github.sha }}.csv
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
+jobs:
+  setup-instance:
+    name: Setup instance (cuda-erc20-benchmarks)
+    runs-on: ubuntu-latest
+    if:  github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: hyperstack
+          profile: 2-h100
+
+  cuda-erc20-benchmarks:
+    name: Execute GPU integer benchmarks
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    strategy:
+      fail-fast: false
+      # explicit include-based build matrix, of known valid options
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
+    steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+      - name: Checkout tfhe-rs repo with tags
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Get benchmark details
+        run: |
+          {
+            echo "BENCH_DATE=$(date --iso-8601=seconds)";
+            echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
+            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+          } >> "${GITHUB_ENV}"
+
+      - name: Set up home
+        # "Install rust" step require root user to have a HOME directory which is not set.
+        run: |
+          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+
+      - name: Install rust
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: nightly
+
+      - name: Export CUDA variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CUDA_PATH=$CUDA_PATH";
+            echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH";
+            echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc";
+          } >> "${GITHUB_ENV}"
+          echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
+
+      # Specify the correct host compilers
+      - name: Export gcc and g++ variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CC=/usr/bin/gcc-${{ matrix.gcc }}";
+            echo "CXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
+          } >> "${GITHUB_ENV}"
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run benchmarks
+        run: |
+          make bench_hlapi_erc20_gpu
+
+      - name: Parse results
+        run: |
+          python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --hardware "n3-H100x2" \
+          --backend gpu \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512
+
+      - name: Parse PBS counts
+        run: |
+          python3 ./ci/benchmark_parser.py tfhe/erc20_pbs_count.csv ${{ env.RESULTS_FILENAME }} \
+          --object-sizes \
+          --append-results
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_erc20
+          path: ${{ env.RESULTS_FILENAME }}
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Send data to Slab
+        shell: bash
+        run: |
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
+
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-erc20-benchmarks ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-erc20-benchmarks.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ needs.cuda-erc20-benchmarks.result }}
+          SLACK_MESSAGE: "ERC20 2xH100 benchmarks finished with status: ${{ needs.cuda-erc20-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-erc20-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-erc20-benchmarks, slack-notify ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (cuda-erc20-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_erc20_8h100.yml

@@ -0,0 +1,195 @@
+# Run ERC20 benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: ERC20 GPU 8xH100 benchmarks
+
+on:
+  workflow_dispatch:
+  schedule:
+    # Weekly benchmarks will be triggered each Saturday at 5a.m.
+    - cron: '0 5 * * 6'
+
+env:
+  CARGO_TERM_COLOR: always
+  RESULTS_FILENAME: parsed_benchmark_results_${{ github.sha }}.json
+  PARSE_INTEGER_BENCH_CSV_FILE: tfhe_rs_integer_benches_${{ github.sha }}.csv
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
+jobs:
+  setup-instance:
+    name: Setup instance (cuda-erc20-benchmarks)
+    runs-on: ubuntu-latest
+    if:  github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: hyperstack
+          profile: multi-h100
+
+  cuda-erc20-benchmarks:
+    name: Execute GPU integer benchmarks
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    strategy:
+      fail-fast: false
+      # explicit include-based build matrix, of known valid options
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
+    steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+      - name: Checkout tfhe-rs repo with tags
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Get benchmark details
+        run: |
+          {
+            echo "BENCH_DATE=$(date --iso-8601=seconds)";
+            echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
+            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+          } >> "${GITHUB_ENV}"
+
+      - name: Set up home
+        # "Install rust" step require root user to have a HOME directory which is not set.
+        run: |
+          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+
+      - name: Install rust
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: nightly
+
+      - name: Export CUDA variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CUDA_PATH=$CUDA_PATH";
+            echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH";
+            echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc";
+          } >> "${GITHUB_ENV}"
+          echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
+
+      # Specify the correct host compilers
+      - name: Export gcc and g++ variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CC=/usr/bin/gcc-${{ matrix.gcc }}";
+            echo "CXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
+          } >> "${GITHUB_ENV}"
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run benchmarks
+        run: |
+          make bench_hlapi_erc20_gpu
+
+      - name: Parse results
+        run: |
+          python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --hardware "n3-H100x8" \
+          --backend gpu \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512
+
+      - name: Parse PBS counts
+        run: |
+          python3 ./ci/benchmark_parser.py tfhe/erc20_pbs_count.csv ${{ env.RESULTS_FILENAME }} \
+          --object-sizes \
+          --append-results
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_erc20
+          path: ${{ env.RESULTS_FILENAME }}
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Send data to Slab
+        shell: bash
+        run: |
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
+
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-erc20-benchmarks ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-erc20-benchmarks.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ needs.cuda-erc20-benchmarks.result }}
+          SLACK_MESSAGE: "ERC20 8xH100 benchmarks finished with status: ${{ needs.cuda-erc20-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-erc20-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-erc20-benchmarks, slack-notify ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (cuda-erc20-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer.yml

@@ -30,7 +30,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -59,7 +59,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -68,9 +68,10 @@ jobs:
           sudo make install
 
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -86,7 +87,7 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -124,7 +125,7 @@ jobs:
             parse_integer_benches
 
       - name: Upload csv results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_csv_integer
           path: ${{ env.PARSE_INTEGER_BENCH_CSV_FILE }}
@@ -144,13 +145,13 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_integer
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -159,16 +160,8 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
   slack-notify:
     name: Slack Notification
@@ -178,7 +171,7 @@ jobs:
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-integer-benchmarks.result }}
           SLACK_MESSAGE: "Integer GPU benchmarks finished with status: ${{ needs.cuda-integer-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
@@ -191,7 +184,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -202,7 +195,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-integer-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_2H100_full.yml

@@ -0,0 +1,194 @@
+# Run integer benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: Integer 2xH100 benchmarks
+
+on:
+  workflow_dispatch:
+  schedule:
+    # Weekly benchmarks will be triggered each Saturday at 1a.m.
+    - cron: '0 1 * * 6'
+
+env:
+  CARGO_TERM_COLOR: always
+  RESULTS_FILENAME: parsed_benchmark_results_${{ github.sha }}.json
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
+jobs:
+  setup-instance:
+    name: Setup instance (cuda-integer-full-2-gpu-benchmarks)
+    runs-on: ubuntu-latest
+    if: github.event_name != 'schedule' ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: hyperstack
+          profile: 2-h100
+
+  cuda-integer-full-2-gpu-benchmarks:
+    name: Execute 2xH100 integer benchmarks
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    timeout-minutes: 1440 # 24 hours
+    continue-on-error: true
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        command: [integer_multi_bit]
+        op_flavor: [default]
+        # explicit include-based build matrix, of known valid options
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
+    steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+      - name: Checkout tfhe-rs repo with tags
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Get benchmark details
+        run: |
+          {
+            echo "BENCH_DATE=$(date --iso-8601=seconds)";
+            echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
+            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+          } >> "${GITHUB_ENV}"
+
+      - name: Set up home
+        # "Install rust" step require root user to have a HOME directory which is not set.
+        run: |
+          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+
+      - name: Install rust
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: nightly
+
+      - name: Export CUDA variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CUDA_PATH=$CUDA_PATH";
+            echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH";
+            echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc";
+          } >> "${GITHUB_ENV}"
+          echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
+
+      # Specify the correct host compilers
+      - name: Export gcc and g++ variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CC=/usr/bin/gcc-${{ matrix.gcc }}";
+            echo "CXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
+          } >> "${GITHUB_ENV}"
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run benchmarks with AVX512
+        run: |
+          make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} bench_${{ matrix.command }}_gpu
+
+      - name: Parse results
+        run: |
+          python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --hardware "n3-H100x2" \
+          --backend gpu \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512 \
+          --throughput
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}
+          path: ${{ env.RESULTS_FILENAME }}
+
+      - name: Send data to Slab
+        shell: bash
+        run: |
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
+
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-integer-full-2-gpu-benchmarks ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-integer-full-2-gpu-benchmarks.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ needs.cuda-integer-full-2-gpu-benchmarks.result }}
+          SLACK_MESSAGE: "Integer GPU 2xH100 benchmarks finished with status: ${{ needs.cuda-integer-full-2-gpu-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-integer-full-2-gpu-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-integer-full-2-gpu-benchmarks ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (cuda-integer-full-2-gpu-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_full.yml

@@ -29,7 +29,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -49,7 +49,7 @@ jobs:
       max-parallel: 1
       matrix:
         command: [integer, integer_multi_bit]
-        op_flavor: [default, unchecked]
+        op_flavor: [default]
         # explicit include-based build matrix, of known valid options
         include:
           - os: ubuntu-22.04
@@ -63,7 +63,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -72,9 +72,10 @@ jobs:
           sudo make install
 
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -90,7 +91,7 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -115,7 +116,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -129,6 +130,12 @@ jobs:
         run: |
           make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} bench_${{ matrix.command }}_gpu
 
+      # Run these benchmarks only once
+      - name: Run compression benchmarks with AVX512
+        if: matrix.op_flavor == 'default' && matrix.command == 'integer'
+        run: |
+          make bench_integer_compression_gpu
+
       - name: Parse results
         run: |
           python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
@@ -144,7 +151,7 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}
           path: ${{ env.RESULTS_FILENAME }}
@@ -152,26 +159,18 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
   slack-notify:
     name: Slack Notification
     needs: [ setup-instance, cuda-integer-full-benchmarks ]
     runs-on: ubuntu-latest
-    if: ${{ !success() && !cancelled() }}
+    if: ${{ always() && needs.cuda-integer-full-benchmarks.result != 'skipped' && failure() }}
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-integer-full-benchmarks.result }}
           SLACK_MESSAGE: "Integer GPU full benchmarks finished with status: ${{ needs.cuda-integer-full-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
@@ -184,7 +183,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -195,7 +194,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-integer-full-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_multi_bit.yml

@@ -42,7 +42,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -72,7 +72,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -81,9 +81,10 @@ jobs:
           sudo make install
 
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -99,7 +100,7 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -147,7 +148,7 @@ jobs:
             parse_integer_benches
 
       - name: Upload csv results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_csv_integer
           path: ${{ env.PARSE_INTEGER_BENCH_CSV_FILE }}
@@ -167,13 +168,13 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_integer
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -182,27 +183,18 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
-
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
   slack-notify:
     name: Slack Notification
     needs: [ setup-instance, cuda-integer-multi-bit-benchmarks ]
     runs-on: ubuntu-latest
-    if: ${{ !success() && !cancelled() }}
+    if: ${{ always() && needs.cuda-integer-multi-bit-benchmarks.result != 'skipped' && failure() }}
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-integer-multi-bit-benchmarks.result }}
           SLACK_MESSAGE: "Integer GPU multi-bit benchmarks finished with status: ${{ needs.cuda-integer-multi-bit-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
@@ -215,7 +207,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -226,7 +218,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-integer-multi-bit-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_multi_bit_multi_gpu.yml

@@ -42,7 +42,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -73,7 +73,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -82,9 +82,10 @@ jobs:
           sudo make install
 
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -100,7 +101,7 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -125,7 +126,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -164,7 +165,7 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_integer
           path: ${{ env.RESULTS_FILENAME }}
@@ -172,26 +173,18 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
   slack-notify:
     name: Slack Notification
     needs: [ setup-instance, cuda-integer-multi-bit-multi-gpu-benchmarks ]
     runs-on: ubuntu-latest
-    if: ${{ !success() && !cancelled() }}
+    if: ${{ always() && needs.cuda-integer-multi-bit-multi-gpu-benchmarks.result != 'skipped' && failure() }}
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-integer-multi-bit-multi-gpu-benchmarks.result }}
           SLACK_MESSAGE: "Integer multi GPU multi-bit benchmarks finished with status: ${{ needs.cuda-integer-multi-bit-multi-gpu-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
@@ -204,7 +197,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -215,7 +208,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-integer-multi-bit-multi-gpu-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_multi_gpu_full.yml

@@ -29,7 +29,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -39,7 +39,7 @@ jobs:
           profile: multi-h100
 
   cuda-integer-full-multi-gpu-benchmarks:
-    name: Execute multi GPU integer benchmarks for all operations flavor
+    name: Execute multi GPU integer benchmarks
     needs: setup-instance
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     timeout-minutes: 1440 # 24 hours
@@ -48,8 +48,8 @@ jobs:
       fail-fast: false
       max-parallel: 1
       matrix:
-        command: [integer, integer_multi_bit]
-        op_flavor: [default, unchecked]
+        command: [integer_multi_bit]
+        op_flavor: [default]
         # explicit include-based build matrix, of known valid options
         include:
           - os: ubuntu-22.04
@@ -63,7 +63,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -72,9 +72,10 @@ jobs:
           sudo make install
 
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -90,7 +91,7 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -115,7 +116,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -144,7 +145,7 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}
           path: ${{ env.RESULTS_FILENAME }}
@@ -152,26 +153,18 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
   slack-notify:
     name: Slack Notification
     needs: [ setup-instance, cuda-integer-full-multi-gpu-benchmarks ]
     runs-on: ubuntu-latest
-    if: ${{ !success() && !cancelled() }}
+    if: ${{ always() && needs.cuda-integer-full-multi-gpu-benchmarks.result != 'skipped' && failure() }}
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-integer-full-multi-gpu-benchmarks.result }}
           SLACK_MESSAGE: "Integer GPU full benchmarks finished with status: ${{ needs.cuda-integer-full-multi-gpu-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
@@ -184,7 +177,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -195,7 +188,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-integer-full-multi-gpu-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_l40.yml

@@ -0,0 +1,206 @@
+# Run benchmarks on an L40 VM and return parsed results to Slab CI bot.
+name: Cuda benchmarks (L40)
+
+on:
+  workflow_dispatch:
+  schedule:
+    # Weekly benchmarks will be triggered each Saturday at 1a.m.
+    - cron: '0 1 * * 6'
+
+env:
+  CARGO_TERM_COLOR: always
+  RESULTS_FILENAME: parsed_benchmark_results_${{ github.sha }}.json
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
+jobs:
+  setup-instance:
+    name: Setup instance (cuda-l40-benchmarks)
+    runs-on: ubuntu-latest
+    if: github.event_name != 'schedule' ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: hyperstack
+          profile: l40 
+
+  cuda-l40-benchmarks:
+    name: Cuda benchmarks (L40)
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    timeout-minutes: 1440 # 24 hours
+    continue-on-error: true
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        command: [integer_multi_bit]
+        op_flavor: [default]
+        # explicit include-based build matrix, of known valid options
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
+    steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+      - name: Checkout tfhe-rs repo with tags
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Get benchmark details
+        run: |
+          {
+            echo "BENCH_DATE=$(date --iso-8601=seconds)";
+            echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
+            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+          } >> "${GITHUB_ENV}"
+
+      - name: Set up home
+        # "Install rust" step require root user to have a HOME directory which is not set.
+        run: |
+          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+
+      - name: Install rust
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: nightly
+
+      - name: Export CUDA variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CUDA_PATH=$CUDA_PATH";
+            echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH";
+            echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc";
+          } >> "${GITHUB_ENV}"
+          echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
+
+      # Specify the correct host compilers
+      - name: Export gcc and g++ variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CC=/usr/bin/gcc-${{ matrix.gcc }}";
+            echo "CXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
+          } >> "${GITHUB_ENV}"
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run benchmarks with AVX512
+        run: |
+          make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} bench_${{ matrix.command }}_gpu
+
+      - name: Run compression benchmarks with AVX512
+        run: |
+          make bench_integer_compression_gpu
+
+      - name: Run PBS benchmarks 
+        run: |
+          make bench_pbs_gpu
+
+      - name: Run KS benchmarks 
+        run: |
+          make bench_ks_gpu
+
+      - name: Parse results
+        run: |
+          python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --hardware "n3-L40x1" \
+          --backend gpu \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512 \
+          --throughput
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}
+          path: ${{ env.RESULTS_FILENAME }}
+
+      - name: Send data to Slab
+        shell: bash
+        run: |
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
+
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-l40-benchmarks ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-l40-benchmarks.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ needs.cuda-l40-benchmarks.result }}
+          SLACK_MESSAGE: "Cuda benchmarks (L40) finished with status: ${{ needs.cuda-l40-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-l40-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-l40-benchmarks, slack-notify ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (cuda-l40-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_integer.yml

@@ -62,7 +62,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -87,9 +87,10 @@ jobs:
         op_flavor: ${{ fromJson(needs.prepare-matrix.outputs.op_flavor) }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -99,18 +100,13 @@ jobs:
             echo "COMMIT_HASH=$(git describe --tags --dirty)";
           } >> "${GITHUB_ENV}"
 
-      - name: Set up home
-        # "Install rust" step require root user to have a HOME directory which is not set.
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
-
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -125,6 +121,12 @@ jobs:
         run: |
           make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} bench_${{ matrix.command }}
 
+      # Run these benchmarks only once
+      - name: Run compression benchmarks with AVX512
+        if: matrix.op_flavor == 'default' && matrix.command == 'integer'
+        run: |
+          make bench_integer_compression
+
       - name: Parse results
         run: |
           python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
@@ -139,7 +141,7 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}
           path: ${{ env.RESULTS_FILENAME }}
@@ -147,21 +149,13 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Integer full benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -174,7 +168,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -185,7 +179,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (integer-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_shortint.yml

@@ -56,7 +56,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -79,9 +79,10 @@ jobs:
         op_flavor: ${{ fromJson(needs.prepare-matrix.outputs.op_flavor) }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -91,18 +92,13 @@ jobs:
             echo "COMMIT_HASH=$(git describe --tags --dirty)";
           } >> "${GITHUB_ENV}"
 
-      - name: Set up home
-        # "Install rust" step require root user to have a HOME directory which is not set.
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
-
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -137,11 +133,11 @@ jobs:
         if: matrix.op_flavor == 'default'
         run: |
           python3 ./ci/benchmark_parser.py tfhe/shortint_key_sizes.csv ${{ env.RESULTS_FILENAME }} \
-          --key-sizes \
+          --object-sizes \
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_shortint_${{ matrix.op_flavor }}
           path: ${{ env.RESULTS_FILENAME }}
@@ -149,21 +145,13 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Shortint full benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -176,7 +164,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -187,7 +175,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (shortint-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_signed_integer.yml

@@ -62,7 +62,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -87,9 +87,10 @@ jobs:
         op_flavor: [ default, unchecked ]
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -99,18 +100,13 @@ jobs:
             echo "COMMIT_HASH=$(git describe --tags --dirty)";
           } >> "${GITHUB_ENV}"
 
-      - name: Set up home
-        # "Install rust" step require root user to have a HOME directory which is not set.
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
-
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -139,7 +135,7 @@ jobs:
           --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}
           path: ${{ env.RESULTS_FILENAME }}
@@ -147,21 +143,13 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Signed integer full benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -174,7 +162,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -185,7 +173,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (signed-integer-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_tfhe_fft.yml

@@ -0,0 +1,151 @@
+# Run FFT benchmarks on an AWS instance and return parsed results to Slab CI bot.
+name: FFT benchmarks
+
+env:
+  CARGO_TERM_COLOR: always
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RESULTS_FILENAME: parsed_benchmark_results_${{ github.sha }}.json
+  RUST_BACKTRACE: "full"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - "main"
+  schedule:
+    # Job will be triggered each Thursday at 11p.m.
+    - cron: '0 23 * * 4'
+
+jobs:
+  setup-ec2:
+    name: Setup EC2 instance (fft-benchmarks)
+    runs-on: ubuntu-latest
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@1dced74825027fe3d481392163ed8fc56813fb5d
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: aws
+          profile: bench
+
+  fft-benchmarks:
+    name: Execute FFT benchmarks in EC2
+    needs: setup-ec2
+    concurrency:
+      group: ${{ github.workflow }}_${{ github.ref }}
+      cancel-in-progress: true
+    runs-on: ${{ needs.setup-ec2.outputs.runner-name }}
+    steps:
+      - name: Checkout tfhe-rs repo with tags
+        uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
+        with:
+          fetch-depth: 0
+
+      - name: Get benchmark details
+        run: |
+          {
+            echo "BENCH_DATE=$(date --iso-8601=seconds)";
+            echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
+            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+          } >> "${GITHUB_ENV}"
+
+      - name: Install rust
+        uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
+        with:
+          toolchain: nightly
+          override: true
+
+      - name: Run benchmarks
+        run: |
+          make FFT128_SUPPORT=ON bench
+
+      - name: Parse results
+        run: |
+          python3 ./ci/fft_benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --database concrete_fft \
+          --hardware "hpc7a.96xlarge" \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}"
+
+          rm -rf target/criterion benchmarks_parameters/
+
+      - name: Run benchmarks with AVX512
+        run: |
+          make FFT128_SUPPORT=ON AVX512_SUPPORT=ON bench
+
+      - name: Parse AVX512 results
+        run: |
+          python3 ./ci/fft_benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --name-suffix avx512 \
+          --append-results
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@0b7f8abb1508181956e8e162db84b466c27e18ce
+        with:
+          name: ${{ github.sha }}_fft
+          path: ${{ env.RESULTS_FILENAME }}
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Send data to Slab
+        shell: bash
+        run: |
+          echo "Computing HMac on downloaded artifact"
+          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
+          echo "Sending results to Slab..."
+          curl -v -k \
+          -H "Content-Type: application/json" \
+          -H "X-Slab-Repository: ${{ github.repository }}" \
+          -H "X-Slab-Command: store_data_v2" \
+          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
+          -d @${{ env.RESULTS_FILENAME }} \
+          ${{ secrets.SLAB_URL }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@b24d75fe0e728a4bf9fc42ee217caa686d141ee8
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "tfhe-fft benchmarks failed. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-ec2:
+    name: Teardown EC2 instance (fft-benchmarks)
+    if: ${{ always() && needs.setup-ec2.result != 'skipped' }}
+    needs: [ setup-ec2, fft-benchmarks ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@1dced74825027fe3d481392163ed8fc56813fb5d
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-ec2.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@b24d75fe0e728a4bf9fc42ee217caa686d141ee8
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "EC2 teardown (fft-benchmarks) failed. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_tfhe_zk_pok.yml

@@ -0,0 +1,173 @@
+# Run benchmarks of the tfhe-zk-pok crate on an instance and return parsed results to Slab CI bot.
+name: tfhe-zk-pok benchmarks
+
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - main
+  schedule:
+    # Weekly benchmarks will be triggered each Saturday at 3a.m.
+    - cron: '0 3 * * 6'
+env:
+  CARGO_TERM_COLOR: always
+  RESULTS_FILENAME: parsed_benchmark_results_${{ github.sha }}.json
+  PARSE_INTEGER_BENCH_CSV_FILE: tfhe_rs_integer_benches_${{ github.sha }}.csv
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
+jobs:
+  should-run:
+    runs-on: ubuntu-latest
+    if: github.event_name == 'workflow_dispatch' ||
+      ((github.event_name == 'push' || github.event_name == 'schedule') && github.repository == 'zama-ai/tfhe-rs')
+    outputs:
+      zk_pok_changed: ${{ steps.changed-files.outputs.zk_pok_any_changed }}
+    steps:
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+
+      - name: Check for file changes
+        id: changed-files
+        uses: tj-actions/changed-files@c3a1bb2c992d77180ae65be6ae6c166cf40f857c
+        with:
+          since_last_remote_commit: true
+          files_yaml: |
+            zk_pok:
+              - tfhe-zk-pok/**
+              - .github/workflows/benchmark_tfhe_zk_pok.yml
+
+  setup-instance:
+    name: Setup instance (tfhe-zk-pok-benchmarks)
+    runs-on: ubuntu-latest
+    needs: should-run
+    if: github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
+      (github.event_name == 'push' &&
+      github.repository == 'zama-ai/tfhe-rs' &&
+      needs.should-run.outputs.zk_pok_changed == 'true')
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: aws
+          profile: bench
+
+  tfhe-zk-pok-benchmarks:
+    name: Execute tfhe-zk-pok benchmarks
+    if: needs.setup-instance.result != 'skipped'
+    needs: setup-instance
+    concurrency:
+      group: ${{ github.workflow }}_${{github.event_name}}_${{ github.ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
+      cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    steps:
+      - name: Checkout tfhe-rs repo with tags
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Get benchmark details
+        run: |
+          {
+            echo "BENCH_DATE=$(date --iso-8601=seconds)";
+            echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
+            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+          } >> "${GITHUB_ENV}"
+
+      - name: Install rust
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: nightly
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Run benchmarks
+        run: |
+          make bench_tfhe_zk_pok
+
+      - name: Parse results
+        run: |
+          python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --crate tfhe-zk-pok \
+          --hardware "hpc7a.96xlarge" \
+          --backend cpu \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512 \
+          --throughput
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_tfhe_zk_pok
+          path: ${{ env.RESULTS_FILENAME }}
+
+      - name: Checkout Slab repo
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          repository: zama-ai/slab
+          path: slab
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Send data to Slab
+        shell: bash
+        run: |
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "tfhe-zk-pok benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (tfhe-zk-pok-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, tfhe-zk-pok-benchmarks ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (tfhe-zk-pok-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_wasm_client.yml

@@ -33,13 +33,13 @@ jobs:
       wasm_bench: ${{ steps.changed-files.outputs.wasm_bench_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -64,7 +64,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -78,11 +78,16 @@ jobs:
     needs: setup-instance
     if: needs.setup-instance.result != 'skipped'
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    strategy:
+      max-parallel: 1
+      matrix:
+        browser: [ chrome, firefox ]
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -92,20 +97,46 @@ jobs:
             echo "COMMIT_HASH=$(git describe --tags --dirty)";
           } >> "${GITHUB_ENV}"
 
-      - name: Set up home
-        # "Install rust" step require root user to have a HOME directory which is not set.
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
-
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
-      - name: Run benchmarks
+      - name: Get Node version
+        run: |
+          echo "NODE_VERSION=$(make node_version)" >> "${GITHUB_ENV}"
+
+      - name: Node cache restoration
+        id: node-cache
+        uses: actions/cache/restore@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
+        with:
+          path: |
+            ~/.nvm
+            ~/.npm
+          key: node-${{ env.NODE_VERSION }}
+
+      - name: Install Node
+        if: steps.node-cache.outputs.cache-hit != 'true'
         run: |
           make install_node
-          make bench_web_js_api_parallel_ci
+
+      - name: Node cache save
+        uses: actions/cache/save@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
+        if: steps.node-cache.outputs.cache-hit != 'true'
+        with:
+          path: |
+            ~/.nvm
+            ~/.npm
+          key: node-${{ env.NODE_VERSION }}
+
+      - name: Install web resources
+        run: |
+          make install_${{ matrix.browser }}_browser
+          make install_${{ matrix.browser }}_web_driver
+
+      - name: Run benchmarks
+        run: |
+          make bench_web_js_api_parallel_${{ matrix.browser }}_ci
 
       - name: Parse results
         run: |
@@ -118,25 +149,29 @@ jobs:
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --key-gen
+          rm tfhe/wasm_pk_gen.csv
 
+      # Run these benchmarks only once
       - name: Measure public key and ciphertext sizes in HL Api
+        if:  matrix.browser == 'chrome'
         run: |
           make measure_hlapi_compact_pk_ct_sizes
 
       - name: Parse key and ciphertext sizes results
+        if:  matrix.browser == 'chrome'
         run: |
           python3 ./ci/benchmark_parser.py tfhe/hlapi_cpk_and_cctl_sizes.csv ${{ env.RESULTS_FILENAME }} \
           --key-gen \
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
-          name: ${{ github.sha }}_wasm
+          name: ${{ github.sha }}_wasm_${{ matrix.browser }}
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -145,24 +180,16 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "WASM benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_MESSAGE: "WASM benchmarks (${{ matrix.browser }}) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
     name: Teardown instance (wasm-client-benchmarks)
@@ -172,7 +199,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -183,7 +210,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (wasm-client-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_zk_pke.yml

@@ -30,19 +30,20 @@ jobs:
       zk_pok_changed: ${{ steps.changed-files.outputs.zk_pok_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             zk_pok:
               - tfhe/Cargo.toml
               - concrete-csprng/**
+              - tfhe-fft/**
               - tfhe-zk-pok/**
               - tfhe/src/core_crypto/**
               - tfhe/src/shortint/**
@@ -65,7 +66,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -79,14 +80,15 @@ jobs:
     if: needs.setup-instance.result != 'skipped'
     needs: setup-instance
     concurrency:
-      group: ${{ github.workflow }}_${{github.event_name}}_${{ github.ref }}
+      group: ${{ github.workflow }}_${{github.event_name}}_${{ github.ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
       cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -96,18 +98,13 @@ jobs:
             echo "COMMIT_HASH=$(git describe --tags --dirty)";
           } >> "${GITHUB_ENV}"
 
-      - name: Set up home
-        # "Install rust" step require root user to have a HOME directory which is not set.
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
-
       - name: Install rust
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -134,17 +131,17 @@ jobs:
       - name: Parse CRS sizes results
         run: |
           python3 ./ci/benchmark_parser.py tfhe/pke_zk_crs_sizes.csv ${{ env.RESULTS_FILENAME }} \
-          --key-sizes \
+          --object-sizes \
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_integer_zk
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -153,21 +150,13 @@ jobs:
       - name: Send data to Slab
         shell: bash
         run: |
-          echo "Computing HMac on results file"
-          SIGNATURE="$(slab/scripts/hmac_calculator.sh ${{ env.RESULTS_FILENAME }} '${{ secrets.JOB_SECRET }}')"
-          echo "Sending results to Slab..."
-          curl -v -k \
-          -H "Content-Type: application/json" \
-          -H "X-Slab-Repository: ${{ github.repository }}" \
-          -H "X-Slab-Command: store_data_v2" \
-          -H "X-Hub-Signature-256: sha256=${SIGNATURE}" \
-          -d @${{ env.RESULTS_FILENAME }} \
-          ${{ secrets.SLAB_URL }}
+          python3 slab/scripts/data_sender.py ${{ env.RESULTS_FILENAME }} "${{ secrets.JOB_SECRET }}" \
+          --slab-url "${{ secrets.SLAB_URL }}"
 
       - name: Slack Notification
-        if: ${{ !success() && !cancelled() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "PKE ZK benchmarks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -180,7 +169,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -191,7 +180,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (pke-zk-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/cargo_build.yml

@@ -1,4 +1,4 @@
-name: Cargo Build
+name: Cargo Build TFHE-rs
 
 on:
   pull_request:
@@ -19,14 +19,21 @@ jobs:
 
     strategy:
       matrix:
-        os: [large_ubuntu_16, macos-latest-large, large_windows_16_latest]
+        # GitHub macos-latest are now M1 macs, so use ours, we limit what runs so it will be fast
+        # even with a few PRs
+        os: [large_ubuntu_16, macos-latest, windows-latest]
       fail-fast: false
 
     steps:
-      - uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+
+      - name: Install latest stable
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: stable
 
       - name: Install and run newline linter checks
-        if: matrix.os == 'ubuntu-latest'
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           wget https://github.com/fernandrone/linelint/releases/download/0.0.6/linelint-linux-amd64
           echo "16b70fb7b471d6f95cbdc0b4e5dc2b0ac9e84ba9ecdc488f7bdf13df823aca4b linelint-linux-amd64" > checksum
@@ -36,27 +43,33 @@ jobs:
           make check_newline
 
       - name: Run pcc checks
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           make pcc
 
       - name: Build concrete-csprng
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           make build_concrete_csprng
 
       - name: Build Release core
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           make build_core AVX512_SUPPORT=ON
           make build_core_experimental AVX512_SUPPORT=ON
 
       - name: Build Release boolean
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           make build_boolean
 
       - name: Build Release shortint
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           make build_shortint
 
       - name: Build Release integer
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           make build_integer
 
@@ -65,10 +78,12 @@ jobs:
           make build_tfhe_full
 
       - name: Build Release c_api
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           make build_c_api
 
       - name: Build coverage tests
+        if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
           make build_tfhe_coverage
 

.github/workflows/cargo_build_tfhe_fft.yml

@@ -0,0 +1,47 @@
+# Build tfhe-fft
+name: Cargo Build tfhe-fft
+
+on:
+  push:
+    branches:
+      - 'main'
+  pull_request:
+
+env:
+  CARGO_TERM_COLOR: always
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref }}
+  cancel-in-progress: true
+
+jobs:
+  cargo-builds-fft:
+    runs-on: ${{ matrix.runner_type }}
+
+    strategy:
+      matrix:
+        runner_type: [ubuntu-latest, macos-latest, windows-latest]
+      fail-fast: false
+
+    steps:
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
+
+      - name: Install Rust
+        uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
+        with:
+          toolchain: stable
+          override: true
+
+      - name: Run pcc checks
+        if: matrix.runner_type == 'ubuntu-latest'
+        run: |
+          sudo apt install -y libfftw3-dev
+          make pcc_fft
+
+      - name: Build release
+        run: |
+          make build_fft
+
+      - name: Build release no-std
+        run: |
+          make build_fft_no_std

.github/workflows/cargo_test_fft.yml

@@ -0,0 +1,74 @@
+# Test tfhe-fft
+name: Cargo Test tfhe-fft
+
+on:
+  push:
+    branches:
+      - "main"
+  pull_request:
+
+env:
+  CARGO_TERM_COLOR: always
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref }}
+  cancel-in-progress: true
+
+jobs:
+  cargo-tests:
+    runs-on: ${{ matrix.runner_type }}
+    strategy:
+      matrix:
+        runner_type: [ubuntu-latest, macos-latest, windows-latest]
+      fail-fast: false
+    steps:
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
+
+      - name: Install Rust
+        uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
+        with:
+          toolchain: stable
+          override: true
+
+      - name: Test debug
+        run: |
+          make test_fft
+
+      - name: Test serialization
+        run: make test_fft_serde
+
+      - name: Test no-std
+        run: |
+          make test_fft_no_std
+
+  cargo-tests-nightly:
+    runs-on: ${{ matrix.runner_type }}
+    strategy:
+      matrix:
+        runner_type: [ubuntu-latest, macos-latest, windows-latest]
+    steps:
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
+
+      - name: Install Rust
+        uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
+        with:
+          toolchain: nightly
+          override: true
+
+      - name: Test nightly
+        run: |
+          make test_fft_nightly
+
+      - name: Test no-std nightly
+        run: |
+          make test_fft_no_std_nightly
+
+  cargo-tests-node-js:
+    runs-on: "ubuntu-latest"
+    steps:
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
+
+      - name: Test node js
+        run: |
+          make install_node
+          make test_fft_node_js_ci

.github/workflows/check_commit.yml

@@ -28,3 +28,6 @@ jobs:
           excludeTitle: "true" # optional: this excludes the title of a pull request
           checkAllCommitMessages: "true" # optional: this checks all commits associated with a pull request
           accessToken: ${{ secrets.GITHUB_TOKEN }} # github access token is only required if checkAllCommitMessages is true
+
+      - name: Check commit signatures
+        uses: 1Password/check-signed-commits-action@ed2885f3ed2577a4f5d3c3fe895432a557d23d52

.github/workflows/ci_lint.yml

@@ -13,7 +13,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
 
       - name: Get actionlint
         run: |
@@ -25,3 +25,9 @@ jobs:
       - name: Lint workflows
         run: |
           make lint_workflow
+
+      - name: Ensure SHA pinned actions
+        uses: zgosalvez/github-actions-ensure-sha-pinned-actions@38608ef4fb69adae7f1eac6eeb88e67b7d083bfd # v3.0.16
+        with:
+          allowlist: |
+            slsa-framework/slsa-github-generator

.github/workflows/code_coverage.yml

@@ -25,7 +25,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -44,20 +44,16 @@ jobs:
     timeout-minutes: 5760 # 4 days
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           files_yaml: |
             tfhe:
@@ -87,7 +83,7 @@ jobs:
           make test_shortint_cov
 
       - name: Upload tfhe coverage to Codecov
-        uses: codecov/codecov-action@e28ff129e5465c2c0dcc6f003fc735cb6ae0c673
+        uses: codecov/codecov-action@5c47607acb93fed5485fdbf7232e8a31425f672a
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
@@ -101,7 +97,7 @@ jobs:
           make test_integer_cov
 
       - name: Upload tfhe coverage to Codecov
-        uses: codecov/codecov-action@e28ff129e5465c2c0dcc6f003fc735cb6ae0c673
+        uses: codecov/codecov-action@5c47607acb93fed5485fdbf7232e8a31425f672a
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
@@ -112,7 +108,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Code coverage finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -125,7 +121,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -136,7 +132,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (code-coverage) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/csprng_randomness_tests.yml

@@ -27,7 +27,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -45,16 +45,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -65,7 +62,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "concrete-csprng randomness check finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -78,7 +75,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -89,7 +86,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (csprng-randomness-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/data_pr_close.yml

@@ -117,7 +117,7 @@ jobs:
     - name: Slack Notification
       if: ${{ always() && job.status == 'failure' }}
       continue-on-error: true
-      uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+      uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
       env:
         SLACK_COLOR: ${{ job.status }}
         SLACK_MESSAGE: "Failed to auto-${{ env.CLOSE_TYPE }} PR on data repo: ${{ fromJson(env.GH_API_RES || env.TARGET_REPO_PR).message }}"

.github/workflows/gpu_4090_tests.yml

@@ -34,12 +34,13 @@ jobs:
 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -74,9 +75,9 @@ jobs:
           github_token: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Slack Notification
-        if: ${{ always() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "CUDA RTX 4090 tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_fast_h100_tests.yml

@@ -1,5 +1,5 @@
 # Compile and test tfhe-cuda-backend on an H100 VM on hyperstack
-name: TFHE Cuda Backend - Base tests on H100
+name: TFHE Cuda Backend - Fast tests on H100
 
 env:
   CARGO_TERM_COLOR: always
@@ -28,13 +28,13 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -44,14 +44,16 @@ jobs:
               - backends/tfhe-cuda-backend/**
               - tfhe/src/core_crypto/gpu/**
               - tfhe/src/integer/gpu/**
-              - tfhe/shortint/parameters/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
+              - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
               - 'tfhe/docs/**.md'
-              - Makefile
-              - '.github/workflows/hyperstack**'
-              - scripts/**
-              - ci/**
+              - '.github/workflows/gpu_fast_h100_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
 
   setup-instance:
     name: Setup instance (cuda-h100-tests)
@@ -65,7 +67,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -99,7 +101,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev 
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -108,14 +110,17 @@ jobs:
           sudo make install
 
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          persist-credentials: 'false'
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -144,33 +149,34 @@ jobs:
 
       - name: Run core crypto and internal CUDA backend tests
         run: |
-          make test_core_crypto_gpu
-          make test_cuda_backend
+          BIG_TESTS_INSTANCE=TRUE make test_core_crypto_gpu
+          BIG_TESTS_INSTANCE=TRUE make test_integer_compression_gpu
+          BIG_TESTS_INSTANCE=TRUE make test_cuda_backend
 
       - name: Run user docs tests
         run: |
-          make test_user_doc_gpu
+          BIG_TESTS_INSTANCE=TRUE make test_user_doc_gpu
 
       - name: Test C API
         run: |
-          make test_c_api_gpu
+          BIG_TESTS_INSTANCE=TRUE make test_c_api_gpu
 
       - name: Run High Level API Tests
         run: |
-          make test_high_level_api_gpu
+          BIG_TESTS_INSTANCE=TRUE make test_high_level_api_gpu
 
   slack-notify:
     name: Slack Notification
     needs: [ setup-instance, cuda-tests-linux ]
     runs-on: ubuntu-latest
-    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' }}
+    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
-          SLACK_MESSAGE: "Base H100 tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_MESSAGE: "Fast H100 tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
     name: Teardown instance (cuda-h100-tests)
@@ -180,7 +186,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -191,7 +197,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-h100-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_fast_tests.yml

@@ -27,13 +27,13 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -42,11 +42,17 @@ jobs:
               - tfhe/build.rs
               - backends/tfhe-cuda-backend/**
               - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
               - tfhe/src/integer/gpu/**
-              - tfhe/shortint/parameters/**
+              - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
               - 'tfhe/docs/**.md'
+              - '.github/workflows/gpu_fast_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
 
   setup-instance:
     name: Setup instance (cuda-tests)
@@ -59,13 +65,13 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
           slab-url: ${{ secrets.SLAB_BASE_URL }}
           job-secret: ${{ secrets.JOB_SECRET }}
-          backend: aws
+          backend: hyperstack
           profile: gpu-test
 
   cuda-tests-linux:
@@ -84,22 +90,35 @@ jobs:
         include:
           - os: ubuntu-22.04
             cuda: "12.2"
-            gcc: 9
+            gcc: 11 
     env:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
-
+      CMAKE_VERSION: 3.29.6
     steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -122,9 +141,14 @@ jobs:
             echo "HOME=/home/ubuntu";
           } >> "${GITHUB_ENV}"
 
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
       - name: Run core crypto and internal CUDA backend tests
         run: |
           make test_core_crypto_gpu
+          make test_integer_compression_gpu
           make test_cuda_backend
 
       - name: Run user docs tests
@@ -139,13 +163,18 @@ jobs:
         run: |
           make test_high_level_api_gpu
 
-      - name: Slack Notification
-        if: ${{ always() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-tests-linux ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "CUDA AWS tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
+          SLACK_MESSAGE: "Base GPU tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
     name: Teardown instance (cuda-tests)
@@ -155,7 +184,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -166,7 +195,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_full_h100_tests.yml

@@ -0,0 +1,156 @@
+# Compile and test tfhe-cuda-backend on an H100 VM on hyperstack
+name: TFHE Cuda Backend - Full tests on H100
+
+env:
+  CARGO_TERM_COLOR: always
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUSTFLAGS: "-C target-cpu=native"
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+  IS_PULL_REQUEST: ${{ github.event_name == 'pull_request' }}
+
+on:
+  workflow_dispatch:
+
+jobs:
+  setup-instance:
+    name: Setup instance (cuda-h100-tests)
+    runs-on: ubuntu-latest
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: hyperstack
+          profile: single-h100
+
+  cuda-tests-linux:
+    name: CUDA H100 tests
+    needs: [ setup-instance ]
+    concurrency:
+      group: ${{ github.workflow }}_${{ github.ref }}
+      cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    strategy:
+      fail-fast: false
+      # explicit include-based build matrix, of known valid options
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
+    steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          persist-credentials: 'false'
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Set up home
+        run: |
+          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+
+      - name: Install latest stable
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: stable
+
+      - name: Export CUDA variables
+        if: ${{ !cancelled() }}
+        run: |
+          echo "CUDA_PATH=$CUDA_PATH" >> "${GITHUB_ENV}"
+          echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
+          echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH" >> "${GITHUB_ENV}"
+          echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc" >> "${GITHUB_ENV}"
+
+      # Specify the correct host compilers
+      - name: Export gcc and g++ variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CC=/usr/bin/gcc-${{ matrix.gcc }}";
+            echo "CXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "HOME=/home/ubuntu";
+          } >> "${GITHUB_ENV}"
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run core crypto, integer and internal CUDA backend tests
+        run: |
+          make test_gpu
+
+      - name: Run user docs tests
+        run: |
+          make test_user_doc_gpu
+
+      - name: Test C API
+        run: |
+          make test_c_api_gpu
+
+      - name: Run High Level API Tests
+        run: |
+          make test_high_level_api_gpu
+
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-tests-linux ]
+    runs-on: ubuntu-latest
+    if: ${{ failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
+          SLACK_MESSAGE: "Full H100 tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-h100-tests)
+    needs: [ setup-instance, cuda-tests-linux ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (cuda-h100-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_full_multi_gpu_tests.yml

@@ -28,13 +28,13 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -43,15 +43,17 @@ jobs:
               - tfhe/build.rs
               - backends/tfhe-cuda-backend/**
               - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
               - tfhe/src/integer/gpu/**
-              - tfhe/shortint/parameters/**
+              - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
               - 'tfhe/docs/**.md'
-              - Makefile
-              - '.github/workflows/aws_tfhe_multi_gpu**'
-              - scripts/**
-              - ci/**
+              - '.github/workflows/**_multi_gpu_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
 
   setup-instance:
     name: Setup instance (cuda-tests-multi-gpu)
@@ -65,13 +67,13 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
           slab-url: ${{ secrets.SLAB_BASE_URL }}
           job-secret: ${{ secrets.JOB_SECRET }}
-          backend: aws
+          backend: hyperstack
           profile: multi-gpu-test
 
   cuda-tests-linux:
@@ -90,20 +92,35 @@ jobs:
         include:
           - os: ubuntu-22.04
             cuda: "12.2"
-            gcc: 9
+            gcc: 11 
     env:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
-
+      CMAKE_VERSION: 3.29.6
     steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          persist-credentials: 'false'
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -126,30 +143,43 @@ jobs:
             echo "HOME=/home/ubuntu";
           } >> "${GITHUB_ENV}"
 
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run multi-bit CUDA integer compression tests
+        run: |
+          BIG_TESTS_INSTANCE=TRUE make test_integer_compression_gpu
+
       # No need to test core_crypto and classic PBS in integer since it's already tested on single GPU.
       - name: Run multi-bit CUDA integer tests
         run: |
-          make test_integer_multi_bit_gpu_ci
+          BIG_TESTS_INSTANCE=TRUE make test_integer_multi_bit_gpu_ci
 
       - name: Run user docs tests
         run: |
-          make test_user_doc_gpu
+          BIG_TESTS_INSTANCE=TRUE make test_user_doc_gpu
 
       - name: Test C API
         run: |
-          make test_c_api_gpu
+          BIG_TESTS_INSTANCE=TRUE make test_c_api_gpu
 
       - name: Run High Level API Tests
         run: |
-          make test_high_level_api_gpu
+          BIG_TESTS_INSTANCE=TRUE make test_high_level_api_gpu
 
-      - name: Slack Notification
-        if: ${{ always() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-tests-linux ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "CUDA AWS multi-GPU tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
+          SLACK_MESSAGE: "Multi-GPU tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
     name: Teardown instance (cuda-tests-multi-gpu)
@@ -159,7 +189,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -170,7 +200,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-tests-multi-gpu) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_pcc.yml

@@ -24,7 +24,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -53,16 +53,17 @@ jobs:
 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -94,9 +95,9 @@ jobs:
           make pcc_gpu
 
       - name: Slack Notification
-        if: ${{ always() }}
+        if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "CUDA AWS post-commit checks finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -109,7 +110,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -120,7 +121,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-pcc) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_signed_integer_classic_tests.yml

@@ -0,0 +1,188 @@
+# Signed integer GPU tests on an RTXA6000 VM on hyperstack with classical PBS
+name: TFHE Cuda Backend - Signed integer tests with classical PBS
+
+env:
+  CARGO_TERM_COLOR: always
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUSTFLAGS: "-C target-cpu=native"
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+  IS_PULL_REQUEST: ${{ github.event_name == 'pull_request' }}
+
+on:
+  # Allows you to run this workflow manually from the Actions tab as an alternative.
+  workflow_dispatch:
+  pull_request:
+      types: [ labeled ]
+
+jobs:
+  should-run:
+    runs-on: ubuntu-latest
+    permissions:
+      pull-requests: write
+    outputs:
+      gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
+    steps:
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Check for file changes
+        id: changed-files
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
+        with:
+          since_last_remote_commit: true
+          files_yaml: |
+            gpu:
+              - tfhe/Cargo.toml
+              - tfhe/build.rs
+              - backends/tfhe-cuda-backend/**
+              - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
+              - tfhe/src/integer/gpu/**
+              - tfhe/src/shortint/parameters/**
+              - tfhe/src/high_level_api/**
+              - tfhe/src/c_api/**
+              - 'tfhe/docs/**.md'
+              - '.github/workflows/gpu_signed_integer_classic_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
+
+  setup-instance:
+    name: Setup instance (cuda-signed-classic-tests)
+    needs: should-run
+    if: github.event_name != 'pull_request' ||
+      (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true') ||
+      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true')
+    runs-on: ubuntu-latest
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: hyperstack
+          profile: gpu-test
+
+  cuda-tests-linux:
+    name: CUDA signed integer tests with classical PBS
+    needs: [ should-run, setup-instance ]
+    if: github.event_name != 'pull_request' ||
+      (github.event_name == 'pull_request' && needs.setup-instance.result != 'skipped')
+    concurrency:
+      group: ${{ github.workflow }}_${{ github.ref }}
+      cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    strategy:
+      fail-fast: false
+      # explicit include-based build matrix, of known valid options
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
+    steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+
+      - name: Set up home
+        run: |
+          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+
+      - name: Install latest stable
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: stable
+
+      - name: Export CUDA variables
+        if: ${{ !cancelled() }}
+        run: |
+          echo "CUDA_PATH=$CUDA_PATH" >> "${GITHUB_ENV}"
+          echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
+          echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH" >> "${GITHUB_ENV}"
+          echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc" >> "${GITHUB_ENV}"
+
+      # Specify the correct host compilers
+      - name: Export gcc and g++ variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CC=/usr/bin/gcc-${{ matrix.gcc }}";
+            echo "CXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "HOME=/home/ubuntu";
+          } >> "${GITHUB_ENV}"
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run signed integer tests
+        run: |
+          BIG_TESTS_INSTANCE=TRUE make test_signed_integer_gpu_ci
+
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-tests-linux ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
+          SLACK_MESSAGE: "Integer GPU signed integer tests with classical PBS finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-signed-classic-tests)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-tests-linux ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (cuda-signed-classic-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_signed_integer_h100_tests.yml

@@ -28,13 +28,14 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -43,15 +44,17 @@ jobs:
               - tfhe/build.rs
               - backends/tfhe-cuda-backend/**
               - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
               - tfhe/src/integer/gpu/**
-              - tfhe/shortint/parameters/**
+              - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
               - 'tfhe/docs/**.md'
-              - Makefile
-              - '.github/workflows/hyperstack**'
-              - scripts/**
-              - ci/**
+              - '.github/workflows/gpu_signed_integer_h100_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
 
   setup-instance:
     name: Setup instance (cuda-h100-tests)
@@ -65,7 +68,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -99,7 +102,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev 
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -107,15 +110,16 @@ jobs:
           make -j"$(nproc)"
           sudo make install
 
+
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -142,23 +146,19 @@ jobs:
         if: ${{ !cancelled() }}
         run: nvidia-smi
 
-      - name: Run signed integer tests
-        run: |
-          make test_signed_integer_gpu_ci
-
       - name: Run signed integer multi-bit tests
         run: |
-          make test_signed_integer_multi_bit_gpu_ci
+          BIG_TESTS_INSTANCE=TRUE make test_signed_integer_multi_bit_gpu_ci
 
   slack-notify:
     name: Slack Notification
     needs: [ setup-instance, cuda-tests-linux ]
     runs-on: ubuntu-latest
-    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' }}
+    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
           SLACK_MESSAGE: "Integer GPU H100 tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
@@ -171,7 +171,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -182,7 +182,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-h100-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_signed_integer_tests.yml

@@ -22,7 +22,6 @@ on:
     types:
       - opened
       - synchronize
-      - labeled
   schedule:
     # Nightly tests @ 1AM after each work day
     - cron: "0 1 * * MON-FRI"
@@ -36,13 +35,13 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -51,11 +50,17 @@ jobs:
               - tfhe/build.rs
               - backends/tfhe-cuda-backend/**
               - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
               - tfhe/src/integer/gpu/**
-              - tfhe/shortint/parameters/**
+              - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
               - 'tfhe/docs/**.md'
+              - '.github/workflows/gpu_signed_integer_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
 
   setup-instance:
     name: Setup instance (cuda-signed-integer-tests)
@@ -63,19 +68,19 @@ jobs:
     needs: should-run
     if: (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
       github.event_name == 'workflow_dispatch' ||
-      (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true')
+      needs.should-run.outputs.gpu_test == 'true'
     outputs:
       runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
           slab-url: ${{ secrets.SLAB_BASE_URL }}
           job-secret: ${{ secrets.JOB_SECRET }}
-          backend: aws
+          backend: hyperstack
           profile: gpu-test
 
   cuda-signed-integer-tests:
@@ -94,20 +99,36 @@ jobs:
         include:
           - os: ubuntu-22.04
             cuda: "12.2"
-            gcc: 9
+            gcc: 11 
     env:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
-
+      CMAKE_VERSION: 3.29.6
     steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          persist-credentials: 'false'
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -138,17 +159,26 @@ jobs:
             echo "NIGHTLY_TESTS=TRUE";
           } >> "${GITHUB_ENV}"
 
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
       - name: Run signed integer multi-bit tests
         run: |
           make test_signed_integer_multi_bit_gpu_ci
 
-      - name: Slack Notification
-        if: ${{ always() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-signed-integer-tests ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-signed-integer-tests.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "CUDA AWS signed integer tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_COLOR: ${{ needs.cuda-signed-integer-tests.result }}
+          SLACK_MESSAGE: "Base GPU tests finished with status: ${{ needs.cuda-signed-integer-tests.result }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
     name: Teardown instance (cuda-tests)
@@ -158,7 +188,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -169,7 +199,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-signed-integer-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_unsigned_integer_classic_tests.yml

@@ -0,0 +1,188 @@
+# Test unsigned integers on an RTXA6000 VM on hyperstack with the classical PBS
+name: TFHE Cuda Backend - Unsigned integer tests with classical PBS
+
+env:
+  CARGO_TERM_COLOR: always
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+  RUSTFLAGS: "-C target-cpu=native"
+  RUST_BACKTRACE: "full"
+  RUST_MIN_STACK: "8388608"
+  SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+  SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+  SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+  SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+  IS_PULL_REQUEST: ${{ github.event_name == 'pull_request' }}
+
+on:
+  # Allows you to run this workflow manually from the Actions tab as an alternative.
+  workflow_dispatch:
+  pull_request:
+      types: [ labeled ]
+
+jobs:
+  should-run:
+    runs-on: ubuntu-latest
+    permissions:
+      pull-requests: write
+    outputs:
+      gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
+    steps:
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Check for file changes
+        id: changed-files
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
+        with:
+          since_last_remote_commit: true
+          files_yaml: |
+            gpu:
+              - tfhe/Cargo.toml
+              - tfhe/build.rs
+              - backends/tfhe-cuda-backend/**
+              - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
+              - tfhe/src/integer/gpu/**
+              - tfhe/src/shortint/parameters/**
+              - tfhe/src/high_level_api/**
+              - tfhe/src/c_api/**
+              - 'tfhe/docs/**.md'
+              - '.github/workflows/gpu_unsigned_integer_classic_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
+
+  setup-instance:
+    name: Setup instance (cuda-unsigned-classic-tests)
+    needs: should-run
+    if: github.event_name != 'pull_request' ||
+      (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true') ||
+      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true')
+    runs-on: ubuntu-latest
+    outputs:
+      runner-name: ${{ steps.start-instance.outputs.label }}
+    steps:
+      - name: Start instance
+        id: start-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: start
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          backend: hyperstack
+          profile: gpu-test
+
+  cuda-tests-linux:
+    name: CUDA unsigned integer tests with classical PBS
+    needs: [ should-run, setup-instance ]
+    if: github.event_name != 'pull_request' ||
+      (github.event_name == 'pull_request' && needs.setup-instance.result != 'skipped')
+    concurrency:
+      group: ${{ github.workflow }}_${{ github.ref }}
+      cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    strategy:
+      fail-fast: false
+      # explicit include-based build matrix, of known valid options
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
+    steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+
+      - name: Set up home
+        run: |
+          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
+
+      - name: Install latest stable
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
+        with:
+          toolchain: stable
+
+      - name: Export CUDA variables
+        if: ${{ !cancelled() }}
+        run: |
+          echo "CUDA_PATH=$CUDA_PATH" >> "${GITHUB_ENV}"
+          echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
+          echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH" >> "${GITHUB_ENV}"
+          echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc" >> "${GITHUB_ENV}"
+
+      # Specify the correct host compilers
+      - name: Export gcc and g++ variables
+        if: ${{ !cancelled() }}
+        run: |
+          {
+            echo "CC=/usr/bin/gcc-${{ matrix.gcc }}";
+            echo "CXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
+            echo "HOME=/home/ubuntu";
+          } >> "${GITHUB_ENV}"
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run unsigned integer tests
+        run: |
+          BIG_TESTS_INSTANCE=TRUE make test_unsigned_integer_gpu_ci
+
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-tests-linux ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
+          SLACK_MESSAGE: "Unsigned integer GPU classic tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-unsigned-classic-tests)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-tests-linux ]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Stop instance
+        id: stop-instance
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
+        with:
+          mode: stop
+          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
+          slab-url: ${{ secrets.SLAB_BASE_URL }}
+          job-secret: ${{ secrets.JOB_SECRET }}
+          label: ${{ needs.setup-instance.outputs.runner-name }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_MESSAGE: "Instance teardown (cuda-unsigned-classic-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_unsigned_integer_h100_tests.yml

@@ -28,13 +28,14 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -43,15 +44,17 @@ jobs:
               - tfhe/build.rs
               - backends/tfhe-cuda-backend/**
               - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
               - tfhe/src/integer/gpu/**
-              - tfhe/shortint/parameters/**
+              - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
               - 'tfhe/docs/**.md'
-              - Makefile
-              - '.github/workflows/hyperstack**'
-              - scripts/**
-              - ci/**
+              - '.github/workflows/gpu_unsigned_integer_h100_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
 
   setup-instance:
     name: Setup instance (cuda-h100-tests)
@@ -65,7 +68,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -99,7 +102,7 @@ jobs:
       - name: Install dependencies
         run: |
           sudo apt update
-          sudo apt install -y checkinstall zlib1g-dev libssl-dev 
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
           wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
           tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
           cd cmake-${{ env.CMAKE_VERSION }}
@@ -107,15 +110,16 @@ jobs:
           make -j"$(nproc)"
           sudo make install
 
+
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -142,23 +146,19 @@ jobs:
         if: ${{ !cancelled() }}
         run: nvidia-smi
 
-      - name: Run unsigned integer tests
-        run: |
-          make test_unsigned_integer_gpu_ci
-
       - name: Run unsigned integer multi-bit tests
         run: |
-          make test_unsigned_integer_multi_bit_gpu_ci
+          BIG_TESTS_INSTANCE=TRUE make test_unsigned_integer_multi_bit_gpu_ci
 
   slack-notify:
     name: Slack Notification
     needs: [ setup-instance, cuda-tests-linux ]
     runs-on: ubuntu-latest
-    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' }}
+    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
     continue-on-error: true
     steps:
       - name: Send message
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
           SLACK_MESSAGE: "Unsigned integer GPU H100 tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.ACTION_RUN_URL }})"
@@ -171,7 +171,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -182,7 +182,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-h100-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_unsigned_integer_tests.yml

@@ -21,7 +21,6 @@ on:
     types:
       - opened
       - synchronize
-      - labeled
   schedule:
     # Nightly tests @ 1AM after each work day
     - cron: "0 1 * * MON-FRI"
@@ -35,13 +34,14 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -50,31 +50,37 @@ jobs:
               - tfhe/build.rs
               - backends/tfhe-cuda-backend/**
               - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_unsigned/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_signed/**
+              - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
               - tfhe/src/integer/gpu/**
-              - tfhe/shortint/parameters/**
+              - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
               - 'tfhe/docs/**.md'
+              - '.github/workflows/gpu_unsigned_integer_tests.yml'
+              - scripts/integer-tests.sh
+              - ci/slab.toml
 
   setup-instance:
     name: Setup instance (cuda-unsigned-integer-tests)
     needs: should-run
     if: (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
       github.event_name == 'workflow_dispatch' ||
-      (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true')
+      needs.should-run.outputs.gpu_test == 'true'
     runs-on: ubuntu-latest
     outputs:
       runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
           slab-url: ${{ secrets.SLAB_BASE_URL }}
           job-secret: ${{ secrets.JOB_SECRET }}
-          backend: aws
+          backend: hyperstack
           profile: gpu-test
 
   cuda-unsigned-integer-tests:
@@ -93,20 +99,33 @@ jobs:
         include:
           - os: ubuntu-22.04
             cuda: "12.2"
-            gcc: 9
+            gcc: 11
     env:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
-
+      CMAKE_VERSION: 3.29.6
     steps:
+      # Mandatory on hyperstack since a bootable volume is not re-usable yet.
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ env.CMAKE_VERSION }}/cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          tar -zxvf cmake-${{ env.CMAKE_VERSION }}.tar.gz
+          cd cmake-${{ env.CMAKE_VERSION }}
+          ./bootstrap
+          make -j"$(nproc)"
+          sudo make install
+
+
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -137,17 +156,26 @@ jobs:
             echo "NIGHTLY_TESTS=TRUE";
           } >> "${GITHUB_ENV}"
 
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
       - name: Run unsigned integer multi-bit tests
         run: |
           make test_unsigned_integer_multi_bit_gpu_ci
 
-      - name: Slack Notification
-        if: ${{ always() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+  slack-notify:
+    name: Slack Notification
+    needs: [ setup-instance, cuda-unsigned-integer-tests ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-unsigned-integer-tests.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "CUDA AWS unsigned integer tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_COLOR: ${{ needs.cuda-unsigned-integer-tests.result }}
+          SLACK_MESSAGE: "Unsigned integer GPU tests finished with status: ${{ needs.cuda-unsigned-integer-tests.result }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
     name: Teardown instance (cuda-tests)
@@ -157,7 +185,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -168,7 +196,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (cuda-unsigned-integer-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/m1_tests.yml

@@ -3,7 +3,7 @@ name: Tests on M1 CPU
 on:
   workflow_dispatch:
   pull_request:
-    types: [ labeled ]
+    types: [labeled]
   # Have a nightly build for M1 tests
   schedule:
     # * is a special character in YAML so you have to quote this string
@@ -27,22 +27,43 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  cargo-builds:
+  cargo-builds-m1:
     if: ${{ (github.event_name == 'schedule' &&  github.repository == 'zama-ai/tfhe-rs') || github.event_name == 'workflow_dispatch' || contains(github.event.label.name, 'm1_test') }}
     runs-on: ["self-hosted", "m1mac"]
     # 12 hours, default is 6 hours, hopefully this is more than enough
     timeout-minutes: 720
 
     steps:
-      - uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
-          persist-credentials: 'false'
+          persist-credentials: "false"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
+      - name: Run pcc FFT checks
+        run: |
+          make pcc_fft
+
+      - name: Build FFT release
+        run: |
+          make build_fft
+
+      - name: Build FFT release no-std
+        run: |
+          make build_fft_no_std
+
+      - name: Run FFT tests
+        run: |
+          make test_fft
+          make test_fft_serde
+          make test_fft_nightly
+          make test_fft_no_std
+          make test_fft_no_std_nightly
+          # we don't run the js stuff here as it's causing issues with the M1 config
+
       - name: Run pcc checks
         run: |
           make pcc
@@ -137,7 +158,7 @@ jobs:
     name: Remove m1_test label
     runs-on: ubuntu-latest
     needs:
-      - cargo-builds
+      - cargo-builds-m1
     if: ${{ always() }}
     steps:
       - uses: actions-ecosystem/action-remove-labels@2ce5d41b4b6aa8503e285553f75ed56e0a40bae0
@@ -147,13 +168,13 @@ jobs:
           github_token: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Slack Notification
-        if: ${{ needs.cargo-builds.result != 'skipped' }}
+        if: ${{ needs.cargo-builds-m1.result != 'skipped' }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
-          SLACK_COLOR: ${{ needs.cargo-builds.result }}
+          SLACK_COLOR: ${{ needs.cargo-builds-m1.result }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
           SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "M1 tests finished with status: ${{ needs.cargo-builds.result }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_MESSAGE: "M1 tests finished with status: ${{ needs.cargo-builds-m1.result }}. (${{ env.ACTION_RUN_URL }})"
           SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
           SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}

.github/workflows/make_release.yml

@@ -30,19 +30,26 @@ env:
   NPM_TAG: ""
 
 jobs:
+  verify_tag:
+    uses: ./.github/workflows/verify_tagged_commit.yml
+    secrets:
+      RELEASE_TEAM: ${{ secrets.RELEASE_TEAM }}
+      READ_ORG_TOKEN: ${{ secrets.READ_ORG_TOKEN }}
+
   package:
     runs-on: ubuntu-latest
+    needs: verify_tag
     outputs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
       - name: Prepare package
         run: |
           cargo package -p tfhe
-      - uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b # v4.3.4
+      - uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882 # v4.4.3
         with:
           name: crate
           path: target/package/*.crate
@@ -74,7 +81,7 @@ jobs:
       id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
       - name: Create NPM version tag
@@ -101,7 +108,7 @@ jobs:
       - name: Slack notification (hashes comparison)
         if: ${{ needs.package.outputs.hash != steps.published_hash.outputs.pub_hash }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: failure
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
@@ -146,7 +153,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}

.github/workflows/make_release_concrete_csprng.yml

@@ -1,4 +1,3 @@
-# Publish new release of tfhe-rs on various platform.
 name: Publish concrete-csprng release
 
 on:
@@ -13,12 +12,19 @@ env:
   ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
 
 jobs:
+  verify_tag:
+    uses: ./.github/workflows/verify_tagged_commit.yml
+    secrets:
+      RELEASE_TEAM: ${{ secrets.RELEASE_TEAM }}
+      READ_ORG_TOKEN: ${{ secrets.READ_ORG_TOKEN }}
+
   publish_release:
     name: Publish concrete-csprng Release
+    needs: verify_tag
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
@@ -32,11 +38,11 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
           SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "concrete-csprng release failed: (${{ env.ACTION_RUN_URL }})"
+          SLACK_MESSAGE: "concrete-csprng release finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
           SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
           SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}

.github/workflows/make_release_cuda.yml

@@ -21,22 +21,29 @@ env:
   SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
 
 jobs:
+  verify_tag:
+    uses: ./.github/workflows/verify_tagged_commit.yml
+    secrets:
+      RELEASE_TEAM: ${{ secrets.RELEASE_TEAM }}
+      READ_ORG_TOKEN: ${{ secrets.READ_ORG_TOKEN }}
+
   setup-instance:
     name: Setup instance (publish-cuda-release)
+    needs: verify_tag
     runs-on: ubuntu-latest
     outputs:
       runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
           slab-url: ${{ secrets.SLAB_BASE_URL }}
           job-secret: ${{ secrets.JOB_SECRET }}
           backend: aws
-          profile: gpu-test
+          profile: gpu-build
 
   publish-cuda-release:
     name: Publish CUDA Release
@@ -54,7 +61,7 @@ jobs:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
     steps:
       - name: Checkout
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
@@ -63,7 +70,7 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -99,7 +106,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "tfhe-cuda-backend release finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
@@ -112,7 +119,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -123,7 +130,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (publish-cuda-release) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/make_release_tfhe_fft.yml

@@ -0,0 +1,48 @@
+# Publish new release of tfhe-fft
+name: Publish tfhe-fft release
+
+on:
+  workflow_dispatch:
+    inputs:
+      dry_run:
+        description: "Dry-run"
+        type: boolean
+        default: true
+
+env:
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+
+jobs:
+  verify_tag:
+    uses: ./.github/workflows/verify_tagged_commit.yml
+    secrets:
+      RELEASE_TEAM: ${{ secrets.RELEASE_TEAM }}
+      READ_ORG_TOKEN: ${{ secrets.READ_ORG_TOKEN }}
+
+  publish_release:
+    name: Publish tfhe-fft Release
+    runs-on: ubuntu-latest
+    needs: verify_tag
+    steps:
+      - name: Checkout
+        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11
+        with:
+          fetch-depth: 0
+
+      - name: Publish crate.io package
+        env:
+          DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
+        run: |
+          cargo publish -p tfhe-fft --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@b24d75fe0e728a4bf9fc42ee217caa686d141ee8
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+          SLACK_MESSAGE: "tfhe-fft release failed: (${{ env.ACTION_RUN_URL }})"
+          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}

.github/workflows/make_release_tfhe_versionable.yml

@@ -0,0 +1,55 @@
+name: Publish tfhe-versionable release
+
+on:
+  workflow_dispatch:
+    inputs:
+      dry_run:
+        description: "Dry-run"
+        type: boolean
+        default: true
+
+env:
+  ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+
+jobs:
+  verify_tag:
+    uses: ./.github/workflows/verify_tagged_commit.yml
+    secrets:
+      RELEASE_TEAM: ${{ secrets.RELEASE_TEAM }}
+      READ_ORG_TOKEN: ${{ secrets.READ_ORG_TOKEN }}
+
+  publish_release:
+    name: Publish tfhe-versionable Release
+    needs: verify_tag
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+
+      - name: Publish proc-macro crate
+        env:
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
+          DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
+        run: |
+          cargo publish -p tfhe-versionable-derive --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
+
+      - name: Publish main crate
+        if: ${{ ! inputs.dry_run }}
+        env:
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
+        run: |
+          cargo publish -p tfhe-versionable --token ${{ env.CRATES_TOKEN }}
+
+      - name: Slack Notification
+        if: ${{ failure() }}
+        continue-on-error: true
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        env:
+          SLACK_COLOR: ${{ job.status }}
+          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
+          SLACK_MESSAGE: "tfhe-versionable release finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
+          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}

.github/workflows/make_release_zk_pok.yml

@@ -13,12 +13,19 @@ env:
   ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
 
 jobs:
+  verify_tag:
+    uses: ./.github/workflows/verify_tagged_commit.yml
+    secrets:
+      RELEASE_TEAM: ${{ secrets.RELEASE_TEAM }}
+      READ_ORG_TOKEN: ${{ secrets.READ_ORG_TOKEN }}
+
   publish_release:
     name: Publish tfhe-zk-pok Release
+    needs: verify_tag
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
 
@@ -32,7 +39,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}

.github/workflows/parameters_check.yml

@@ -14,17 +14,17 @@ on:
 
 jobs:
   params-curves-security-check:
-    runs-on: ubuntu-latest
+    runs-on: large_ubuntu_16
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
 
       - name: Checkout lattice-estimator
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: malb/lattice-estimator
           path: lattice_estimator
-          ref: '53508253629d3b5d31a2ad110e85dc69391ccb95'
+          ref: 'e80ec6bbbba212428b0e92d0467c18629cf9ed67'
 
       - name: Install Sage
         run: |
@@ -42,7 +42,7 @@ jobs:
       - name: Slack Notification
         if: ${{ always() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}

.github/workflows/sync_on_push.yml

@@ -13,7 +13,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout repo
-        uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
       - name: git-sync

.github/workflows/verify_tagged_commit.yml

@@ -0,0 +1,32 @@
+# Verify a tagged commit
+name: Verify tagged commit
+
+on:
+  workflow_call:
+    secrets:
+      RELEASE_TEAM:
+        required: true
+      READ_ORG_TOKEN:
+        required: true
+
+jobs:
+  checks:
+    runs-on: ubuntu-latest
+    if: startsWith(github.ref, 'refs/tags/')
+    steps:
+      # Check triggering actor membership
+      - name: Actor verification
+        id: actor_check
+        uses: morfien101/actions-authorized-user@4a3cfbf0bcb3cafe4a71710a278920c5d94bb38b
+        with:
+          username: ${{ github.triggering_actor }}
+          org: ${{ github.repository_owner }}
+          team: ${{ secrets.RELEASE_TEAM }}
+          github_token: ${{ secrets.READ_ORG_TOKEN }}
+
+      - name: Actor authorized
+        run: |
+          if [ "${{ steps.actor_check.outputs.authorized }}" == "false" ]; then
+            echo "Actor '${{ github.triggering_actor }}' is not authorized to perform release"
+            exit 1
+          fi

.gitignore

@@ -13,6 +13,7 @@ target/
 
 # Some of our bench outputs
 /tfhe/benchmarks_parameters
+/tfhe-zk-pok/benchmarks_parameters
 **/*.csv
 
 # dieharder run log
@@ -26,6 +27,10 @@ backends/tfhe-cuda-backend/cuda/cmake-build-debug/
 
 # WASM tests
 tfhe/web_wasm_parallel_tests/server.PID
+venv/
+web-test-runner/
+node_modules/
+package-lock.json
 
 # Dir used for backward compatibility test data
 tfhe/tfhe-backward-compat-data/

Cargo.toml

@@ -2,6 +2,7 @@
 resolver = "2"
 members = [
     "tfhe",
+    "tfhe-fft",
     "tfhe-zk-pok",
     "tasks",
     "apps/trivium",
@@ -16,6 +17,14 @@ exclude = [
     "utils/cargo-tfhe-lints-inner",
     "utils/cargo-tfhe-lints"
 ]
+[workspace.dependencies]
+aligned-vec = { version = "0.5", default-features = false }
+bytemuck = "1.14.3"
+dyn-stack = { version = "0.10", default-features = false }
+num-complex = "0.4"
+pulp = { version = "0.18.22", default-features = false }
+serde = { version = "1.0", default-features = false }
+wasm-bindgen = ">=0.2.86,<0.2.94"
 
 [profile.bench]
 lto = "fat"

Makefile

@@ -18,23 +18,20 @@ FAST_TESTS?=FALSE
 FAST_BENCH?=FALSE
 NIGHTLY_TESTS?=FALSE
 BENCH_OP_FLAVOR?=DEFAULT
-NODE_VERSION=22.4
-FORWARD_COMPAT?=OFF
+NODE_VERSION=22.6
 BACKWARD_COMPAT_DATA_URL=https://github.com/zama-ai/tfhe-backward-compat-data.git
-BACKWARD_COMPAT_DATA_BRANCH?=v0.1
+BACKWARD_COMPAT_DATA_BRANCH?=v0.4
 BACKWARD_COMPAT_DATA_PROJECT=tfhe-backward-compat-data
 BACKWARD_COMPAT_DATA_DIR=$(BACKWARD_COMPAT_DATA_PROJECT)
 TFHE_SPEC:=tfhe
+# We are kind of hacking the cut here, the version cannot contain a quote '"'
+WASM_BINDGEN_VERSION:=$(shell grep '^wasm-bindgen[[:space:]]*=' Cargo.toml | cut -d '"' -f 2 | xargs)
+WEB_RUNNER_DIR=web-test-runner
+WEB_SERVER_DIR=tfhe/web_wasm_parallel_tests
 # This is done to avoid forgetting it, we still precise the RUSTFLAGS in the commands to be able to
 # copy paste the command in the terminal and change them if required without forgetting the flags
 export RUSTFLAGS?=-C target-cpu=native
 
-ifeq ($(AVX512_SUPPORT),ON)
-		AVX512_FEATURE=nightly-avx512
-else
-		AVX512_FEATURE=
-endif
-
 ifeq ($(GEN_KEY_CACHE_MULTI_BIT_ONLY),TRUE)
 		MULTI_BIT_ONLY=--multi-bit-only
 else
@@ -47,12 +44,6 @@ else
 		COVERAGE_ONLY=
 endif
 
-ifeq ($(FORWARD_COMPAT),ON)
-		FORWARD_COMPAT_FEATURE=forward_compatibility
-else
-		FORWARD_COMPAT_FEATURE=
-endif
-
 # Variables used only for regex_engine example
 REGEX_STRING?=''
 REGEX_PATTERN?=''
@@ -102,12 +93,26 @@ install_rs_build_toolchain:
 	( echo "Unable to install $(RS_BUILD_TOOLCHAIN) toolchain, check your rustup installation. \
 	Rustup can be downloaded at https://rustup.rs/" && exit 1 )
 
+.PHONY: install_build_wasm32_target # Install the wasm32 toolchain used for builds
+install_build_wasm32_target: install_rs_build_toolchain
+	rustup +$(RS_BUILD_TOOLCHAIN) target add wasm32-unknown-unknown || \
+	( echo "Unable to install wasm32-unknown-unknown target toolchain, check your rustup installation. \
+	Rustup can be downloaded at https://rustup.rs/" && exit 1 )
+
 .PHONY: install_cargo_nextest # Install cargo nextest used for shortint tests
 install_cargo_nextest: install_rs_build_toolchain
 	@cargo nextest --version > /dev/null 2>&1 || \
 	cargo $(CARGO_RS_BUILD_TOOLCHAIN) install cargo-nextest --locked || \
 	( echo "Unable to install cargo nextest, unknown error." && exit 1 )
 
+# The installation should use the ^ symbol if the specified version in the root Cargo.toml is of the
+# form "0.2.96" then we get ^0.2.96 e.g., as we don't lock those dependencies
+# this allows to get the matching CLI
+# If a version range is specified no need to add the leading ^
+.PHONY: install_wasm_bindgen_cli # Install wasm-bindgen-cli to get access to the test runner
+install_wasm_bindgen_cli: install_rs_build_toolchain
+	cargo +$(RS_BUILD_TOOLCHAIN) install --locked wasm-bindgen-cli --version "$(WASM_BINDGEN_VERSION)"
+
 .PHONY: install_wasm_pack # Install wasm-pack to build JS packages
 install_wasm_pack: install_rs_build_toolchain
 	@wasm-pack --version > /dev/null 2>&1 || \
@@ -126,6 +131,10 @@ install_node:
 	$(SHELL) -i -c 'nvm install $(NODE_VERSION)' || \
 	( echo "Unable to install node, unknown error." && exit 1 )
 
+.PHONY: node_version  # Return Node version that will be installed
+node_version:
+	@echo "$(NODE_VERSION)"
+
 .PHONY: install_dieharder # Install dieharder for apt distributions or macOS
 install_dieharder:
 	@dieharder -h > /dev/null 2>&1 || \
@@ -146,6 +155,63 @@ install_tfhe_lints:
 	(cd utils/cargo-tfhe-lints-inner && cargo install --path .) && \
 	cd utils/cargo-tfhe-lints && cargo install --path .
 
+.PHONY: install_typos_checker # Install typos checker
+install_typos_checker: install_rs_build_toolchain
+	@typos --version > /dev/null 2>&1 || \
+	cargo $(CARGO_RS_BUILD_TOOLCHAIN) install typos-cli || \
+	( echo "Unable to install typos-cli, unknown error." && exit 1 )
+
+.PHONY: setup_venv # Setup Python virtualenv for wasm tests
+setup_venv:
+	python3 -m venv venv
+	@source venv/bin/activate && \
+	pip3 install -r ci/webdriver_requirements.txt
+
+# This is an internal target, not meant to be called on its own.
+install_web_resource:
+	wget -P $(dest) $(url)
+	@cd $(dest) && \
+	echo "$(checksum) $(filename)" > checksum && \
+	sha256sum -c checksum && \
+	rm checksum && \
+	$(decompress_cmd) $(filename)
+
+install_chrome_browser: url = "https://storage.googleapis.com/chrome-for-testing-public/130.0.6723.69/linux64/chrome-linux64.zip"
+install_chrome_browser: checksum = "f789d53911a50cfa4a2bc1f09cde57567247f52515436d92b1aa9de93c2787d0"
+install_chrome_browser: dest = "$(WEB_RUNNER_DIR)/chrome"
+install_chrome_browser: filename = "chrome-linux64.zip"
+install_chrome_browser: decompress_cmd = unzip
+
+.PHONY: install_chrome_browser # Install Chrome browser for Linux
+install_chrome_browser: install_web_resource
+
+install_chrome_web_driver: url = "https://storage.googleapis.com/chrome-for-testing-public/130.0.6723.69/linux64/chromedriver-linux64.zip"
+install_chrome_web_driver: checksum = "90fe8dedf33eefe4b72704f626fa9f5834427c042235cfeb4251f18c9f0336ea"
+install_chrome_web_driver: dest = "$(WEB_RUNNER_DIR)/chrome"
+install_chrome_web_driver: filename = "chromedriver-linux64.zip"
+install_chrome_web_driver: decompress_cmd = unzip
+
+.PHONY: install_chrome_web_driver # Install Chrome web driver for Linux
+install_chrome_web_driver: install_web_resource
+
+install_firefox_browser: url = "https://download-installer.cdn.mozilla.net/pub/firefox/releases/131.0/linux-x86_64/en-US/firefox-131.0.tar.bz2"
+install_firefox_browser: checksum = "4ca8504a62a31472ecb8c3a769d4301dd4ac692d4cc5d51b8fe2cf41e7b11106"
+install_firefox_browser: dest = "$(WEB_RUNNER_DIR)/firefox"
+install_firefox_browser: filename = "firefox-131.0.tar.bz2"
+install_firefox_browser: decompress_cmd = tar -xvf
+
+.PHONY: install_firefox_browser # Install firefox browser for Linux
+install_firefox_browser: install_web_resource
+
+install_firefox_web_driver: url = "https://github.com/mozilla/geckodriver/releases/download/v0.35.0/geckodriver-v0.35.0-linux64.tar.gz"
+install_firefox_web_driver: checksum = "ac26e9ba8f3b8ce0fbf7339b9c9020192f6dcfcbf04a2bcd2af80dfe6bb24260"
+install_firefox_web_driver: dest = "$(WEB_RUNNER_DIR)/firefox"
+install_firefox_web_driver: filename = "geckodriver-v0.35.0-linux64.tar.gz"
+install_firefox_web_driver: decompress_cmd = tar -xvf
+
+.PHONY: install_firefox_web_driver # Install firefox web driver for Linux
+install_firefox_web_driver: install_web_resource
+
 .PHONY: check_linelint_installed # Check if linelint newline linter is installed
 check_linelint_installed:
 	@printf "\n" | linelint - > /dev/null 2>&1 || \
@@ -207,6 +273,10 @@ check_fmt_js: check_nvm_installed
 	nvm use $(NODE_VERSION) && \
 	$(MAKE) -C tfhe/web_wasm_parallel_tests check_fmt
 
+.PHONY: check_typos # Check for typos in codebase
+check_typos: install_typos_checker
+	@typos && echo "No typos found"
+
 .PHONY: clippy_gpu # Run clippy lints on tfhe with "gpu" enabled
 clippy_gpu: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
@@ -214,6 +284,13 @@ clippy_gpu: install_rs_check_toolchain
 		--all-targets \
 		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
+.PHONY: check_gpu # Run check on tfhe with "gpu" enabled
+check_gpu: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" check \
+		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,internal-keycache,gpu \
+		--all-targets \
+		-p $(TFHE_SPEC)
+
 .PHONY: fix_newline # Fix newline at end of file issues to be UNIX compliant
 fix_newline: check_linelint_installed
 	linelint -a .
@@ -240,6 +317,9 @@ clippy_core: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=$(TARGET_ARCH_FEATURE),experimental,nightly-avx512 \
 		-p $(TFHE_SPEC) -- --no-deps -D warnings
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
+		--features=$(TARGET_ARCH_FEATURE),zk-pok \
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_boolean # Run clippy lints enabling the boolean features
 clippy_boolean: install_rs_check_toolchain
@@ -252,12 +332,21 @@ clippy_shortint: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=$(TARGET_ARCH_FEATURE),shortint \
 		-p $(TFHE_SPEC) -- --no-deps -D warnings
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
+		--features=$(TARGET_ARCH_FEATURE),shortint,experimental \
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
+		--features=$(TARGET_ARCH_FEATURE),zk-pok,shortint \
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_integer # Run clippy lints enabling the integer features
 clippy_integer: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=$(TARGET_ARCH_FEATURE),integer \
 		-p $(TFHE_SPEC) -- --no-deps -D warnings
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
+		--features=$(TARGET_ARCH_FEATURE),integer,experimental \
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy # Run clippy lints enabling the boolean, shortint, integer
 clippy: install_rs_check_toolchain
@@ -284,6 +373,9 @@ clippy_c_api: install_rs_check_toolchain
 
 .PHONY: clippy_js_wasm_api # Run clippy lints enabling the boolean, shortint, integer and the js wasm API
 clippy_js_wasm_api: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
+		--features=boolean-client-js-wasm-api,shortint-client-js-wasm-api,integer-client-js-wasm-api,high-level-client-js-wasm-api,zk-pok \
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=boolean-client-js-wasm-api,shortint-client-js-wasm-api,integer-client-js-wasm-api,high-level-client-js-wasm-api \
 		-p $(TFHE_SPEC) -- --no-deps -D warnings
@@ -303,6 +395,9 @@ clippy_all_targets: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,internal-keycache,zk-pok \
 		-p $(TFHE_SPEC) -- --no-deps -D warnings
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
+		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,internal-keycache,zk-pok,experimental \
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_concrete_csprng # Run clippy lints on concrete-csprng
 clippy_concrete_csprng: install_rs_check_toolchain
@@ -315,9 +410,17 @@ clippy_zk_pok: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		-p tfhe-zk-pok -- --no-deps -D warnings
 
+.PHONY: clippy_versionable # Run clippy lints on tfhe-versionable
+clippy_versionable: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
+		-p tfhe-versionable-derive -- --no-deps -D warnings
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
+		-p tfhe-versionable -- --no-deps -D warnings
+
 .PHONY: clippy_all # Run all clippy targets
 clippy_all: clippy_rustdoc clippy clippy_boolean clippy_shortint clippy_integer clippy_all_targets \
-clippy_c_api clippy_js_wasm_api clippy_tasks clippy_core clippy_concrete_csprng clippy_zk_pok clippy_trivium
+clippy_c_api clippy_js_wasm_api clippy_tasks clippy_core clippy_concrete_csprng clippy_zk_pok clippy_trivium \
+clippy_versionable
 
 .PHONY: clippy_fast # Run main clippy targets
 clippy_fast: clippy_rustdoc clippy clippy_all_targets clippy_c_api clippy_js_wasm_api clippy_tasks \
@@ -328,10 +431,18 @@ clippy_cuda_backend: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		-p tfhe-cuda-backend -- --no-deps -D warnings
 
+.PHONY: check_rust_bindings_did_not_change # Check rust bindings are up to date for tfhe-cuda-backend
+check_rust_bindings_did_not_change:
+	cargo build -p tfhe-cuda-backend && "$(MAKE)" fmt_gpu && \
+	git diff --quiet HEAD -- backends/tfhe-cuda-backend/src/bindings.rs || \
+	( echo "Generated bindings have changed! Please run 'git add backends/tfhe-cuda-backend/src/bindings.rs' \
+	and commit the changes." && exit 1 ) 
+
+
 .PHONY: tfhe_lints # Run custom tfhe-rs lints
 tfhe_lints: install_tfhe_lints
 	cd tfhe && RUSTFLAGS="$(RUSTFLAGS)" cargo tfhe-lints \
-		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer -- -D warnings
+		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,zk-pok -- -D warnings
 
 .PHONY: build_core # Build core_crypto without experimental features
 build_core: install_rs_build_toolchain install_rs_check_toolchain
@@ -339,7 +450,7 @@ build_core: install_rs_build_toolchain install_rs_check_toolchain
 		--features=$(TARGET_ARCH_FEATURE) -p $(TFHE_SPEC)
 	@if [[ "$(AVX512_SUPPORT)" == "ON" ]]; then \
 		RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-			--features=$(TARGET_ARCH_FEATURE),$(AVX512_FEATURE) -p $(TFHE_SPEC); \
+			--features=$(TARGET_ARCH_FEATURE),nightly-avx512 -p $(TFHE_SPEC); \
 	fi
 
 .PHONY: build_core_experimental # Build core_crypto with experimental features
@@ -348,7 +459,7 @@ build_core_experimental: install_rs_build_toolchain install_rs_check_toolchain
 		--features=$(TARGET_ARCH_FEATURE),experimental -p $(TFHE_SPEC)
 	@if [[ "$(AVX512_SUPPORT)" == "ON" ]]; then \
 		RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-			--features=$(TARGET_ARCH_FEATURE),experimental,$(AVX512_FEATURE) -p $(TFHE_SPEC); \
+			--features=$(TARGET_ARCH_FEATURE),experimental,nightly-avx512 -p $(TFHE_SPEC); \
 	fi
 
 .PHONY: build_boolean # Build with boolean enabled
@@ -376,32 +487,23 @@ build_tfhe_coverage: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS) --cfg tarpaulin" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
 		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,internal-keycache -p $(TFHE_SPEC) --tests
 
-.PHONY: symlink_c_libs_without_fingerprint # Link the .a and .so files without the changing hash part in target
-symlink_c_libs_without_fingerprint:
-	@./scripts/symlink_c_libs_without_fingerprint.sh \
-		--cargo-profile "$(CARGO_PROFILE)" \
-		--lib-name tfhe-c-api-dynamic-buffer
-
 .PHONY: build_c_api # Build the C API for boolean, shortint and integer
 build_c_api: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		--features=$(TARGET_ARCH_FEATURE),boolean-c-api,shortint-c-api,high-level-c-api,zk-pok,$(FORWARD_COMPAT_FEATURE) \
+		--features=$(TARGET_ARCH_FEATURE),boolean-c-api,shortint-c-api,high-level-c-api,zk-pok \
 		-p $(TFHE_SPEC)
-	@"$(MAKE)" symlink_c_libs_without_fingerprint
 
 .PHONY: build_c_api_gpu # Build the C API for boolean, shortint and integer
 build_c_api_gpu: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
 		--features=$(TARGET_ARCH_FEATURE),boolean-c-api,shortint-c-api,high-level-c-api,zk-pok,gpu \
 		-p $(TFHE_SPEC)
-	@"$(MAKE)" symlink_c_libs_without_fingerprint
 
 .PHONY: build_c_api_experimental_deterministic_fft # Build the C API for boolean, shortint and integer with experimental deterministic FFT
 build_c_api_experimental_deterministic_fft: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		--features=$(TARGET_ARCH_FEATURE),boolean-c-api,shortint-c-api,high-level-c-api,zk-pok,experimental-force_fft_algo_dif4,$(FORWARD_COMPAT_FEATURE) \
+		--features=$(TARGET_ARCH_FEATURE),boolean-c-api,shortint-c-api,high-level-c-api,zk-pok,experimental-force_fft_algo_dif4 \
 		-p $(TFHE_SPEC)
-	@"$(MAKE)" symlink_c_libs_without_fingerprint
 
 .PHONY: build_web_js_api # Build the js API targeting the web browser
 build_web_js_api: install_rs_build_toolchain install_wasm_pack
@@ -419,6 +521,7 @@ build_web_js_api_parallel: install_rs_check_toolchain install_wasm_pack
 		-- --features=boolean-client-js-wasm-api,shortint-client-js-wasm-api,integer-client-js-wasm-api,parallel-wasm-api,zk-pok \
 		-Z build-std=panic_abort,std && \
 	find pkg/snippets -type f -iname workerHelpers.worker.js -exec sed -i "s|from '..\/..\/..\/';|from '..\/..\/..\/tfhe.js';|" {} \;
+	jq '.files += ["snippets"]' tfhe/pkg/package.json > tmp_pkg.json && mv -f tmp_pkg.json tfhe/pkg/package.json
 
 .PHONY: build_node_js_api # Build the js API targeting nodejs
 build_node_js_api: install_rs_build_toolchain install_wasm_pack
@@ -438,7 +541,7 @@ test_core_crypto: install_rs_build_toolchain install_rs_check_toolchain
 		--features=$(TARGET_ARCH_FEATURE),experimental,zk-pok -p $(TFHE_SPEC) -- core_crypto::
 	@if [[ "$(AVX512_SUPPORT)" == "ON" ]]; then \
 		RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-			--features=$(TARGET_ARCH_FEATURE),experimental,zk-pok,$(AVX512_FEATURE) -p $(TFHE_SPEC) -- core_crypto::; \
+			--features=$(TARGET_ARCH_FEATURE),experimental,zk-pok,nightly-avx512 -p $(TFHE_SPEC) -- core_crypto::; \
 	fi
 
 .PHONY: test_core_crypto_cov # Run the tests of the core_crypto module with code coverage
@@ -452,7 +555,7 @@ test_core_crypto_cov: install_rs_build_toolchain install_rs_check_toolchain inst
 		RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) tarpaulin --profile $(CARGO_PROFILE) \
 			--out xml --output-dir coverage/core_crypto_avx512 --line --engine llvm --timeout 500 \
 			--implicit-test-threads $(COVERAGE_EXCLUDED_FILES) \
-			--features=$(TARGET_ARCH_FEATURE),experimental,internal-keycache,$(AVX512_FEATURE) \
+			--features=$(TARGET_ARCH_FEATURE),experimental,internal-keycache,nightly-avx512 \
 			-p $(TFHE_SPEC) -- -Z unstable-options --report-time core_crypto::; \
 	fi
 
@@ -481,6 +584,20 @@ test_integer_gpu: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --doc --profile $(CARGO_PROFILE) \
 		--features=$(TARGET_ARCH_FEATURE),integer,gpu -p $(TFHE_SPEC) -- integer::gpu::server_key::
 
+.PHONY: test_integer_compression
+test_integer_compression: install_rs_build_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
+		--features=$(TARGET_ARCH_FEATURE),integer -p $(TFHE_SPEC) -- integer::ciphertext::compressed_ciphertext_list::tests::
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --doc --profile $(CARGO_PROFILE) \
+		--features=$(TARGET_ARCH_FEATURE),integer -p $(TFHE_SPEC) -- integer::ciphertext::compress
+
+.PHONY: test_integer_compression_gpu
+test_integer_compression_gpu: install_rs_build_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
+		--features=$(TARGET_ARCH_FEATURE),integer,gpu -p $(TFHE_SPEC) -- integer::gpu::ciphertext::compressed_ciphertext_list::tests::
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --doc --profile $(CARGO_PROFILE) \
+		--features=$(TARGET_ARCH_FEATURE),integer,gpu -p $(TFHE_SPEC) -- integer::gpu::ciphertext::compress
+
 .PHONY: test_integer_gpu_ci # Run the tests for integer ci on gpu backend
 test_integer_gpu_ci: install_rs_check_toolchain install_cargo_nextest
 	BIG_TESTS_INSTANCE="$(BIG_TESTS_INSTANCE)" \
@@ -647,10 +764,15 @@ test_signed_integer_multi_bit_ci: install_rs_check_toolchain install_cargo_nexte
 		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --avx512-support "$(AVX512_SUPPORT)" \
 		--signed-only --tfhe-package "$(TFHE_SPEC)"
 
-.PHONY: test_safe_deserialization # Run the tests for safe deserialization
-test_safe_deserialization: install_rs_build_toolchain install_cargo_nextest
+.PHONY: test_safe_serialization # Run the tests for safe serialization
+test_safe_serialization: install_rs_build_toolchain install_cargo_nextest
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,internal-keycache -p $(TFHE_SPEC) -- safe_deserialization::
+		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,internal-keycache -p $(TFHE_SPEC) -- safe_serialization::
+
+.PHONY: test_zk # Run the tests for the zk module of the TFHE-rs crate
+test_zk: install_rs_build_toolchain install_cargo_nextest
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
+		--features=$(TARGET_ARCH_FEATURE),shortint,zk-pok -p $(TFHE_SPEC) -- zk::
 
 .PHONY: test_integer # Run all the tests for integer
 test_integer: install_rs_build_toolchain
@@ -677,6 +799,13 @@ test_high_level_api_gpu: install_rs_build_toolchain install_cargo_nextest
 		--features=$(TARGET_ARCH_FEATURE),integer,internal-keycache,gpu -p $(TFHE_SPEC) \
 		-E "test(/high_level_api::.*gpu.*/)"
 
+.PHONY: test_strings # Run the tests for strings ci
+test_strings: install_rs_build_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
+		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,strings -p $(TFHE_SPEC) \
+		-- strings::
+
+
 .PHONY: test_user_doc # Run tests from the .md documentation
 test_user_doc: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) --doc \
@@ -690,11 +819,7 @@ test_user_doc_gpu: install_rs_build_toolchain
 		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,integer,internal-keycache,gpu,zk-pok -p $(TFHE_SPEC) \
 		-- test_user_docs::
 
-.PHONY: test_fhe_strings # Run tests for fhe_strings example
-test_fhe_strings: install_rs_build_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--example fhe_strings \
-		--features=$(TARGET_ARCH_FEATURE),integer
+
 
 .PHONY: test_regex_engine # Run tests for regex_engine example
 test_regex_engine: install_rs_build_toolchain
@@ -731,10 +856,23 @@ test_zk_pok: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
 		-p tfhe-zk-pok
 
+.PHONY: test_zk_wasm_x86_compat_ci
+test_zk_wasm_x86_compat_ci: check_nvm_installed
+	source ~/.nvm/nvm.sh && \
+	nvm install $(NODE_VERSION) && \
+	nvm use $(NODE_VERSION) && \
+	$(MAKE) test_zk_wasm_x86_compat
+
+.PHONY: test_zk_wasm_x86_compat # Check compatibility between wasm and x86_64 proofs
+test_zk_wasm_x86_compat: install_rs_build_toolchain build_node_js_api
+	cd tfhe/tests/zk_wasm_x86_test && npm install
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
+		-p tfhe --test zk_wasm_x86_test --features=$(TARGET_ARCH_FEATURE),integer,zk-pok
+
 .PHONY: test_versionable # Run tests for tfhe-versionable subcrate
 test_versionable: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		-p tfhe-versionable
+		--all-targets -p tfhe-versionable
 
 # The backward compat data repo holds historical binary data but also rust code to generate and load them.
 # Here we use the "patch" functionality of Cargo to make sure the repo used for the data is the same as the one used for the code.
@@ -742,7 +880,7 @@ test_versionable: install_rs_build_toolchain
 test_backward_compatibility_ci: install_rs_build_toolchain
 	TFHE_BACKWARD_COMPAT_DATA_DIR="$(BACKWARD_COMPAT_DATA_DIR)" RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
 		--config "patch.'$(BACKWARD_COMPAT_DATA_URL)'.$(BACKWARD_COMPAT_DATA_PROJECT).path=\"tfhe/$(BACKWARD_COMPAT_DATA_DIR)\"" \
-		--features=$(TARGET_ARCH_FEATURE),shortint,integer -p $(TFHE_SPEC) test_backward_compatibility -- --nocapture
+		--features=$(TARGET_ARCH_FEATURE),shortint,integer,zk-pok -p $(TFHE_SPEC) test_backward_compatibility -- --nocapture
 
 .PHONY: test_backward_compatibility # Same as test_backward_compatibility_ci but tries to clone the data repo first if needed
 test_backward_compatibility: tfhe/$(BACKWARD_COMPAT_DATA_DIR) test_backward_compatibility_ci
@@ -815,36 +953,58 @@ check_compile_tests_benches_gpu: install_rs_build_toolchain
 		cmake .. -DCMAKE_BUILD_TYPE=Debug -DTFHE_CUDA_BACKEND_BUILD_TESTS=ON -DTFHE_CUDA_BACKEND_BUILD_BENCHMARKS=ON && \
 		"$(MAKE)" -j "$(CPU_COUNT)"
 
-.PHONY: build_nodejs_test_docker # Build a docker image with tools to run nodejs tests for wasm API
-build_nodejs_test_docker:
-	DOCKER_BUILDKIT=1 docker build --build-arg RUST_TOOLCHAIN="$(RS_BUILD_TOOLCHAIN)" \
-		-f docker/Dockerfile.wasm_tests --build-arg NODE_VERSION=$(NODE_VERSION) -t tfhe-wasm-tests .
-
-.PHONY: test_nodejs_wasm_api_in_docker # Run tests for the nodejs on wasm API in a docker container
-test_nodejs_wasm_api_in_docker: build_nodejs_test_docker
-	if [[ -t 1 ]]; then RUN_FLAGS="-it"; else RUN_FLAGS="-i"; fi && \
-	docker run --rm "$${RUN_FLAGS}" \
-		-v "$$(pwd)":/tfhe-wasm-tests/tfhe-rs \
-		-v tfhe-rs-root-target-cache:/root/tfhe-rs-target \
-		-v tfhe-rs-pkg-cache:/tfhe-wasm-tests/tfhe-rs/tfhe/pkg \
-		-v tfhe-rs-root-cargo-registry-cache:/root/.cargo/registry \
-		-v tfhe-rs-root-cache:/root/.cache \
-		tfhe-wasm-tests /bin/bash -i -c 'make test_nodejs_wasm_api'
-
 .PHONY: test_nodejs_wasm_api # Run tests for the nodejs on wasm API
 test_nodejs_wasm_api: build_node_js_api
-	cd tfhe/js_on_wasm_tests && npm run test
+	cd tfhe/js_on_wasm_tests && npm install && npm run test
 
-.PHONY: test_web_js_api_parallel # Run tests for the web wasm api
-test_web_js_api_parallel: build_web_js_api_parallel
-	$(MAKE) -C tfhe/web_wasm_parallel_tests test
-
-.PHONY: test_web_js_api_parallel_ci # Run tests for the web wasm api
-test_web_js_api_parallel_ci: build_web_js_api_parallel
+.PHONY: test_nodejs_wasm_api_ci # Run tests for the nodejs on wasm API
+test_nodejs_wasm_api_ci: build_node_js_api
 	source ~/.nvm/nvm.sh && \
 	nvm install $(NODE_VERSION) && \
 	nvm use $(NODE_VERSION) && \
-	$(MAKE) -C tfhe/web_wasm_parallel_tests test-ci
+	$(MAKE) test_nodejs_wasm_api
+
+# This is an internal target, not meant to be called on its own.
+run_web_js_api_parallel: build_web_js_api_parallel setup_venv
+	cd $(WEB_SERVER_DIR) && npm install && npm run build
+	source venv/bin/activate && \
+	python ci/webdriver.py \
+	--browser-path $(browser_path) \
+	--driver-path $(driver_path) \
+	--browser-kind  $(browser_kind) \
+	--server-cmd "npm run server" \
+	--server-workdir "$(WEB_SERVER_DIR)" \
+	--id-pattern $(filter)
+
+test_web_js_api_parallel_chrome: browser_path = "$(WEB_RUNNER_DIR)/chrome/chrome-linux64/chrome"
+test_web_js_api_parallel_chrome: driver_path = "$(WEB_RUNNER_DIR)/chrome/chromedriver-linux64/chromedriver"
+test_web_js_api_parallel_chrome: browser_kind = chrome
+test_web_js_api_parallel_chrome: filter = Test
+
+.PHONY: test_web_js_api_parallel_chrome # Run tests for the web wasm api on Chrome
+test_web_js_api_parallel_chrome: run_web_js_api_parallel
+
+.PHONY: test_web_js_api_parallel_chrome_ci # Run tests for the web wasm api on Chrome
+test_web_js_api_parallel_chrome_ci: setup_venv
+	source ~/.nvm/nvm.sh && \
+	nvm install $(NODE_VERSION) && \
+	nvm use $(NODE_VERSION) && \
+	$(MAKE) test_web_js_api_parallel_chrome
+
+test_web_js_api_parallel_firefox: browser_path = "$(WEB_RUNNER_DIR)/firefox/firefox/firefox"
+test_web_js_api_parallel_firefox: driver_path = "$(WEB_RUNNER_DIR)/firefox/geckodriver"
+test_web_js_api_parallel_firefox: browser_kind = firefox
+test_web_js_api_parallel_firefox: filter = Test
+
+.PHONY: test_web_js_api_parallel_firefox # Run tests for the web wasm api on Firefox
+test_web_js_api_parallel_firefox: run_web_js_api_parallel
+
+.PHONY: test_web_js_api_parallel_firefox_ci # Run tests for the web wasm api on Firefox
+test_web_js_api_parallel_firefox_ci: setup_venv
+	source ~/.nvm/nvm.sh && \
+	nvm install $(NODE_VERSION) && \
+	nvm use $(NODE_VERSION) && \
+	$(MAKE) test_web_js_api_parallel_firefox
 
 .PHONY: no_tfhe_typo # Check we did not invert the h and f in tfhe
 no_tfhe_typo:
@@ -862,6 +1022,11 @@ dieharder_csprng: install_dieharder build_concrete_csprng
 # Benchmarks
 #
 
+.PHONY: print_doc_bench_parameters # Print parameters used in doc benchmarks
+print_doc_bench_parameters:
+	RUSTFLAGS="" cargo run --example print_doc_bench_parameters \
+	--features=$(TARGET_ARCH_FEATURE),shortint,internal-keycache -p tfhe
+
 .PHONY: bench_integer # Run benchmarks for unsigned integer
 bench_integer: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_OP_FLAVOR=$(BENCH_OP_FLAVOR) __TFHE_RS_FAST_BENCH=$(FAST_BENCH) \
@@ -883,6 +1048,18 @@ bench_integer_gpu: install_rs_check_toolchain
 	--bench integer-bench \
 	--features=$(TARGET_ARCH_FEATURE),integer,gpu,internal-keycache,nightly-avx512 -p $(TFHE_SPEC) --
 
+.PHONY: bench_integer_compression # Run benchmarks for unsigned integer compression
+bench_integer_compression: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
+	--bench	glwe_packing_compression-integer-bench \
+	--features=$(TARGET_ARCH_FEATURE),integer,internal-keycache,nightly-avx512 -p $(TFHE_SPEC) --
+
+.PHONY: bench_integer_compression_gpu
+bench_integer_compression_gpu: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
+	--bench	glwe_packing_compression-integer-bench \
+	--features=$(TARGET_ARCH_FEATURE),integer,internal-keycache,gpu -p $(TFHE_SPEC) --
+
 .PHONY: bench_integer_multi_bit # Run benchmarks for unsigned integer using multi-bit parameters
 bench_integer_multi_bit: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=MULTI_BIT \
@@ -981,15 +1158,52 @@ bench_ks_gpu: install_rs_check_toolchain
 	--bench ks-bench \
 	--features=$(TARGET_ARCH_FEATURE),boolean,shortint,gpu,internal-keycache,nightly-avx512 -p $(TFHE_SPEC)
 
-.PHONY: bench_web_js_api_parallel # Run benchmarks for the web wasm api
-bench_web_js_api_parallel: build_web_js_api_parallel
-	$(MAKE) -C tfhe/web_wasm_parallel_tests bench
+bench_web_js_api_parallel_chrome: browser_path = "$(WEB_RUNNER_DIR)/chrome/chrome-linux64/chrome"
+bench_web_js_api_parallel_chrome: driver_path = "$(WEB_RUNNER_DIR)/chrome/chromedriver-linux64/chromedriver"
+bench_web_js_api_parallel_chrome: browser_kind = chrome
+bench_web_js_api_parallel_chrome: filter = Bench
 
-.PHONY: bench_web_js_api_parallel_ci # Run benchmarks for the web wasm api
-bench_web_js_api_parallel_ci: build_web_js_api_parallel
+.PHONY: bench_web_js_api_parallel_chrome # Run benchmarks for the web wasm api
+bench_web_js_api_parallel_chrome: run_web_js_api_parallel
+
+.PHONY: bench_web_js_api_parallel_chrome_ci # Run benchmarks for the web wasm api
+bench_web_js_api_parallel_chrome_ci: setup_venv
 	source ~/.nvm/nvm.sh && \
+	nvm install $(NODE_VERSION) && \
 	nvm use $(NODE_VERSION) && \
-	$(MAKE) -C tfhe/web_wasm_parallel_tests bench-ci
+	$(MAKE) bench_web_js_api_parallel_chrome
+
+bench_web_js_api_parallel_firefox: browser_path = "$(WEB_RUNNER_DIR)/firefox/firefox/firefox"
+bench_web_js_api_parallel_firefox: driver_path = "$(WEB_RUNNER_DIR)/firefox/geckodriver"
+bench_web_js_api_parallel_firefox: browser_kind = firefox
+bench_web_js_api_parallel_firefox: filter = Bench
+
+.PHONY: bench_web_js_api_parallel_firefox # Run benchmarks for the web wasm api
+bench_web_js_api_parallel_firefox: run_web_js_api_parallel
+
+.PHONY: bench_web_js_api_parallel_firefox_ci # Run benchmarks for the web wasm api
+bench_web_js_api_parallel_firefox_ci: setup_venv
+	source ~/.nvm/nvm.sh && \
+	nvm install $(NODE_VERSION) && \
+	nvm use $(NODE_VERSION) && \
+	$(MAKE) bench_web_js_api_parallel_firefox
+
+.PHONY: bench_hlapi_erc20 # Run benchmarks for ECR20 operations
+bench_hlapi_erc20: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
+	--bench hlapi-erc20 \
+	--features=$(TARGET_ARCH_FEATURE),integer,internal-keycache,pbs-stats,nightly-avx512 -p $(TFHE_SPEC) --
+
+.PHONY: bench_hlapi_erc20_gpu # Run benchmarks for ECR20 operations on GPU
+bench_hlapi_erc20_gpu: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
+	--bench hlapi-erc20 \
+	--features=$(TARGET_ARCH_FEATURE),integer,gpu,internal-keycache,pbs-stats,nightly-avx512 -p $(TFHE_SPEC) --
+
+.PHONY: bench_tfhe_zk_pok # Run benchmarks for the tfhe_zk_pok crate
+bench_tfhe_zk_pok: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" \
+	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench -p tfhe-zk-pok --
 
 #
 # Utility tools
@@ -999,8 +1213,8 @@ bench_web_js_api_parallel_ci: build_web_js_api_parallel
 gen_key_cache: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS) --cfg tarpaulin" cargo $(CARGO_RS_BUILD_TOOLCHAIN) run --profile $(CARGO_PROFILE) \
 		--example generates_test_keys \
-		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,internal-keycache -- \
-		$(MULTI_BIT_ONLY) $(COVERAGE_ONLY)
+		--features=$(TARGET_ARCH_FEATURE),boolean,shortint,experimental,internal-keycache -p $(TFHE_SPEC) \
+		-- $(MULTI_BIT_ONLY) $(COVERAGE_ONLY)
 
 .PHONY: gen_key_cache_core_crypto # Run function to generate keys and cache them for core_crypto tests
 gen_key_cache_core_crypto: install_rs_build_toolchain
@@ -1037,7 +1251,7 @@ parse_wasm_benchmarks: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) run --profile $(CARGO_PROFILE) \
 	--example wasm_benchmarks_parser \
 	--features=$(TARGET_ARCH_FEATURE),shortint,internal-keycache \
-	-- web_wasm_parallel_tests/test/benchmark_results
+	-- wasm_benchmark_results.json
 
 .PHONY: write_params_to_file # Gather all crypto parameters into a file with a Sage readable format.
 write_params_to_file: install_rs_check_toolchain
@@ -1076,19 +1290,129 @@ sha256_bool: install_rs_check_toolchain
 	--features=$(TARGET_ARCH_FEATURE),boolean
 
 .PHONY: pcc # pcc stands for pre commit checks (except GPU)
-pcc: no_tfhe_typo no_dbg_log check_fmt lint_doc check_md_docs_are_tested check_intra_md_links \
+pcc: no_tfhe_typo no_dbg_log check_fmt check_typos lint_doc check_md_docs_are_tested check_intra_md_links \
 clippy_all tfhe_lints check_compile_tests
 
 .PHONY: pcc_gpu # pcc stands for pre commit checks for GPU compilation
-pcc_gpu: clippy_gpu clippy_cuda_backend check_compile_tests_benches_gpu
+pcc_gpu: clippy_gpu clippy_cuda_backend check_compile_tests_benches_gpu check_rust_bindings_did_not_change
 
 .PHONY: fpcc # pcc stands for pre commit checks, the f stands for fast
-fpcc: no_tfhe_typo no_dbg_log check_fmt lint_doc check_md_docs_are_tested clippy_fast \
+fpcc: no_tfhe_typo no_dbg_log check_fmt check_typos lint_doc check_md_docs_are_tested clippy_fast \
 check_compile_tests
 
 .PHONY: conformance # Automatically fix problems that can be fixed
 conformance: fix_newline fmt
 
+#=============================== FFT Section ==================================
+.PHONY: doc_fft # Build rust doc for tfhe-fft
+doc_fft: install_rs_check_toolchain
+	@# Even though we are not in docs.rs, this allows to "just" build the doc
+	DOCS_RS=1 \
+	RUSTDOCFLAGS="--html-in-header katex-header.html" \
+	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" doc \
+		--all-features --no-deps -p tfhe-fft
+
+.PHONY: docs_fft # Build rust doc tfhe-fft, alias for doc
+docs_fft: doc_fft
+
+.PHONY: lint_doc_fft # Build rust doc for tfhe-fft with linting enabled
+lint_doc_fft: install_rs_check_toolchain
+	@# Even though we are not in docs.rs, this allows to "just" build the doc
+	DOCS_RS=1 \
+	RUSTDOCFLAGS="--html-in-header katex-header.html -Dwarnings" \
+	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" doc \
+		--all-features --no-deps -p tfhe-fft
+
+.PHONY: lint_docs_fft # Build rust doc for tfhe-fft with linting enabled, alias for lint_doc
+lint_docs_fft: lint_doc_fft
+
+.PHONY: clippy_fft # Run clippy lints on tfhe-fft
+clippy_fft: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
+		--all-features -p tfhe-fft -- --no-deps -D warnings
+
+.PHONY: pcc_fft # pcc stands for pre commit checks
+pcc_fft: check_fmt lint_doc_fft clippy_fft
+
+.PHONY: build_fft
+build_fft: install_rs_build_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --release -p tfhe-fft
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --release -p tfhe-fft \
+		--features=fft128
+
+.PHONY: build_fft_no_std
+buildfft__no_std: install_rs_build_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --release -p tfhe-fft \
+		--no-default-features
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --release -p tfhe-fft \
+		--no-default-features \
+		--features=fft128
+
+##### Tests #####
+
+.PHONY: test_fft
+test_fft: install_rs_build_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --release -p tfhe-fft
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --release -p tfhe-fft \
+		--features=fft128
+
+.PHONY: test_fft_serde
+test_fft_serde: install_rs_build_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --release -p tfhe-fft \
+		--features=serde
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --release -p tfhe-fft \
+		--features=serde,fft128
+
+.PHONY: test_fft_nightly
+test_fft_nightly: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --release -p tfhe-fft \
+		--features=nightly
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --release -p tfhe-fft \
+		--features=nightly,fft128
+
+.PHONY: test_fft_no_std
+test_fft_no_std: install_rs_build_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --release -p tfhe-fft \
+		--no-default-features 
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --release -p tfhe-fft \
+		--no-default-features \
+		--features=fft128
+
+.PHONY: test_fft_no_std_nightly
+test_fft_no_std_nightly: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --release -p tfhe-fft \
+		--no-default-features \
+		--features=nightly
+	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --release -p tfhe-fft \
+		--no-default-features \
+		--features=nightly,fft128
+
+.PHONY: test_fft_node_js
+test_fft_node_js: install_rs_build_toolchain install_build_wasm32_target install_wasm_bindgen_cli
+	RUSTFLAGS="" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --release \
+		--features=serde --target wasm32-unknown-unknown -p tfhe-fft
+
+.PHONY: test_fft_node_js_ci
+test_fft_node_js_ci: check_nvm_installed
+	source ~/.nvm/nvm.sh && \
+	nvm install $(NODE_VERSION) && \
+	nvm use $(NODE_VERSION) && \
+	"$(MAKE)" test_fft_node_js
+
+.PHONY: test_fft_all
+test_fft_all: test_fft test_fft_serde test_fft_nightly test_fft_no_std test_fft_no_std_nightly \
+test_fft_node_js_ci
+
+##### Bench #####
+
+.PHONY: bench_fft # Run FFT benchmarks
+bench_fft: install_rs_check_toolchain
+	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" bench --bench fft -p tfhe-fft \
+		--features=serde \
+		--features=nightly \
+		--features=fft128
+#============================End FFT Section ==================================
+
 .PHONY: help # Generate list of targets with descriptions
 help:
 	@grep '^\.PHONY: .* #' Makefile | sed 's/\.PHONY: \(.*\) # \(.*\)/\1\t\2/' | expand -t30 | sort

README.md

@@ -89,7 +89,7 @@ tfhe = { version = "*", features = ["boolean", "shortint", "integer", "x86_64"]
 ```
 
 > [!Note]
-> Note: You need to use a Rust version >= 1.73 to compile TFHE-rs.
+> Note: You need to use a Rust version >= 1.81 to compile TFHE-rs.
 
 > [!Note]
 > Note: aarch64-based machines are not yet supported for Windows as it's currently missing an entropy source to be able to seed the [CSPRNGs](https://en.wikipedia.org/wiki/Cryptographically_secure_pseudorandom_number_generator) used in TFHE-rs.
@@ -159,7 +159,7 @@ To run this code, use the following command:
 > Note that when running code that uses `TFHE-rs`, it is highly recommended
 to run in release mode with cargo's `--release` flag to have the best performances possible.
 
-*Find an example with more explanations in [this part of the documentation](https://docs.zama.ai/tfhe-rs/getting-started/quick_start)*
+*Find an example with more explanations in [this part of the documentation](https://docs.zama.ai/tfhe-rs/get-started/quick_start)*
 
 <p align="right">
   <a href="#about" > ↑ Back to top </a> 

_typos.toml

@@ -0,0 +1,15 @@
+[default]
+extend-ignore-identifiers-re = [
+    # Related to serialized object
+    "ser",
+    "unser",
+    # Used when dumping tfhe-rs parameters set into Sage format
+    "ND.*",
+    # Related to FHE strings example handling "banana"
+    "ba",
+    "enc_ba",
+    # Example with string replacing "hello" with "herlo"
+    "herlo",
+    # Example in trivium
+    "C9217BA0D762ACA1"
+]

apps/trivium/README.md

@@ -7,7 +7,7 @@ using multithreading to accelerate the computation.
 
 
 Quite similarly, the function `TriviumStream::<FheBool>::new` will return a very similar object running in FHE space. Its arguments are
-2 arrays of 80 FheBool representing the encrypted Trivium key, and the encrypted IV. It also requires a reference to the the server key of the 
+2 arrays of 80 FheBool representing the encrypted Trivium key, and the encrypted IV. It also requires a reference to the server key of the 
 current scheme. This means that any user of this feature must also have the `tfhe-rs` crate as a dependency.
 
 

apps/trivium/benches/kreyvium_shortint.rs

@@ -1,5 +1,6 @@
 use criterion::Criterion;
 use tfhe::prelude::*;
+use tfhe::shortint::parameters::PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64;
 use tfhe::shortint::prelude::*;
 use tfhe::{generate_keys, ConfigBuilder, FheUint64};
 use tfhe_trivium::{KreyviumStreamShortint, TransCiphering};
@@ -10,7 +11,8 @@ pub fn kreyvium_shortint_warmup(c: &mut Criterion) {
     let underlying_ck: tfhe::shortint::ClientKey = (*hl_client_key.as_ref()).clone().into();
     let underlying_sk: tfhe::shortint::ServerKey = (*hl_server_key.as_ref()).clone().into();
 
-    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+    let (client_key, server_key): (ClientKey, ServerKey) =
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
@@ -60,7 +62,8 @@ pub fn kreyvium_shortint_gen(c: &mut Criterion) {
     let underlying_ck: tfhe::shortint::ClientKey = (*hl_client_key.as_ref()).clone().into();
     let underlying_sk: tfhe::shortint::ServerKey = (*hl_server_key.as_ref()).clone().into();
 
-    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+    let (client_key, server_key): (ClientKey, ServerKey) =
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
@@ -105,7 +108,8 @@ pub fn kreyvium_shortint_trans(c: &mut Criterion) {
     let underlying_ck: tfhe::shortint::ClientKey = (*hl_client_key.as_ref()).clone().into();
     let underlying_sk: tfhe::shortint::ServerKey = (*hl_server_key.as_ref()).clone().into();
 
-    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+    let (client_key, server_key): (ClientKey, ServerKey) =
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),

apps/trivium/benches/trivium_shortint.rs

@@ -1,5 +1,6 @@
 use criterion::Criterion;
 use tfhe::prelude::*;
+use tfhe::shortint::parameters::PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64;
 use tfhe::shortint::prelude::*;
 use tfhe::{generate_keys, ConfigBuilder, FheUint64};
 use tfhe_trivium::{TransCiphering, TriviumStreamShortint};
@@ -10,7 +11,8 @@ pub fn trivium_shortint_warmup(c: &mut Criterion) {
     let underlying_ck: tfhe::shortint::ClientKey = (*hl_client_key.as_ref()).clone().into();
     let underlying_sk: tfhe::shortint::ServerKey = (*hl_server_key.as_ref()).clone().into();
 
-    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+    let (client_key, server_key): (ClientKey, ServerKey) =
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
@@ -60,7 +62,8 @@ pub fn trivium_shortint_gen(c: &mut Criterion) {
     let underlying_ck: tfhe::shortint::ClientKey = (*hl_client_key.as_ref()).clone().into();
     let underlying_sk: tfhe::shortint::ServerKey = (*hl_server_key.as_ref()).clone().into();
 
-    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+    let (client_key, server_key): (ClientKey, ServerKey) =
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
@@ -105,7 +108,8 @@ pub fn trivium_shortint_trans(c: &mut Criterion) {
     let underlying_ck: tfhe::shortint::ClientKey = (*hl_client_key.as_ref()).clone().into();
     let underlying_sk: tfhe::shortint::ServerKey = (*hl_server_key.as_ref()).clone().into();
 
-    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+    let (client_key, server_key): (ClientKey, ServerKey) =
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),

apps/trivium/src/kreyvium/kreyvium.rs

@@ -148,10 +148,9 @@ where
 
     /// Computes one turn of the stream, updating registers and outputting the new bit.
     pub fn next_bool(&mut self) -> T {
-        match &self.fhe_key {
-            Some(sk) => set_server_key(sk.clone()),
-            None => (),
-        };
+        if let Some(sk) = &self.fhe_key {
+            set_server_key(sk.clone());
+        }
 
         let [o, a, b, c] = self.get_output_and_values(0);
 
@@ -226,18 +225,12 @@ where
     /// Computes 64 turns of the stream, outputting the 64 bits all at once in a
     /// Vec (first value is oldest, last is newest)
     pub fn next_64(&mut self) -> Vec<T> {
-        match &self.fhe_key {
-            Some(sk) => {
-                rayon::broadcast(|_| set_server_key(sk.clone()));
-            }
-            None => (),
+        if let Some(sk) = &self.fhe_key {
+            rayon::broadcast(|_| set_server_key(sk.clone()));
         }
         let mut values = self.get_64_output_and_values();
-        match &self.fhe_key {
-            Some(_) => {
-                rayon::broadcast(|_| unset_server_key());
-            }
-            None => (),
+        if self.fhe_key.is_some() {
+            rayon::broadcast(|_| unset_server_key());
         }
 
         let mut ret = Vec::<T>::with_capacity(64);

apps/trivium/src/kreyvium/kreyvium_byte.rs

@@ -237,18 +237,12 @@ where
     /// Computes 64 turns of the stream, outputting the 64 bits (in 8 bytes) all at once in a
     /// Vec (first value is oldest, last is newest)
     pub fn next_64(&mut self) -> Vec<T> {
-        match &self.fhe_key {
-            Some(sk) => {
-                rayon::broadcast(|_| set_server_key(sk.clone()));
-            }
-            None => (),
+        if let Some(sk) = &self.fhe_key {
+            rayon::broadcast(|_| set_server_key(sk.clone()));
         }
         let values = self.get_64_output_and_values();
-        match &self.fhe_key {
-            Some(_) => {
-                rayon::broadcast(|_| unset_server_key());
-            }
-            None => (),
+        if self.fhe_key.is_some() {
+            rayon::broadcast(|_| unset_server_key());
         }
 
         let mut bytes = Vec::<T>::with_capacity(8);

apps/trivium/src/kreyvium/test.rs

@@ -1,7 +1,7 @@
 use crate::{KreyviumStream, KreyviumStreamByte, KreyviumStreamShortint, TransCiphering};
 use tfhe::prelude::*;
+use tfhe::shortint::parameters::PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64;
 use tfhe::{generate_keys, ConfigBuilder, FheBool, FheUint64, FheUint8};
-
 // Values for these tests come from the github repo renaud1239/Kreyvium,
 // commit fd6828f68711276c25f55e605935028f5e843f43
 
@@ -221,7 +221,8 @@ fn kreyvium_test_shortint_long() {
     let underlying_ck: tfhe::shortint::ClientKey = (*hl_client_key.as_ref()).clone().into();
     let underlying_sk: tfhe::shortint::ServerKey = (*hl_server_key.as_ref()).clone().into();
 
-    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+    let (client_key, server_key): (ClientKey, ServerKey) =
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),

apps/trivium/src/lib.rs

@@ -1,3 +1,5 @@
+#![allow(clippy::too_long_first_doc_paragraph)]
+
 mod static_deque;
 
 mod kreyvium;

apps/trivium/src/trivium/test.rs

@@ -1,7 +1,7 @@
 use crate::{TransCiphering, TriviumStream, TriviumStreamByte, TriviumStreamShortint};
 use tfhe::prelude::*;
+use tfhe::shortint::parameters::PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64;
 use tfhe::{generate_keys, ConfigBuilder, FheBool, FheUint64, FheUint8};
-
 // Values for these tests come from the github repo cantora/avr-crypto-lib, commit 2a5b018,
 // file testvectors/trivium-80.80.test-vectors
 
@@ -357,7 +357,8 @@ fn trivium_test_shortint_long() {
     let underlying_ck: tfhe::shortint::ClientKey = (*hl_client_key.as_ref()).clone().into();
     let underlying_sk: tfhe::shortint::ServerKey = (*hl_server_key.as_ref()).clone().into();
 
-    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+    let (client_key, server_key): (ClientKey, ServerKey) =
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),

apps/trivium/src/trivium/trivium_bool.rs

@@ -120,10 +120,9 @@ where
 
     /// Computes one turn of the stream, updating registers and outputting the new bit.
     pub fn next_bool(&mut self) -> T {
-        match &self.fhe_key {
-            Some(sk) => set_server_key(sk.clone()),
-            None => (),
-        };
+        if let Some(sk) = &self.fhe_key {
+            set_server_key(sk.clone());
+        }
 
         let [o, a, b, c] = self.get_output_and_values(0);
 
@@ -196,18 +195,12 @@ where
     /// Computes 64 turns of the stream, outputting the 64 bits all at once in a
     /// Vec (first value is oldest, last is newest)
     pub fn next_64(&mut self) -> Vec<T> {
-        match &self.fhe_key {
-            Some(sk) => {
-                rayon::broadcast(|_| set_server_key(sk.clone()));
-            }
-            None => (),
+        if let Some(sk) = &self.fhe_key {
+            rayon::broadcast(|_| set_server_key(sk.clone()));
         }
         let mut values = self.get_64_output_and_values();
-        match &self.fhe_key {
-            Some(_) => {
-                rayon::broadcast(|_| unset_server_key());
-            }
-            None => (),
+        if self.fhe_key.is_some() {
+            rayon::broadcast(|_| unset_server_key());
         }
 
         let mut ret = Vec::<T>::with_capacity(64);

apps/trivium/src/trivium/trivium_byte.rs

@@ -187,18 +187,12 @@ where
     /// Computes 64 turns of the stream, outputting the 64 bits (in 8 bytes) all at once in a
     /// Vec (first value is oldest, last is newest)
     pub fn next_64(&mut self) -> Vec<T> {
-        match &self.fhe_key {
-            Some(sk) => {
-                rayon::broadcast(|_| set_server_key(sk.clone()));
-            }
-            None => (),
+        if let Some(sk) = &self.fhe_key {
+            rayon::broadcast(|_| set_server_key(sk.clone()));
         }
         let values = self.get_64_output_and_values();
-        match &self.fhe_key {
-            Some(_) => {
-                rayon::broadcast(|_| unset_server_key());
-            }
-            None => (),
+        if self.fhe_key.is_some() {
+            rayon::broadcast(|_| unset_server_key());
         }
 
         let mut bytes = Vec::<T>::with_capacity(8);

backends/tfhe-cuda-backend/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tfhe-cuda-backend"
-version = "0.4.0-alpha.0"
+version = "0.6.0"
 edition = "2021"
 authors = ["Zama team"]
 license = "BSD-3-Clause-Clear"
@@ -14,3 +14,4 @@ keywords = ["fully", "homomorphic", "encryption", "fhe", "cryptography"]
 [build-dependencies]
 cmake = { version = "0.1" }
 pkg-config = { version = "0.3" }
+bindgen = "0.70.1"

backends/tfhe-cuda-backend/build.rs

@@ -1,8 +1,8 @@
-use std::env;
+use std::path::PathBuf;
 use std::process::Command;
 
 fn main() {
-    if let Ok(val) = env::var("DOCS_RS") {
+    if let Ok(val) = std::env::var("DOCS_RS") {
         if val.parse::<u32>() == Ok(1) {
             return;
         }
@@ -26,7 +26,8 @@ fn main() {
     println!("cargo::rerun-if-changed=cuda/tests_and_benchmarks");
     println!("cargo::rerun-if-changed=cuda/CMakeLists.txt");
     println!("cargo::rerun-if-changed=src");
-    if env::consts::OS == "linux" {
+
+    if std::env::consts::OS == "linux" {
         let output = Command::new("./get_os_name.sh").output().unwrap();
         let distribution = String::from_utf8(output.stdout).unwrap();
         if distribution != "Ubuntu\n" {
@@ -35,6 +36,7 @@ fn main() {
                 Only Ubuntu is supported by tfhe-cuda-backend at this time. Build may fail\n"
             );
         }
+
         let dest = cmake::build("cuda");
         println!("cargo:rustc-link-search=native={}", dest.display());
         println!("cargo:rustc-link-lib=static=tfhe_cuda_backend");
@@ -51,6 +53,57 @@ fn main() {
         println!("cargo:rustc-link-lib=cudart");
         println!("cargo:rustc-link-search=native=/usr/lib/x86_64-linux-gnu/");
         println!("cargo:rustc-link-lib=stdc++");
+
+        let header_path = "wrapper.h";
+        let headers = vec![
+            "wrapper.h",
+            "cuda/include/ciphertext.h",
+            "cuda/include/integer/compression/compression.h",
+            "cuda/include/integer/integer.h",
+            "cuda/include/keyswitch.h",
+            "cuda/include/linear_algebra.h",
+            "cuda/include/pbs/programmable_bootstrap.h",
+            "cuda/include/pbs/programmable_bootstrap_multibit.h",
+        ];
+        let out_path = PathBuf::from("src").join("bindings.rs");
+        let bindings_modified = if out_path.exists() {
+            std::fs::metadata(&out_path).unwrap().modified().unwrap()
+        } else {
+            std::time::SystemTime::UNIX_EPOCH // If bindings file doesn't exist, consider it older
+        };
+        let mut headers_modified = bindings_modified;
+        for header in headers {
+            println!("cargo:rerun-if-changed={}", header);
+            // Check modification times
+            let header_modified = std::fs::metadata(header).unwrap().modified().unwrap();
+            if header_modified > headers_modified {
+                headers_modified = header_modified;
+            }
+        }
+
+        // Regenerate bindings only if header has been modified
+        if headers_modified > bindings_modified {
+            let bindings = bindgen::Builder::default()
+                .header(header_path)
+                // allow only what we are interested in, the custom types appearing in the interface
+                .allowlist_type("PBS_TYPE")
+                .allowlist_type("SHIFT_OR_ROTATE_TYPE")
+                // and the functions reachable from the headers included in wrapper.h
+                .allowlist_function(".*")
+                .clang_arg("-x")
+                .clang_arg("c++")
+                .clang_arg("-std=c++17")
+                .clang_arg("-I/usr/include")
+                .clang_arg("-I/usr/local/include")
+                .ctypes_prefix("ffi")
+                .raw_line("use crate::ffi;")
+                .generate()
+                .expect("Unable to generate bindings");
+
+            bindings
+                .write_to_file(&out_path)
+                .expect("Couldn't write bindings!");
+        }
     } else {
         panic!(
             "Error: platform not supported, tfhe-cuda-backend not built (only Linux is supported)"

backends/tfhe-cuda-backend/cuda/CMakeLists.txt

@@ -67,9 +67,21 @@ endif()
 
 add_compile_definitions(CUDA_ARCH=${CUDA_ARCH})
 
+# Check if the DEBUG flag is defined
+if(CMAKE_BUILD_TYPE STREQUAL "Debug")
+  # Debug mode
+  message("Compiling in Debug mode")
+  add_definitions(-DDEBUG)
+  set(OPTIMIZATION_FLAGS "${OPTIMIZATION_FLAGS} -O0 -G -g")
+else()
+  # Release mode
+  message("Compiling in Release mode")
+  set(OPTIMIZATION_FLAGS "${OPTIMIZATION_FLAGS} -O3")
+endif()
+
 # in production, should use -arch=sm_70 --ptxas-options=-v to see register spills -lineinfo for better debugging
 set(CMAKE_CUDA_FLAGS
-    "${CMAKE_CUDA_FLAGS} -ccbin ${CMAKE_CXX_COMPILER} -O3 \
+    "${CMAKE_CUDA_FLAGS} -ccbin ${CMAKE_CXX_COMPILER} ${OPTIMIZATION_FLAGS}\
   -std=c++17 --no-exceptions  --expt-relaxed-constexpr -rdc=true \
   --use_fast_math -Xcompiler -fPIC")
 

backends/tfhe-cuda-backend/cuda/include/ciphertext.h

@@ -1,18 +1,24 @@
 #ifndef CUDA_CIPHERTEXT_H
 #define CUDA_CIPHERTEXT_H
 
-#include <cstdint>
+#include "stdint.h"
 
 extern "C" {
 void cuda_convert_lwe_ciphertext_vector_to_gpu_64(void *stream,
                                                   uint32_t gpu_index,
-                                                  void *dest, void *src,
+                                                  void *dest, void const *src,
                                                   uint32_t number_of_cts,
                                                   uint32_t lwe_dimension);
 void cuda_convert_lwe_ciphertext_vector_to_cpu_64(void *stream,
                                                   uint32_t gpu_index,
-                                                  void *dest, void *src,
+                                                  void *dest, void const *src,
                                                   uint32_t number_of_cts,
                                                   uint32_t lwe_dimension);
-};
+
+void cuda_glwe_sample_extract_64(void *stream, uint32_t gpu_index,
+                                 void *lwe_array_out, void const *glwe_array_in,
+                                 uint32_t const *nth_array, uint32_t num_nths,
+                                 uint32_t glwe_dimension,
+                                 uint32_t polynomial_size);
+}
 #endif

backends/tfhe-cuda-backend/cuda/include/device.h

@@ -39,16 +39,15 @@ void *cuda_malloc_async(uint64_t size, cudaStream_t stream, uint32_t gpu_index);
 
 void cuda_check_valid_malloc(uint64_t size, uint32_t gpu_index);
 
-bool cuda_check_support_cooperative_groups();
-
-bool cuda_check_support_thread_block_clusters();
-
 void cuda_memcpy_async_to_gpu(void *dest, void *src, uint64_t size,
                               cudaStream_t stream, uint32_t gpu_index);
 
-void cuda_memcpy_async_gpu_to_gpu(void *dest, void *src, uint64_t size,
+void cuda_memcpy_async_gpu_to_gpu(void *dest, void const *src, uint64_t size,
                                   cudaStream_t stream, uint32_t gpu_index);
 
+void cuda_memcpy_gpu_to_gpu(void *dest, void *src, uint64_t size,
+                            uint32_t gpu_index);
+
 void cuda_memcpy_async_to_cpu(void *dest, const void *src, uint64_t size,
                               cudaStream_t stream, uint32_t gpu_index);
 
@@ -62,9 +61,13 @@ void cuda_synchronize_device(uint32_t gpu_index);
 void cuda_drop(void *ptr, uint32_t gpu_index);
 
 void cuda_drop_async(void *ptr, cudaStream_t stream, uint32_t gpu_index);
+}
 
 int cuda_get_max_shared_memory(uint32_t gpu_index);
-}
+
+bool cuda_check_support_cooperative_groups();
+
+bool cuda_check_support_thread_block_clusters();
 
 template <typename Torus>
 void cuda_set_value_async(cudaStream_t stream, uint32_t gpu_index,

backends/tfhe-cuda-backend/cuda/include/helper_multi_gpu.h

@@ -8,7 +8,7 @@ extern std::mutex m;
 extern bool p2p_enabled;
 
 extern "C" {
-int cuda_setup_multi_gpu();
+int32_t cuda_setup_multi_gpu();
 }
 
 // Define a variant type that can be either a vector or a single pointer

backends/tfhe-cuda-backend/cuda/include/integer/compression/compression.h

@@ -0,0 +1,45 @@
+#ifndef CUDA_INTEGER_COMPRESSION_H
+#define CUDA_INTEGER_COMPRESSION_H
+
+#include "../../pbs/pbs_enums.h"
+
+extern "C" {
+void scratch_cuda_integer_compress_radix_ciphertext_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t compression_glwe_dimension,
+    uint32_t compression_polynomial_size, uint32_t lwe_dimension,
+    uint32_t ks_level, uint32_t ks_base_log, uint32_t num_radix_blocks,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    uint32_t lwe_per_glwe, uint32_t storage_log_modulus,
+    bool allocate_gpu_memory);
+
+void scratch_cuda_integer_decompress_radix_ciphertext_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t encryption_glwe_dimension,
+    uint32_t encryption_polynomial_size, uint32_t compression_glwe_dimension,
+    uint32_t compression_polynomial_size, uint32_t lwe_dimension,
+    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t num_radix_blocks,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    uint32_t storage_log_modulus, uint32_t body_count,
+    bool allocate_gpu_memory);
+
+void cuda_integer_compress_radix_ciphertext_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *glwe_array_out, void const *lwe_array_in, void *const *fp_ksk,
+    uint32_t num_nths, int8_t *mem_ptr);
+
+void cuda_integer_decompress_radix_ciphertext_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array_out, void const *glwe_in, uint32_t const *indexes_array,
+    uint32_t indexes_array_size, void *const *bsks, int8_t *mem_ptr);
+
+void cleanup_cuda_integer_compress_radix_ciphertext_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr_void);
+
+void cleanup_cuda_integer_decompress_radix_ciphertext_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr_void);
+}
+
+#endif

backends/tfhe-cuda-backend/cuda/include/integer/compression/compression_utilities.h

@@ -0,0 +1,124 @@
+#ifndef CUDA_INTEGER_COMPRESSION_UTILITIES_H
+#define CUDA_INTEGER_COMPRESSION_UTILITIES_H
+
+#include "../integer_utilities.h"
+
+template <typename Torus> struct int_compression {
+  int_radix_params compression_params;
+  uint32_t storage_log_modulus;
+  uint32_t lwe_per_glwe;
+
+  uint32_t body_count;
+
+  // Compression
+  int8_t *fp_ks_buffer;
+  Torus *tmp_lwe;
+  Torus *tmp_glwe_array_out;
+
+  int_compression(cudaStream_t const *streams, uint32_t const *gpu_indexes,
+                  uint32_t gpu_count, int_radix_params compression_params,
+                  uint32_t num_radix_blocks, uint32_t lwe_per_glwe,
+                  uint32_t storage_log_modulus, bool allocate_gpu_memory) {
+    this->compression_params = compression_params;
+    this->lwe_per_glwe = lwe_per_glwe;
+    this->storage_log_modulus = storage_log_modulus;
+    this->body_count = num_radix_blocks;
+
+    if (allocate_gpu_memory) {
+      Torus glwe_accumulator_size = (compression_params.glwe_dimension + 1) *
+                                    compression_params.polynomial_size;
+
+      tmp_lwe = (Torus *)cuda_malloc_async(
+          num_radix_blocks * (compression_params.small_lwe_dimension + 1) *
+              sizeof(Torus),
+          streams[0], gpu_indexes[0]);
+      tmp_glwe_array_out = (Torus *)cuda_malloc_async(
+          lwe_per_glwe * glwe_accumulator_size * sizeof(Torus), streams[0],
+          gpu_indexes[0]);
+
+      scratch_packing_keyswitch_lwe_list_to_glwe_64(
+          streams[0], gpu_indexes[0], &fp_ks_buffer,
+          compression_params.glwe_dimension, compression_params.polynomial_size,
+          num_radix_blocks, true);
+    }
+  }
+  void release(cudaStream_t const *streams, uint32_t const *gpu_indexes,
+               uint32_t gpu_count) {
+    cuda_drop_async(tmp_lwe, streams[0], gpu_indexes[0]);
+    cuda_drop_async(tmp_glwe_array_out, streams[0], gpu_indexes[0]);
+    cleanup_packing_keyswitch_lwe_list_to_glwe(streams[0], gpu_indexes[0],
+                                               &fp_ks_buffer);
+  }
+};
+
+template <typename Torus> struct int_decompression {
+  int_radix_params encryption_params;
+  int_radix_params compression_params;
+
+  uint32_t storage_log_modulus;
+
+  uint32_t num_radix_blocks;
+  uint32_t body_count;
+
+  Torus *tmp_extracted_glwe;
+  Torus *tmp_extracted_lwe;
+  uint32_t *tmp_indexes_array;
+
+  int_radix_lut<Torus> *carry_extract_lut;
+
+  int_decompression(cudaStream_t const *streams, uint32_t const *gpu_indexes,
+                    uint32_t gpu_count, int_radix_params encryption_params,
+                    int_radix_params compression_params,
+                    uint32_t num_radix_blocks, uint32_t body_count,
+                    uint32_t storage_log_modulus, bool allocate_gpu_memory) {
+    this->encryption_params = encryption_params;
+    this->compression_params = compression_params;
+    this->storage_log_modulus = storage_log_modulus;
+    this->num_radix_blocks = num_radix_blocks;
+    this->body_count = body_count;
+
+    if (allocate_gpu_memory) {
+      Torus glwe_accumulator_size = (compression_params.glwe_dimension + 1) *
+                                    compression_params.polynomial_size;
+      Torus lwe_accumulator_size = (compression_params.glwe_dimension *
+                                        compression_params.polynomial_size +
+                                    1);
+      carry_extract_lut = new int_radix_lut<Torus>(
+          streams, gpu_indexes, gpu_count, encryption_params, 1,
+          num_radix_blocks, allocate_gpu_memory);
+
+      tmp_extracted_glwe = (Torus *)cuda_malloc_async(
+          num_radix_blocks * glwe_accumulator_size * sizeof(Torus), streams[0],
+          gpu_indexes[0]);
+      tmp_indexes_array = (uint32_t *)cuda_malloc_async(
+          num_radix_blocks * sizeof(uint32_t), streams[0], gpu_indexes[0]);
+      tmp_extracted_lwe = (Torus *)cuda_malloc_async(
+          num_radix_blocks * lwe_accumulator_size * sizeof(Torus), streams[0],
+          gpu_indexes[0]);
+
+      // Carry extract LUT
+      auto carry_extract_f = [encryption_params](Torus x) -> Torus {
+        return x / encryption_params.message_modulus;
+      };
+
+      generate_device_accumulator<Torus>(
+          streams[0], gpu_indexes[0],
+          carry_extract_lut->get_lut(gpu_indexes[0], 0),
+          encryption_params.glwe_dimension, encryption_params.polynomial_size,
+          encryption_params.message_modulus, encryption_params.carry_modulus,
+          carry_extract_f);
+
+      carry_extract_lut->broadcast_lut(streams, gpu_indexes, gpu_indexes[0]);
+    }
+  }
+  void release(cudaStream_t const *streams, uint32_t const *gpu_indexes,
+               uint32_t gpu_count) {
+    cuda_drop_async(tmp_extracted_glwe, streams[0], gpu_indexes[0]);
+    cuda_drop_async(tmp_extracted_lwe, streams[0], gpu_indexes[0]);
+    cuda_drop_async(tmp_indexes_array, streams[0], gpu_indexes[0]);
+
+    carry_extract_lut->release(streams, gpu_indexes, gpu_count);
+    delete carry_extract_lut;
+  }
+};
+#endif

backends/tfhe-cuda-backend/cuda/include/integer/integer.h

@@ -0,0 +1,440 @@
+#ifndef CUDA_INTEGER_H
+#define CUDA_INTEGER_H
+
+#include "../pbs/pbs_enums.h"
+#include <stdint.h>
+
+enum OUTPUT_CARRY { NONE = 0, GENERATED = 1, PROPAGATED = 2 };
+enum SHIFT_OR_ROTATE_TYPE {
+  LEFT_SHIFT = 0,
+  RIGHT_SHIFT = 1,
+  LEFT_ROTATE = 2,
+  RIGHT_ROTATE = 3
+};
+enum BITOP_TYPE {
+  BITAND = 0,
+  BITOR = 1,
+  BITXOR = 2,
+  SCALAR_BITAND = 3,
+  SCALAR_BITOR = 4,
+  SCALAR_BITXOR = 5,
+};
+
+enum COMPARISON_TYPE {
+  EQ = 0,
+  NE = 1,
+  GT = 2,
+  GE = 3,
+  LT = 4,
+  LE = 5,
+  MAX = 6,
+  MIN = 7,
+};
+
+enum CMP_ORDERING { IS_INFERIOR = 0, IS_EQUAL = 1, IS_SUPERIOR = 2 };
+
+enum SIGNED_OPERATION { ADDITION = 1, SUBTRACTION = -1 };
+
+extern "C" {
+void scratch_cuda_apply_univariate_lut_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, void const *input_lut, uint32_t lwe_dimension,
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t input_lwe_ciphertext_count,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    bool allocate_gpu_memory);
+
+void cuda_apply_univariate_lut_kb_64(void *const *streams,
+                                     uint32_t const *gpu_indexes,
+                                     uint32_t gpu_count, void *output_radix_lwe,
+                                     void const *input_radix_lwe,
+                                     int8_t *mem_ptr, void *const *ksks,
+                                     void *const *bsks, uint32_t num_blocks);
+
+void cleanup_cuda_apply_univariate_lut_kb_64(void *const *streams,
+                                             uint32_t const *gpu_indexes,
+                                             uint32_t gpu_count,
+                                             int8_t **mem_ptr_void);
+
+void scratch_cuda_apply_bivariate_lut_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, void const *input_lut, uint32_t lwe_dimension,
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t input_lwe_ciphertext_count,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    bool allocate_gpu_memory);
+
+void cuda_apply_bivariate_lut_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *output_radix_lwe, void const *input_radix_lwe_1,
+    void const *input_radix_lwe_2, int8_t *mem_ptr, void *const *ksks,
+    void *const *bsks, uint32_t num_blocks, uint32_t shift);
+
+void cleanup_cuda_apply_bivariate_lut_kb_64(void *const *streams,
+                                            uint32_t const *gpu_indexes,
+                                            uint32_t gpu_count,
+                                            int8_t **mem_ptr_void);
+
+void cuda_apply_many_univariate_lut_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *output_radix_lwe, void const *input_radix_lwe, int8_t *mem_ptr,
+    void *const *ksks, void *const *bsks, uint32_t num_blocks,
+    uint32_t num_luts, uint32_t lut_stride);
+
+void scratch_cuda_full_propagation_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t lwe_dimension, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t ks_level, uint32_t ks_base_log,
+    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    bool allocate_gpu_memory);
+
+void cuda_full_propagation_64_inplace(void *const *streams,
+                                      uint32_t const *gpu_indexes,
+                                      uint32_t gpu_count, void *input_blocks,
+                                      int8_t *mem_ptr, void *const *ksks,
+                                      void *const *bsks, uint32_t num_blocks);
+
+void cleanup_cuda_full_propagation(void *const *streams,
+                                   uint32_t const *gpu_indexes,
+                                   uint32_t gpu_count, int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_mult_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t message_modulus, uint32_t carry_modulus,
+    uint32_t glwe_dimension, uint32_t lwe_dimension, uint32_t polynomial_size,
+    uint32_t pbs_base_log, uint32_t pbs_level, uint32_t ks_base_log,
+    uint32_t ks_level, uint32_t grouping_factor, uint32_t num_blocks,
+    PBS_TYPE pbs_type, bool allocate_gpu_memory);
+
+void cuda_integer_mult_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *radix_lwe_out, void const *radix_lwe_left,
+    void const *radix_lwe_right, void *const *bsks, void *const *ksks,
+    int8_t *mem_ptr, uint32_t polynomial_size, uint32_t num_blocks);
+
+void cleanup_cuda_integer_mult(void *const *streams,
+                               uint32_t const *gpu_indexes, uint32_t gpu_count,
+                               int8_t **mem_ptr_void);
+
+void cuda_negate_integer_radix_ciphertext_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array_out, void const *lwe_array_in, uint32_t lwe_dimension,
+    uint32_t lwe_ciphertext_count, uint32_t message_modulus,
+    uint32_t carry_modulus);
+
+void cuda_scalar_addition_integer_radix_ciphertext_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array, void const *scalar_input, uint32_t lwe_dimension,
+    uint32_t lwe_ciphertext_count, uint32_t message_modulus,
+    uint32_t carry_modulus);
+
+void scratch_cuda_integer_radix_logical_scalar_shift_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
+    uint32_t carry_modulus, PBS_TYPE pbs_type, SHIFT_OR_ROTATE_TYPE shift_type,
+    bool allocate_gpu_memory);
+
+void cuda_integer_radix_logical_scalar_shift_kb_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array, uint32_t shift, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t num_blocks);
+
+void scratch_cuda_integer_radix_arithmetic_scalar_shift_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
+    uint32_t carry_modulus, PBS_TYPE pbs_type, SHIFT_OR_ROTATE_TYPE shift_type,
+    bool allocate_gpu_memory);
+
+void cuda_integer_radix_arithmetic_scalar_shift_kb_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array, uint32_t shift, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t num_blocks);
+
+void cleanup_cuda_integer_radix_logical_scalar_shift(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr_void);
+
+void cleanup_cuda_integer_radix_arithmetic_scalar_shift(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_radix_shift_and_rotate_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
+    uint32_t carry_modulus, PBS_TYPE pbs_type, SHIFT_OR_ROTATE_TYPE shift_type,
+    bool is_signed, bool allocate_gpu_memory);
+
+void cuda_integer_radix_shift_and_rotate_kb_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array, void const *lwe_shift, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t num_blocks);
+
+void cleanup_cuda_integer_radix_shift_and_rotate(void *const *streams,
+                                                 uint32_t const *gpu_indexes,
+                                                 uint32_t gpu_count,
+                                                 int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_radix_comparison_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t lwe_ciphertext_count,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    COMPARISON_TYPE op_type, bool is_signed, bool allocate_gpu_memory);
+
+void cuda_comparison_integer_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array_out, void const *lwe_array_1, void const *lwe_array_2,
+    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
+    uint32_t lwe_ciphertext_count);
+
+void cuda_scalar_comparison_integer_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array_out, void const *lwe_array_in, void const *scalar_blocks,
+    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
+    uint32_t lwe_ciphertext_count, uint32_t num_scalar_blocks);
+
+void cleanup_cuda_integer_comparison(void *const *streams,
+                                     uint32_t const *gpu_indexes,
+                                     uint32_t gpu_count, int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_radix_bitop_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t lwe_ciphertext_count,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    BITOP_TYPE op_type, bool allocate_gpu_memory);
+
+void cuda_bitop_integer_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array_out, void const *lwe_array_1, void const *lwe_array_2,
+    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
+    uint32_t lwe_ciphertext_count);
+
+void cuda_scalar_bitop_integer_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array_out, void const *lwe_array_input, void const *clear_blocks,
+    uint32_t num_clear_blocks, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t lwe_ciphertext_count, BITOP_TYPE op);
+
+void cleanup_cuda_integer_bitop(void *const *streams,
+                                uint32_t const *gpu_indexes, uint32_t gpu_count,
+                                int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_radix_cmux_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t lwe_ciphertext_count,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    bool allocate_gpu_memory);
+
+void cuda_cmux_integer_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array_out, void const *lwe_condition, void const *lwe_array_true,
+    void const *lwe_array_false, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t lwe_ciphertext_count);
+
+void cleanup_cuda_integer_radix_cmux(void *const *streams,
+                                     uint32_t const *gpu_indexes,
+                                     uint32_t gpu_count, int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_radix_scalar_rotate_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
+    uint32_t carry_modulus, PBS_TYPE pbs_type, SHIFT_OR_ROTATE_TYPE shift_type,
+    bool allocate_gpu_memory);
+
+void cuda_integer_radix_scalar_rotate_kb_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array, uint32_t n, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t num_blocks);
+
+void cleanup_cuda_integer_radix_scalar_rotate(void *const *streams,
+                                              uint32_t const *gpu_indexes,
+                                              uint32_t gpu_count,
+                                              int8_t **mem_ptr_void);
+
+void scratch_cuda_propagate_single_carry_kb_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
+    uint32_t carry_modulus, PBS_TYPE pbs_type, bool allocate_gpu_memory);
+
+void cuda_propagate_single_carry_kb_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array, void *carry_out, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t num_blocks);
+
+void cuda_propagate_single_carry_get_input_carries_kb_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array, void *carry_out, void *input_carries, int8_t *mem_ptr,
+    void *const *bsks, void *const *ksks, uint32_t num_blocks);
+
+void cleanup_cuda_propagate_single_carry(void *const *streams,
+                                         uint32_t const *gpu_indexes,
+                                         uint32_t gpu_count,
+                                         int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_radix_partial_sum_ciphertexts_vec_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
+    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
+    uint32_t num_blocks_in_radix, uint32_t max_num_radix_in_vec,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    bool allocate_gpu_memory);
+
+void cuda_integer_radix_partial_sum_ciphertexts_vec_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *radix_lwe_out, void *radix_lwe_vec, uint32_t num_radix_in_vec,
+    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
+    uint32_t num_blocks_in_radix);
+
+void cleanup_cuda_integer_radix_partial_sum_ciphertexts_vec(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_radix_overflowing_sub_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
+    uint32_t carry_modulus, PBS_TYPE pbs_type, bool allocate_gpu_memory);
+
+void cuda_integer_radix_overflowing_sub_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *radix_lwe_out, void *radix_lwe_overflowed, void const *radix_lwe_left,
+    void const *radix_lwe_right, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t num_blocks_in_radix);
+
+void cleanup_cuda_integer_radix_overflowing_sub(void *const *streams,
+                                                uint32_t const *gpu_indexes,
+                                                uint32_t gpu_count,
+                                                int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_scalar_mul_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
+    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
+    uint32_t num_blocks, uint32_t message_modulus, uint32_t carry_modulus,
+    PBS_TYPE pbs_type, bool allocate_gpu_memory);
+
+void cuda_scalar_multiplication_integer_radix_ciphertext_64_inplace(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lwe_array, uint64_t const *decomposed_scalar,
+    uint64_t const *has_at_least_one_set, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t lwe_dimension, uint32_t polynomial_size,
+    uint32_t message_modulus, uint32_t num_blocks, uint32_t num_scalars);
+
+void cleanup_cuda_integer_radix_scalar_mul(void *const *streams,
+                                           uint32_t const *gpu_indexes,
+                                           uint32_t gpu_count,
+                                           int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_div_rem_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
+    uint32_t carry_modulus, PBS_TYPE pbs_type, bool allocate_gpu_memory);
+
+void cuda_integer_div_rem_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *quotient, void *remainder, void const *numerator, void const *divisor,
+    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
+    uint32_t num_blocks_in_radix);
+
+void cleanup_cuda_integer_div_rem(void *const *streams,
+                                  uint32_t const *gpu_indexes,
+                                  uint32_t gpu_count, int8_t **mem_ptr_void);
+
+void scratch_cuda_signed_overflowing_add_or_sub_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t big_lwe_dimension, uint32_t small_lwe_dimension, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_blocks, int8_t signed_operation,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    bool allocate_gpu_memory);
+
+void cuda_signed_overflowing_add_or_sub_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *lhs, void const *rhs, void *overflowed, int8_t signed_operation,
+    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
+    uint32_t num_blocks_in_radix);
+
+void cleanup_signed_overflowing_add_or_sub(void *const *streams,
+                                           uint32_t const *gpu_indexes,
+                                           uint32_t gpu_count,
+                                           int8_t **mem_ptr_void);
+
+void scratch_cuda_integer_compute_prefix_sum_hillis_steele_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, void const *input_lut, uint32_t lwe_dimension,
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t ks_level,
+    uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
+    uint32_t grouping_factor, uint32_t num_radix_blocks,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    bool allocate_gpu_memory);
+
+void cuda_integer_compute_prefix_sum_hillis_steele_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *output_radix_lwe, void *generates_or_propagates, int8_t *mem_ptr,
+    void *const *ksks, void *const *bsks, uint32_t num_blocks, uint32_t shift);
+
+void cleanup_cuda_integer_compute_prefix_sum_hillis_steele_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr_void);
+
+void cuda_integer_reverse_blocks_64_inplace(void *const *streams,
+                                            uint32_t const *gpu_indexes,
+                                            uint32_t gpu_count, void *lwe_array,
+                                            uint32_t num_blocks,
+                                            uint32_t lwe_size);
+
+void scratch_cuda_integer_abs_inplace_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    int8_t **mem_ptr, bool is_signed, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t big_lwe_dimension,
+    uint32_t small_lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
+    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
+    uint32_t num_blocks, uint32_t message_modulus, uint32_t carry_modulus,
+    PBS_TYPE pbs_type, bool allocate_gpu_memory);
+
+void cuda_integer_abs_inplace_radix_ciphertext_kb_64(
+    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
+    void *ct, int8_t *mem_ptr, bool is_signed, void *const *bsks,
+    void *const *ksks, uint32_t num_blocks);
+
+void cleanup_cuda_integer_abs_inplace(void *const *streams,
+                                      uint32_t const *gpu_indexes,
+                                      uint32_t gpu_count,
+                                      int8_t **mem_ptr_void);
+
+} // extern C
+#endif // CUDA_INTEGER_H

backends/tfhe-cuda-backend/cuda/include/keyswitch.h

@@ -1,21 +1,39 @@
 #ifndef CNCRT_KS_H_
 #define CNCRT_KS_H_
 
-#include <cstdint>
+#include <stdint.h>
 
 extern "C" {
 
 void cuda_keyswitch_lwe_ciphertext_vector_32(
     void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void *lwe_output_indexes, void *lwe_array_in, void *lwe_input_indexes,
-    void *ksk, uint32_t lwe_dimension_in, uint32_t lwe_dimension_out,
-    uint32_t base_log, uint32_t level_count, uint32_t num_samples);
+    void const *lwe_output_indexes, void const *lwe_array_in,
+    void const *lwe_input_indexes, void const *ksk, uint32_t lwe_dimension_in,
+    uint32_t lwe_dimension_out, uint32_t base_log, uint32_t level_count,
+    uint32_t num_samples);
 
 void cuda_keyswitch_lwe_ciphertext_vector_64(
     void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void *lwe_output_indexes, void *lwe_array_in, void *lwe_input_indexes,
-    void *ksk, uint32_t lwe_dimension_in, uint32_t lwe_dimension_out,
-    uint32_t base_log, uint32_t level_count, uint32_t num_samples);
+    void const *lwe_output_indexes, void const *lwe_array_in,
+    void const *lwe_input_indexes, void const *ksk, uint32_t lwe_dimension_in,
+    uint32_t lwe_dimension_out, uint32_t base_log, uint32_t level_count,
+    uint32_t num_samples);
+
+void scratch_packing_keyswitch_lwe_list_to_glwe_64(
+    void *stream, uint32_t gpu_index, int8_t **fp_ks_buffer,
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t num_lwes,
+    bool allocate_gpu_memory);
+
+void cuda_packing_keyswitch_lwe_list_to_glwe_64(
+    void *stream, uint32_t gpu_index, void *glwe_array_out,
+    void const *lwe_array_in, void const *fp_ksk_array, int8_t *fp_ks_buffer,
+    uint32_t input_lwe_dimension, uint32_t output_glwe_dimension,
+    uint32_t output_polynomial_size, uint32_t base_log, uint32_t level_count,
+    uint32_t num_lwes);
+
+void cleanup_packing_keyswitch_lwe_list_to_glwe(void *stream,
+                                                uint32_t gpu_index,
+                                                int8_t **fp_ks_buffer);
 }
 
 #endif // CNCRT_KS_H_

backends/tfhe-cuda-backend/cuda/include/linear_algebra.h

@@ -1,50 +1,48 @@
 #ifndef CUDA_LINALG_H_
 #define CUDA_LINALG_H_
 
-#include "programmable_bootstrap.h"
-#include <cstdint>
-#include <device.h>
+#include <stdint.h>
 
 extern "C" {
 
 void cuda_negate_lwe_ciphertext_vector_32(void *stream, uint32_t gpu_index,
                                           void *lwe_array_out,
-                                          void *lwe_array_in,
+                                          void const *lwe_array_in,
                                           uint32_t input_lwe_dimension,
                                           uint32_t input_lwe_ciphertext_count);
 void cuda_negate_lwe_ciphertext_vector_64(void *stream, uint32_t gpu_index,
                                           void *lwe_array_out,
-                                          void *lwe_array_in,
+                                          void const *lwe_array_in,
                                           uint32_t input_lwe_dimension,
                                           uint32_t input_lwe_ciphertext_count);
 void cuda_add_lwe_ciphertext_vector_32(void *stream, uint32_t gpu_index,
                                        void *lwe_array_out,
-                                       void *lwe_array_in_1,
-                                       void *lwe_array_in_2,
+                                       void const *lwe_array_in_1,
+                                       void const *lwe_array_in_2,
                                        uint32_t input_lwe_dimension,
                                        uint32_t input_lwe_ciphertext_count);
 void cuda_add_lwe_ciphertext_vector_64(void *stream, uint32_t gpu_index,
                                        void *lwe_array_out,
-                                       void *lwe_array_in_1,
-                                       void *lwe_array_in_2,
+                                       void const *lwe_array_in_1,
+                                       void const *lwe_array_in_2,
                                        uint32_t input_lwe_dimension,
                                        uint32_t input_lwe_ciphertext_count);
 void cuda_add_lwe_ciphertext_vector_plaintext_vector_32(
-    void *stream, uint32_t gpu_index, void *lwe_array_out, void *lwe_array_in,
-    void *plaintext_array_in, uint32_t input_lwe_dimension,
-    uint32_t input_lwe_ciphertext_count);
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_array_in, void const *plaintext_array_in,
+    uint32_t input_lwe_dimension, uint32_t input_lwe_ciphertext_count);
 void cuda_add_lwe_ciphertext_vector_plaintext_vector_64(
-    void *stream, uint32_t gpu_index, void *lwe_array_out, void *lwe_array_in,
-    void *plaintext_array_in, uint32_t input_lwe_dimension,
-    uint32_t input_lwe_ciphertext_count);
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_array_in, void const *plaintext_array_in,
+    uint32_t input_lwe_dimension, uint32_t input_lwe_ciphertext_count);
 void cuda_mult_lwe_ciphertext_vector_cleartext_vector_32(
-    void *stream, uint32_t gpu_index, void *lwe_array_out, void *lwe_array_in,
-    void *cleartext_array_in, uint32_t input_lwe_dimension,
-    uint32_t input_lwe_ciphertext_count);
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_array_in, void const *cleartext_array_in,
+    uint32_t input_lwe_dimension, uint32_t input_lwe_ciphertext_count);
 void cuda_mult_lwe_ciphertext_vector_cleartext_vector_64(
-    void *stream, uint32_t gpu_index, void *lwe_array_out, void *lwe_array_in,
-    void *cleartext_array_in, uint32_t input_lwe_dimension,
-    uint32_t input_lwe_ciphertext_count);
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_array_in, void const *cleartext_array_in,
+    uint32_t input_lwe_dimension, uint32_t input_lwe_ciphertext_count);
 }
 
 #endif // CUDA_LINALG_H_

backends/tfhe-cuda-backend/cuda/include/pbs/pbs_enums.h

@@ -0,0 +1,7 @@
+#ifndef CUDA_PBS_ENUMS_H
+#define CUDA_PBS_ENUMS_H
+
+enum PBS_TYPE { MULTI_BIT = 0, CLASSICAL = 1 };
+enum PBS_VARIANT { DEFAULT = 0, CG = 1, TBC = 2 };
+
+#endif // CUDA_PBS_ENUMS_H

backends/tfhe-cuda-backend/cuda/include/pbs/pbs_multibit_utilities.h

@@ -1,161 +1,112 @@
-#ifndef CUDA_MULTI_BIT_H
-#define CUDA_MULTI_BIT_H
+#ifndef CUDA_MULTI_BIT_UTILITIES_H
+#define CUDA_MULTI_BIT_UTILITIES_H
 
-#include "programmable_bootstrap.h"
-#include <cstdint>
-
-extern "C" {
-
-bool has_support_to_cuda_programmable_bootstrap_cg_multi_bit(
-    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
-    uint32_t num_samples, uint32_t max_shared_memory);
-
-void cuda_convert_lwe_multi_bit_programmable_bootstrap_key_64(
-    void *stream, uint32_t gpu_index, void *dest, void *src,
-    uint32_t input_lwe_dim, uint32_t glwe_dim, uint32_t level_count,
-    uint32_t polynomial_size, uint32_t grouping_factor);
-
-void scratch_cuda_multi_bit_programmable_bootstrap_64(
-    void *stream, uint32_t gpu_index, int8_t **pbs_buffer,
-    uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t level_count, uint32_t grouping_factor,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory, uint32_t chunk_size = 0);
-
-void cuda_multi_bit_programmable_bootstrap_lwe_ciphertext_vector_64(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void *lwe_output_indexes, void *lut_vector, void *lut_vector_indexes,
-    void *lwe_array_in, void *lwe_input_indexes, void *bootstrapping_key,
-    int8_t *buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t grouping_factor, uint32_t base_log,
-    uint32_t level_count, uint32_t num_samples, uint32_t max_shared_memory,
-    uint32_t lwe_chunk_size = 0);
-
-void cleanup_cuda_multi_bit_programmable_bootstrap(void *stream,
-                                                   uint32_t gpu_index,
-                                                   int8_t **pbs_buffer);
-}
+#include "pbs_utilities.h"
 
 template <typename Torus>
-__host__ bool
-supports_distributed_shared_memory_on_multibit_programmable_bootstrap(
-    uint32_t polynomial_size, uint32_t max_shared_memory);
+bool supports_distributed_shared_memory_on_multibit_programmable_bootstrap(
+    uint32_t polynomial_size);
 
 template <typename Torus>
 bool has_support_to_cuda_programmable_bootstrap_tbc_multi_bit(
     uint32_t num_samples, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t level_count, uint32_t max_shared_memory);
+    uint32_t level_count);
 
 #if CUDA_ARCH >= 900
 template <typename Torus>
 void scratch_cuda_tbc_multi_bit_programmable_bootstrap(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, MULTI_BIT> **buffer,
-    uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t level_count, uint32_t grouping_factor,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory, uint32_t lwe_chunk_size);
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
 
 template <typename Torus>
 void cuda_tbc_multi_bit_programmable_bootstrap_lwe_ciphertext_vector(
     void *stream, uint32_t gpu_index, Torus *lwe_array_out,
-    Torus *lwe_output_indexes, Torus *lut_vector, Torus *lut_vector_indexes,
-    Torus *lwe_array_in, Torus *lwe_input_indexes, Torus *bootstrapping_key,
+    Torus const *lwe_output_indexes, Torus const *lut_vector,
+    Torus const *lut_vector_indexes, Torus const *lwe_array_in,
+    Torus const *lwe_input_indexes, Torus const *bootstrapping_key,
     pbs_buffer<Torus, MULTI_BIT> *pbs_buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t grouping_factor,
     uint32_t base_log, uint32_t level_count, uint32_t num_samples,
-    uint32_t max_shared_memory, uint32_t lwe_chunk_size);
+    uint32_t lut_count, uint32_t lut_stride);
 #endif
 
-template <typename Torus>
-void scratch_cuda_cg_multi_bit_programmable_bootstrap(
-    void *stream, uint32_t gpu_index, pbs_buffer<Torus, MULTI_BIT> **pbs_buffer,
-    uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t level_count, uint32_t grouping_factor,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory, uint32_t lwe_chunk_size = 0);
-
 template <typename Torus>
 void scratch_cuda_cg_multi_bit_programmable_bootstrap(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, MULTI_BIT> **pbs_buffer,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory, uint32_t lwe_chunk_size = 0);
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
 
 template <typename Torus>
 void cuda_cg_multi_bit_programmable_bootstrap_lwe_ciphertext_vector(
     void *stream, uint32_t gpu_index, Torus *lwe_array_out,
-    Torus *lwe_output_indexes, Torus *lut_vector, Torus *lut_vector_indexes,
-    Torus *lwe_array_in, Torus *lwe_input_indexes, Torus *bootstrapping_key,
+    Torus const *lwe_output_indexes, Torus const *lut_vector,
+    Torus const *lut_vector_indexes, Torus const *lwe_array_in,
+    Torus const *lwe_input_indexes, Torus const *bootstrapping_key,
     pbs_buffer<Torus, MULTI_BIT> *pbs_buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t grouping_factor,
     uint32_t base_log, uint32_t level_count, uint32_t num_samples,
-    uint32_t max_shared_memory, uint32_t lwe_chunk_size = 0);
+    uint32_t lut_count, uint32_t lut_stride);
 
 template <typename Torus>
 void scratch_cuda_multi_bit_programmable_bootstrap(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, MULTI_BIT> **pbs_buffer,
-    uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t level_count, uint32_t grouping_factor,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory, uint32_t lwe_chunk_size = 0);
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
 
 template <typename Torus>
 void cuda_multi_bit_programmable_bootstrap_lwe_ciphertext_vector(
     void *stream, uint32_t gpu_index, Torus *lwe_array_out,
-    Torus *lwe_output_indexes, Torus *lut_vector, Torus *lut_vector_indexes,
-    Torus *lwe_array_in, Torus *lwe_input_indexes, Torus *bootstrapping_key,
+    Torus const *lwe_output_indexes, Torus const *lut_vector,
+    Torus const *lut_vector_indexes, Torus const *lwe_array_in,
+    Torus const *lwe_input_indexes, Torus const *bootstrapping_key,
     pbs_buffer<Torus, MULTI_BIT> *pbs_buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t grouping_factor,
     uint32_t base_log, uint32_t level_count, uint32_t num_samples,
-    uint32_t max_shared_memory, uint32_t lwe_chunk_size = 0);
+    uint32_t lut_count, uint32_t lut_stride);
 
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_full_sm_multibit_programmable_bootstrap_keybundle(
+uint64_t get_buffer_size_full_sm_multibit_programmable_bootstrap_keybundle(
     uint32_t polynomial_size);
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_full_sm_multibit_programmable_bootstrap_step_one(
+uint64_t get_buffer_size_full_sm_multibit_programmable_bootstrap_step_one(
     uint32_t polynomial_size);
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_full_sm_multibit_programmable_bootstrap_step_two(
+uint64_t get_buffer_size_full_sm_multibit_programmable_bootstrap_step_two(
     uint32_t polynomial_size);
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_partial_sm_multibit_programmable_bootstrap_step_one(
+uint64_t get_buffer_size_partial_sm_multibit_programmable_bootstrap_step_one(
     uint32_t polynomial_size);
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_full_sm_cg_multibit_programmable_bootstrap(
+uint64_t get_buffer_size_full_sm_cg_multibit_programmable_bootstrap(
     uint32_t polynomial_size);
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_partial_sm_cg_multibit_programmable_bootstrap(
+uint64_t get_buffer_size_partial_sm_cg_multibit_programmable_bootstrap(
     uint32_t polynomial_size);
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_sm_dsm_plus_tbc_multibit_programmable_bootstrap(
+uint64_t get_buffer_size_sm_dsm_plus_tbc_multibit_programmable_bootstrap(
     uint32_t polynomial_size);
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_partial_sm_tbc_multibit_programmable_bootstrap(
+uint64_t get_buffer_size_partial_sm_tbc_multibit_programmable_bootstrap(
     uint32_t polynomial_size);
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_full_sm_tbc_multibit_programmable_bootstrap(
+uint64_t get_buffer_size_full_sm_tbc_multibit_programmable_bootstrap(
     uint32_t polynomial_size);
 
+template <typename Torus, class params>
+uint32_t get_lwe_chunk_size(uint32_t gpu_index, uint32_t max_num_pbs,
+                            uint32_t polynomial_size);
+
 template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::MULTI_BIT> {
   int8_t *d_mem_keybundle = NULL;
   int8_t *d_mem_acc_step_one = NULL;
   int8_t *d_mem_acc_step_two = NULL;
   int8_t *d_mem_acc_cg = NULL;
   int8_t *d_mem_acc_tbc = NULL;
-
+  uint32_t lwe_chunk_size;
   double2 *keybundle_fft;
   Torus *global_accumulator;
-  double2 *global_accumulator_fft;
+  double2 *global_join_buffer;
 
   PBS_VARIANT pbs_variant;
 
@@ -164,6 +115,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::MULTI_BIT> {
              uint32_t input_lwe_ciphertext_count, uint32_t lwe_chunk_size,
              PBS_VARIANT pbs_variant, bool allocate_gpu_memory) {
     this->pbs_variant = pbs_variant;
+    this->lwe_chunk_size = lwe_chunk_size;
     auto max_shared_memory = cuda_get_max_shared_memory(gpu_index);
 
     // default
@@ -273,10 +225,12 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::MULTI_BIT> {
           num_blocks_keybundle * (polynomial_size / 2) * sizeof(double2),
           stream, gpu_index);
       global_accumulator = (Torus *)cuda_malloc_async(
-          num_blocks_acc_step_one * polynomial_size * sizeof(Torus), stream,
-          gpu_index);
-      global_accumulator_fft = (double2 *)cuda_malloc_async(
-          num_blocks_acc_step_one * (polynomial_size / 2) * sizeof(double2),
+          input_lwe_ciphertext_count * (glwe_dimension + 1) * polynomial_size *
+              sizeof(Torus),
+          stream, gpu_index);
+      global_join_buffer = (double2 *)cuda_malloc_async(
+          level_count * (glwe_dimension + 1) * input_lwe_ciphertext_count *
+              (polynomial_size / 2) * sizeof(double2),
           stream, gpu_index);
     }
   }
@@ -308,13 +262,8 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::MULTI_BIT> {
 
     cuda_drop_async(keybundle_fft, stream, gpu_index);
     cuda_drop_async(global_accumulator, stream, gpu_index);
-    cuda_drop_async(global_accumulator_fft, stream, gpu_index);
+    cuda_drop_async(global_join_buffer, stream, gpu_index);
   }
 };
 
-template <typename Torus, class params>
-__host__ uint32_t get_lwe_chunk_size(uint32_t gpu_index, uint32_t max_num_pbs,
-                                     uint32_t polynomial_size,
-                                     uint32_t max_shared_memory);
-
-#endif // CUDA_MULTI_BIT_H
+#endif // CUDA_MULTI_BIT_UTILITIES_H

backends/tfhe-cuda-backend/cuda/include/pbs/pbs_utilities.h

@@ -1,123 +1,32 @@
-#ifndef CUDA_BOOTSTRAP_H
-#define CUDA_BOOTSTRAP_H
+#ifndef CUDA_BOOTSTRAP_UTILITIES_H
+#define CUDA_BOOTSTRAP_UTILITIES_H
 
 #include "device.h"
-#include <cstdint>
-
-enum PBS_TYPE { MULTI_BIT = 0, CLASSICAL = 1 };
-enum PBS_VARIANT { DEFAULT = 0, CG = 1, TBC = 2 };
-
-extern "C" {
-void cuda_fourier_polynomial_mul(cudaStream_t stream, uint32_t gpu_index,
-                                 void *input1, void *input2, void *output,
-                                 uint32_t polynomial_size,
-                                 uint32_t total_polynomials);
-
-void cuda_convert_lwe_programmable_bootstrap_key_32(
-    void *stream, uint32_t gpu_index, void *dest, void *src,
-    uint32_t input_lwe_dim, uint32_t glwe_dim, uint32_t level_count,
-    uint32_t polynomial_size);
-
-void cuda_convert_lwe_programmable_bootstrap_key_64(
-    void *stream, uint32_t gpu_index, void *dest, void *src,
-    uint32_t input_lwe_dim, uint32_t glwe_dim, uint32_t level_count,
-    uint32_t polynomial_size);
-
-void scratch_cuda_programmable_bootstrap_amortized_32(
-    void *stream, uint32_t gpu_index, int8_t **pbs_buffer,
-    uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory);
-
-void scratch_cuda_programmable_bootstrap_amortized_64(
-    void *stream, uint32_t gpu_index, int8_t **pbs_buffer,
-    uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory);
-
-void cuda_programmable_bootstrap_amortized_lwe_ciphertext_vector_32(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void *lwe_output_indexes, void *lut_vector, void *lut_vector_indexes,
-    void *lwe_array_in, void *lwe_input_indexes, void *bootstrapping_key,
-    int8_t *pbs_buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
-    uint32_t num_samples, uint32_t max_shared_memory);
-
-void cuda_programmable_bootstrap_amortized_lwe_ciphertext_vector_64(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void *lwe_output_indexes, void *lut_vector, void *lut_vector_indexes,
-    void *lwe_array_in, void *lwe_input_indexes, void *bootstrapping_key,
-    int8_t *pbs_buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
-    uint32_t num_samples, uint32_t max_shared_memory);
-
-void cleanup_cuda_programmable_bootstrap_amortized(void *stream,
-                                                   uint32_t gpu_index,
-                                                   int8_t **pbs_buffer);
-
-void scratch_cuda_programmable_bootstrap_32(
-    void *stream, uint32_t gpu_index, int8_t **buffer, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory);
-
-void scratch_cuda_programmable_bootstrap_64(
-    void *stream, uint32_t gpu_index, int8_t **buffer, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory);
-
-void cuda_programmable_bootstrap_lwe_ciphertext_vector_32(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void *lwe_output_indexes, void *lut_vector, void *lut_vector_indexes,
-    void *lwe_array_in, void *lwe_input_indexes, void *bootstrapping_key,
-    int8_t *buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
-    uint32_t num_samples, uint32_t max_shared_memory);
-
-void cuda_programmable_bootstrap_lwe_ciphertext_vector_64(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void *lwe_output_indexes, void *lut_vector, void *lut_vector_indexes,
-    void *lwe_array_in, void *lwe_input_indexes, void *bootstrapping_key,
-    int8_t *buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
-    uint32_t num_samples, uint32_t max_shared_memory);
-
-void cleanup_cuda_programmable_bootstrap(void *stream, uint32_t gpu_index,
-                                         int8_t **pbs_buffer);
-
-uint64_t get_buffer_size_programmable_bootstrap_amortized_64(
-    uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory);
-
-uint64_t get_buffer_size_programmable_bootstrap_64(
-    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory);
-}
+#include "pbs_enums.h"
+#include "vector_types.h"
+#include <stdint.h>
 
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_full_sm_programmable_bootstrap_step_one(
+uint64_t get_buffer_size_full_sm_programmable_bootstrap_step_one(
     uint32_t polynomial_size) {
   return sizeof(Torus) * polynomial_size +      // accumulator_rotated
          sizeof(double2) * polynomial_size / 2; // accumulator fft
 }
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_full_sm_programmable_bootstrap_step_two(
+uint64_t get_buffer_size_full_sm_programmable_bootstrap_step_two(
     uint32_t polynomial_size) {
   return sizeof(Torus) * polynomial_size +      // accumulator
          sizeof(double2) * polynomial_size / 2; // accumulator fft
 }
 
 template <typename Torus>
-__host__ __device__ uint64_t
+uint64_t
 get_buffer_size_partial_sm_programmable_bootstrap(uint32_t polynomial_size) {
   return sizeof(double2) * polynomial_size / 2; // accumulator fft
 }
 
 template <typename Torus>
-__host__ __device__ uint64_t
+uint64_t
 get_buffer_size_full_sm_programmable_bootstrap_tbc(uint32_t polynomial_size) {
   return sizeof(Torus) * polynomial_size +      // accumulator_rotated
          sizeof(Torus) * polynomial_size +      // accumulator
@@ -125,21 +34,19 @@ get_buffer_size_full_sm_programmable_bootstrap_tbc(uint32_t polynomial_size) {
 }
 
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_partial_sm_programmable_bootstrap_tbc(
+uint64_t get_buffer_size_partial_sm_programmable_bootstrap_tbc(
     uint32_t polynomial_size) {
   return sizeof(double2) * polynomial_size / 2; // accumulator fft mask & body
 }
 
 template <typename Torus>
-__host__ __device__ uint64_t
-get_buffer_size_sm_dsm_plus_tbc_classic_programmable_bootstrap(
+uint64_t get_buffer_size_sm_dsm_plus_tbc_classic_programmable_bootstrap(
     uint32_t polynomial_size) {
   return sizeof(double2) * polynomial_size / 2; // tbc
 }
 
 template <typename Torus>
-__host__ __device__ uint64_t
+uint64_t
 get_buffer_size_full_sm_programmable_bootstrap_cg(uint32_t polynomial_size) {
   return sizeof(Torus) * polynomial_size +      // accumulator_rotated
          sizeof(Torus) * polynomial_size +      // accumulator
@@ -147,15 +54,14 @@ get_buffer_size_full_sm_programmable_bootstrap_cg(uint32_t polynomial_size) {
 }
 
 template <typename Torus>
-__host__ __device__ uint64_t
+uint64_t
 get_buffer_size_partial_sm_programmable_bootstrap_cg(uint32_t polynomial_size) {
   return sizeof(double2) * polynomial_size / 2; // accumulator fft mask & body
 }
 
 template <typename Torus>
-__host__ bool
-supports_distributed_shared_memory_on_classic_programmable_bootstrap(
-    uint32_t polynomial_size, uint32_t max_shared_memory);
+bool supports_distributed_shared_memory_on_classic_programmable_bootstrap(
+    uint32_t polynomial_size);
 
 template <typename Torus, PBS_TYPE pbs_type> struct pbs_buffer;
 
@@ -163,7 +69,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
   int8_t *d_mem;
 
   Torus *global_accumulator;
-  double2 *global_accumulator_fft;
+  double2 *global_join_buffer;
 
   PBS_VARIANT pbs_variant;
 
@@ -174,7 +80,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
 
     this->pbs_variant = pbs_variant;
 
-    auto max_shared_memory = cuda_get_max_shared_memory(gpu_index);
+    auto max_shared_memory = cuda_get_max_shared_memory(0);
 
     if (allocate_gpu_memory) {
       switch (pbs_variant) {
@@ -208,7 +114,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
         // Otherwise, both kernels run all in shared memory
         d_mem = (int8_t *)cuda_malloc_async(device_mem, stream, gpu_index);
 
-        global_accumulator_fft = (double2 *)cuda_malloc_async(
+        global_join_buffer = (double2 *)cuda_malloc_async(
             (glwe_dimension + 1) * level_count * input_lwe_ciphertext_count *
                 (polynomial_size / 2) * sizeof(double2),
             stream, gpu_index);
@@ -241,7 +147,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
         // Otherwise, both kernels run all in shared memory
         d_mem = (int8_t *)cuda_malloc_async(device_mem, stream, gpu_index);
 
-        global_accumulator_fft = (double2 *)cuda_malloc_async(
+        global_join_buffer = (double2 *)cuda_malloc_async(
             (glwe_dimension + 1) * level_count * input_lwe_ciphertext_count *
                 polynomial_size / 2 * sizeof(double2),
             stream, gpu_index);
@@ -251,7 +157,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
 
         bool supports_dsm =
             supports_distributed_shared_memory_on_classic_programmable_bootstrap<
-                Torus>(polynomial_size, max_shared_memory);
+                Torus>(polynomial_size);
 
         uint64_t full_sm =
             get_buffer_size_full_sm_programmable_bootstrap_tbc<Torus>(
@@ -288,7 +194,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
         // Otherwise, both kernels run all in shared memory
         d_mem = (int8_t *)cuda_malloc_async(device_mem, stream, gpu_index);
 
-        global_accumulator_fft = (double2 *)cuda_malloc_async(
+        global_join_buffer = (double2 *)cuda_malloc_async(
             (glwe_dimension + 1) * level_count * input_lwe_ciphertext_count *
                 polynomial_size / 2 * sizeof(double2),
             stream, gpu_index);
@@ -302,7 +208,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
 
   void release(cudaStream_t stream, uint32_t gpu_index) {
     cuda_drop_async(d_mem, stream, gpu_index);
-    cuda_drop_async(global_accumulator_fft, stream, gpu_index);
+    cuda_drop_async(global_join_buffer, stream, gpu_index);
 
     if (pbs_variant == DEFAULT)
       cuda_drop_async(global_accumulator, stream, gpu_index);
@@ -310,10 +216,10 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
 };
 
 template <typename Torus>
-__host__ __device__ uint64_t get_buffer_size_programmable_bootstrap_cg(
+uint64_t get_buffer_size_programmable_bootstrap_cg(
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory) {
-
+    uint32_t input_lwe_ciphertext_count) {
+  int max_shared_memory = cuda_get_max_shared_memory(0);
   uint64_t full_sm =
       get_buffer_size_full_sm_programmable_bootstrap_cg<Torus>(polynomial_size);
   uint64_t partial_sm =
@@ -339,65 +245,66 @@ template <typename Torus>
 bool has_support_to_cuda_programmable_bootstrap_cg(uint32_t glwe_dimension,
                                                    uint32_t polynomial_size,
                                                    uint32_t level_count,
-                                                   uint32_t num_samples,
-                                                   uint32_t max_shared_memory);
+                                                   uint32_t num_samples);
 
 template <typename Torus>
 void cuda_programmable_bootstrap_cg_lwe_ciphertext_vector(
     void *stream, uint32_t gpu_index, Torus *lwe_array_out,
-    Torus *lwe_output_indexes, Torus *lut_vector, Torus *lut_vector_indexes,
-    Torus *lwe_array_in, Torus *lwe_input_indexes, double2 *bootstrapping_key,
+    Torus const *lwe_output_indexes, Torus const *lut_vector,
+    Torus const *lut_vector_indexes, Torus const *lwe_array_in,
+    Torus const *lwe_input_indexes, double2 const *bootstrapping_key,
     pbs_buffer<Torus, CLASSICAL> *buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t base_log,
-    uint32_t level_count, uint32_t num_samples, uint32_t max_shared_memory);
+    uint32_t level_count, uint32_t num_samples, uint32_t lut_count,
+    uint32_t lut_stride);
 
 template <typename Torus>
 void cuda_programmable_bootstrap_lwe_ciphertext_vector(
     void *stream, uint32_t gpu_index, Torus *lwe_array_out,
-    Torus *lwe_output_indexes, Torus *lut_vector, Torus *lut_vector_indexes,
-    Torus *lwe_array_in, Torus *lwe_input_indexes, double2 *bootstrapping_key,
+    Torus const *lwe_output_indexes, Torus const *lut_vector,
+    Torus const *lut_vector_indexes, Torus const *lwe_array_in,
+    Torus const *lwe_input_indexes, double2 const *bootstrapping_key,
     pbs_buffer<Torus, CLASSICAL> *buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t base_log,
-    uint32_t level_count, uint32_t num_samples, uint32_t max_shared_memory);
+    uint32_t level_count, uint32_t num_samples, uint32_t lut_count,
+    uint32_t lut_stride);
 
 #if (CUDA_ARCH >= 900)
 template <typename Torus>
 void cuda_programmable_bootstrap_tbc_lwe_ciphertext_vector(
     void *stream, uint32_t gpu_index, Torus *lwe_array_out,
-    Torus *lwe_output_indexes, Torus *lut_vector, Torus *lut_vector_indexes,
-    Torus *lwe_array_in, Torus *lwe_input_indexes, double2 *bootstrapping_key,
+    Torus const *lwe_output_indexes, Torus const *lut_vector,
+    Torus const *lut_vector_indexes, Torus const *lwe_array_in,
+    Torus const *lwe_input_indexes, double2 const *bootstrapping_key,
     pbs_buffer<Torus, CLASSICAL> *buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t base_log,
-    uint32_t level_count, uint32_t num_samples, uint32_t max_shared_memory);
+    uint32_t level_count, uint32_t num_samples, uint32_t lut_count,
+    uint32_t lut_stride);
 
 template <typename Torus>
 void scratch_cuda_programmable_bootstrap_tbc(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, CLASSICAL> **pbs_buffer,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory);
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
 #endif
 
 template <typename Torus>
 void scratch_cuda_programmable_bootstrap_cg(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, CLASSICAL> **pbs_buffer,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory);
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
 
 template <typename Torus>
 void scratch_cuda_programmable_bootstrap(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, CLASSICAL> **buffer,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, uint32_t max_shared_memory,
-    bool allocate_gpu_memory);
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
 
 template <typename Torus>
 bool has_support_to_cuda_programmable_bootstrap_tbc(uint32_t num_samples,
                                                     uint32_t glwe_dimension,
                                                     uint32_t polynomial_size,
-                                                    uint32_t level_count,
-                                                    uint32_t max_shared_memory);
+                                                    uint32_t level_count);
 
 #ifdef __CUDACC__
 __device__ inline int get_start_ith_ggsw(int i, uint32_t polynomial_size,
@@ -427,4 +334,4 @@ __device__ const T *get_multi_bit_ith_lwe_gth_group_kth_block(
 
 #endif
 
-#endif // CUDA_BOOTSTRAP_H
+#endif // CUDA_BOOTSTRAP_UTILITIES_H

backends/tfhe-cuda-backend/cuda/include/pbs/programmable_bootstrap.h

@@ -0,0 +1,86 @@
+#ifndef CUDA_BOOTSTRAP_H
+#define CUDA_BOOTSTRAP_H
+
+#include "pbs_enums.h"
+#include <stdint.h>
+
+extern "C" {
+void cuda_fourier_polynomial_mul(void *stream, uint32_t gpu_index,
+                                 void const *input1, void const *input2,
+                                 void *output, uint32_t polynomial_size,
+                                 uint32_t total_polynomials);
+
+void cuda_convert_lwe_programmable_bootstrap_key_32(
+    void *stream, uint32_t gpu_index, void *dest, void const *src,
+    uint32_t input_lwe_dim, uint32_t glwe_dim, uint32_t level_count,
+    uint32_t polynomial_size);
+
+void cuda_convert_lwe_programmable_bootstrap_key_64(
+    void *stream, uint32_t gpu_index, void *dest, void const *src,
+    uint32_t input_lwe_dim, uint32_t glwe_dim, uint32_t level_count,
+    uint32_t polynomial_size);
+
+void scratch_cuda_programmable_bootstrap_amortized_32(
+    void *stream, uint32_t gpu_index, int8_t **pbs_buffer,
+    uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
+
+void scratch_cuda_programmable_bootstrap_amortized_64(
+    void *stream, uint32_t gpu_index, int8_t **pbs_buffer,
+    uint32_t glwe_dimension, uint32_t polynomial_size,
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
+
+void cuda_programmable_bootstrap_amortized_lwe_ciphertext_vector_32(
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_output_indexes, void const *lut_vector,
+    void const *lut_vector_indexes, void const *lwe_array_in,
+    void const *lwe_input_indexes, void const *bootstrapping_key,
+    int8_t *pbs_buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
+    uint32_t num_samples);
+
+void cuda_programmable_bootstrap_amortized_lwe_ciphertext_vector_64(
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_output_indexes, void const *lut_vector,
+    void const *lut_vector_indexes, void const *lwe_array_in,
+    void const *lwe_input_indexes, void const *bootstrapping_key,
+    int8_t *pbs_buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
+    uint32_t num_samples);
+
+void cleanup_cuda_programmable_bootstrap_amortized(void *stream,
+                                                   uint32_t gpu_index,
+                                                   int8_t **pbs_buffer);
+
+void scratch_cuda_programmable_bootstrap_32(
+    void *stream, uint32_t gpu_index, int8_t **buffer, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t level_count,
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
+
+void scratch_cuda_programmable_bootstrap_64(
+    void *stream, uint32_t gpu_index, int8_t **buffer, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t level_count,
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
+
+void cuda_programmable_bootstrap_lwe_ciphertext_vector_32(
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_output_indexes, void const *lut_vector,
+    void const *lut_vector_indexes, void const *lwe_array_in,
+    void const *lwe_input_indexes, void const *bootstrapping_key,
+    int8_t *buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
+    uint32_t num_samples, uint32_t lut_count, uint32_t lut_stride);
+
+void cuda_programmable_bootstrap_lwe_ciphertext_vector_64(
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_output_indexes, void const *lut_vector,
+    void const *lut_vector_indexes, void const *lwe_array_in,
+    void const *lwe_input_indexes, void const *bootstrapping_key,
+    int8_t *buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
+    uint32_t num_samples, uint32_t lut_count, uint32_t lut_stride);
+
+void cleanup_cuda_programmable_bootstrap(void *stream, uint32_t gpu_index,
+                                         int8_t **pbs_buffer);
+}
+#endif // CUDA_BOOTSTRAP_H

backends/tfhe-cuda-backend/cuda/include/pbs/programmable_bootstrap_multibit.h

@@ -0,0 +1,38 @@
+#ifndef CUDA_MULTI_BIT_H
+#define CUDA_MULTI_BIT_H
+
+#include "pbs_enums.h"
+#include "stdint.h"
+
+extern "C" {
+
+bool has_support_to_cuda_programmable_bootstrap_cg_multi_bit(
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
+    uint32_t num_samples);
+
+void cuda_convert_lwe_multi_bit_programmable_bootstrap_key_64(
+    void *stream, uint32_t gpu_index, void *dest, void const *src,
+    uint32_t input_lwe_dim, uint32_t glwe_dim, uint32_t level_count,
+    uint32_t polynomial_size, uint32_t grouping_factor);
+
+void scratch_cuda_multi_bit_programmable_bootstrap_64(
+    void *stream, uint32_t gpu_index, int8_t **pbs_buffer,
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
+    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
+
+void cuda_multi_bit_programmable_bootstrap_lwe_ciphertext_vector_64(
+    void *stream, uint32_t gpu_index, void *lwe_array_out,
+    void const *lwe_output_indexes, void const *lut_vector,
+    void const *lut_vector_indexes, void const *lwe_array_in,
+    void const *lwe_input_indexes, void const *bootstrapping_key,
+    int8_t *buffer, uint32_t lwe_dimension, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t grouping_factor, uint32_t base_log,
+    uint32_t level_count, uint32_t num_samples, uint32_t lut_count,
+    uint32_t lut_stride);
+
+void cleanup_cuda_multi_bit_programmable_bootstrap(void *stream,
+                                                   uint32_t gpu_index,
+                                                   int8_t **pbs_buffer);
+}
+
+#endif // CUDA_MULTI_BIT_H

backends/tfhe-cuda-backend/cuda/src/CMakeLists.txt

@@ -1,17 +1,3 @@
-set(SOURCES
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/bit_extraction.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/bitwise_ops.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/bootstrap.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/bootstrap_multibit.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/ciphertext.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/circuit_bootstrap.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/device.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/integer.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/keyswitch.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/linear_algebra.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/shifts.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/vertical_packing.h
-    ${CMAKE_SOURCE_DIR}/${INCLUDE_DIR}/helper_multi_gpu.h)
 file(GLOB_RECURSE SOURCES "*.cu")
 add_library(tfhe_cuda_backend STATIC ${SOURCES})
 set_target_properties(tfhe_cuda_backend PROPERTIES CUDA_SEPARABLE_COMPILATION ON CUDA_RESOLVE_DEVICE_SYMBOLS ON)

backends/tfhe-cuda-backend/cuda/src/crypto/ciphertext.cu

@@ -1,4 +1,5 @@
 #include "ciphertext.cuh"
+#include "polynomial/parameters.cuh"
 
 void cuda_convert_lwe_ciphertext_vector_to_gpu_64(void *stream,
                                                   uint32_t gpu_index,
@@ -19,3 +20,58 @@ void cuda_convert_lwe_ciphertext_vector_to_cpu_64(void *stream,
       static_cast<cudaStream_t>(stream), gpu_index, (uint64_t *)dest,
       (uint64_t *)src, number_of_cts, lwe_dimension);
 }
+
+void cuda_glwe_sample_extract_64(void *stream, uint32_t gpu_index,
+                                 void *lwe_array_out, void const *glwe_array_in,
+                                 uint32_t const *nth_array, uint32_t num_nths,
+                                 uint32_t glwe_dimension,
+                                 uint32_t polynomial_size) {
+
+  switch (polynomial_size) {
+  case 256:
+    host_sample_extract<uint64_t, AmortizedDegree<256>>(
+        static_cast<cudaStream_t>(stream), gpu_index, (uint64_t *)lwe_array_out,
+        (uint64_t const *)glwe_array_in, (uint32_t const *)nth_array, num_nths,
+        glwe_dimension);
+    break;
+  case 512:
+    host_sample_extract<uint64_t, AmortizedDegree<512>>(
+        static_cast<cudaStream_t>(stream), gpu_index, (uint64_t *)lwe_array_out,
+        (uint64_t const *)glwe_array_in, (uint32_t const *)nth_array, num_nths,
+        glwe_dimension);
+    break;
+  case 1024:
+    host_sample_extract<uint64_t, AmortizedDegree<1024>>(
+        static_cast<cudaStream_t>(stream), gpu_index, (uint64_t *)lwe_array_out,
+        (uint64_t const *)glwe_array_in, (uint32_t const *)nth_array, num_nths,
+        glwe_dimension);
+    break;
+  case 2048:
+    host_sample_extract<uint64_t, AmortizedDegree<2048>>(
+        static_cast<cudaStream_t>(stream), gpu_index, (uint64_t *)lwe_array_out,
+        (uint64_t const *)glwe_array_in, (uint32_t const *)nth_array, num_nths,
+        glwe_dimension);
+    break;
+  case 4096:
+    host_sample_extract<uint64_t, AmortizedDegree<4096>>(
+        static_cast<cudaStream_t>(stream), gpu_index, (uint64_t *)lwe_array_out,
+        (uint64_t const *)glwe_array_in, (uint32_t const *)nth_array, num_nths,
+        glwe_dimension);
+    break;
+  case 8192:
+    host_sample_extract<uint64_t, AmortizedDegree<8192>>(
+        static_cast<cudaStream_t>(stream), gpu_index, (uint64_t *)lwe_array_out,
+        (uint64_t const *)glwe_array_in, (uint32_t const *)nth_array, num_nths,
+        glwe_dimension);
+    break;
+  case 16384:
+    host_sample_extract<uint64_t, AmortizedDegree<16384>>(
+        static_cast<cudaStream_t>(stream), gpu_index, (uint64_t *)lwe_array_out,
+        (uint64_t const *)glwe_array_in, (uint32_t const *)nth_array, num_nths,
+        glwe_dimension);
+    break;
+  default:
+    PANIC("Cuda error: unsupported polynomial size. Supported "
+          "N's are powers of two in the interval [256..16384].")
+  }
+}

backends/tfhe-cuda-backend/cuda/src/crypto/ciphertext.cuh

@@ -3,6 +3,7 @@
 
 #include "ciphertext.h"
 #include "device.h"
+#include "polynomial/functions.cuh"
 #include <cstdint>
 
 template <typename T>
@@ -25,4 +26,42 @@ void cuda_convert_lwe_ciphertext_vector_to_cpu(cudaStream_t stream,
   cuda_memcpy_async_to_cpu(dest, src, size, stream, gpu_index);
 }
 
+template <typename Torus, class params>
+__global__ void sample_extract(Torus *lwe_array_out, Torus const *glwe_array_in,
+                               uint32_t const *nth_array,
+                               uint32_t glwe_dimension) {
+
+  const int input_id = blockIdx.x;
+
+  const int glwe_input_size = (glwe_dimension + 1) * params::degree;
+  const int lwe_output_size = glwe_dimension * params::degree + 1;
+
+  auto lwe_out = lwe_array_out + input_id * lwe_output_size;
+
+  // We assume each GLWE will store the first polynomial_size inputs
+  uint32_t lwe_per_glwe = params::degree;
+  auto glwe_in = glwe_array_in + (input_id / lwe_per_glwe) * glwe_input_size;
+
+  // nth is ensured to be in [0, lwe_per_glwe)
+  auto nth = nth_array[input_id] % lwe_per_glwe;
+
+  sample_extract_mask<Torus, params>(lwe_out, glwe_in, glwe_dimension, nth);
+  sample_extract_body<Torus, params>(lwe_out, glwe_in, glwe_dimension, nth);
+}
+
+template <typename Torus, class params>
+__host__ void host_sample_extract(cudaStream_t stream, uint32_t gpu_index,
+                                  Torus *lwe_array_out,
+                                  Torus const *glwe_array_in,
+                                  uint32_t const *nth_array, uint32_t num_nths,
+                                  uint32_t glwe_dimension) {
+  cudaSetDevice(gpu_index);
+
+  dim3 grid(num_nths);
+  dim3 thds(params::degree / params::opt);
+  sample_extract<Torus, params><<<grid, thds, 0, stream>>>(
+      lwe_array_out, glwe_array_in, nth_array, glwe_dimension);
+  check_cuda_error(cudaGetLastError());
+}
+
 #endif

backends/tfhe-cuda-backend/cuda/src/crypto/gadget.cuh

@@ -1,6 +1,7 @@
 #ifndef CNCRT_CRYPTO_CUH
 #define CNCRT_CRPYTO_CUH
 
+#include "crypto/torus.cuh"
 #include "device.h"
 #include <cstdint>
 
@@ -21,7 +22,6 @@ private:
   uint32_t base_log;
   uint32_t mask;
   uint32_t num_poly;
-  int current_level;
   T mask_mod_b;
   T *state;
 
@@ -32,13 +32,6 @@ public:
         state(state) {
 
     mask_mod_b = (1ll << base_log) - 1ll;
-    current_level = level_count;
-    int tid = threadIdx.x;
-    for (int i = 0; i < num_poly * params::opt; i++) {
-      state[tid] >>= (sizeof(T) * 8 - base_log * level_count);
-      tid += params::degree / params::opt;
-    }
-    synchronize_threads_in_block();
   }
 
   // Decomposes all polynomials at once
@@ -52,28 +45,30 @@ public:
   // Decomposes a single polynomial
   __device__ void decompose_and_compress_next_polynomial(double2 *result,
                                                          int j) {
-    if (j == 0)
-      current_level -= 1;
-
-    int tid = threadIdx.x;
-    auto state_slice = state + j * params::degree;
+    uint32_t tid = threadIdx.x;
+    auto state_slice = &state[j * params::degree];
     for (int i = 0; i < params::opt / 2; i++) {
-      T res_re = state_slice[tid] & mask_mod_b;
-      T res_im = state_slice[tid + params::degree / 2] & mask_mod_b;
-      state_slice[tid] >>= base_log;
-      state_slice[tid + params::degree / 2] >>= base_log;
-      T carry_re = ((res_re - 1ll) | state_slice[tid]) & res_re;
-      T carry_im =
-          ((res_im - 1ll) | state_slice[tid + params::degree / 2]) & res_im;
+      auto input1 = &state_slice[tid];
+      auto input2 = &state_slice[tid + params::degree / 2];
+      T res_re = *input1 & mask_mod_b;
+      T res_im = *input2 & mask_mod_b;
+
+      *input1 >>= base_log; // Update state
+      *input2 >>= base_log; // Update state
+
+      T carry_re = ((res_re - 1ll) | *input1) & res_re;
+      T carry_im = ((res_im - 1ll) | *input2) & res_im;
       carry_re >>= (base_log - 1);
       carry_im >>= (base_log - 1);
-      state_slice[tid] += carry_re;
-      state_slice[tid + params::degree / 2] += carry_im;
+
+      *input1 += carry_re; // Update state
+      *input2 += carry_im; // Update state
+
       res_re -= carry_re << base_log;
       res_im -= carry_im << base_log;
 
-      result[tid].x = (int32_t)res_re;
-      result[tid].y = (int32_t)res_im;
+      typecast_torus_to_double(res_re, result[tid].x);
+      typecast_torus_to_double(res_im, result[tid].y);
 
       tid += params::degree / params::opt;
     }

backends/tfhe-cuda-backend/cuda/src/crypto/keyswitch.cu

@@ -10,7 +10,7 @@ void cuda_keyswitch_lwe_ciphertext_vector_32(
     void *lwe_output_indexes, void *lwe_array_in, void *lwe_input_indexes,
     void *ksk, uint32_t lwe_dimension_in, uint32_t lwe_dimension_out,
     uint32_t base_log, uint32_t level_count, uint32_t num_samples) {
-  cuda_keyswitch_lwe_ciphertext_vector(
+  host_keyswitch_lwe_ciphertext_vector<uint32_t>(
       static_cast<cudaStream_t>(stream), gpu_index,
       static_cast<uint32_t *>(lwe_array_out),
       static_cast<uint32_t *>(lwe_output_indexes),
@@ -37,14 +37,49 @@ void cuda_keyswitch_lwe_ciphertext_vector_32(
  */
 void cuda_keyswitch_lwe_ciphertext_vector_64(
     void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void *lwe_output_indexes, void *lwe_array_in, void *lwe_input_indexes,
-    void *ksk, uint32_t lwe_dimension_in, uint32_t lwe_dimension_out,
-    uint32_t base_log, uint32_t level_count, uint32_t num_samples) {
-  cuda_keyswitch_lwe_ciphertext_vector(
+    void const *lwe_output_indexes, void const *lwe_array_in,
+    void const *lwe_input_indexes, void const *ksk, uint32_t lwe_dimension_in,
+    uint32_t lwe_dimension_out, uint32_t base_log, uint32_t level_count,
+    uint32_t num_samples) {
+  host_keyswitch_lwe_ciphertext_vector<uint64_t>(
       static_cast<cudaStream_t>(stream), gpu_index,
       static_cast<uint64_t *>(lwe_array_out),
-      static_cast<uint64_t *>(lwe_output_indexes),
-      static_cast<uint64_t *>(lwe_array_in),
-      static_cast<uint64_t *>(lwe_input_indexes), static_cast<uint64_t *>(ksk),
-      lwe_dimension_in, lwe_dimension_out, base_log, level_count, num_samples);
+      static_cast<const uint64_t *>(lwe_output_indexes),
+      static_cast<const uint64_t *>(lwe_array_in),
+      static_cast<const uint64_t *>(lwe_input_indexes),
+      static_cast<const uint64_t *>(ksk), lwe_dimension_in, lwe_dimension_out,
+      base_log, level_count, num_samples);
+}
+
+void scratch_packing_keyswitch_lwe_list_to_glwe_64(
+    void *stream, uint32_t gpu_index, int8_t **fp_ks_buffer,
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t num_lwes,
+    bool allocate_gpu_memory) {
+  scratch_packing_keyswitch_lwe_list_to_glwe<uint64_t>(
+      static_cast<cudaStream_t>(stream), gpu_index, fp_ks_buffer,
+      glwe_dimension, polynomial_size, num_lwes, allocate_gpu_memory);
+}
+/* Perform functional packing keyswitch on a batch of 64 bits input LWE
+ * ciphertexts.
+ */
+void cuda_packing_keyswitch_lwe_list_to_glwe_64(
+    void *stream, uint32_t gpu_index, void *glwe_array_out,
+    void const *lwe_array_in, void const *fp_ksk_array, int8_t *fp_ks_buffer,
+    uint32_t input_lwe_dimension, uint32_t output_glwe_dimension,
+    uint32_t output_polynomial_size, uint32_t base_log, uint32_t level_count,
+    uint32_t num_lwes) {
+
+  host_packing_keyswitch_lwe_list_to_glwe<uint64_t>(
+      static_cast<cudaStream_t>(stream), gpu_index,
+      static_cast<uint64_t *>(glwe_array_out),
+      static_cast<const uint64_t *>(lwe_array_in),
+      static_cast<const uint64_t *>(fp_ksk_array), fp_ks_buffer,
+      input_lwe_dimension, output_glwe_dimension, output_polynomial_size,
+      base_log, level_count, num_lwes);
+}
+
+void cleanup_packing_keyswitch_lwe_list_to_glwe(void *stream,
+                                                uint32_t gpu_index,
+                                                int8_t **fp_ks_buffer) {
+  cuda_drop_async(*fp_ks_buffer, static_cast<cudaStream_t>(stream), gpu_index);
 }

backends/tfhe-cuda-backend/cuda/src/crypto/keyswitch.cuh

@@ -7,6 +7,7 @@
 #include "polynomial/functions.cuh"
 #include "polynomial/polynomial_math.cuh"
 #include "torus.cuh"
+#include "utils/helper.cuh"
 #include "utils/kernel_dimensions.cuh"
 #include <thread>
 #include <vector>
@@ -70,9 +71,8 @@ keyswitch(Torus *lwe_array_out, const Torus *__restrict__ lwe_output_indexes,
 
     // This loop distribution seems to benefit the global mem reads
     for (int i = start_i; i < end_i; i++) {
-      Torus a_i = round_to_closest_multiple(block_lwe_array_in[i], base_log,
-                                            level_count);
-      Torus state = a_i >> (sizeof(Torus) * 8 - base_log * level_count);
+      Torus state =
+          init_decomposer_state(block_lwe_array_in[i], base_log, level_count);
 
       for (int j = 0; j < level_count; j++) {
         auto ksk_block =
@@ -98,11 +98,12 @@ keyswitch(Torus *lwe_array_out, const Torus *__restrict__ lwe_output_indexes,
 }
 
 template <typename Torus>
-__host__ void cuda_keyswitch_lwe_ciphertext_vector(
+__host__ void host_keyswitch_lwe_ciphertext_vector(
     cudaStream_t stream, uint32_t gpu_index, Torus *lwe_array_out,
-    Torus *lwe_output_indexes, Torus *lwe_array_in, Torus *lwe_input_indexes,
-    Torus *ksk, uint32_t lwe_dimension_in, uint32_t lwe_dimension_out,
-    uint32_t base_log, uint32_t level_count, uint32_t num_samples) {
+    Torus const *lwe_output_indexes, Torus const *lwe_array_in,
+    Torus const *lwe_input_indexes, Torus const *ksk, uint32_t lwe_dimension_in,
+    uint32_t lwe_dimension_out, uint32_t base_log, uint32_t level_count,
+    uint32_t num_samples) {
 
   cudaSetDevice(gpu_index);
 
@@ -123,13 +124,13 @@ __host__ void cuda_keyswitch_lwe_ciphertext_vector(
 }
 
 template <typename Torus>
-void execute_keyswitch_async(cudaStream_t *streams, uint32_t *gpu_indexes,
-                             uint32_t gpu_count,
+void execute_keyswitch_async(cudaStream_t const *streams,
+                             uint32_t const *gpu_indexes, uint32_t gpu_count,
                              const LweArrayVariant<Torus> &lwe_array_out,
                              const LweArrayVariant<Torus> &lwe_output_indexes,
                              const LweArrayVariant<Torus> &lwe_array_in,
                              const LweArrayVariant<Torus> &lwe_input_indexes,
-                             Torus **ksks, uint32_t lwe_dimension_in,
+                             Torus *const *ksks, uint32_t lwe_dimension_in,
                              uint32_t lwe_dimension_out, uint32_t base_log,
                              uint32_t level_count, uint32_t num_samples) {
 
@@ -146,7 +147,7 @@ void execute_keyswitch_async(cudaStream_t *streams, uint32_t *gpu_indexes,
         GET_VARIANT_ELEMENT(lwe_input_indexes, i);
 
     // Compute Keyswitch
-    cuda_keyswitch_lwe_ciphertext_vector<Torus>(
+    host_keyswitch_lwe_ciphertext_vector<Torus>(
         streams[i], gpu_indexes[i], current_lwe_array_out,
         current_lwe_output_indexes, current_lwe_array_in,
         current_lwe_input_indexes, ksks[i], lwe_dimension_in, lwe_dimension_out,
@@ -154,4 +155,158 @@ void execute_keyswitch_async(cudaStream_t *streams, uint32_t *gpu_indexes,
   }
 }
 
+template <typename Torus>
+__host__ void scratch_packing_keyswitch_lwe_list_to_glwe(
+    cudaStream_t stream, uint32_t gpu_index, int8_t **fp_ks_buffer,
+    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t num_lwes,
+    bool allocate_gpu_memory) {
+  cudaSetDevice(gpu_index);
+
+  int glwe_accumulator_size = (glwe_dimension + 1) * polynomial_size;
+
+  if (allocate_gpu_memory)
+    *fp_ks_buffer = (int8_t *)cuda_malloc_async(
+        2 * num_lwes * glwe_accumulator_size * sizeof(Torus), stream,
+        gpu_index);
+}
+
+// public functional packing keyswitch for a single LWE ciphertext
+//
+// Assumes there are (glwe_dimension+1) * polynomial_size threads split through
+// different thread blocks at the x-axis to work on that input.
+template <typename Torus>
+__device__ void packing_keyswitch_lwe_ciphertext_into_glwe_ciphertext(
+    Torus *glwe_out, Torus const *lwe_in, Torus const *fp_ksk,
+    uint32_t lwe_dimension_in, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count) {
+
+  const int tid = threadIdx.x + blockIdx.x * blockDim.x;
+  size_t glwe_size = (glwe_dimension + 1);
+
+  if (tid < glwe_size * polynomial_size) {
+    const int local_index = threadIdx.x;
+    // the output_glwe is split in polynomials and each x-block takes one of
+    // them
+    size_t poly_id = blockIdx.x;
+    size_t coef_per_block = blockDim.x;
+
+    // number of coefficients inside fp-ksk block for each lwe_input coefficient
+    size_t ksk_block_size = glwe_size * polynomial_size * level_count;
+
+    // initialize accumulator to 0
+    glwe_out[tid] = SEL(0, lwe_in[lwe_dimension_in],
+                        tid == glwe_dimension * polynomial_size);
+
+    // Iterate through all lwe elements
+    for (int i = 0; i < lwe_dimension_in; i++) {
+      // Round and prepare decomposition
+      Torus state = init_decomposer_state(lwe_in[i], base_log, level_count);
+
+      Torus mod_b_mask = (1ll << base_log) - 1ll;
+
+      // block of key for current lwe coefficient (cur_input_lwe[i])
+      auto ksk_block = &fp_ksk[i * ksk_block_size];
+      for (int j = 0; j < level_count; j++) {
+        auto ksk_glwe = &ksk_block[j * glwe_size * polynomial_size];
+        // Iterate through each level and multiply by the ksk piece
+        auto ksk_glwe_chunk = &ksk_glwe[poly_id * coef_per_block];
+        Torus decomposed = decompose_one<Torus>(state, mod_b_mask, base_log);
+        glwe_out[tid] -= decomposed * ksk_glwe_chunk[local_index];
+      }
+    }
+  }
+}
+
+// public functional packing keyswitch for a batch of LWE ciphertexts
+//
+// Selects the input each thread is working on using the y-block index.
+//
+// Assumes there are (glwe_dimension+1) * polynomial_size threads split through
+// different thread blocks at the x-axis to work on that input.
+template <typename Torus>
+__global__ void packing_keyswitch_lwe_list_to_glwe(
+    Torus *glwe_array_out, Torus const *lwe_array_in, Torus const *fp_ksk,
+    uint32_t lwe_dimension_in, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
+    Torus *d_mem) {
+  const int tid = threadIdx.x + blockIdx.x * blockDim.x;
+
+  const int glwe_accumulator_size = (glwe_dimension + 1) * polynomial_size;
+  const int lwe_size = (lwe_dimension_in + 1);
+
+  const int input_id = blockIdx.y;
+  const int degree = input_id;
+
+  // Select an input
+  auto lwe_in = lwe_array_in + input_id * lwe_size;
+  auto ks_glwe_out = d_mem + input_id * glwe_accumulator_size;
+  auto glwe_out = glwe_array_out + input_id * glwe_accumulator_size;
+  // KS LWE to GLWE
+  packing_keyswitch_lwe_ciphertext_into_glwe_ciphertext<Torus>(
+      ks_glwe_out, lwe_in, fp_ksk, lwe_dimension_in, glwe_dimension,
+      polynomial_size, base_log, level_count);
+
+  // P * x ^degree
+  auto in_poly = ks_glwe_out + (tid / polynomial_size) * polynomial_size;
+  auto out_result = glwe_out + (tid / polynomial_size) * polynomial_size;
+  polynomial_accumulate_monic_monomial_mul<Torus>(out_result, in_poly, degree,
+                                                  tid % polynomial_size,
+                                                  polynomial_size, 1, true);
+}
+
+/// To-do: Rewrite this kernel for efficiency
+template <typename Torus>
+__global__ void accumulate_glwes(Torus *glwe_out, Torus *glwe_array_in,
+                                 uint32_t glwe_dimension,
+                                 uint32_t polynomial_size, uint32_t num_lwes) {
+  const int tid = threadIdx.x + blockIdx.x * blockDim.x;
+  if (tid < (glwe_dimension + 1) * polynomial_size) {
+    glwe_out[tid] = glwe_array_in[tid];
+
+    // Accumulate
+    for (int i = 1; i < num_lwes; i++) {
+      auto glwe_in = glwe_array_in + i * (glwe_dimension + 1) * polynomial_size;
+      glwe_out[tid] += glwe_in[tid];
+    }
+  }
+}
+
+template <typename Torus>
+__host__ void host_packing_keyswitch_lwe_list_to_glwe(
+    cudaStream_t stream, uint32_t gpu_index, Torus *glwe_out,
+    Torus const *lwe_array_in, Torus const *fp_ksk_array, int8_t *fp_ks_buffer,
+    uint32_t lwe_dimension_in, uint32_t glwe_dimension,
+    uint32_t polynomial_size, uint32_t base_log, uint32_t level_count,
+    uint32_t num_lwes) {
+
+  if (num_lwes > polynomial_size)
+    PANIC("Cuda error: too many LWEs to pack. The number of LWEs should be "
+          "smaller than "
+          "polynomial_size.")
+
+  cudaSetDevice(gpu_index);
+  int glwe_accumulator_size = (glwe_dimension + 1) * polynomial_size;
+
+  int num_blocks = 0, num_threads = 0;
+  getNumBlocksAndThreads(glwe_accumulator_size, 128, num_blocks, num_threads);
+
+  dim3 grid(num_blocks, num_lwes);
+  dim3 threads(num_threads);
+
+  auto d_mem = (Torus *)fp_ks_buffer;
+  auto d_tmp_glwe_array_out = d_mem + num_lwes * glwe_accumulator_size;
+
+  // individually keyswitch each lwe
+  packing_keyswitch_lwe_list_to_glwe<Torus><<<grid, threads, 0, stream>>>(
+      d_tmp_glwe_array_out, lwe_array_in, fp_ksk_array, lwe_dimension_in,
+      glwe_dimension, polynomial_size, base_log, level_count, d_mem);
+  check_cuda_error(cudaGetLastError());
+
+  // accumulate to a single glwe
+  accumulate_glwes<Torus><<<num_blocks, threads, 0, stream>>>(
+      glwe_out, d_tmp_glwe_array_out, glwe_dimension, polynomial_size,
+      num_lwes);
+  check_cuda_error(cudaGetLastError());
+}
+
 #endif

backends/tfhe-cuda-backend/cuda/src/crypto/torus.cuh

@@ -1,9 +1,22 @@
 #ifndef CNCRT_TORUS_CUH
 #define CNCRT_TORUS_CUH
 
+#include "device.h"
+#include "polynomial/parameters.cuh"
 #include "types/int128.cuh"
+#include "utils/kernel_dimensions.cuh"
 #include <limits>
 
+template <typename T>
+__host__ __device__ __forceinline__ constexpr double get_two_pow_torus_bits() {
+  return (sizeof(T) == 4) ? 4294967296.0 : 18446744073709551616.0;
+}
+
+template <typename T>
+__host__ __device__ __forceinline__ constexpr T scalar_max() {
+  return std::numeric_limits<T>::max();
+}
+
 template <typename T>
 __device__ inline void typecast_double_to_torus(double x, T &r) {
   r = T(x);
@@ -27,22 +40,52 @@ __device__ inline void typecast_double_to_torus<uint64_t>(double x,
 }
 
 template <typename T>
-__device__ inline T round_to_closest_multiple(T x, uint32_t base_log,
-                                              uint32_t level_count) {
-  T shift = sizeof(T) * 8 - level_count * base_log;
-  T mask = 1ll << (shift - 1);
-  T b = (x & mask) >> (shift - 1);
-  T res = x >> shift;
-  res += b;
-  res <<= shift;
-  return res;
+__device__ inline void typecast_double_round_to_torus(double x, T &r) {
+  constexpr double mx = get_two_pow_torus_bits<T>();
+  // floor must be used here because round has an issue with rounding .5,
+  // as it rounds away from zero.
+  double frac = x - floor(x);
+  frac *= mx;
+  typecast_double_to_torus(round(frac), r);
+}
+
+template <typename T>
+__device__ inline void typecast_torus_to_double(T x, double &r);
+
+template <>
+__device__ inline void typecast_torus_to_double<uint32_t>(uint32_t x,
+                                                          double &r) {
+  r = __int2double_rn(x);
+}
+
+template <>
+__device__ inline void typecast_torus_to_double<uint64_t>(uint64_t x,
+                                                          double &r) {
+  r = __ll2double_rn(x);
+}
+
+template <typename T>
+__device__ inline T init_decomposer_state(T input, uint32_t base_log,
+                                          uint32_t level_count) {
+  const T rep_bit_count = level_count * base_log;
+  const T non_rep_bit_count = sizeof(T) * 8 - rep_bit_count;
+  T res = input >> (non_rep_bit_count - 1);
+  T rounding_bit = res & (T)(1);
+  res++;
+  res >>= 1;
+  T torus_max = scalar_max<T>();
+  T mod_mask = torus_max >> non_rep_bit_count;
+  res &= mod_mask;
+  T shifted_random = rounding_bit << (rep_bit_count - 1);
+  T need_balance =
+      (((res - (T)(1)) | shifted_random) & res) >> (rep_bit_count - 1);
+  return res - (need_balance << rep_bit_count);
 }
 
 template <typename T>
 __device__ __forceinline__ void modulus_switch(T input, T &output,
                                                uint32_t log_modulus) {
   constexpr uint32_t BITS = sizeof(T) * 8;
-
   output = input + (((T)1) << (BITS - log_modulus - 1));
   output >>= (BITS - log_modulus);
 }
@@ -54,4 +97,27 @@ __device__ __forceinline__ T modulus_switch(T input, uint32_t log_modulus) {
   return output;
 }
 
+template <typename Torus>
+__global__ void modulus_switch_inplace(Torus *array, int size,
+                                       uint32_t log_modulus) {
+  const int tid = threadIdx.x + blockIdx.x * blockDim.x;
+  if (tid < size) {
+    array[tid] = modulus_switch(array[tid], log_modulus);
+  }
+}
+
+template <typename Torus>
+__host__ void host_modulus_switch_inplace(cudaStream_t stream,
+                                          uint32_t gpu_index, Torus *array,
+                                          int size, uint32_t log_modulus) {
+  cudaSetDevice(gpu_index);
+
+  int num_threads = 0, num_blocks = 0;
+  getNumBlocksAndThreads(size, 1024, num_blocks, num_threads);
+
+  modulus_switch_inplace<<<num_blocks, num_threads, 0, stream>>>(array, size,
+                                                                 log_modulus);
+  check_cuda_error(cudaGetLastError());
+}
+
 #endif // CNCRT_TORUS_H

backends/tfhe-cuda-backend/cuda/src/device.cu

@@ -113,7 +113,7 @@ void cuda_memcpy_async_to_gpu(void *dest, void *src, uint64_t size,
 }
 
 /// Copy memory within a GPU asynchronously
-void cuda_memcpy_async_gpu_to_gpu(void *dest, void *src, uint64_t size,
+void cuda_memcpy_async_gpu_to_gpu(void *dest, void const *src, uint64_t size,
                                   cudaStream_t stream, uint32_t gpu_index) {
   if (size == 0)
     return;
@@ -137,6 +137,30 @@ void cuda_memcpy_async_gpu_to_gpu(void *dest, void *src, uint64_t size,
   }
 }
 
+/// Copy memory within a GPU
+void cuda_memcpy_gpu_to_gpu(void *dest, void *src, uint64_t size,
+                            uint32_t gpu_index) {
+  if (size == 0)
+    return;
+  cudaPointerAttributes attr_dest;
+  check_cuda_error(cudaPointerGetAttributes(&attr_dest, dest));
+  if (attr_dest.type != cudaMemoryTypeDevice) {
+    PANIC("Cuda error: invalid dest device pointer in copy from GPU to GPU.")
+  }
+  cudaPointerAttributes attr_src;
+  check_cuda_error(cudaPointerGetAttributes(&attr_src, src));
+  if (attr_src.type != cudaMemoryTypeDevice) {
+    PANIC("Cuda error: invalid src device pointer in copy from GPU to GPU.")
+  }
+  check_cuda_error(cudaSetDevice(gpu_index));
+  if (attr_src.device == attr_dest.device) {
+    check_cuda_error(cudaMemcpy(dest, src, size, cudaMemcpyDeviceToDevice));
+  } else {
+    check_cuda_error(
+        cudaMemcpyPeer(dest, attr_dest.device, src, attr_src.device, size));
+  }
+}
+
 /// Synchronizes device
 void cuda_synchronize_device(uint32_t gpu_index) {
   check_cuda_error(cudaSetDevice(gpu_index));
@@ -177,8 +201,8 @@ void cuda_set_value_async(cudaStream_t stream, uint32_t gpu_index,
     int num_blocks = (n + block_size - 1) / block_size;
 
     // Launch the kernel
-    cuda_set_value_kernel<<<num_blocks, block_size, 0, stream>>>(d_array, value,
-                                                                 n);
+    cuda_set_value_kernel<Torus>
+        <<<num_blocks, block_size, 0, stream>>>(d_array, value, n);
     check_cuda_error(cudaGetLastError());
   }
 }
@@ -243,10 +267,21 @@ void cuda_drop_async(void *ptr, cudaStream_t stream, uint32_t gpu_index) {
 
 /// Get the maximum size for the shared memory
 int cuda_get_max_shared_memory(uint32_t gpu_index) {
-  check_cuda_error(cudaSetDevice(gpu_index));
   int max_shared_memory = 0;
+#if CUDA_ARCH == 900
+  max_shared_memory = 226000;
+#elif CUDA_ARCH == 890
+  max_shared_memory = 100000;
+#elif CUDA_ARCH == 860
+  max_shared_memory = 100000;
+#elif CUDA_ARCH == 800
+  max_shared_memory = 163000;
+#elif CUDA_ARCH == 700
+  max_shared_memory = 95000;
+#else
   cudaDeviceGetAttribute(&max_shared_memory, cudaDevAttrMaxSharedMemoryPerBlock,
                          gpu_index);
   check_cuda_error(cudaGetLastError());
+#endif
   return max_shared_memory;
 }

backends/tfhe-cuda-backend/cuda/src/fft/bnsmfft.cuh

@@ -6,6 +6,7 @@
 #include "twiddles.cuh"
 #include "types/complex/operations.cuh"
 
+using Index = unsigned;
 /*
  * Direct negacyclic FFT:
  *   - before the FFT the N real coefficients are stored into a
@@ -31,290 +32,81 @@ template <class params> __device__ void NSMFFT_direct(double2 *A) {
    *  full loop, which should increase performance
    */
 
-  size_t tid = threadIdx.x;
-  size_t twid_id;
-  size_t i1, i2;
-  double2 u, v, w;
+  __syncthreads();
+  constexpr Index BUTTERFLY_DEPTH = params::opt >> 1;
+  constexpr Index LOG2_DEGREE = params::log2_degree;
+  constexpr Index HALF_DEGREE = params::degree >> 1;
+  constexpr Index STRIDE = params::degree / params::opt;
+
+  Index tid = threadIdx.x;
+  double2 u[BUTTERFLY_DEPTH], v[BUTTERFLY_DEPTH], w;
+
+  // load into registers
+#pragma unroll
+  for (Index i = 0; i < BUTTERFLY_DEPTH; ++i) {
+    u[i] = A[tid];
+    v[i] = A[tid + HALF_DEGREE];
+
+    tid += STRIDE;
+  }
+
   // level 1
   // we don't make actual complex multiplication on level1 since we have only
   // one twiddle, it's real and image parts are equal, so we can multiply
   // it with simpler operations
 #pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    i1 = tid;
-    i2 = tid + params::degree / 2;
-
-    u = A[i1];
-    v = A[i2] * (double2){0.707106781186547461715008466854,
-                          0.707106781186547461715008466854};
-
-    A[i1] += v;
-    A[i2] = u - v;
-
-    tid += params::degree / params::opt;
+  for (Index i = 0; i < BUTTERFLY_DEPTH; ++i) {
+    w = v[i] * (double2){0.707106781186547461715008466854,
+                         0.707106781186547461715008466854};
+    v[i] = u[i] - w;
+    u[i] = u[i] + w;
   }
-  __syncthreads();
 
-  // level 2
-  // from this level there are more than one twiddles and none of them has equal
-  // real and imag parts, so complete complex multiplication is needed
-  // for each level params::degree / 2^level represents number of coefficients
-  // inside divided chunk of specific level
-  //
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 4);
-    i1 = 2 * (params::degree / 4) * twid_id + (tid & (params::degree / 4 - 1));
-    i2 = i1 + params::degree / 4;
+  Index twiddle_shift = 1;
+  for (Index l = LOG2_DEGREE - 1; l >= 1; --l) {
+    Index lane_mask = 1 << (l - 1);
+    Index thread_mask = (1 << l) - 1;
+    twiddle_shift <<= 1;
 
-    w = negtwiddles[twid_id + 2];
-    u = A[i1];
-    v = A[i2] * w;
-
-    A[i1] += v;
-    A[i2] = u - v;
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 3
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 8);
-    i1 = 2 * (params::degree / 8) * twid_id + (tid & (params::degree / 8 - 1));
-    i2 = i1 + params::degree / 8;
-
-    w = negtwiddles[twid_id + 4];
-    u = A[i1];
-    v = A[i2] * w;
-
-    A[i1] += v;
-    A[i2] = u - v;
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 4
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 16);
-    i1 =
-        2 * (params::degree / 16) * twid_id + (tid & (params::degree / 16 - 1));
-    i2 = i1 + params::degree / 16;
-
-    w = negtwiddles[twid_id + 8];
-    u = A[i1];
-    v = A[i2] * w;
-
-    A[i1] += v;
-    A[i2] = u - v;
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 5
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 32);
-    i1 =
-        2 * (params::degree / 32) * twid_id + (tid & (params::degree / 32 - 1));
-    i2 = i1 + params::degree / 32;
-
-    w = negtwiddles[twid_id + 16];
-    u = A[i1];
-    v = A[i2] * w;
-
-    A[i1] += v;
-    A[i2] = u - v;
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 6
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 64);
-    i1 =
-        2 * (params::degree / 64) * twid_id + (tid & (params::degree / 64 - 1));
-    i2 = i1 + params::degree / 64;
-
-    w = negtwiddles[twid_id + 32];
-    u = A[i1];
-    v = A[i2] * w;
-
-    A[i1] += v;
-    A[i2] = u - v;
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 7
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 128);
-    i1 = 2 * (params::degree / 128) * twid_id +
-         (tid & (params::degree / 128 - 1));
-    i2 = i1 + params::degree / 128;
-
-    w = negtwiddles[twid_id + 64];
-    u = A[i1];
-    v = A[i2] * w;
-
-    A[i1] += v;
-    A[i2] = u - v;
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // from level 8, we need to check size of params degree, because we support
-  // minimum actual polynomial size = 256,  when compressed size is halfed and
-  // minimum supported compressed size is 128, so we always need first 7
-  // levels of butterfly operation, since butterfly levels are hardcoded
-  // we need to check if polynomial size is big enough to require specific level
-  // of butterfly.
-  if constexpr (params::degree >= 256) {
-    // level 8
     tid = threadIdx.x;
+    __syncthreads();
 #pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 256);
-      i1 = 2 * (params::degree / 256) * twid_id +
-           (tid & (params::degree / 256 - 1));
-      i2 = i1 + params::degree / 256;
-
-      w = negtwiddles[twid_id + 128];
-      u = A[i1];
-      v = A[i2] * w;
-
-      A[i1] += v;
-      A[i2] = u - v;
-
-      tid += params::degree / params::opt;
+    for (Index i = 0; i < BUTTERFLY_DEPTH; i++) {
+      Index rank = tid & thread_mask;
+      bool u_stays_in_register = rank < lane_mask;
+      A[tid] = (u_stays_in_register) ? v[i] : u[i];
+      tid = tid + STRIDE;
     }
     __syncthreads();
-  }
 
-  if constexpr (params::degree >= 512) {
-    // level 9
     tid = threadIdx.x;
 #pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 512);
-      i1 = 2 * (params::degree / 512) * twid_id +
-           (tid & (params::degree / 512 - 1));
-      i2 = i1 + params::degree / 512;
+    for (Index i = 0; i < BUTTERFLY_DEPTH; i++) {
+      Index rank = tid & thread_mask;
+      bool u_stays_in_register = rank < lane_mask;
+      w = A[tid ^ lane_mask];
+      u[i] = (u_stays_in_register) ? u[i] : w;
+      v[i] = (u_stays_in_register) ? w : v[i];
+      w = negtwiddles[tid / lane_mask + twiddle_shift];
 
-      w = negtwiddles[twid_id + 256];
-      u = A[i1];
-      v = A[i2] * w;
+      w *= v[i];
 
-      A[i1] += v;
-      A[i2] = u - v;
-
-      tid += params::degree / params::opt;
+      v[i] = u[i] - w;
+      u[i] = u[i] + w;
+      tid = tid + STRIDE;
     }
-    __syncthreads();
   }
+  __syncthreads();
 
-  if constexpr (params::degree >= 1024) {
-    // level 10
-    tid = threadIdx.x;
+  // store registers in SM
+  tid = threadIdx.x;
 #pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 1024);
-      i1 = 2 * (params::degree / 1024) * twid_id +
-           (tid & (params::degree / 1024 - 1));
-      i2 = i1 + params::degree / 1024;
-
-      w = negtwiddles[twid_id + 512];
-      u = A[i1];
-      v = A[i2] * w;
-
-      A[i1] += v;
-      A[i2] = u - v;
-
-      tid += params::degree / params::opt;
-    }
-    __syncthreads();
-  }
-
-  if constexpr (params::degree >= 2048) {
-    // level 11
-    tid = threadIdx.x;
-#pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 2048);
-      i1 = 2 * (params::degree / 2048) * twid_id +
-           (tid & (params::degree / 2048 - 1));
-      i2 = i1 + params::degree / 2048;
-
-      w = negtwiddles[twid_id + 1024];
-      u = A[i1];
-      v = A[i2] * w;
-
-      A[i1] += v;
-      A[i2] = u - v;
-
-      tid += params::degree / params::opt;
-    }
-    __syncthreads();
-  }
-
-  if constexpr (params::degree >= 4096) {
-    // level 12
-    tid = threadIdx.x;
-#pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 4096);
-      i1 = 2 * (params::degree / 4096) * twid_id +
-           (tid & (params::degree / 4096 - 1));
-      i2 = i1 + params::degree / 4096;
-
-      w = negtwiddles[twid_id + 2048];
-      u = A[i1];
-      v = A[i2] * w;
-
-      A[i1] += v;
-      A[i2] = u - v;
-
-      tid += params::degree / params::opt;
-    }
-    __syncthreads();
-  }
-
-  if constexpr (params::degree >= 8192) {
-    // level 13
-    tid = threadIdx.x;
-#pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 8192);
-      i1 = 2 * (params::degree / 8192) * twid_id +
-           (tid & (params::degree / 8192 - 1));
-      i2 = i1 + params::degree / 8192;
-
-      w = negtwiddles[twid_id + 4096];
-      u = A[i1];
-      v = A[i2] * w;
-
-      A[i1] += v;
-      A[i2] = u - v;
-
-      tid += params::degree / params::opt;
-    }
-    __syncthreads();
+  for (Index i = 0; i < BUTTERFLY_DEPTH; i++) {
+    A[tid * 2] = u[i];
+    A[tid * 2 + 1] = v[i];
+    tid = tid + STRIDE;
   }
+  __syncthreads();
 }
 
 /*
@@ -329,284 +121,82 @@ template <class params> __device__ void NSMFFT_inverse(double2 *A) {
    *  full loop, which should increase performance
    */
 
+  __syncthreads();
+  constexpr Index BUTTERFLY_DEPTH = params::opt >> 1;
+  constexpr Index LOG2_DEGREE = params::log2_degree;
+  constexpr Index DEGREE = params::degree;
+  constexpr Index HALF_DEGREE = params::degree >> 1;
+  constexpr Index STRIDE = params::degree / params::opt;
+
   size_t tid = threadIdx.x;
-  size_t twid_id;
-  size_t i1, i2;
-  double2 u, w;
+  double2 u[BUTTERFLY_DEPTH], v[BUTTERFLY_DEPTH], w;
 
-  // divide input by compressed polynomial size
-  tid = threadIdx.x;
-  for (size_t i = 0; i < params::opt; ++i) {
-    A[tid] /= params::degree;
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // none of the twiddles have equal real and imag part, so
-  // complete complex multiplication has to be done
-  // here we have more than one twiddle
-  // mapping in backward fft is reversed
-  // butterfly operation is started from last level
-
-  if constexpr (params::degree >= 8192) {
-    // level 13
-    tid = threadIdx.x;
+  // load into registers and divide by compressed polynomial size
 #pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 8192);
-      i1 = 2 * (params::degree / 8192) * twid_id +
-           (tid & (params::degree / 8192 - 1));
-      i2 = i1 + params::degree / 8192;
+  for (Index i = 0; i < BUTTERFLY_DEPTH; ++i) {
+    u[i] = A[2 * tid];
+    v[i] = A[2 * tid + 1];
 
-      w = negtwiddles[twid_id + 4096];
-      u = A[i1] - A[i2];
+    u[i] /= DEGREE;
+    v[i] /= DEGREE;
 
-      A[i1] += A[i2];
-      A[i2] = u * conjugate(w);
+    tid += STRIDE;
+  }
 
-      tid += params::degree / params::opt;
+  Index twiddle_shift = DEGREE;
+  for (Index l = 1; l <= LOG2_DEGREE - 1; ++l) {
+    Index lane_mask = 1 << (l - 1);
+    Index thread_mask = (1 << l) - 1;
+    tid = threadIdx.x;
+    twiddle_shift >>= 1;
+
+    // at this point registers are ready for the  butterfly
+    tid = threadIdx.x;
+    __syncthreads();
+#pragma unroll
+    for (Index i = 0; i < BUTTERFLY_DEPTH; ++i) {
+      w = (u[i] - v[i]);
+      u[i] += v[i];
+      v[i] = w * conjugate(negtwiddles[tid / lane_mask + twiddle_shift]);
+
+      // keep one of the register for next iteration and store another one in sm
+      Index rank = tid & thread_mask;
+      bool u_stays_in_register = rank < lane_mask;
+      A[tid] = (u_stays_in_register) ? v[i] : u[i];
+
+      tid = tid + STRIDE;
     }
     __syncthreads();
-  }
 
-  if constexpr (params::degree >= 4096) {
-    // level 12
+    // prepare registers for next butterfly iteration
     tid = threadIdx.x;
 #pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 4096);
-      i1 = 2 * (params::degree / 4096) * twid_id +
-           (tid & (params::degree / 4096 - 1));
-      i2 = i1 + params::degree / 4096;
+    for (Index i = 0; i < BUTTERFLY_DEPTH; ++i) {
+      Index rank = tid & thread_mask;
+      bool u_stays_in_register = rank < lane_mask;
+      w = A[tid ^ lane_mask];
+      u[i] = (u_stays_in_register) ? u[i] : w;
+      v[i] = (u_stays_in_register) ? w : v[i];
 
-      w = negtwiddles[twid_id + 2048];
-      u = A[i1] - A[i2];
-
-      A[i1] += A[i2];
-      A[i2] = u * conjugate(w);
-
-      tid += params::degree / params::opt;
+      tid = tid + STRIDE;
     }
-    __syncthreads();
   }
 
-  if constexpr (params::degree >= 2048) {
-    // level 11
-    tid = threadIdx.x;
-#pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 2048);
-      i1 = 2 * (params::degree / 2048) * twid_id +
-           (tid & (params::degree / 2048 - 1));
-      i2 = i1 + params::degree / 2048;
-
-      w = negtwiddles[twid_id + 1024];
-      u = A[i1] - A[i2];
-
-      A[i1] += A[i2];
-      A[i2] = u * conjugate(w);
-
-      tid += params::degree / params::opt;
-    }
-    __syncthreads();
-  }
-
-  if constexpr (params::degree >= 1024) {
-    // level 10
-    tid = threadIdx.x;
-#pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 1024);
-      i1 = 2 * (params::degree / 1024) * twid_id +
-           (tid & (params::degree / 1024 - 1));
-      i2 = i1 + params::degree / 1024;
-
-      w = negtwiddles[twid_id + 512];
-      u = A[i1] - A[i2];
-
-      A[i1] += A[i2];
-      A[i2] = u * conjugate(w);
-
-      tid += params::degree / params::opt;
-    }
-    __syncthreads();
-  }
-
-  if constexpr (params::degree >= 512) {
-    // level 9
-    tid = threadIdx.x;
-#pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 512);
-      i1 = 2 * (params::degree / 512) * twid_id +
-           (tid & (params::degree / 512 - 1));
-      i2 = i1 + params::degree / 512;
-
-      w = negtwiddles[twid_id + 256];
-      u = A[i1] - A[i2];
-
-      A[i1] += A[i2];
-      A[i2] = u * conjugate(w);
-
-      tid += params::degree / params::opt;
-    }
-    __syncthreads();
-  }
-
-  if constexpr (params::degree >= 256) {
-    // level 8
-    tid = threadIdx.x;
-#pragma unroll
-    for (size_t i = 0; i < params::opt / 2; ++i) {
-      twid_id = tid / (params::degree / 256);
-      i1 = 2 * (params::degree / 256) * twid_id +
-           (tid & (params::degree / 256 - 1));
-      i2 = i1 + params::degree / 256;
-
-      w = negtwiddles[twid_id + 128];
-      u = A[i1] - A[i2];
-
-      A[i1] += A[i2];
-      A[i2] = u * conjugate(w);
-
-      tid += params::degree / params::opt;
-    }
-    __syncthreads();
-  }
-
-  // below level 8, we don't need to check size of params degree, because we
-  // support minimum actual polynomial size = 256,  when compressed size is
-  // halfed and minimum supported compressed size is 128, so we always need
-  // last 7 levels of butterfly operation, since butterfly levels are hardcoded
-  // we don't need to check if polynomial size is big enough to require
-  // specific level of butterfly.
-  // level 7
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 128);
-    i1 = 2 * (params::degree / 128) * twid_id +
-         (tid & (params::degree / 128 - 1));
-    i2 = i1 + params::degree / 128;
-
-    w = negtwiddles[twid_id + 64];
-    u = A[i1] - A[i2];
-
-    A[i1] += A[i2];
-    A[i2] = u * conjugate(w);
-
-    tid += params::degree / params::opt;
+  // last iteration
+  for (Index i = 0; i < BUTTERFLY_DEPTH; ++i) {
+    w = (u[i] - v[i]);
+    u[i] = u[i] + v[i];
+    v[i] = w * (double2){0.707106781186547461715008466854,
+                         -0.707106781186547461715008466854};
   }
   __syncthreads();
-
-  // level 6
+  // store registers in SM
   tid = threadIdx.x;
 #pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 64);
-    i1 =
-        2 * (params::degree / 64) * twid_id + (tid & (params::degree / 64 - 1));
-    i2 = i1 + params::degree / 64;
-
-    w = negtwiddles[twid_id + 32];
-    u = A[i1] - A[i2];
-
-    A[i1] += A[i2];
-    A[i2] = u * conjugate(w);
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 5
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 32);
-    i1 =
-        2 * (params::degree / 32) * twid_id + (tid & (params::degree / 32 - 1));
-    i2 = i1 + params::degree / 32;
-
-    w = negtwiddles[twid_id + 16];
-    u = A[i1] - A[i2];
-
-    A[i1] += A[i2];
-    A[i2] = u * conjugate(w);
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 4
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 16);
-    i1 =
-        2 * (params::degree / 16) * twid_id + (tid & (params::degree / 16 - 1));
-    i2 = i1 + params::degree / 16;
-
-    w = negtwiddles[twid_id + 8];
-    u = A[i1] - A[i2];
-
-    A[i1] += A[i2];
-    A[i2] = u * conjugate(w);
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 3
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 8);
-    i1 = 2 * (params::degree / 8) * twid_id + (tid & (params::degree / 8 - 1));
-    i2 = i1 + params::degree / 8;
-
-    w = negtwiddles[twid_id + 4];
-    u = A[i1] - A[i2];
-
-    A[i1] += A[i2];
-    A[i2] = u * conjugate(w);
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 2
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 4);
-    i1 = 2 * (params::degree / 4) * twid_id + (tid & (params::degree / 4 - 1));
-    i2 = i1 + params::degree / 4;
-
-    w = negtwiddles[twid_id + 2];
-    u = A[i1] - A[i2];
-
-    A[i1] += A[i2];
-    A[i2] = u * conjugate(w);
-
-    tid += params::degree / params::opt;
-  }
-  __syncthreads();
-
-  // level 1
-  tid = threadIdx.x;
-#pragma unroll
-  for (size_t i = 0; i < params::opt / 2; ++i) {
-    twid_id = tid / (params::degree / 2);
-    i1 = 2 * (params::degree / 2) * twid_id + (tid & (params::degree / 2 - 1));
-    i2 = i1 + params::degree / 2;
-
-    w = negtwiddles[twid_id + 1];
-    u = A[i1] - A[i2];
-
-    A[i1] += A[i2];
-    A[i2] = u * conjugate(w);
-
-    tid += params::degree / params::opt;
+  for (Index i = 0; i < BUTTERFLY_DEPTH; i++) {
+    A[tid] = u[i];
+    A[tid + HALF_DEGREE] = v[i];
+    tid = tid + STRIDE;
   }
   __syncthreads();
 }