update gpu cost model

limit a bit more the number of base log 1 cases

.editorconfig

@@ -8,14 +8,8 @@ root = true
 end_of_line = lf
 insert_final_newline = true
 
-# 4 space indentation for rust and toml
-[*.{rs,toml}]
+# 4 space indentation
+[*.rs]
 charset = utf-8
 indent_style = space
 indent_size = 4
-
-# 2 for c and js
-[*.{js,json,c,h}]
-charset = utf-8
-indent_style = space
-indent_size = 2

.github/actionlint.yaml

@@ -5,7 +5,6 @@ self-hosted-runner:
     - 4090-desktop
     - large_windows_16_latest
     - large_ubuntu_16
-    - large_ubuntu_16-22.04
 # Configuration variables in array of strings defined in your repository or
 # organization. `null` means disabling configuration variables check.
 # Empty array means no configuration variable is allowed.

.github/actions/hyperstack_setup/action.yml

@@ -1,53 +0,0 @@
-name: Setup Cuda
-description: Setup Cuda on Hyperstack instance
-
-inputs:
-  cuda-version:
-    description: Version of Cuda to use
-    required: true
-  gcc-version:
-    description: Version of GCC to use
-    required: true
-  cmake-version:
-    description: Version of cmake to use
-    default: 3.29.6
-
-runs:
-  using: "composite"
-  steps:
-    # Mandatory on hyperstack since a bootable volume is not re-usable yet.
-    - name: Install dependencies
-      shell: bash
-      run: |
-        sudo apt update
-        sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
-        wget https://github.com/Kitware/CMake/releases/download/v${{ inputs.cmake-version }}/cmake-${{ inputs.cmake-version }}.tar.gz
-        tar -zxvf cmake-${{ inputs.cmake-version }}.tar.gz
-        cd cmake-${{ inputs.cmake-version }}
-        ./bootstrap
-        make -j"$(nproc)"
-        sudo make install
-
-    - name: Export CUDA variables
-      shell: bash
-      run: |
-        CUDA_PATH=/usr/local/cuda-${{ inputs.cuda-version }}
-        echo "CUDA_PATH=$CUDA_PATH" >> "${GITHUB_ENV}"
-        echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}"
-        echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH" >> "${GITHUB_ENV}"
-        echo "CUDACXX=/usr/local/cuda-${{ inputs.cuda-version }}/bin/nvcc" >> "${GITHUB_ENV}"
-
-    # Specify the correct host compilers
-    - name: Export gcc and g++ variables
-      shell: bash
-      run: |
-        {
-          echo "CC=/usr/bin/gcc-${{ inputs.gcc-version }}";
-          echo "CXX=/usr/bin/g++-${{ inputs.gcc-version }}";
-          echo "CUDAHOSTCXX=/usr/bin/g++-${{ inputs.gcc-version }}";
-          echo "HOME=/home/ubuntu";
-        } >> "${GITHUB_ENV}"
-
-    - name: Check device is detected
-      shell: bash
-      run: nvidia-smi

.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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -50,7 +50,7 @@ jobs:
           token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -76,7 +76,7 @@ jobs:
         with:
           persist-credentials: 'false'
           repository: zama-ai/tfhe-backward-compat-data
-          path: tests/tfhe-backward-compat-data
+          path: tfhe/tfhe-backward-compat-data
           lfs: 'true'
           ref: ${{ steps.backward_compat_branch.outputs.branch }}
 
@@ -94,13 +94,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (backward-compat-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, backward-compat-tests ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/aws_tfhe_fast_tests.yml

@@ -11,26 +11,12 @@ env:
   SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
   SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
   SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
-  MSG_MINIMAL: event,action url,commit
-  BRANCH: ${{ github.head_ref || github.ref }}
-  IS_PULL_REQUEST: ${{ github.event_name == 'pull_request' || github.event_name == 'pull_request_target' }}
-  REF: ${{ github.event.pull_request.head.sha || github.sha }}
+  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:
-  # Trigger pull_request event on CI files to be able to test changes before merging to main branch.
-  # Workflow would fail if changes come from a forked repository since secrets are not available with this event.
   pull_request:
-    paths:
-      - '.github/**'
-      - 'ci/**'
-  # General entry point for Zama's pull request as well as contribution from forks.
-  pull_request_target:
-    paths:
-      - '**'
-      - '!.github/**'
-      - '!ci/**'
 
 jobs:
   should-run:
@@ -62,31 +48,28 @@ jobs:
       user_docs_test: ${{ env.IS_PULL_REQUEST == 'false' ||
         steps.changed-files.outputs.user_docs_any_changed ||
         steps.changed-files.outputs.dependencies_any_changed }}
-      ci_file_changed: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.ci_any_changed }}
       any_file_changed: ${{ env.IS_PULL_REQUEST == 'false' || steps.aggregated-changes.outputs.any_changed }}
     steps:
       - name: Checkout tfhe-rs
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
-          ref: ${{ env.REF }}
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             dependencies:
               - tfhe/Cargo.toml
-              - tfhe-csprng/**
+              - concrete-csprng/**
               - tfhe-fft/**
               - tfhe-zk-pok/**
               - utils/tfhe-versionable/**
               - utils/tfhe-versionable-derive/**
             csprng:
-              - tfhe-csprng/**
+              - concrete-csprng/**
             zk_pok:
               - tfhe-zk-pok/**
             versionable:
@@ -119,15 +102,11 @@ jobs:
             user_docs:
               - tfhe/src/**
               - '!tfhe/src/c_api/**'
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - README.md
-            ci:
-              - .github/**
-              - ci/**
 
       - name: Aggregate file changes
         id: aggregated-changes
-        # CI files are not included in this aggregator.
         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' ||
@@ -142,28 +121,18 @@ jobs:
         run: |
           echo "any_changed=true" >> "$GITHUB_OUTPUT"
 
-  # Fail if the triggering actor is not part of Zama organization.
-  # If pull_request_target is emitted and CI files have changed, skip this job. This would skip following jobs.
-  check-user-permission:
-    needs: should-run
-    if: github.event_name != 'pull_request_target' ||
-      (github.event_name == 'pull_request_target' && needs.should-run.outputs.ci_file_changed == 'false')
-    uses: ./.github/workflows/check_triggering_actor.yml
-    secrets:
-      TOKEN: ${{ secrets.GITHUB_TOKEN }}
-
   setup-instance:
     name: Setup instance (fast-tests)
-    if: github.event_name == 'workflow_dispatch' ||
-      (github.event_name != 'workflow_dispatch' && needs.should-run.outputs.any_file_changed == 'true')
-    needs: [ should-run, check-user-permission ]
+    if: github.event_name != 'pull_request' ||
+      needs.should-run.outputs.any_file_changed == 'true'
+    needs: should-run
     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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -174,9 +143,11 @@ jobs:
 
   fast-tests:
     name: Fast CPU tests
+    if: github.event_name != 'pull_request' ||
+      (github.event_name == 'pull_request' && needs.setup-instance.result != 'skipped')
     needs: [ should-run, setup-instance ]
     concurrency:
-      group: ${{ github.workflow }}_${{ github.head_ref || github.ref }}
+      group: ${{ github.workflow }}_${{ github.ref }}
       cancel-in-progress: true
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
@@ -184,18 +155,17 @@ jobs:
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
-          token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
-          ref: ${{ env.REF }}
+          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
-      - name: Run tfhe-csprng tests
+      - name: Run concrete-csprng tests
         if: needs.should-run.outputs.csprng_test == 'true'
         run: |
-          make test_tfhe_csprng
+          make test_concrete_csprng
 
       - name: Run tfhe-zk-pok tests
         if: needs.should-run.outputs.zk_pok_test == 'true'
@@ -228,7 +198,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@1bd1e32a3bdc45362d1e726936510720a7c30a57 #v4.2.0
+        uses: actions/cache/restore@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
         with:
           path: |
             ~/.nvm
@@ -241,7 +211,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@1bd1e32a3bdc45362d1e726936510720a7c30a57 #v4.2.0
+        uses: actions/cache/save@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |
@@ -289,17 +259,17 @@ jobs:
         uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "Fast AWS tests finished with status: ${{ job.status }} on '${{ env.BRANCH }}'. (${{ env.ACTION_RUN_URL }})"
+          SLACK_MESSAGE: "Fast AWS tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
     name: Teardown instance (fast-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, fast-tests ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -313,4 +283,4 @@ jobs:
         uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "Instance teardown (fast-tests) finished with status: ${{ job.status }} on '${{ env.BRANCH }}'. (${{ env.ACTION_RUN_URL }})"
+          SLACK_MESSAGE: "Instance teardown (fast-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/aws_tfhe_integer_tests.yml

@@ -42,24 +42,22 @@ jobs:
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
           persist-credentials: "false"
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             integer:
               - tfhe/Cargo.toml
-              - tfhe-csprng/**
+              - concrete-csprng/**
               - tfhe-fft/**
               - tfhe-zk-pok/**
               - tfhe/src/core_crypto/**
               - tfhe/src/shortint/**
               - tfhe/src/integer/**
-              - .github/workflows/aws_tfhe_integer_tests.yml
 
   setup-instance:
     name: Setup instance (unsigned-integer-tests)
@@ -75,7 +73,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -99,7 +97,7 @@ jobs:
           token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -134,13 +132,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (unsigned-integer-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [setup-instance, unsigned-integer-tests]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/aws_tfhe_signed_integer_tests.yml

@@ -42,24 +42,22 @@ jobs:
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
           persist-credentials: "false"
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             integer:
               - tfhe/Cargo.toml
-              - tfhe-csprng/**
+              - concrete-csprng/**
               - tfhe-fft/**
               - tfhe-zk-pok/**
               - tfhe/src/core_crypto/**
               - tfhe/src/shortint/**
               - tfhe/src/integer/**
-              - .github/workflows/aws_tfhe_signed_integer_tests.yml
 
   setup-instance:
     name: Setup instance (unsigned-integer-tests)
@@ -75,7 +73,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -99,7 +97,7 @@ jobs:
           token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -138,13 +136,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (signed-integer-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [setup-instance, signed-integer-tests]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/aws_tfhe_tests.yml

@@ -40,9 +40,6 @@ jobs:
       shortint_test: ${{ env.IS_PULL_REQUEST == 'false' ||
         steps.changed-files.outputs.shortint_any_changed ||
         steps.changed-files.outputs.dependencies_any_changed }}
-      strings_test: ${{ env.IS_PULL_REQUEST == 'false' ||
-        steps.changed-files.outputs.strings_any_changed ||
-        steps.changed-files.outputs.dependencies_any_changed }}
       high_level_api_test: ${{ env.IS_PULL_REQUEST == 'false' ||
         steps.changed-files.outputs.high_level_api_any_changed ||
         steps.changed-files.outputs.dependencies_any_changed }}
@@ -63,21 +60,20 @@ jobs:
         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@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             dependencies:
               - tfhe/Cargo.toml
-              - tfhe-csprng/**
+              - concrete-csprng/**
               - tfhe-fft/**
               - tfhe-zk-pok/**
             csprng:
-              - tfhe-csprng/**
+              - concrete-csprng/**
             zk_pok:
               - tfhe-zk-pok/**
             core_crypto:
@@ -88,11 +84,6 @@ jobs:
             shortint:
               - tfhe/src/core_crypto/**
               - tfhe/src/shortint/**
-            strings:
-              - tfhe/src/core_crypto/**
-              - tfhe/src/shortint/**
-              - tfhe/src/integer/**
-              - tfhe/src/strings/**
             high_level_api:
               - tfhe/src/**
               - '!tfhe/src/c_api/**'
@@ -111,7 +102,7 @@ jobs:
             user_docs:
               - tfhe/src/**
               - '!tfhe/src/c_api/**'
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - README.md
 
       - name: Aggregate file changes
@@ -122,7 +113,6 @@ jobs:
           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' ||
-          steps.changed-files.outputs.strings_any_changed == 'true' ||
           steps.changed-files.outputs.high_level_api_any_changed == 'true' ||
           steps.changed-files.outputs.c_api_any_changed == 'true' ||
           steps.changed-files.outputs.examples_any_changed == 'true' ||
@@ -142,7 +132,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -168,14 +158,14 @@ jobs:
           token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
-      - name: Run tfhe-csprng tests
+      - name: Run concrete-csprng tests
         if: needs.should-run.outputs.csprng_test == 'true'
         run: |
-          make test_tfhe_csprng
+          make test_concrete_csprng
 
       - name: Run tfhe-zk-pok tests
         if: needs.should-run.outputs.zk_pok_test == 'true'
@@ -212,11 +202,6 @@ jobs:
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_shortint_ci
 
-      - name: Run strings tests
-        if: needs.should-run.outputs.strings_test == 'true'
-        run: |
-          BIG_TESTS_INSTANCE=TRUE make test_strings
-
       - name: Run high-level API tests
         if: needs.should-run.outputs.high_level_api_test == 'true'
         run: |
@@ -244,13 +229,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cpu-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, cpu-tests ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -51,7 +51,7 @@ jobs:
           token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -61,7 +61,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@1bd1e32a3bdc45362d1e726936510720a7c30a57 #v4.2.0
+        uses: actions/cache/restore@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
         with:
           path: |
             ~/.nvm
@@ -74,7 +74,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@1bd1e32a3bdc45362d1e726936510720a7c30a57 #v4.2.0
+        uses: actions/cache/save@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |
@@ -113,13 +113,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (wasm-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, wasm-tests ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_boolean.yml

@@ -29,7 +29,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -62,7 +62,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -80,7 +80,8 @@ jobs:
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
-          --name-suffix avx512
+          --name-suffix avx512 \
+          --throughput
 
       - name: Measure key sizes
         run: |
@@ -93,7 +94,7 @@ jobs:
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_boolean
           path: ${{ env.RESULTS_FILENAME }}
@@ -121,13 +122,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (boolean-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, boolean-benchmarks ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_core_crypto.yml

@@ -3,9 +3,6 @@ name: Core crypto benchmarks
 
 on:
   workflow_dispatch:
-  schedule:
-    # Weekly benchmarks will be triggered each Saturday at 5a.m.
-    - cron: '0 5 * * 6'
 
 env:
   CARGO_TERM_COLOR: always
@@ -29,7 +26,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -61,7 +58,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -81,10 +78,11 @@ jobs:
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --name-suffix avx512 \
-          --walk-subdirs
+          --walk-subdirs \
+          --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_core_crypto
           path: ${{ env.RESULTS_FILENAME }}
@@ -112,13 +110,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (core-crypto-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, core-crypto-benchmarks ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_erc20.yml

@@ -29,7 +29,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -63,7 +63,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -97,7 +97,7 @@ jobs:
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_erc20
           path: ${{ env.RESULTS_FILENAME }}
@@ -118,13 +118,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (erc20-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_gpu_4090.yml

@@ -54,7 +54,7 @@ jobs:
           echo "FAST_BENCH=TRUE" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -79,10 +79,11 @@ jobs:
           --branch ${{ github.ref_name }} \
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
-          --walk-subdirs
+          --walk-subdirs \
+          --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_integer_multi_bit_gpu_default
           path: ${{ env.RESULTS_FILENAME }}
@@ -116,7 +117,6 @@ jobs:
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Get benchmark details
         run: |
@@ -127,7 +127,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -154,10 +154,10 @@ jobs:
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
-      
+          --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_core_crypto
           path: ${{ env.RESULTS_FILENAME }}

.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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -48,19 +48,29 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Get benchmark details
         run: |
           {
@@ -75,10 +85,31 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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 }}";
+            echo "HOME=/home/ubuntu";
+          } >> "${GITHUB_ENV}"
+
       - name: Run benchmarks with AVX512
         run: |
           make bench_pbs_gpu
@@ -95,10 +126,11 @@ jobs:
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --name-suffix avx512 \
-          --walk-subdirs
+          --walk-subdirs \
+          --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_core_crypto
           path: ${{ env.RESULTS_FILENAME }}
@@ -131,13 +163,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-integer-full-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, cuda-core-crypto-benchmarks, slack-notify ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_gpu_erc20.yml

@@ -1,44 +1,195 @@
-# Run CUDA ERC20 benchmarks on a Hyperstack VM and return parsed results to Slab CI bot.
-name: Cuda ERC20 benchmarks
+# Run ERC20 benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: ERC20 GPU H100 benchmarks
 
 on:
   workflow_dispatch:
-    inputs:
-      profile:
-        description: "Instance type"
-        required: true
-        type: choice
-        options:
-          - "l40 (n3-L40x1)"
-          - "single-h100 (n3-H100x1)"
-          - "2-h100 (n3-H100x2)"
-          - "4-h100 (n3-H100x4)"
-          - "multi-h100 (n3-H100x8)"
-          - "multi-h100-nvlink (n3-H100x8-NVLink)"
-          - "multi-h100-sxm5 (n3-H100x8-SXM5)"
+  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:
-  parse-inputs:
+  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:
-      profile: ${{ steps.parse_profile.outputs.profile }}
-      hardware_name: ${{ steps.parse_hardware_name.outputs.name }}
+      runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
-      - name: Parse profile
-        id: parse_profile
-        run: |
-          echo "profile=$(echo '${{ inputs.profile }}' | sed 's|\(.*\)[[:space:]](.*)|\1|')" >> "${GITHUB_OUTPUT}"
+      - 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
 
-      - name: Parse hardware name
-        id: parse_hardware_name
+  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: |
-          echo "name=$(echo '${{ inputs.profile }}' | sed 's|.*[[:space:]](\(.*\))|\1|')" >> "${GITHUB_OUTPUT}"
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ 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
 
-  run-benchmarks:
-    name: Run benchmarks
-    needs: parse-inputs
-    uses: ./.github/workflows/benchmark_gpu_erc20_common.yml
-    with:
-      profile: ${{ needs.parse-inputs.outputs.profile }}
-      hardware_name: ${{ needs.parse-inputs.outputs.hardware_name }}
-    secrets: inherit
+      - 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

@@ -1,35 +1,11 @@
 # Run ERC20 benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
-name: Cuda ERC20 benchmarks - common
+name: ERC20 GPU 2xH100 benchmarks
 
 on:
-  workflow_call:
-    inputs:
-      backend:
-        type: string
-        default: hyperstack
-      profile:
-        type: string
-        required: true
-      hardware_name:
-        type: string
-        required: true
-    secrets:
-      FHE_ACTIONS_TOKEN:
-        required: true
-      SLAB_ACTION_TOKEN:
-        required: true
-      SLAB_BASE_URL:
-        required: true
-      SLAB_URL:
-        required: true
-      JOB_SECRET:
-        required: true
-      SLACK_CHANNEL:
-        required: true
-      BOT_USERNAME:
-        required: true
-      SLACK_WEBHOOK:
-        required: true
+  workflow_dispatch:
+  schedule:
+    # Weekly benchmarks will be triggered each Saturday at 5a.m.
+    - cron: '0 5 * * 6'
 
 env:
   CARGO_TERM_COLOR: always
@@ -54,17 +30,17 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        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: ${{ inputs.backend }}
-          profile: ${{ inputs.profile }}
+          backend: hyperstack
+          profile: 2-h100
 
   cuda-erc20-benchmarks:
-    name: Cuda ERC20 benchmarks (${{ inputs.profile }})
+    name: Execute GPU integer benchmarks
     needs: setup-instance
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     strategy:
@@ -75,19 +51,28 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Get benchmark details
         run: |
           {
@@ -102,10 +87,34 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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
@@ -114,7 +123,7 @@ jobs:
         run: |
           python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
           --database tfhe_rs \
-          --hardware "${{ inputs.hardware_name }}" \
+          --hardware "n3-H100x2" \
           --backend gpu \
           --project-version "${{ env.COMMIT_HASH }}" \
           --branch ${{ github.ref_name }} \
@@ -123,10 +132,16 @@ jobs:
           --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@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
-          name: ${{ github.sha }}_erc20_${{ inputs.profile }}
+          name: ${{ github.sha }}_erc20
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
@@ -153,17 +168,17 @@ jobs:
         uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ needs.cuda-erc20-benchmarks.result }}
-          SLACK_MESSAGE: "Cuda ERC20 benchmarks (${{ inputs.profile }}) finished with status: ${{ needs.cuda-erc20-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_MESSAGE: "ERC20 2xH100 benchmarks finished with status: ${{ needs.cuda-erc20-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
-    name: Teardown instance (cuda-erc20-${{ inputs.profile }}-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -177,4 +192,4 @@ jobs:
         uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "Instance teardown (cuda-erc20-${{ inputs.profile }}-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+          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_erc20_weekly.yml

@@ -1,35 +0,0 @@
-# Run CUDA ERC20 benchmarks on multiple Hyperstack VMs and return parsed results to Slab CI bot.
-name: Cuda ERC20 weekly benchmarks
-
-on:
-  schedule:
-    # Weekly benchmarks will be triggered each Saturday at 5a.m.
-    - cron: '0 5 * * 6'
-
-jobs:
-  run-benchmarks-1-h100:
-    name: Run benchmarks (1xH100)
-    if: github.repository == 'zama-ai/tfhe-rs'
-    uses: ./.github/workflows/benchmark_gpu_erc20_common.yml
-    with:
-      profile: single-h100
-      hardware_name: n3-H100x1
-    secrets: inherit
-
-  run-benchmarks-2-h100:
-    name: Run benchmarks (2xH100)
-    if: github.repository == 'zama-ai/tfhe-rs'
-    uses: ./.github/workflows/benchmark_gpu_erc20_common.yml
-    with:
-      profile: 2-h100
-      hardware_name: n3-H100x2
-    secrets: inherit
-
-  run-benchmarks-8-h100:
-    name: Run benchmarks (8xH100)
-    if: github.repository == 'zama-ai/tfhe-rs'
-    uses: ./.github/workflows/benchmark_gpu_erc20_common.yml
-    with:
-      profile: multi-h100
-      hardware_name: n3-H100x8
-    secrets: inherit

.github/workflows/benchmark_gpu_integer.yml

@@ -1,79 +1,201 @@
-# Run CUDA benchmarks on a Hyperstack VM and return parsed results to Slab CI bot.
-name: Cuda benchmarks
+# Run integer benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: Integer GPU benchmarks
 
 on:
   workflow_dispatch:
-    inputs:
-      profile:
-        description: "Instance type"
-        required: true
-        type: choice
-        options:
-          - "l40 (n3-L40x1)"
-          - "single-h100 (n3-H100x1)"
-          - "2-h100 (n3-H100x2)"
-          - "4-h100 (n3-H100x4)"
-          - "multi-h100 (n3-H100x8)"
-          - "multi-h100-nvlink (n3-H100x8-NVLink)"
-          - "multi-h100-sxm5 (n3-H100x8-SXM5)"
-          - "multi-a100-nvlink (n3-A100x8-NVLink)"
-      command:
-        description: "Benchmark command to run"
-        type: choice
-        default: integer_multi_bit
-        options:
-          - integer
-          - integer_multi_bit
-          - integer_compression
-          - pbs
-          - ks
-      op_flavor:
-        description: "Operations set to run"
-        type: choice
-        default: default
-        options:
-          - default
-          - fast_default
-          - unchecked
-      all_precisions:
-        description: "Run all precisions"
-        type: boolean
-        default: false
-      bench_type:
-        description: "Benchmarks type"
-        type: choice
-        default: latency
-        options:
-          - latency
-          - throughput
-          - both
+  push:
+    branches:
+      - main
+
+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:
-  parse-inputs:
+  setup-instance:
+    name: Setup instance (cuda-integer-benchmarks)
     runs-on: ubuntu-latest
+    if:  github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'push' && github.repository == 'zama-ai/tfhe-rs')
     outputs:
-      profile: ${{ steps.parse_profile.outputs.profile }}
-      hardware_name: ${{ steps.parse_hardware_name.outputs.name }}
+      runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
-      - name: Parse profile
-        id: parse_profile
-        run: |
-          echo "profile=$(echo '${{ inputs.profile }}' | sed 's|\(.*\)[[:space:]](.*)|\1|')" >> "${GITHUB_OUTPUT}"
+      - 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
 
-      - name: Parse hardware name
-        id: parse_hardware_name
+  cuda-integer-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: |
-          echo "name=$(echo '${{ inputs.profile }}' | sed 's|.*[[:space:]](\(.*\))|\1|')" >> "${GITHUB_OUTPUT}"
+          sudo apt update
+          sudo apt install -y checkinstall zlib1g-dev libssl-dev libclang-dev
+          wget https://github.com/Kitware/CMake/releases/download/v${{ 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
 
-  run-benchmarks:
-    name: Run benchmarks
-    needs: parse-inputs
-    uses: ./.github/workflows/benchmark_gpu_integer_common.yml
-    with:
-      profile: ${{ needs.parse-inputs.outputs.profile }}
-      hardware_name: ${{ needs.parse-inputs.outputs.hardware_name }}
-      command: ${{ inputs.command }}
-      op_flavor: ${{ inputs.op_flavor }}
-      bench_type: ${{ inputs.bench_type }}
-      all_precisions: ${{ inputs.all_precisions }}
-    secrets: inherit
+      - 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: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run benchmarks with AVX512
+        run: |
+          make FAST_BENCH=TRUE BENCH_OP_FLAVOR=default bench_integer_gpu
+
+      - name: Parse benchmarks to csv
+        run: |
+          make PARSE_INTEGER_BENCH_CSV_FILE=${{ env.PARSE_INTEGER_BENCH_CSV_FILE }} \
+            parse_integer_benches
+
+      - name: Upload csv results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_csv_integer
+          path: ${{ env.PARSE_INTEGER_BENCH_CSV_FILE }}
+
+      - 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 \
+          --throughput
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_integer
+          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-integer-benchmarks ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-integer-benchmarks.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        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 }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-integer-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-integer-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-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_common.yml

@@ -1,256 +0,0 @@
-# Run integer benchmarks on CUDA instance and return parsed results to Slab CI bot.
-name: Cuda benchmarks - common
-
-on:
-  workflow_call:
-    inputs:
-      backend:
-        type: string
-        default: hyperstack
-      profile:
-        type: string
-        required: true
-      hardware_name:
-        type: string
-        required: true
-      command: # Use a comma separated values to generate an array
-        type: string
-        required: true
-      op_flavor: # Use a comma separated values to generate an array
-        type: string
-        required: true
-      bench_type:
-        type: string
-        default: latency
-      all_precisions:
-        type: boolean
-        default: false
-    secrets:
-      FHE_ACTIONS_TOKEN:
-        required: true
-      SLAB_ACTION_TOKEN:
-        required: true
-      SLAB_BASE_URL:
-        required: true
-      SLAB_URL:
-        required: true
-      JOB_SECRET:
-        required: true
-      SLACK_CHANNEL:
-        required: true
-      BOT_USERNAME:
-        required: true
-      SLACK_WEBHOOK:
-        required: true
-
-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 }}
-  FAST_BENCH: TRUE
-
-jobs:
-  prepare-matrix:
-    name: Prepare operations matrix
-    runs-on: ubuntu-latest
-    outputs:
-      command: ${{ steps.set_command.outputs.command }}
-      op_flavor: ${{ steps.set_op_flavor.outputs.op_flavor }}
-      bench_type: ${{ steps.set_bench_type.outputs.bench_type }}
-    steps:
-      - name: Set single command
-        if: ${{ !contains(inputs.command, ',')}}
-        run: |
-          echo "COMMAND=[\"${{ inputs.command }}\"]" >> "${GITHUB_ENV}"
-
-      - name: Set multiple commands
-        if: ${{ contains(inputs.command, ',')}}
-        run: |
-          PARSED_COMMAND=$(echo "${{ inputs.command }}" | sed 's/[[:space:]]*,[[:space:]]*/\\", \\"/g')
-          echo "COMMAND=[\"${PARSED_COMMAND}\"]" >> "${GITHUB_ENV}"
-
-      - name: Set single operations flavor
-        if: ${{ !contains(inputs.op_flavor, ',')}}
-        run: |
-          echo "OP_FLAVOR=[\"${{ inputs.op_flavor }}\"]" >> "${GITHUB_ENV}"
-
-      - name: Set multiple operations flavors
-        if: ${{ contains(inputs.op_flavor, ',')}}
-        run: |
-          PARSED_OP_FLAVOR=$(echo "${{ inputs.op_flavor }}" | sed 's/[[:space:]]*,[[:space:]]*/", "/g')
-          echo "OP_FLAVOR=[\"${PARSED_OP_FLAVOR}\"]" >> "${GITHUB_ENV}"
-
-      - name: Set benchmark types
-        run: |
-          if [[ "${{ inputs.bench_type }}" == "both" ]]; then
-            echo "BENCH_TYPE=[\"latency\", \"throughput\"]" >> "${GITHUB_ENV}"
-          else
-            echo "BENCH_TYPE=[\"${{ inputs.bench_type }}\"]" >> "${GITHUB_ENV}"
-          fi
-
-      - name: Set command output
-        id: set_command
-        run: |
-          echo "command=${{ toJSON(env.COMMAND) }}" >> "${GITHUB_OUTPUT}"
-
-      - name: Set operation flavor output
-        id: set_op_flavor
-        run: |
-          echo "op_flavor=${{ toJSON(env.OP_FLAVOR) }}" >> "${GITHUB_OUTPUT}"
-
-      - name: Set benchmark types output
-        id: set_bench_type
-        run: |
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
-
-  setup-instance:
-    name: Setup instance (cuda-${{ inputs.profile }}-benchmarks)
-    needs: prepare-matrix
-    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@79939325c3c429837c10d6041e4fd8589d328bac
-        with:
-          mode: start
-          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
-          slab-url: ${{ secrets.SLAB_BASE_URL }}
-          job-secret: ${{ secrets.JOB_SECRET }}
-          backend: ${{ inputs.backend }}
-          profile: ${{ inputs.profile }}
-
-  cuda-benchmarks:
-    name: Cuda benchmarks (${{ inputs.profile }})
-    needs: [ prepare-matrix, 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: ${{ fromJSON(needs.prepare-matrix.outputs.command) }}
-        op_flavor: ${{ fromJSON(needs.prepare-matrix.outputs.op_flavor) }}
-        bench_type: ${{ fromJSON(needs.prepare-matrix.outputs.bench_type) }}
-        # explicit include-based build matrix, of known valid options
-        include:
-          - os: ubuntu-22.04
-            cuda: "12.2"
-            gcc: 11
-    steps:
-      - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-        with:
-          fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
-
-      - name: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
-      - name: Get benchmark details
-        run: |
-          {
-            echo "BENCH_DATE=$(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@a54c7afa936fefeb4456b2dd8068152669aa8203
-        with:
-          toolchain: nightly
-
-      - name: Should run benchmarks with all precisions
-        if: inputs.all_precisions
-        run: |
-          echo "FAST_BENCH=FALSE" >> "${GITHUB_ENV}"
-
-      - name: Run benchmarks
-        run: |
-          make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} BENCH_TYPE=${{ matrix.bench_type }} bench_${{ matrix.command }}_gpu
-
-      - name: Parse results
-        run: |
-          python3 ./ci/benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
-          --database tfhe_rs \
-          --hardware "${{ inputs.hardware_name }}" \
-          --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 \
-          --bench-type ${{ matrix.bench_type }}
-
-      - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
-        with:
-          name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}_${{ inputs.profile }}
-          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-benchmarks ]
-    runs-on: ubuntu-latest
-    if: ${{ always() && needs.cuda-benchmarks.result != 'skipped' && failure() }}
-    continue-on-error: true
-    steps:
-      - name: Send message
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
-        env:
-          SLACK_COLOR: ${{ needs.cuda-benchmarks.result }}
-          SLACK_MESSAGE: "Cuda benchmarks (${{ inputs.profile }}) finished with status: ${{ needs.cuda-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
-
-  teardown-instance:
-    name: Teardown instance (cuda-${{ inputs.profile }}-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
-    needs: [ setup-instance, cuda-benchmarks, slack-notify ]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Stop instance
-        id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
-        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: ${{ job.status }}
-          SLACK_MESSAGE: "Instance teardown (cuda-${{ inputs.profile }}-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_full.yml

@@ -0,0 +1,200 @@
+# Run all integer benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: Integer GPU full 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-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: single-h100
+
+  cuda-integer-full-benchmarks:
+    name: Execute GPU integer benchmarks for all operations flavor
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    timeout-minutes: 1440 # 24 hours
+    continue-on-error: true
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        command: [integer, integer_multi_bit]
+        op_flavor: [default]
+        # explicit include-based build matrix, of known valid options
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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
+
+      # 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 ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --hardware "n3-H100x1" \
+          --backend gpu \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512 \
+          --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-benchmarks ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-integer-full-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-benchmarks.result }}
+          SLACK_MESSAGE: "Integer GPU full benchmarks finished with status: ${{ needs.cuda-integer-full-benchmarks.result }}. (${{ env.ACTION_RUN_URL }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-integer-full-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-integer-full-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-integer-full-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_multi_bit.yml

@@ -0,0 +1,224 @@
+# Run integer benchmarks with multi-bit cryptographic parameters on an instance and return parsed results to Slab CI bot.
+name: Integer GPU Multi-bit benchmarks
+
+on:
+  workflow_dispatch:
+    inputs:
+      all_precisions:
+        description: "Run all precisions"
+        type: boolean
+        default: false
+      fast_default:
+        description: "Run only deduplicated default operations without scalar variants"
+        type: boolean
+        default: false
+
+  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_${{ github.sha }}.json
+  PARSE_INTEGER_BENCH_CSV_FILE: tfhe_rs_integer_benches_${{ 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 }}
+  FAST_BENCH: TRUE
+  BENCH_OP_FLAVOR: default
+
+jobs:
+  setup-instance:
+    name: Setup instance (cuda-integer-multi-bit-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: single-h100
+
+  cuda-integer-multi-bit-benchmarks:
+    name: Execute GPU integer multi-bit benchmarks
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    timeout-minutes: 1440 # 24 hours
+    strategy:
+      fail-fast: false
+      # explicit include-based build matrix, of known valid options
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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 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 }}";
+          } >> "${GITHUB_ENV}"
+
+      - name: Should run benchmarks with all precisions
+        if: inputs.all_precisions
+        run: |
+          echo "FAST_BENCH=FALSE" >> "${GITHUB_ENV}"
+
+      - name: Should run fast subset benchmarks
+        if: inputs.fast_default
+        run: |
+          echo "BENCH_OP_FLAVOR=fast_default" >> "${GITHUB_ENV}"
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run multi-bit benchmarks with AVX512
+        run: |
+          make bench_unsigned_integer_multi_bit_gpu
+
+      - name: Parse benchmarks to csv
+        run: |
+          make PARSE_INTEGER_BENCH_CSV_FILE=${{ env.PARSE_INTEGER_BENCH_CSV_FILE }} \
+            parse_integer_benches
+
+      - name: Upload csv results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_csv_integer
+          path: ${{ env.PARSE_INTEGER_BENCH_CSV_FILE }}
+
+      - 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 \
+          --throughput
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_integer
+          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-integer-multi-bit-benchmarks ]
+    runs-on: ubuntu-latest
+    if: ${{ always() && needs.cuda-integer-multi-bit-benchmarks.result != 'skipped' && failure() }}
+    continue-on-error: true
+    steps:
+      - name: Send message
+        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 }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-integer-full-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-integer-multi-bit-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-integer-multi-bit-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_multi_bit_multi_gpu.yml

@@ -0,0 +1,214 @@
+# Run 64-bit multi-bit integer benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: Integer multi GPU Multi-bit benchmarks
+
+on:
+  workflow_dispatch:
+    inputs:
+      all_precisions:
+        description: "Run all precisions"
+        type: boolean
+        default: false
+      fast_default:
+        description: "Run only deduplicated default operations without scalar variants"
+        type: boolean
+        default: false
+
+  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 }}
+  FAST_BENCH: TRUE
+  BENCH_OP_FLAVOR: default
+
+jobs:
+  setup-instance:
+    name: Setup instance (cuda-integer-multi-bit-multi-gpu-benchmarks)
+    runs-on: ubuntu-latest
+    if: ${{ (github.event_name == 'push' && github.repository == 'zama-ai/tfhe-rs') ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
+      github.event_name == 'workflow_dispatch' }}
+    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-integer-multi-bit-multi-gpu-benchmarks:
+    name: Execute multi GPU integer multi-bit benchmarks
+    needs: setup-instance
+    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
+    timeout-minutes: 1440 # 24 hours
+    continue-on-error: true
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cuda: "12.2"
+            gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+      - name: Should run benchmarks with all precisions
+        if: inputs.all_precisions
+        run: |
+          echo "FAST_BENCH=FALSE" >> "${GITHUB_ENV}"
+
+      - name: Should run fast subset benchmarks
+        if: inputs.fast_default
+        run: |
+          echo "BENCH_OP_FLAVOR=fast_default" >> "${GITHUB_ENV}"
+
+      - name: Check device is detected
+        if: ${{ !cancelled() }}
+        run: nvidia-smi
+
+      - name: Run multi-bit benchmarks with AVX512
+        run: |
+          make bench_unsigned_integer_multi_bit_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 \
+          --throughput
+
+      - name: Upload parsed results artifact
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
+        with:
+          name: ${{ github.sha }}_integer
+          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-multi-bit-multi-gpu-benchmarks ]
+    runs-on: ubuntu-latest
+    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@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 }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-integer-multi-bit-multi-gpu-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-integer-multi-bit-multi-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-multi-bit-multi-gpu-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_multi_gpu_full.yml

@@ -0,0 +1,194 @@
+# Run all integer benchmarks on an instance with CUDA and return parsed results to Slab CI bot.
+name: Integer multi GPU full 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-multi-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: multi-h100
+
+  cuda-integer-full-multi-gpu-benchmarks:
+    name: Execute multi GPU 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 ${{ env.RESULTS_FILENAME }} \
+          --database tfhe_rs \
+          --hardware "n3-H100x8" \
+          --backend gpu \
+          --project-version "${{ env.COMMIT_HASH }}" \
+          --branch ${{ github.ref_name }} \
+          --commit-date "${{ env.COMMIT_DATE }}" \
+          --bench-date "${{ env.BENCH_DATE }}" \
+          --walk-subdirs \
+          --name-suffix avx512 \
+          --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-multi-gpu-benchmarks ]
+    runs-on: ubuntu-latest
+    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@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 }})"
+
+  teardown-instance:
+    name: Teardown instance (cuda-integer-full-multi-gpu-benchmarks)
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
+    needs: [ setup-instance, cuda-integer-full-multi-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-multi-gpu-benchmarks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_gpu_integer_weekly.yml

@@ -1,60 +0,0 @@
-# Run CUDA benchmarks on multiple Hyperstack VMs and return parsed results to Slab CI bot.
-name: Cuda weekly benchmarks
-
-on:
-  schedule:
-    # Weekly benchmarks will be triggered each Saturday at 1a.m.
-    - cron: '0 1 * * 6'
-
-jobs:
-  run-benchmarks-1-h100:
-    name: Run benchmarks (1xH100)
-    if: github.repository == 'zama-ai/tfhe-rs'
-    uses: ./.github/workflows/benchmark_gpu_integer_common.yml
-    with:
-      profile: single-h100
-      hardware_name: n3-H100x1
-      command: integer,integer_multi_bit
-      op_flavor: default
-      bench_type: latency
-      all_precisions: true
-    secrets: inherit
-
-  run-benchmarks-2-h100:
-    name: Run benchmarks (2xH100)
-    if: github.repository == 'zama-ai/tfhe-rs'
-    uses: ./.github/workflows/benchmark_gpu_integer_common.yml
-    with:
-      profile: 2-h100
-      hardware_name: n3-H100x2
-      command: integer_multi_bit
-      op_flavor: default
-      bench_type: latency
-      all_precisions: true
-    secrets: inherit
-
-  run-benchmarks-8-h100:
-    name: Run benchmarks (8xH100)
-    if: github.repository == 'zama-ai/tfhe-rs'
-    uses: ./.github/workflows/benchmark_gpu_integer_common.yml
-    with:
-      profile: multi-h100
-      hardware_name: n3-H100x8
-      command: integer_multi_bit
-      op_flavor: default
-      bench_type: latency
-      all_precisions: true
-    secrets: inherit
-
-  run-benchmarks-l40:
-    name: Run benchmarks (L40)
-    if: github.repository == 'zama-ai/tfhe-rs'
-    uses: ./.github/workflows/benchmark_gpu_integer_common.yml
-    with:
-      profile: l40
-      hardware_name: n3-L40x1
-      command: integer_multi_bit,integer_compression,pbs,ks
-      op_flavor: default
-      bench_type: latency
-      all_precisions: true
-    secrets: inherit

.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

@@ -8,14 +8,6 @@ on:
         description: "Run all precisions"
         type: boolean
         default: false
-      bench_type:
-        description: "Benchmarks type"
-        type: choice
-        default: latency
-        options:
-          - latency
-          - throughput
-          - both
 
   schedule:
     # Weekly benchmarks will be triggered each Saturday at 1a.m.
@@ -44,10 +36,10 @@ jobs:
       (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
     outputs:
       op_flavor: ${{ steps.set_op_flavor.outputs.op_flavor }}
-      bench_type: ${{ steps.set_bench_type.outputs.bench_type }}
     steps:
       - name: Weekly benchmarks
-        if: github.event.schedule == '0 1 * * 6'
+        if: github.event_name == 'workflow_dispatch' ||
+          github.event.schedule == '0 1 * * 6'
         run: |
           echo "OP_FLAVOR=[\"default\"]" >> "${GITHUB_ENV}"
 
@@ -56,31 +48,11 @@ jobs:
         run: |
           echo "OP_FLAVOR=[\"default\", \"smart\", \"unchecked\", \"misc\"]" >> "${GITHUB_ENV}"
 
-      - name: Set benchmark types
-        if: github.event_name == 'workflow_dispatch'
-        run: |
-          echo "OP_FLAVOR=[\"default\"]" >> "${GITHUB_ENV}"
-          if [[ "${{ inputs.bench_type }}" == "both" ]]; then
-            echo "BENCH_TYPE=[\"latency\", \"throughput\"]" >> "${GITHUB_ENV}"
-          else
-            echo "BENCH_TYPE=[\"${{ inputs.bench_type }}\"]" >> "${GITHUB_ENV}"
-          fi
-
-      - name: Default benchmark type
-        if: github.event_name != 'workflow_dispatch'
-        run: |
-          echo "BENCH_TYPE=[\"latency\"]" >> "${GITHUB_ENV}"
-
-      - name: Set operation flavor output
-        id: set_op_flavor
-        run: |
+      -  name: Set operation flavor output
+         id: set_op_flavor
+         run: |
           echo "op_flavor=${{ toJSON(env.OP_FLAVOR) }}" >> "${GITHUB_OUTPUT}"
 
-      - name: Set benchmark types output
-        id: set_bench_type
-        run: |
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
-
   setup-instance:
     name: Setup instance (integer-benchmarks)
     needs: prepare-matrix
@@ -90,7 +62,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -113,7 +85,6 @@ jobs:
       matrix:
         command: [ integer, integer_multi_bit]
         op_flavor: ${{ fromJson(needs.prepare-matrix.outputs.op_flavor) }}
-        bench_type: ${{ fromJSON(needs.prepare-matrix.outputs.bench_type) }}
     steps:
       - name: Checkout tfhe-rs repo with tags
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
@@ -130,7 +101,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -148,13 +119,13 @@ jobs:
 
       - name: Run benchmarks with AVX512
         run: |
-          make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} BENCH_TYPE=${{ matrix.bench_type }} bench_${{ matrix.command }}
+          make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} bench_${{ matrix.command }}
 
-      # Run these benchmarks only once per benchmark type
+      # Run these benchmarks only once
       - name: Run compression benchmarks with AVX512
         if: matrix.op_flavor == 'default' && matrix.command == 'integer'
         run: |
-          make BENCH_TYPE=${{ matrix.bench_type }} bench_integer_compression
+          make bench_integer_compression
 
       - name: Parse results
         run: |
@@ -167,12 +138,12 @@ jobs:
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
           --name-suffix avx512 \
-          --bench-type ${{ matrix.bench_type }}
+          --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
-          name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}_${{ matrix.bench_type }}
+          name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Send data to Slab
@@ -191,13 +162,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (integer-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, integer-benchmarks ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_shortint.yml

@@ -56,7 +56,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -93,7 +93,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -120,7 +120,8 @@ jobs:
           --commit-date "${COMMIT_DATE}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
-          --name-suffix avx512
+          --name-suffix avx512 \
+          --throughput
 
       # This small benchmark needs to be executed only once.
       - name: Measure key sizes
@@ -136,7 +137,7 @@ jobs:
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_shortint_${{ matrix.op_flavor }}
           path: ${{ env.RESULTS_FILENAME }}
@@ -157,13 +158,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (shortint-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, shortint-benchmarks ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_signed_integer.yml

@@ -8,14 +8,6 @@ on:
         description: "Run all precisions"
         type: boolean
         default: false
-      bench_type:
-        description: "Benchmarks type"
-        type: choice
-        default: latency
-        options:
-          - latency
-          - throughput
-          - both
 
   schedule:
     # Weekly benchmarks will be triggered each Saturday at 1a.m.
@@ -44,10 +36,10 @@ jobs:
       (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
     outputs:
       op_flavor: ${{ steps.set_op_flavor.outputs.op_flavor }}
-      bench_type: ${{ steps.set_bench_type.outputs.bench_type }}
     steps:
       - name: Weekly benchmarks
-        if: github.event.schedule == '0 1 * * 6'
+        if: github.event_name == 'workflow_dispatch' ||
+          github.event.schedule == '0 1 * * 6'
         run: |
           echo "OP_FLAVOR=[\"default\"]" >> "${GITHUB_ENV}"
 
@@ -56,31 +48,11 @@ jobs:
         run: |
           echo "OP_FLAVOR=[\"default\", \"unchecked\"]" >> "${GITHUB_ENV}"
 
-      - name: Set benchmark types
-        if: github.event_name == 'workflow_dispatch'
-        run: |
-          echo "OP_FLAVOR=[\"default\"]" >> "${GITHUB_ENV}"
-          if [[ "${{ inputs.bench_type }}" == "both" ]]; then
-            echo "BENCH_TYPE=[\"latency\", \"throughput\"]" >> "${GITHUB_ENV}"
-          else
-            echo "BENCH_TYPE=[\"${{ inputs.bench_type }}\"]" >> "${GITHUB_ENV}"
-          fi
-
-      - name: Default benchmark type
-        if: github.event_name != 'workflow_dispatch'
-        run: |
-          echo "BENCH_TYPE=[\"latency\"]" >> "${GITHUB_ENV}"
-
       - name: Set operation flavor output
         id: set_op_flavor
         run: |
           echo "op_flavor=${{ toJSON(env.OP_FLAVOR) }}" >> "${GITHUB_OUTPUT}"
 
-      - name: Set benchmark types output
-        id: set_bench_type
-        run: |
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
-
   setup-instance:
     name: Setup instance (signed-integer-benchmarks)
     needs: prepare-matrix
@@ -90,7 +62,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -112,8 +84,7 @@ jobs:
       max-parallel: 1
       matrix:
         command: [ integer, integer_multi_bit ]
-        op_flavor: ${{ fromJSON(needs.prepare-matrix.outputs.op_flavor) }}
-        bench_type: ${{ fromJSON(needs.prepare-matrix.outputs.bench_type) }}
+        op_flavor: [ default, unchecked ]
     steps:
       - name: Checkout tfhe-rs repo with tags
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
@@ -130,7 +101,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -148,7 +119,7 @@ jobs:
 
       - name: Run benchmarks with AVX512
         run: |
-          make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} BENCH_TYPE=${{ matrix.bench_type }} bench_signed_${{ matrix.command }}
+          make BENCH_OP_FLAVOR=${{ matrix.op_flavor }} bench_signed_${{ matrix.command }}
 
       - name: Parse results
         run: |
@@ -161,12 +132,12 @@ jobs:
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
           --name-suffix avx512 \
-          --bench-type ${{ matrix.bench_type }}
+          --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
-          name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}_${{ matrix.bench_type }}
+          name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Send data to Slab
@@ -185,13 +156,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (integer-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, signed-integer-benchmarks ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_tfhe_fft.yml

@@ -16,9 +16,6 @@ on:
   push:
     branches:
       - "main"
-    paths:
-      - tfhe-fft/**
-      - .github/workflows/benchmark_tfhe_fft.yml
   schedule:
     # Job will be triggered each Thursday at 11p.m.
     - cron: '0 23 * * 4'
@@ -32,7 +29,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@1dced74825027fe3d481392163ed8fc56813fb5d
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -50,7 +47,7 @@ jobs:
     runs-on: ${{ needs.setup-ec2.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
         with:
           fetch-depth: 0
 
@@ -68,11 +65,11 @@ jobs:
           toolchain: nightly
           override: true
 
-      - name: Run benchmarks with AVX512
+      - name: Run benchmarks
         run: |
-          make bench_fft
+          make FFT128_SUPPORT=ON bench
 
-      - name: Parse AVX512 results
+      - name: Parse results
         run: |
           python3 ./ci/fft_benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
           --database concrete_fft \
@@ -80,17 +77,28 @@ jobs:
           --project-version "${{ env.COMMIT_HASH }}" \
           --branch ${{ github.ref_name }} \
           --commit-date "${{ env.COMMIT_DATE }}" \
-          --bench-date "${{ env.BENCH_DATE }}" \
-          --name-suffix avx512
+          --bench-date "${{ 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@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@0b7f8abb1508181956e8e162db84b466c27e18ce
         with:
           name: ${{ github.sha }}_fft
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
         with:
           repository: zama-ai/slab
           path: slab
@@ -113,7 +121,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        uses: rtCamp/action-slack-notify@b24d75fe0e728a4bf9fc42ee217caa686d141ee8
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "tfhe-fft benchmarks failed. (${{ env.ACTION_RUN_URL }})"
@@ -126,7 +134,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@1dced74825027fe3d481392163ed8fc56813fb5d
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -137,7 +145,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
+        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_ntt.yml

@@ -1,143 +0,0 @@
-# Run NTT benchmarks on an AWS instance and return parsed results to Slab CI bot.
-name: NTT 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"
-    paths:
-      - tfhe-ntt/**
-      - .github/workflows/benchmark_tfhe_ntt.yml
-  schedule:
-    # Job will be triggered each Friday at 11p.m.
-    - cron: "0 23 * * 5"
-
-jobs:
-  setup-ec2:
-    name: Setup EC2 instance (ntt-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@79939325c3c429837c10d6041e4fd8589d328bac
-        with:
-          mode: start
-          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
-          slab-url: ${{ secrets.SLAB_BASE_URL }}
-          job-secret: ${{ secrets.JOB_SECRET }}
-          backend: aws
-          profile: bench
-
-  ntt-benchmarks:
-    name: Execute NTT benchmarks in EC2
-    needs: setup-ec2
-    concurrency:
-      group: ${{ github.workflow }}_${{ github.ref }}
-      cancel-in-progress: true
-    runs-on: ${{ needs.setup-ec2.outputs.runner-name }}
-    steps:
-      - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-        with:
-          fetch-depth: 0
-
-      - name: Get benchmark details
-        run: |
-          {
-            echo "BENCH_DATE=$(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 bench_ntt
-
-      - name: Parse results
-        run: |
-          python3 ./ci/ntt_benchmark_parser.py target/criterion ${{ env.RESULTS_FILENAME }} \
-          --database concrete_ntt \
-          --hardware "hpc7a.96xlarge" \
-          --project-version "${{ env.COMMIT_HASH }}" \
-          --branch ${{ github.ref_name }} \
-          --commit-date "${{ env.COMMIT_DATE }}" \
-          --bench-date "${{ env.BENCH_DATE }}" \
-          --name-suffix avx512
-
-      - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
-        with:
-          name: ${{ github.sha }}_ntt
-          path: ${{ env.RESULTS_FILENAME }}
-
-      - name: Checkout Slab repo
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-        with:
-          repository: zama-ai/slab
-          path: slab
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
-
-      - name: Send data to Slab
-        shell: bash
-        run: |
-          echo "Computing HMac on downloaded artifact"
-          SIGNATURE="$(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@c33737706dea87cd7784c687dadc9adf1be59990
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "tfhe-ntt benchmarks failed. (${{ env.ACTION_RUN_URL }})"
-
-  teardown-ec2:
-    name: Teardown EC2 instance (ntt-benchmarks)
-    if: ${{ always() && needs.setup-ec2.result != 'skipped' }}
-    needs: [setup-ec2, ntt-benchmarks]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Stop instance
-        id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
-        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@c33737706dea87cd7784c687dadc9adf1be59990
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "EC2 teardown (ntt-benchmarks) failed. (${{ env.ACTION_RUN_URL }})"

.github/workflows/benchmark_tfhe_zk_pok.yml

@@ -36,7 +36,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@c3a1bb2c992d77180ae65be6ae6c166cf40f857c
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -58,7 +58,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -91,7 +91,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -118,10 +118,11 @@ jobs:
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
-          --name-suffix avx512
+          --name-suffix avx512 \
+          --throughput
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_tfhe_zk_pok
           path: ${{ env.RESULTS_FILENAME }}
@@ -149,13 +150,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (tfhe-zk-pok-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_wasm_client.yml

@@ -36,17 +36,16 @@ jobs:
         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@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             wasm_bench:
               - tfhe/Cargo.toml
-              - tfhe-csprng/**
+              - concrete-csprng/**
               - tfhe-zk-pok/**
               - tfhe/src/**
               - '!tfhe/src/c_api/**'
@@ -65,7 +64,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -99,7 +98,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -109,7 +108,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@1bd1e32a3bdc45362d1e726936510720a7c30a57 #v4.2.0
+        uses: actions/cache/restore@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
         with:
           path: |
             ~/.nvm
@@ -122,7 +121,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@1bd1e32a3bdc45362d1e726936510720a7c30a57 #v4.2.0
+        uses: actions/cache/save@6849a6489940f00c2f30c0fb92c6274307ccb58a #v4.1.2
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |
@@ -166,7 +165,7 @@ jobs:
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_wasm_${{ matrix.browser }}
           path: ${{ env.RESULTS_FILENAME }}
@@ -194,13 +193,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (wasm-client-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, wasm-client-benchmarks ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/benchmark_zk_pke.yml

@@ -3,16 +3,6 @@ name: PKE ZK benchmarks
 
 on:
   workflow_dispatch:
-    inputs:
-      bench_type:
-        description: "Benchmarks type"
-        type: choice
-        default: latency
-        options:
-          - latency
-          - throughput
-          - both
-
   push:
     branches:
       - main
@@ -43,17 +33,16 @@ jobs:
         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@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
             zk_pok:
               - tfhe/Cargo.toml
-              - tfhe-csprng/**
+              - concrete-csprng/**
               - tfhe-fft/**
               - tfhe-zk-pok/**
               - tfhe/src/core_crypto/**
@@ -63,37 +52,10 @@ jobs:
               - tfhe/benches/integer/zk_pke.rs
               - .github/workflows/zk_pke_benchmark.yml
 
-  prepare-matrix:
-    name: Prepare operations matrix
-    runs-on: ubuntu-latest
-    if: github.event_name != 'schedule' ||
-      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
-    outputs:
-      bench_type: ${{ steps.set_bench_type.outputs.bench_type }}
-    steps:
-      - name: Set benchmark types
-        if: github.event_name == 'workflow_dispatch'
-        run: |
-          if [[ "${{ inputs.bench_type }}" == "both" ]]; then
-            echo "BENCH_TYPE=[\"latency\", \"throughput\"]" >> "${GITHUB_ENV}"
-          else
-            echo "BENCH_TYPE=[\"${{ inputs.bench_type }}\"]" >> "${GITHUB_ENV}"
-          fi
-
-      - name: Default benchmark type
-        if: github.event_name != 'workflow_dispatch'
-        run: |
-          echo "BENCH_TYPE=[\"latency\"]" >> "${GITHUB_ENV}"
-
-      - name: Set benchmark types output
-        id: set_bench_type
-        run: |
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
-
   setup-instance:
     name: Setup instance (pke-zk-benchmarks)
     runs-on: ubuntu-latest
-    needs: [ should-run, prepare-matrix ]
+    needs: should-run
     if: github.event_name == 'workflow_dispatch' ||
       (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
       (github.event_name == 'push' &&
@@ -104,7 +66,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -116,15 +78,11 @@ jobs:
   pke-zk-benchmarks:
     name: Execute PKE ZK benchmarks
     if: needs.setup-instance.result != 'skipped'
-    needs: [ prepare-matrix, setup-instance ]
+    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 }}
-    strategy:
-      max-parallel: 1
-      matrix:
-        bench_type: ${{ fromJSON(needs.prepare-matrix.outputs.bench_type) }}
     steps:
       - name: Checkout tfhe-rs repo with tags
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
@@ -141,7 +99,7 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: nightly
 
@@ -154,7 +112,7 @@ jobs:
 
       - name: Run benchmarks with AVX512
         run: |
-          make BENCH_TYPE=${{ matrix.bench_type }} bench_integer_zk
+          make bench_integer_zk
 
       - name: Parse results
         run: |
@@ -168,7 +126,7 @@ jobs:
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
           --name-suffix avx512 \
-          --bench-type ${{ matrix.bench_type }}
+          --throughput
 
       - name: Parse CRS sizes results
         run: |
@@ -177,7 +135,7 @@ jobs:
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08
+        uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
         with:
           name: ${{ github.sha }}_integer_zk
           path: ${{ env.RESULTS_FILENAME }}
@@ -205,13 +163,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (pke-zk-benchmarks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, pke-zk-benchmarks ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/cargo_build.yml

@@ -28,7 +28,7 @@ jobs:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -47,10 +47,10 @@ jobs:
         run: |
           make pcc
 
-      - name: Build tfhe-csprng
+      - name: Build concrete-csprng
         if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |
-          make build_tfhe_csprng
+          make build_concrete_csprng
 
       - name: Build Release core
         if: ${{ contains(matrix.os, 'ubuntu') }}

.github/workflows/cargo_build_tfhe_fft.yml

@@ -2,6 +2,9 @@
 name: Cargo Build tfhe-fft
 
 on:
+  push:
+    branches:
+      - 'main'
   pull_request:
 
 env:
@@ -21,7 +24,7 @@ jobs:
       fail-fast: false
 
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
 
       - name: Install Rust
         uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af

.github/workflows/cargo_build_tfhe_ntt.yml

@@ -1,40 +0,0 @@
-# Build tfhe-ntt
-name: Cargo Build tfhe-ntt
-
-on:
-  pull_request:
-
-env:
-  CARGO_TERM_COLOR: always
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref }}
-  cancel-in-progress: true
-
-jobs:
-  cargo-builds-ntt:
-    runs-on: ${{ matrix.os }}
-    strategy:
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
-      fail-fast: false
-    steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-
-      - name: Install Rust
-        uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
-        with:
-          toolchain: stable
-          override: true
-
-      - name: Run pcc checks
-        run: |
-          make pcc_ntt
-
-      - name: Build release
-        run: |
-          make build_ntt
-
-      - name: Build release no-std
-        run: |
-          make build_ntt_no_std

.github/workflows/cargo_test_fft.yml

@@ -2,6 +2,9 @@
 name: Cargo Test tfhe-fft
 
 on:
+  push:
+    branches:
+      - "main"
   pull_request:
 
 env:
@@ -12,14 +15,14 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  cargo-tests-fft:
+  cargo-tests:
     runs-on: ${{ matrix.runner_type }}
     strategy:
       matrix:
         runner_type: [ubuntu-latest, macos-latest, windows-latest]
       fail-fast: false
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
 
       - name: Install Rust
         uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
@@ -38,13 +41,13 @@ jobs:
         run: |
           make test_fft_no_std
 
-  cargo-tests-fft-nightly:
+  cargo-tests-nightly:
     runs-on: ${{ matrix.runner_type }}
     strategy:
       matrix:
         runner_type: [ubuntu-latest, macos-latest, windows-latest]
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
 
       - name: Install Rust
         uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
@@ -60,10 +63,10 @@ jobs:
         run: |
           make test_fft_no_std_nightly
 
-  cargo-tests-fft-node-js:
+  cargo-tests-node-js:
     runs-on: "ubuntu-latest"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c
 
       - name: Test node js
         run: |

.github/workflows/cargo_test_ntt.yml

@@ -1,54 +0,0 @@
-# Test tfhe-ntt
-name: Cargo Test tfhe-ntt
-
-on:
-  pull_request:
-
-env:
-  CARGO_TERM_COLOR: always
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref }}
-  cancel-in-progress: true
-
-jobs:
-  cargo-tests-ntt:
-    runs-on: ${{ matrix.os }}
-    strategy:
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
-      fail-fast: false
-    steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-
-      - name: Install Rust
-        uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
-        with:
-          toolchain: stable
-          override: true
-
-      - name: Test debug
-        run: make test_ntt
-
-      - name: Test no-std
-        run: make test_ntt_no_std
-
-  cargo-tests-ntt-nightly:
-    runs-on: ${{ matrix.os }}
-    strategy:
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
-    steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-
-      - name: Install Rust
-        uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af
-        with:
-          toolchain: nightly
-          override: true
-
-      - name: Test nightly
-        run: make test_ntt_nightly
-
-      - name: Test no-std nightly
-        run: make test_ntt_no_std_nightly

.github/workflows/check_triggering_actor.yml

@@ -1,29 +0,0 @@
-# Check if triggering actor is a collaborator and has write access
-name: Check Triggering Actor
-
-on:
-  workflow_call:
-    secrets:
-      TOKEN:
-        required: true
-
-jobs:
-  check-actor-permission:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Get User Permission
-        id: check-access
-        uses: actions-cool/check-user-permission@7b90a27f92f3961b368376107661682c441f6103 # v2.3.0
-        with:
-          require: write
-          username: ${{ github.triggering_actor }}
-        env:
-          GITHUB_TOKEN: ${{ secrets.TOKEN }}
-
-      - name: Check User Permission
-        if: steps.check-access.outputs.require-result == 'false'
-        run: |
-          echo "${{ github.triggering_actor }} does not have permissions on this repo."
-          echo "Current permission level is ${{ steps.check-access.outputs.user-permission }}"
-          echo "Job originally triggered by ${{ github.actor }}"
-          exit 1

.github/workflows/ci_lint.yml

@@ -27,7 +27,7 @@ jobs:
           make lint_workflow
 
       - name: Ensure SHA pinned actions
-        uses: zgosalvez/github-actions-ensure-sha-pinned-actions@c3a2b64f69b7a1542a68f44d9edbd9ec3fc1455e # v3.0.20
+        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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -47,19 +47,19 @@ jobs:
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           files_yaml: |
             tfhe:
               - tfhe/src/**
-            tfhe_csprng:
-              - tfhe-csprng/src/**
+            concrete_csprng:
+              - concrete-csprng/src/**
 
       - name: Generate Keys
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
@@ -83,7 +83,7 @@ jobs:
           make test_shortint_cov
 
       - name: Upload tfhe coverage to Codecov
-        uses: codecov/codecov-action@1e68e06f1dbfde0e4cefc87efeba9e4643565303
+        uses: codecov/codecov-action@5c47607acb93fed5485fdbf7232e8a31425f672a
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
@@ -97,7 +97,7 @@ jobs:
           make test_integer_cov
 
       - name: Upload tfhe coverage to Codecov
-        uses: codecov/codecov-action@1e68e06f1dbfde0e4cefc87efeba9e4643565303
+        uses: codecov/codecov-action@5c47607acb93fed5485fdbf7232e8a31425f672a
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
@@ -115,13 +115,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (code-coverage)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, code-coverage ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/csprng_randomness_tests.yml

@@ -27,7 +27,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -51,7 +51,7 @@ jobs:
           token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -65,17 +65,17 @@ jobs:
         uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "tfhe-csprng randomness check finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
+          SLACK_MESSAGE: "concrete-csprng randomness check finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
 
   teardown-instance:
     name: Teardown instance (csprng-randomness-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, csprng-randomness-tests ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_4090_tests.yml

@@ -40,7 +40,7 @@ jobs:
           token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 

.github/workflows/gpu_fast_h100_tests.yml

@@ -31,11 +31,10 @@ jobs:
         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@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -51,7 +50,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/gpu_fast_h100_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -68,7 +67,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -94,28 +93,60 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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 and internal CUDA backend tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_core_crypto_gpu
@@ -149,13 +180,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-h100-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_fast_tests.yml

@@ -30,11 +30,10 @@ jobs:
         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@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -50,7 +49,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/gpu_fast_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -66,7 +65,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -92,28 +91,60 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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 and internal CUDA backend tests
         run: |
           make test_core_crypto_gpu
@@ -147,13 +178,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_full_h100_tests.yml

@@ -25,7 +25,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -49,6 +49,9 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
+      CMAKE_VERSION: 3.29.6
     steps:
       # Mandatory on hyperstack since a bootable volume is not re-usable yet.
       - name: Install dependencies
@@ -68,21 +71,38 @@ jobs:
           persist-credentials: 'false'
           token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
-      - name: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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
@@ -119,7 +139,7 @@ jobs:
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_full_multi_gpu_tests.yml

@@ -31,11 +31,10 @@ jobs:
         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@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -51,7 +50,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/**_multi_gpu_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -68,7 +67,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -94,28 +93,60 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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 multi-bit CUDA integer compression tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_integer_compression_gpu
@@ -152,13 +183,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-tests-multi-gpu)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_integer_long_run_tests.yml

@@ -1,115 +0,0 @@
-name: Long Run Tests on GPU
-
-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 }}
-
-on:
-  # Allows you to run this workflow manually from the Actions tab as an alternative.
-  workflow_dispatch:
-  schedule:
-    # Weekly tests will be triggered each Friday at 9p.m.
-    - cron: "0 21 * * 5"
-
-jobs:
-  setup-instance:
-    name: Setup instance (gpu-tests)
-    if: github.event_name != 'schedule' ||
-      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
-    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@79939325c3c429837c10d6041e4fd8589d328bac
-        with:
-          mode: start
-          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
-          slab-url: ${{ secrets.SLAB_BASE_URL }}
-          job-secret: ${{ secrets.JOB_SECRET }}
-          backend: hyperstack
-          profile: multi-gpu-test
-
-  cuda-tests:
-    name: Long run GPU tests
-    needs: [ setup-instance ]
-    concurrency:
-      group: ${{ github.workflow }}_${{github.event_name}}_${{ github.ref }}
-      cancel-in-progress: true
-    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 
-    timeout-minutes: 4320 # 72 hours
-    steps:
-      - name: Checkout tfhe-rs
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-
-      - name: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
-      - name: Set up home
-        run: |
-          echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
-
-      - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
-        with:
-          toolchain: stable
-
-      - name: Run tests
-        run: |
-          make test_integer_long_run_gpu
-
-  slack-notify:
-    name: Slack Notification
-    needs: [ setup-instance, cuda-tests ]
-    runs-on: ubuntu-latest
-    if: ${{ always() && needs.cuda-tests.result != 'skipped' && failure() }}
-    continue-on-error: true
-    steps:
-      - name: Send message
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
-        env:
-          SLACK_COLOR: ${{ needs.cuda-tests.result }}
-          SLACK_MESSAGE: "Integer GPU long run tests finished with status: ${{ needs.cuda-tests.result }}. (${{ env.ACTION_RUN_URL }})"
-
-  teardown-instance:
-    name: Teardown instance (gpu-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
-    needs: [ setup-instance, cuda-tests ]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Stop instance
-        id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
-        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 (gpu-long-run-tests) 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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -63,7 +63,7 @@ jobs:
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -104,13 +104,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-pcc)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, cuda-pcc ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_signed_integer_classic_tests.yml

@@ -35,7 +35,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -51,7 +51,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/gpu_signed_integer_classic_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -68,7 +68,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -94,25 +94,58 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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
@@ -132,13 +165,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-signed-classic-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_signed_integer_h100_tests.yml

@@ -35,7 +35,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -51,7 +51,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/gpu_signed_integer_h100_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -68,7 +68,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -94,25 +94,58 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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 multi-bit tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_signed_integer_multi_bit_gpu_ci
@@ -132,13 +165,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-h100-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_signed_integer_tests.yml

@@ -38,11 +38,10 @@ jobs:
         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@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -58,7 +57,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/gpu_signed_integer_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -75,7 +74,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -101,28 +100,57 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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";
+          } >> "${GITHUB_ENV}"
+
       - name: Should run nightly tests
         if: github.event_name == 'schedule'
         run: |
@@ -131,6 +159,10 @@ jobs:
             echo "NIGHTLY_TESTS=TRUE";
           } >> "${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
@@ -150,13 +182,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, cuda-signed-integer-tests ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_unsigned_integer_classic_tests.yml

@@ -35,7 +35,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -51,7 +51,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/gpu_unsigned_integer_classic_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -68,7 +68,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -94,25 +94,58 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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
@@ -132,13 +165,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-unsigned-classic-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_unsigned_integer_h100_tests.yml

@@ -35,7 +35,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -51,7 +51,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/gpu_unsigned_integer_h100_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -68,7 +68,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -94,25 +94,58 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11 
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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 multi-bit tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_unsigned_integer_multi_bit_gpu_ci
@@ -132,13 +165,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-h100-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    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@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/gpu_unsigned_integer_tests.yml

@@ -41,7 +41,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@d6e91a2266cdb9d62096cebf1e8546899c6aa18f
+        uses: tj-actions/changed-files@4edd678ac3f81e2dc578756871e4d00c19191daf
         with:
           since_last_remote_commit: true
           files_yaml: |
@@ -57,7 +57,7 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
+              - 'tfhe/docs/**.md'
               - '.github/workflows/gpu_unsigned_integer_tests.yml'
               - scripts/integer-tests.sh
               - ci/slab.toml
@@ -74,7 +74,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -100,25 +100,54 @@ jobs:
           - os: ubuntu-22.04
             cuda: "12.2"
             gcc: 11
+    env:
+      CUDA_PATH: /usr/local/cuda-${{ 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: Setup Hyperstack dependencies
-        uses: ./.github/actions/hyperstack_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        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";
+          } >> "${GITHUB_ENV}"
+
       - name: Should run nightly tests
         if: github.event_name == 'schedule'
         run: |
@@ -127,6 +156,10 @@ jobs:
             echo "NIGHTLY_TESTS=TRUE";
           } >> "${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
@@ -146,13 +179,13 @@ jobs:
 
   teardown-instance:
     name: Teardown instance (cuda-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, cuda-unsigned-integer-tests ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/integer_long_run_tests.yml

@@ -1,95 +0,0 @@
-name: AWS Long Run Tests on CPU
-
-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 }}
-
-on:
-  # Allows you to run this workflow manually from the Actions tab as an alternative.
-  workflow_dispatch:
-  schedule:
-    # Weekly tests will be triggered each Friday at 9p.m.
-    - cron: "0 21 * * 5"
-
-jobs:
-  setup-instance:
-    name: Setup instance (cpu-tests)
-    if: github.event_name != 'schedule' ||
-      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
-    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@79939325c3c429837c10d6041e4fd8589d328bac
-        with:
-          mode: start
-          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
-          slab-url: ${{ secrets.SLAB_BASE_URL }}
-          job-secret: ${{ secrets.JOB_SECRET }}
-          backend: aws
-          profile: cpu-big
-
-  cpu-tests:
-    name: Long run CPU tests
-    needs: [ setup-instance ]
-    concurrency:
-      group: ${{ github.workflow }}_${{github.event_name}}_${{ github.ref }}
-      cancel-in-progress: true
-    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
-    timeout-minutes: 4320 # 72 hours
-    steps:
-      - name: Checkout tfhe-rs
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-        with:
-          persist-credentials: 'false'
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
-
-      - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
-        with:
-          toolchain: stable
-
-      - name: Run tests
-        run: |
-          make test_integer_long_run
-
-      - name: Slack Notification
-        if: ${{ failure() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "CPU long run tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
-
-  teardown-instance:
-    name: Teardown instance (cpu-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
-    needs: [ setup-instance, cpu-tests ]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Stop instance
-        id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
-        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 (cpu-long-run-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/m1_tests.yml

@@ -39,7 +39,7 @@ jobs:
           persist-credentials: "false"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -64,29 +64,13 @@ jobs:
           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 NTT checks
-        run: |
-          make pcc_ntt
-
-      - name: Build NTT release
-        run: |
-          make build_ntt
-
-      - name: Build NTT release no-std
-        run: |
-          make build_ntt_no_std
-
-      - name: Run NTT tests
-        run: |
-          make test_ntt_all
-
       - name: Run pcc checks
         run: |
           make pcc
 
-      - name: Build tfhe-csprng
+      - name: Build concrete-csprng
         run: |
-          make build_tfhe_csprng
+          make build_concrete_csprng
 
       - name: Build Release core
         run: |
@@ -112,9 +96,9 @@ jobs:
         run: |
           make build_c_api
 
-      - name: Run tfhe-csprng tests
+      - name: Run concrete-csprng tests
         run: |
-          make test_tfhe_csprng
+          make test_concrete_csprng
 
       - name: Run tfhe-zk-pok tests
         run: |

.github/workflows/make_release.yml

@@ -43,14 +43,13 @@ jobs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
       - name: Prepare package
         run: |
           cargo package -p tfhe
-      - uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
+      - uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882 # v4.4.3
         with:
           name: crate
           path: target/package/*.crate
@@ -77,12 +76,14 @@ jobs:
     name: Publish Release
     needs: [package] # for comparing hashes
     runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
       - name: Create NPM version tag
         if: ${{ inputs.npm_latest_tag }}
         run: |
@@ -107,7 +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@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: failure
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
@@ -152,7 +153,7 @@ jobs:
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}

.github/workflows/make_release_concrete_csprng.yml

@@ -0,0 +1,48 @@
+name: Publish concrete-csprng 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 concrete-csprng Release
+    needs: verify_tag
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          fetch-depth: 0
+
+      - name: Publish crate.io package
+        env:
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
+          DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
+        run: |
+          cargo publish -p concrete-csprng --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
+
+      - 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: "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

@@ -1,3 +1,4 @@
+# Publish new release of tfhe-cuda-backend on crates.io.
 name: Publish CUDA release
 
 on:
@@ -7,6 +8,10 @@ on:
         description: "Dry-run"
         type: boolean
         default: true
+      push_to_crates:
+        description: "Push to crate"
+        type: boolean
+        default: true
 
 env:
   ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
@@ -31,7 +36,7 @@ jobs:
     steps:
       - name: Start instance
         id: start-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: start
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
@@ -40,12 +45,10 @@ jobs:
           backend: aws
           profile: gpu-build
 
-  package:
-    name: Package CUDA Release for provenance
+  publish-cuda-release:
+    name: Publish CUDA Release
     needs: setup-instance
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
-    outputs:
-      hash: ${{ steps.hash.outputs.hash }}
     strategy:
       fail-fast: false
       # explicit include-based build matrix, of known valid options
@@ -58,17 +61,16 @@ jobs:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
     steps:
       - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Set up home
         run: |
           echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}"
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
+        uses: dtolnay/rust-toolchain@7b1c307e0dcbda6122208f10795a713336a9b35a
         with:
           toolchain: stable
 
@@ -92,81 +94,32 @@ jobs:
             echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}";
             echo "HOME=/home/ubuntu";
           } >> "${GITHUB_ENV}"
-      - name: Prepare package
-        run: |
-          cargo package -p tfhe-cuda-backend
-      - name: generate hash
-        id: hash
-        run: cd target/package && echo "hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
 
-  provenance:
-    if: ${{ !inputs.dry_run  }}
-    needs: [package]
-    uses: slsa-framework/slsa-github-generator/.github/workflows/generator_generic_slsa3.yml@v2.0.0
-    permissions:
-      # Needed to detect the GitHub Actions environment
-      actions: read
-      # Needed to create the provenance via GitHub OIDC
-      id-token: write
-      # Needed to upload assets/artifacts
-      contents: write
-    with:
-      # SHA-256 hashes of the Crate package.
-      base64-subjects: ${{ needs.package.outputs.hash }}
-
-  publish-cuda-release:
-    name: Publish CUDA Release
-    needs: [setup-instance, package] # for comparing hashes
-    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: 9
-    steps:
       - name: Publish crate.io package
+        if: ${{ inputs.push_to_crates }}
         env:
           CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           cargo publish -p tfhe-cuda-backend --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
 
-      - name: Generate hash
-        id: published_hash
-        run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-      - 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@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        env:
-          SLACK_COLOR: failure
-          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "SLSA tfhe-cuda-backend crate - hash comparison failure: (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
-
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
+        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 }})"
 
   teardown-instance:
     name: Teardown instance (publish-release)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
+    if: ${{ always() && needs.setup-instance.result != 'skipped' }}
     needs: [ setup-instance, publish-cuda-release ]
     runs-on: ubuntu-latest
     steps:
       - name: Stop instance
         id: stop-instance
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
+        uses: zama-ai/slab-github-runner@801df0b8db5ea2b06128b7476c652f5ed5f193a8
         with:
           mode: stop
           github-token: ${{ secrets.SLAB_ACTION_TOKEN }}

.github/workflows/make_release_tfhe_csprng.yml

@@ -1,103 +0,0 @@
-name: Publish tfhe-csprng 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 }}
-
-  package:
-    runs-on: ubuntu-latest
-    outputs:
-      hash: ${{ steps.hash.outputs.hash }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-      - name: Prepare package
-        run: |
-          cargo package -p tfhe-csprng
-      - uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
-        with:
-          name: crate-tfhe-csprng
-          path: target/package/*.crate
-      - name: generate hash
-        id: hash
-        run: cd target/package && echo "hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-
-  provenance:
-    if: ${{ !inputs.dry_run  }}
-    needs: [package]
-    uses: slsa-framework/slsa-github-generator/.github/workflows/generator_generic_slsa3.yml@v2.0.0
-    permissions:
-      # Needed to detect the GitHub Actions environment
-      actions: read
-      # Needed to create the provenance via GitHub OIDC
-      id-token: write
-      # Needed to upload assets/artifacts
-      contents: write
-    with:
-      # SHA-256 hashes of the Crate package.
-      base64-subjects: ${{ needs.package.outputs.hash }}
-
-
-  publish_release:
-    name: Publish tfhe-csprng Release
-    needs: [verify_tag, package]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
-      - name: Download artifact
-        uses: actions/download-artifact@fa0a91b85d4f404e444e00e005971372dc801d16 # v4.1.8
-        with:
-          name: crate-tfhe-csprng
-          path: target/package
-      - name: Publish crate.io package
-        env:
-          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
-          DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
-        run: |
-          cargo publish -p tfhe-csprng --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
-      - name: Generate hash
-        id: published_hash
-        run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-      - 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@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        env:
-          SLACK_COLOR: failure
-          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "SLSA tfhe-csprng - hash comparison failure: (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
-      - name: Slack Notification
-        if: ${{ failure() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        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-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_tfhe_fft.yml

@@ -19,81 +19,26 @@ jobs:
       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@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
-      - name: Prepare package
-        run: |
-          cargo package -p tfhe-fft
-      - uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
-        with:
-          name: crate
-          path: target/package/*.crate
-      - name: generate hash
-        id: hash
-        run: cd target/package && echo "hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-  provenance:
-    if: ${{ !inputs.dry_run  }}
-    needs: [package]
-    uses: slsa-framework/slsa-github-generator/.github/workflows/generator_generic_slsa3.yml@v2.0.0
-    permissions:
-      # Needed to detect the GitHub Actions environment
-      actions: read
-      # Needed to create the provenance via GitHub OIDC
-      id-token: write
-      # Needed to upload assets/artifacts
-      contents: write
-    with:
-      # SHA-256 hashes of the Crate package.
-      base64-subjects: ${{ needs.package.outputs.hash }}
-
   publish_release:
     name: Publish tfhe-fft Release
     runs-on: ubuntu-latest
-    needs: [verify_tag, package] # for comparing hashes
+    needs: verify_tag
     steps:
       - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
 
       - name: Publish crate.io package
         env:
-          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           cargo publish -p tfhe-fft --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
 
-      - name: Generate hash
-        id: published_hash
-        run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-      - 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@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        env:
-          SLACK_COLOR: failure
-          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "SLSA tfhe-fft crate - hash comparison failure: (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
-
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
+        uses: rtCamp/action-slack-notify@b24d75fe0e728a4bf9fc42ee217caa686d141ee8
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}

.github/workflows/make_release_tfhe_ntt.yml

@@ -1,102 +0,0 @@
-# Publish new release of tfhe-ntt
-name: Publish tfhe-ntt 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 }}
-
-  package:
-    runs-on: ubuntu-latest
-    needs: verify_tag
-    outputs:
-      hash: ${{ steps.hash.outputs.hash }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
-      - name: Prepare package
-        run: |
-          cargo package -p tfhe-ntt
-      - uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
-        with:
-          name: crate
-          path: target/package/*.crate
-      - name: generate hash
-        id: hash
-        run: cd target/package && echo "hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-  provenance:
-    if: ${{ !inputs.dry_run  }}
-    needs: [package]
-    uses: slsa-framework/slsa-github-generator/.github/workflows/generator_generic_slsa3.yml@v2.0.0
-    permissions:
-      # Needed to detect the GitHub Actions environment
-      actions: read
-      # Needed to create the provenance via GitHub OIDC
-      id-token: write
-      # Needed to upload assets/artifacts
-      contents: write
-    with:
-      # SHA-256 hashes of the Crate package.
-      base64-subjects: ${{ needs.package.outputs.hash }}
-
-  publish_release:
-    name: Publish tfhe-ntt Release
-    runs-on: ubuntu-latest
-    needs: [verify_tag, package] # for comparing hashes
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-
-      - name: Publish crate.io package
-        env:
-          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
-          DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
-        run: |
-          cargo publish -p tfhe-ntt --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
-
-      - name: Generate hash
-        id: published_hash
-        run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-      - 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@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        env:
-          SLACK_COLOR: failure
-          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "SLSA tfhe-ntt crate - hash comparison failure: (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
-
-      - name: Slack Notification
-        if: ${{ failure() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "tfhe-ntt release failed: (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}

.github/workflows/make_release_tfhe_versionable.yml

@@ -18,159 +18,34 @@ jobs:
       RELEASE_TEAM: ${{ secrets.RELEASE_TEAM }}
       READ_ORG_TOKEN: ${{ secrets.READ_ORG_TOKEN }}
 
-  package-derive:
-    runs-on: ubuntu-latest
-    outputs:
-      hash: ${{ steps.hash.outputs.hash }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-      - name: Prepare package
-        run: |
-          cargo package -p tfhe-versionable-derive
-      - uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
-        with:
-          name: crate-tfhe-versionable-derive
-          path: target/package/*.crate
-      - name: generate hash
-        id: hash
-        run: cd target/package && echo "hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-  provenance-derive:
-    needs: [package-derive]
-    uses: slsa-framework/slsa-github-generator/.github/workflows/generator_generic_slsa3.yml@v2.0.0
-    permissions:
-      # Needed to detect the GitHub Actions environment
-      actions: read
-      # Needed to create the provenance via GitHub OIDC
-      id-token: write
-      # Needed to upload assets/artifacts
-      contents: write
-    with:
-      # SHA-256 hashes of the Crate package.
-      base64-subjects: ${{ needs.package-derive.outputs.hash }}
-
-  publish_release-derive:
-    name: Publish tfhe-versionable Release
-    needs: [verify_tag, package-derive] # for comparing hashes
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-      - name: Download artifact
-        uses: actions/download-artifact@fa0a91b85d4f404e444e00e005971372dc801d16 # v4.1.8
-        with:
-          name: crate-tfhe-versionable-derive
-          path: target/package
-      - name: Publish crate.io package
-        env:
-          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
-        run: |
-          cargo publish -p tfhe-versionable-derive --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
-      - name: Generate hash
-        id: published_hash
-        run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-      - name: Slack notification (hashes comparison)
-        if: ${{ needs.package-derive.outputs.hash != steps.published_hash.outputs.pub_hash }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        env:
-          SLACK_COLOR: failure
-          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "SLSA tfhe-versionable-derive - hash comparison failure: (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
-      - name: Slack Notification
-        if: ${{ failure() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        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-derive release finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
-
-  package:
-    runs-on: ubuntu-latest
-    outputs:
-      hash: ${{ steps.hash.outputs.hash }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-        with:
-          fetch-depth: 0
-      - name: Prepare package
-        run: |
-          cargo package -p tfhe-versionable
-      - uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
-        with:
-          name: crate-tfhe-versionable
-          path: target/package/*.crate
-      - name: generate hash
-        id: hash
-        run: cd target/package && echo "hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-  provenance:
-    needs: [package]
-    uses: slsa-framework/slsa-github-generator/.github/workflows/generator_generic_slsa3.yml@v2.0.0
-    permissions:
-      # Needed to detect the GitHub Actions environment
-      actions: read
-      # Needed to create the provenance via GitHub OIDC
-      id-token: write
-      # Needed to upload assets/artifacts
-      contents: write
-    with:
-      # SHA-256 hashes of the Crate package.
-      base64-subjects: ${{ needs.package.outputs.hash }}
-
   publish_release:
     name: Publish tfhe-versionable Release
-    needs: [package] # for comparing hashes
+    needs: verify_tag
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-      - name: Download artifact
-        uses: actions/download-artifact@fa0a91b85d4f404e444e00e005971372dc801d16 # v4.1.8
-        with:
-          name: crate-tfhe-versionable
-          path: target/package
-      - name: Publish crate.io package
+
+      - 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 }} ${{ env.DRY_RUN }}
-
-      - name: Generate hash
-        id: published_hash
-        run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-
-      - 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@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        env:
-          SLACK_COLOR: failure
-          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "SLSA tfhe-versionable - hash comparison failure: (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+          cargo publish -p tfhe-versionable --token ${{ env.CRATES_TOKEN }}
 
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}

.github/workflows/make_release_zk_pok.yml

@@ -1,3 +1,4 @@
+# Publish new release of tfhe-zk-pok on crates.io.
 name: Publish tfhe-zk-pok release
 
 on:
@@ -12,40 +13,6 @@ env:
   ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
 
 jobs:
-  package:
-      runs-on: ubuntu-latest
-      outputs:
-        hash: ${{ steps.hash.outputs.hash }}
-      steps:
-        - name: Checkout
-          uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-          with:
-            fetch-depth: 0
-        - name: Prepare package
-          run: |
-            cargo package -p tfhe-zk-pok
-        - uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
-          with:
-            name: crate-zk-pok
-            path: target/package/*.crate
-        - name: generate hash
-          id: hash
-          run: cd target/package && echo "hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-  provenance:
-    if: ${{ !inputs.dry_run  }}
-    needs: [package]
-    uses: slsa-framework/slsa-github-generator/.github/workflows/generator_generic_slsa3.yml@v2.0.0
-    permissions:
-      # Needed to detect the GitHub Actions environment
-      actions: read
-      # Needed to create the provenance via GitHub OIDC
-      id-token: write
-      # Needed to upload assets/artifacts
-      contents: write
-    with:
-      # SHA-256 hashes of the Crate package.
-      base64-subjects: ${{ needs.package.outputs.hash }}
-
   verify_tag:
     uses: ./.github/workflows/verify_tagged_commit.yml
     secrets:
@@ -54,43 +21,25 @@ jobs:
 
   publish_release:
     name: Publish tfhe-zk-pok Release
-    needs: [verify_tag, package] # for comparing hashes
+    needs: verify_tag
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
-      - name: Download artifact
-        uses: actions/download-artifact@fa0a91b85d4f404e444e00e005971372dc801d16 # v4.1.8
-        with:
-          name: crate-zk-pok
-          path: target/package
+
       - name: Publish crate.io package
         env:
           CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           cargo publish -p tfhe-zk-pok --token ${{ env.CRATES_TOKEN }} ${{ env.DRY_RUN }}
-      - name: Verify hash
-        id: published_hash
-        run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
-      - 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@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
-        env:
-          SLACK_COLOR: failure
-          SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-          SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png
-          SLACK_MESSAGE: "SLSA tfhe-zk-pok crate - hash comparison failure: (${{ env.ACTION_RUN_URL }})"
-          SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
-          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+
       - name: Slack Notification
         if: ${{ failure() }}
         continue-on-error: true
-        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990 # v2.3.2
+        uses: rtCamp/action-slack-notify@c33737706dea87cd7784c687dadc9adf1be59990
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}

.github/workflows/parameters_check.yml

@@ -14,7 +14,7 @@ on:
 
 jobs:
   params-curves-security-check:
-    runs-on: large_ubuntu_16-22.04
+    runs-on: large_ubuntu_16
     steps:
       - name: Checkout tfhe-rs
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683

.github/workflows/sync_on_push.yml

@@ -16,7 +16,6 @@ jobs:
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          token: ${{ secrets.FHE_ACTIONS_TOKEN }}
       - name: git-sync
         uses: wei/git-sync@55c6b63b4f21607da0e9877ca9b4d11a29fc6d83
         with:

.gitignore

@@ -33,4 +33,8 @@ node_modules/
 package-lock.json
 
 # Dir used for backward compatibility test data
-tests/tfhe-backward-compat-data/
+tfhe/tfhe-backward-compat-data/
+
+# Sampling tool stuff
+/venv/
+**/*.algo_sample_acquistion

.linelint.yml

@@ -1,15 +1,11 @@
 ignore:
   - .git
   - target
-  - tfhe/build
-  - venv
-  - web-test-runner
   - tfhe/benchmarks_parameters
   - tfhe/web_wasm_parallel_tests/node_modules
   - tfhe/web_wasm_parallel_tests/dist
   - keys
   - coverage
-  - utils/tfhe-lints/ui/main.stderr
 
 rules:
   # checks if file ends in a newline character

Cargo.toml

@@ -3,29 +3,31 @@ resolver = "2"
 members = [
     "tfhe",
     "tfhe-fft",
-    "tfhe-ntt",
     "tfhe-zk-pok",
     "tasks",
     "apps/trivium",
-    "tfhe-csprng",
+    "concrete-csprng",
     "backends/tfhe-cuda-backend",
+    "concrete-cpu-noise-model",
+    "tfhe-rs-cost-model",
+    "concrete-security-curves-rust",
     "utils/tfhe-versionable",
     "utils/tfhe-versionable-derive",
-    "tests",
 ]
 
-exclude = ["tests/backward_compatibility_tests", "utils/tfhe-lints"]
+exclude = [
+    "tfhe/backward_compatibility_tests",
+    "utils/cargo-tfhe-lints-inner",
+    "utils/cargo-tfhe-lints"
+]
 [workspace.dependencies]
-aligned-vec = { version = "0.6", default-features = false }
+aligned-vec = { version = "0.5", default-features = false }
 bytemuck = "1.14.3"
-dyn-stack = { version = "0.11", default-features = false }
-itertools = "0.14"
+dyn-stack = { version = "0.10", default-features = false }
 num-complex = "0.4"
-pulp = { version = "0.20", default-features = false }
-rand = "0.8"
-rayon = "1"
+pulp = { version = "0.18.22", default-features = false }
 serde = { version = "1.0", default-features = false }
-wasm-bindgen = "0.2.100"
+wasm-bindgen = ">=0.2.86,<0.2.94"
 
 [profile.bench]
 lto = "fat"
@@ -43,6 +45,3 @@ inherits = "dev"
 opt-level = 3
 lto = "off"
 debug-assertions = false
-
-[workspace.metadata.dylint]
-libraries = [{ path = "utils/tfhe-lints" }]

README.md

@@ -70,8 +70,22 @@ production-ready library for all the advanced features of TFHE.
 ### Cargo.toml configuration
 To use the latest version of `TFHE-rs` in your project, you first need to add it as a dependency in your `Cargo.toml`:
 
++ For x86_64-based machines running Unix-like OSes:
+
 ```toml
-tfhe = { version = "*", features = ["boolean", "shortint", "integer"] }
+tfhe = { version = "*", features = ["boolean", "shortint", "integer", "x86_64-unix"] }
+```
+
++ For Apple Silicon or aarch64-based machines running Unix-like OSes:
+
+```toml
+tfhe = { version = "*", features = ["boolean", "shortint", "integer", "aarch64-unix"] }
+```
+
++ For x86_64-based machines with the [`rdseed instruction`](https://en.wikipedia.org/wiki/RDRAND) running Windows:
+
+```toml
+tfhe = { version = "*", features = ["boolean", "shortint", "integer", "x86_64"] }
 ```
 
 > [!Note]

apps/trivium/Cargo.toml

@@ -6,8 +6,15 @@ edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-rayon = { workspace = true }
-tfhe = { path = "../../tfhe", features = [ "boolean", "shortint", "integer" ] }
+rayon = { version = "1.7.0"}
+
+[target.'cfg(target_arch = "x86_64")'.dependencies.tfhe]
+path = "../../tfhe"
+features = [ "boolean", "shortint", "integer", "x86_64" ]
+
+[target.'cfg(target_arch = "aarch64")'.dependencies.tfhe]
+path = "../../tfhe"
+features = [ "boolean", "shortint", "integer", "aarch64-unix" ]
 
 [dev-dependencies]
 criterion = { version = "0.5.1", features = [ "html_reports" ]}

apps/trivium/README.md

@@ -95,7 +95,7 @@ fn main() {
 			val >>= 1;
 		}
 	}
-
+	
 	let output_0_63    = "F4CD954A717F26A7D6930830C4E7CF0819F80E03F25F342C64ADC66ABA7F8A8E6EAA49F23632AE3CD41A7BD290A0132F81C6D4043B6E397D7388F3A03B5FE358".to_string();
 
 	let cipher_key = key.map(|x| FheBool::encrypt(x, &client_key));
@@ -129,36 +129,24 @@ Other sizes than 64 bit are expected to be available in the future.
 
 # FHE shortint Trivium implementation
 
-The same implementation is also available for generic Ciphertexts representing bits (meant to be used with parameters `V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64`).
-It uses a lower level API of tfhe-rs, so the syntax is a little bit different. It also implements the `TransCiphering` trait. For optimization purposes, it does not internally run
-on the same cryptographic parameters as the high level API of tfhe-rs. As such, it requires the usage of a casting key, to switch from one parameter space to another, which makes
+The same implementation is also available for generic Ciphertexts representing bits (meant to be used with parameters `PARAM_MESSAGE_1_CARRY_1_KS_PBS`). It uses a lower level API 
+of tfhe-rs, so the syntax is a little bit different. It also implements the `TransCiphering` trait. For optimization purposes, it does not internally run on the same 
+cryptographic parameters as the high level API of tfhe-rs. As such, it requires the usage of a casting key, to switch from one parameter space to another, which makes 
 its setup a little more intricate.
 
 Example code:
 ```rust
 use tfhe::shortint::prelude::*;
-use tfhe::shortint::parameters::{
-    V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64,
-    V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64,
-};
+use tfhe::shortint::CastingKey;
 use tfhe::{ConfigBuilder, generate_keys, FheUint64};
 use tfhe::prelude::*;
 use tfhe_trivium::TriviumStreamShortint;
 
 fn test_shortint() {
-	let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+	let config = ConfigBuilder::default().build();
 	let (hl_client_key, hl_server_key) = generate_keys(config);
-    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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
-    let ksk = KeySwitchingKey::new(
-        (&client_key, Some(&server_key)),
-        (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
-    );
+	let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS);
+	let ksk = CastingKey::new((&client_key, &server_key), (&hl_client_key, &hl_server_key));
 
 	let key_string = "0053A6F94C9FF24598EB".to_string();
 	let mut key = [0; 80];

apps/trivium/benches/kreyvium_shortint.rs

@@ -1,28 +1,23 @@
 use criterion::Criterion;
 use tfhe::prelude::*;
-use tfhe::shortint::parameters::{
-    V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64,
-    V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64,
-};
+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};
 
 pub fn kreyvium_shortint_warmup(c: &mut Criterion) {
-    let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+    let config = ConfigBuilder::default().build();
     let (hl_client_key, hl_server_key) = generate_keys(config);
     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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
+        PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
     );
 
     let key_string = "0053A6F94C9FF24598EB000000000000".to_string();
@@ -62,20 +57,18 @@ pub fn kreyvium_shortint_warmup(c: &mut Criterion) {
 }
 
 pub fn kreyvium_shortint_gen(c: &mut Criterion) {
-    let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+    let config = ConfigBuilder::default().build();
     let (hl_client_key, hl_server_key) = generate_keys(config);
     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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
+        PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
     );
 
     let key_string = "0053A6F94C9FF24598EB000000000000".to_string();
@@ -110,20 +103,18 @@ pub fn kreyvium_shortint_gen(c: &mut Criterion) {
 }
 
 pub fn kreyvium_shortint_trans(c: &mut Criterion) {
-    let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+    let config = ConfigBuilder::default().build();
     let (hl_client_key, hl_server_key) = generate_keys(config);
     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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
+        PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
     );
 
     let key_string = "0053A6F94C9FF24598EB000000000000".to_string();

apps/trivium/benches/trivium_shortint.rs

@@ -1,28 +1,23 @@
 use criterion::Criterion;
 use tfhe::prelude::*;
-use tfhe::shortint::parameters::{
-    V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64,
-    V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64,
-};
+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};
 
 pub fn trivium_shortint_warmup(c: &mut Criterion) {
-    let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+    let config = ConfigBuilder::default().build();
     let (hl_client_key, hl_server_key) = generate_keys(config);
     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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
+        PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();
@@ -62,20 +57,18 @@ pub fn trivium_shortint_warmup(c: &mut Criterion) {
 }
 
 pub fn trivium_shortint_gen(c: &mut Criterion) {
-    let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+    let config = ConfigBuilder::default().build();
     let (hl_client_key, hl_server_key) = generate_keys(config);
     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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
+        PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();
@@ -110,20 +103,18 @@ pub fn trivium_shortint_gen(c: &mut Criterion) {
 }
 
 pub fn trivium_shortint_trans(c: &mut Criterion) {
-    let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+    let config = ConfigBuilder::default().build();
     let (hl_client_key, hl_server_key) = generate_keys(config);
     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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
+        PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();

apps/trivium/src/kreyvium/test.rs

@@ -1,9 +1,6 @@
 use crate::{KreyviumStream, KreyviumStreamByte, KreyviumStreamShortint, TransCiphering};
 use tfhe::prelude::*;
-use tfhe::shortint::parameters::{
-    V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64,
-    V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64,
-};
+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
@@ -219,20 +216,18 @@ use tfhe::shortint::prelude::*;
 
 #[test]
 fn kreyvium_test_shortint_long() {
-    let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+    let config = ConfigBuilder::default().build();
     let (hl_client_key, hl_server_key) = generate_keys(config);
     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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
+        PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
     );
 
     let key_string = "0053A6F94C9FF24598EB000000000000".to_string();

apps/trivium/src/trivium/test.rs

@@ -1,9 +1,6 @@
 use crate::{TransCiphering, TriviumStream, TriviumStreamByte, TriviumStreamShortint};
 use tfhe::prelude::*;
-use tfhe::shortint::parameters::{
-    V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64,
-    V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64,
-};
+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
@@ -355,20 +352,18 @@ use tfhe::shortint::prelude::*;
 
 #[test]
 fn trivium_test_shortint_long() {
-    let config = ConfigBuilder::default()
-        .use_custom_parameters(V0_11_PARAM_MESSAGE_2_CARRY_2_PBS_KS_GAUSSIAN_2M64)
-        .build();
+    let config = ConfigBuilder::default().build();
     let (hl_client_key, hl_server_key) = generate_keys(config);
     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(V0_11_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
+        gen_keys(PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M64);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V0_11_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
+        PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();

backends/tfhe-cuda-backend/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tfhe-cuda-backend"
-version = "0.7.0"
+version = "0.6.0"
 edition = "2021"
 authors = ["Zama team"]
 license = "BSD-3-Clause-Clear"
@@ -14,4 +14,4 @@ keywords = ["fully", "homomorphic", "encryption", "fhe", "cryptography"]
 [build-dependencies]
 cmake = { version = "0.1" }
 pkg-config = { version = "0.3" }
-bindgen = "0.71"
+bindgen = "0.70.1"

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

@@ -27,23 +27,12 @@ inline void cuda_error(cudaError_t code, const char *file, int line) {
     std::abort();                                                              \
   }
 
-cudaEvent_t cuda_create_event(uint32_t gpu_index);
-
-void cuda_event_record(cudaEvent_t event, cudaStream_t stream,
-                       uint32_t gpu_index);
-void cuda_stream_wait_event(cudaStream_t stream, cudaEvent_t event,
-                            uint32_t gpu_index);
-
-void cuda_event_destroy(cudaEvent_t event, uint32_t gpu_index);
-
 cudaStream_t cuda_create_stream(uint32_t gpu_index);
 
 void cuda_destroy_stream(cudaStream_t stream, uint32_t gpu_index);
 
 void cuda_synchronize_stream(cudaStream_t stream, uint32_t gpu_index);
 
-uint32_t cuda_is_available();
-
 void *cuda_malloc(uint64_t size, uint32_t gpu_index);
 
 void *cuda_malloc_async(uint64_t size, cudaStream_t stream, uint32_t gpu_index);

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

@@ -38,7 +38,6 @@ template <typename Torus> struct int_compression {
 
       scratch_packing_keyswitch_lwe_list_to_glwe_64(
           streams[0], gpu_indexes[0], &fp_ks_buffer,
-          compression_params.small_lwe_dimension,
           compression_params.glwe_dimension, compression_params.polynomial_size,
           num_radix_blocks, true);
     }
@@ -65,7 +64,7 @@ template <typename Torus> struct int_decompression {
   Torus *tmp_extracted_lwe;
   uint32_t *tmp_indexes_array;
 
-  int_radix_lut<Torus> *decompression_rescale_lut;
+  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,
@@ -84,7 +83,7 @@ template <typename Torus> struct int_decompression {
       Torus lwe_accumulator_size = (compression_params.glwe_dimension *
                                         compression_params.polynomial_size +
                                     1);
-      decompression_rescale_lut = new int_radix_lut<Torus>(
+      carry_extract_lut = new int_radix_lut<Torus>(
           streams, gpu_indexes, gpu_count, encryption_params, 1,
           num_radix_blocks, allocate_gpu_memory);
 
@@ -97,28 +96,19 @@ template <typename Torus> struct int_decompression {
           num_radix_blocks * lwe_accumulator_size * sizeof(Torus), streams[0],
           gpu_indexes[0]);
 
-      // Rescale is done using an identity LUT
-      // Here we do not divide by message_modulus
-      // Example: in the 2_2 case we are mapping a 2 bits message onto a 4 bits
-      // space, we want to keep the original 2 bits value in the 4 bits space,
-      // so we apply the identity and the encoding will rescale it for us.
-      auto decompression_rescale_f = [encryption_params](Torus x) -> Torus {
-        return x;
+      // Carry extract LUT
+      auto carry_extract_f = [encryption_params](Torus x) -> Torus {
+        return x / encryption_params.message_modulus;
       };
 
-      auto effective_compression_message_modulus =
-          encryption_params.carry_modulus;
-      auto effective_compression_carry_modulus = 1;
-
-      generate_device_accumulator_with_encoding<Torus>(
-          streams[0], gpu_indexes[0], decompression_rescale_lut->get_lut(0, 0),
+      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,
-          effective_compression_message_modulus,
-          effective_compression_carry_modulus,
           encryption_params.message_modulus, encryption_params.carry_modulus,
-          decompression_rescale_f);
+          carry_extract_f);
 
-      decompression_rescale_lut->broadcast_lut(streams, gpu_indexes, 0);
+      carry_extract_lut->broadcast_lut(streams, gpu_indexes, gpu_indexes[0]);
     }
   }
   void release(cudaStream_t const *streams, uint32_t const *gpu_indexes,
@@ -127,8 +117,8 @@ template <typename Torus> struct int_decompression {
     cuda_drop_async(tmp_extracted_lwe, streams[0], gpu_indexes[0]);
     cuda_drop_async(tmp_indexes_array, streams[0], gpu_indexes[0]);
 
-    decompression_rescale_lut->release(streams, gpu_indexes, gpu_count);
-    delete decompression_rescale_lut;
+    carry_extract_lut->release(streams, gpu_indexes, gpu_count);
+    delete carry_extract_lut;
   }
 };
 #endif

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

@@ -35,8 +35,6 @@ enum CMP_ORDERING { IS_INFERIOR = 0, IS_EQUAL = 1, IS_SUPERIOR = 2 };
 
 enum SIGNED_OPERATION { ADDITION = 1, SUBTRACTION = -1 };
 
-enum outputFlag { FLAG_NONE = 0, FLAG_OVERFLOW = 1, FLAG_CARRY = 2 };
-
 extern "C" {
 void scratch_cuda_apply_univariate_lut_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -46,14 +44,7 @@ void scratch_cuda_apply_univariate_lut_kb_64(
     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 scratch_cuda_apply_many_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 num_radix_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    uint32_t num_many_lut, 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,
@@ -112,19 +103,17 @@ void cleanup_cuda_full_propagation(void *const *streams,
 
 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, bool const is_boolean_left, bool const is_boolean_right,
-    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);
+    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, bool const is_bool_left,
-    void const *radix_lwe_right, bool const is_bool_right, void *const *bsks,
-    void *const *ksks, int8_t *mem_ptr, uint32_t polynomial_size,
-    uint32_t num_blocks);
+    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,
@@ -291,61 +280,23 @@ void scratch_cuda_propagate_single_carry_kb_64_inplace(
     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, uint32_t requested_flag,
-    uint32_t uses_carry, bool allocate_gpu_memory);
-
-void scratch_cuda_add_and_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, uint32_t requested_flag,
-    uint32_t uses_carry, bool allocate_gpu_memory);
+    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, const void *carry_in, int8_t *mem_ptr,
-    void *const *bsks, void *const *ksks, uint32_t num_blocks,
-    uint32_t requested_flag, uint32_t uses_carry);
+    void *lwe_array, void *carry_out, int8_t *mem_ptr, void *const *bsks,
+    void *const *ksks, uint32_t num_blocks);
 
-void cuda_add_and_propagate_single_carry_kb_64_inplace(
+void cuda_propagate_single_carry_get_input_carries_kb_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    void *lhs_array, const void *rhs_array, void *carry_out,
-    const void *carry_in, int8_t *mem_ptr, void *const *bsks, void *const *ksks,
-    uint32_t num_blocks, uint32_t requested_flag, uint32_t uses_carry);
+    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 cleanup_cuda_add_and_propagate_single_carry(void *const *streams,
-                                                 uint32_t const *gpu_indexes,
-                                                 uint32_t gpu_count,
-                                                 int8_t **mem_ptr_void);
-
-void scratch_cuda_integer_overflowing_sub_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, uint32_t compute_overflow,
-    bool allocate_gpu_memory);
-
-void cuda_integer_overflowing_sub_kb_64_inplace(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    void *lhs_array, const void *rhs_array, void *overflow_block,
-    const void *input_borrow, int8_t *mem_ptr, void *const *bsks,
-    void *const *ksks, uint32_t num_blocks, uint32_t compute_overflow,
-    uint32_t uses_input_borrow);
-
-void cleanup_cuda_integer_overflowing_sub(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,
@@ -365,6 +316,25 @@ 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,
@@ -387,23 +357,42 @@ void cleanup_cuda_integer_radix_scalar_mul(void *const *streams,
 
 void scratch_cuda_integer_div_rem_radix_ciphertext_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    bool is_signed, 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);
+    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,
-    bool is_signed, int8_t *mem_ptr, void *const *bsks, void *const *ksks,
+    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,
@@ -447,41 +436,5 @@ void cleanup_cuda_integer_abs_inplace(void *const *streams,
                                       uint32_t gpu_count,
                                       int8_t **mem_ptr_void);
 
-void scratch_cuda_integer_are_all_comparisons_block_true_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_radix_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory);
-
-void cuda_integer_are_all_comparisons_block_true_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    void *lwe_array_out, void const *lwe_array_in, int8_t *mem_ptr,
-    void *const *bsks, void *const *ksks, uint32_t num_radix_blocks);
-
-void cleanup_cuda_integer_are_all_comparisons_block_true(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void);
-
-void scratch_cuda_integer_is_at_least_one_comparisons_block_true_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_radix_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory);
-
-void cuda_integer_is_at_least_one_comparisons_block_true_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    void *lwe_array_out, void const *lwe_array_in, int8_t *mem_ptr,
-    void *const *bsks, void *const *ksks, uint32_t num_radix_blocks);
-
-void cleanup_cuda_integer_is_at_least_one_comparisons_block_true(
-    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

@@ -21,8 +21,8 @@ void cuda_keyswitch_lwe_ciphertext_vector_64(
 
 void scratch_packing_keyswitch_lwe_list_to_glwe_64(
     void *stream, uint32_t gpu_index, int8_t **fp_ks_buffer,
-    uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t num_lwes, bool allocate_gpu_memory);
+    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,

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

@@ -5,50 +5,44 @@
 
 extern "C" {
 
-void cuda_negate_lwe_ciphertext_vector_32(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void const *lwe_array_in, const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
-void cuda_negate_lwe_ciphertext_vector_64(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void const *lwe_array_in, const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
-void cuda_add_lwe_ciphertext_vector_32(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void const *lwe_array_in_1, void const *lwe_array_in_2,
-    const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
-void cuda_add_lwe_ciphertext_vector_64(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void const *lwe_array_in_1, void const *lwe_array_in_2,
-    const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
-
+void cuda_negate_lwe_ciphertext_vector_32(void *stream, uint32_t gpu_index,
+                                          void *lwe_array_out,
+                                          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 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 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 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 const *lwe_array_in, void const *plaintext_array_in,
-    const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
+    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 const *lwe_array_in, void const *plaintext_array_in,
-    const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
+    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 const *lwe_array_in, void const *cleartext_array_in,
-    const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
+    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 const *lwe_array_in, void const *cleartext_array_in,
-    const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
-void cuda_add_lwe_ciphertext_vector_plaintext_64(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void const *lwe_array_in, const uint64_t plaintext_in,
-    const uint32_t input_lwe_dimension,
-    const uint32_t input_lwe_ciphertext_count);
+    uint32_t input_lwe_dimension, uint32_t input_lwe_ciphertext_count);
 }
 
 #endif // CUDA_LINALG_H_

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

@@ -28,7 +28,7 @@ void cuda_tbc_multi_bit_programmable_bootstrap_lwe_ciphertext_vector(
     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 num_many_lut, uint32_t lut_stride);
+    uint32_t lut_count, uint32_t lut_stride);
 #endif
 
 template <typename Torus>
@@ -46,7 +46,7 @@ void cuda_cg_multi_bit_programmable_bootstrap_lwe_ciphertext_vector(
     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 num_many_lut, uint32_t lut_stride);
+    uint32_t lut_count, uint32_t lut_stride);
 
 template <typename Torus>
 void scratch_cuda_multi_bit_programmable_bootstrap(
@@ -63,7 +63,7 @@ void cuda_multi_bit_programmable_bootstrap_lwe_ciphertext_vector(
     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 num_many_lut, uint32_t lut_stride);
+    uint32_t lut_count, uint32_t lut_stride);
 
 template <typename Torus>
 uint64_t get_buffer_size_full_sm_multibit_programmable_bootstrap_keybundle(

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

@@ -255,7 +255,7 @@ void cuda_programmable_bootstrap_cg_lwe_ciphertext_vector(
     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 num_many_lut,
+    uint32_t level_count, uint32_t num_samples, uint32_t lut_count,
     uint32_t lut_stride);
 
 template <typename Torus>
@@ -266,7 +266,7 @@ void cuda_programmable_bootstrap_lwe_ciphertext_vector(
     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 num_many_lut,
+    uint32_t level_count, uint32_t num_samples, uint32_t lut_count,
     uint32_t lut_stride);
 
 #if (CUDA_ARCH >= 900)
@@ -278,7 +278,7 @@ void cuda_programmable_bootstrap_tbc_lwe_ciphertext_vector(
     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 num_many_lut,
+    uint32_t level_count, uint32_t num_samples, uint32_t lut_count,
     uint32_t lut_stride);
 
 template <typename Torus>

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

@@ -69,7 +69,7 @@ void cuda_programmable_bootstrap_lwe_ciphertext_vector_32(
     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 num_many_lut, uint32_t lut_stride);
+    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,
@@ -78,7 +78,7 @@ void cuda_programmable_bootstrap_lwe_ciphertext_vector_64(
     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 num_many_lut, uint32_t lut_stride);
+    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);

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

@@ -27,7 +27,7 @@ void cuda_multi_bit_programmable_bootstrap_lwe_ciphertext_vector_64(
     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 num_many_lut,
+    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,

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

@@ -1,356 +0,0 @@
-#ifndef CNCRT_FAST_KS_CUH
-#define CNCRT_FAST_KS_CUH
-
-#undef NDEBUG
-#include <assert.h>
-
-#include "device.h"
-#include "gadget.cuh"
-#include "helper_multi_gpu.h"
-#include "keyswitch.cuh"
-#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>
-
-#define CEIL_DIV(M, N) ((M) + (N)-1) / (N)
-
-const int BLOCK_SIZE_GEMM = 64;
-const int THREADS_GEMM = 8;
-const int BLOCK_SIZE_DECOMP = 8;
-
-template <typename Torus> uint64_t get_shared_mem_size_tgemm() {
-  return BLOCK_SIZE_GEMM * THREADS_GEMM * 2 * sizeof(Torus);
-}
-
-// Initialize decomposition by performing rounding
-// and decomposing one level of an array of Torus LWEs. Only
-// decomposes the mask elements of the incoming LWEs.
-template <typename Torus, typename TorusVec>
-__global__ void decompose_vectorize_init(Torus const *lwe_in, Torus *lwe_out,
-                                         uint32_t lwe_dimension,
-                                         uint32_t num_lwe, uint32_t base_log,
-                                         uint32_t level_count) {
-
-  // index of this LWE ct in the buffer
-  auto lwe_idx = blockIdx.x * blockDim.x + threadIdx.x;
-  // index of the LWE sample in the LWE ct
-  auto lwe_sample_idx = blockIdx.y * blockDim.y + threadIdx.y;
-
-  if (lwe_idx >= num_lwe || lwe_sample_idx >= lwe_dimension)
-    return;
-
-  // Input LWE array is [mask_0, .., mask_lwe_dim, message] and
-  // we only decompose the mask. Thus the stride for reading
-  // is lwe_dimension + 1, while for writing it is lwe_dimension
-  auto read_val_idx = lwe_idx * (lwe_dimension + 1) + lwe_sample_idx;
-  auto write_val_idx = lwe_idx * lwe_dimension + lwe_sample_idx;
-  auto write_state_idx =
-      num_lwe * lwe_dimension + lwe_idx * lwe_dimension + lwe_sample_idx;
-
-  Torus a_i = lwe_in[read_val_idx];
-
-  Torus state = init_decomposer_state(a_i, base_log, level_count);
-
-  Torus mod_b_mask = (1ll << base_log) - 1ll;
-  lwe_out[write_val_idx] = decompose_one<Torus>(state, mod_b_mask, base_log);
-  synchronize_threads_in_block();
-  lwe_out[write_state_idx] = state;
-}
-
-// Continue decomposiion of an array of Torus elements in place. Supposes
-// that the array contains already decomposed elements and
-// computes the new decomposed level in place.
-template <typename Torus, typename TorusVec>
-__global__ void
-decompose_vectorize_step_inplace(Torus *buffer_in, uint32_t lwe_dimension,
-                                 uint32_t num_lwe, uint32_t base_log,
-                                 uint32_t level_count) {
-
-  // index of this LWE ct in the buffer
-  auto lwe_idx = blockIdx.x * blockDim.x + threadIdx.x;
-  // index of the LWE sample in the LWE ct
-  auto lwe_sample_idx = blockIdx.y * blockDim.y + threadIdx.y;
-
-  if (lwe_idx >= num_lwe || lwe_sample_idx >= lwe_dimension)
-    return;
-
-  auto val_idx = lwe_idx * lwe_dimension + lwe_sample_idx;
-  auto state_idx = num_lwe * lwe_dimension + val_idx;
-
-  Torus state = buffer_in[state_idx];
-  synchronize_threads_in_block();
-
-  Torus mod_b_mask = (1ll << base_log) - 1ll;
-
-  buffer_in[val_idx] = decompose_one<Torus>(state, mod_b_mask, base_log);
-  synchronize_threads_in_block();
-  buffer_in[state_idx] = state;
-}
-
-// Multiply matrices A, B of size (M, K), (K, N) respectively
-// with K as the inner dimension.
-//
-// A block of threads processeds blocks of size (BLOCK_SIZE_GEMM,
-// BLOCK_SIZE_GEMM) splitting them in multiple tiles: (BLOCK_SIZE_GEMM,
-// THREADS_GEMM)-shaped tiles of values from A, and a (THREADS_GEMM,
-// BLOCK_SIZE_GEMM)-shaped tiles of values from B.
-//
-// This code is adapted by generalizing the 1d block-tiling
-// kernel from https://github.com/siboehm/SGEMM_CUDA
-// to any matrix dimension
-template <typename Torus, typename TorusVec>
-__global__ void tgemm(int M, int N, int K, const Torus *A, const Torus *B,
-                      int stride_B, Torus *C) {
-
-  const int BM = BLOCK_SIZE_GEMM;
-  const int BN = BLOCK_SIZE_GEMM;
-  const int BK = THREADS_GEMM;
-  const int TM = THREADS_GEMM;
-
-  const uint cRow = blockIdx.y;
-  const uint cCol = blockIdx.x;
-
-  const int threadCol = threadIdx.x % BN;
-  const int threadRow = threadIdx.x / BN;
-
-  // Allocate space for the current block tile in shared memory
-  __shared__ Torus As[BM * BK];
-  __shared__ Torus Bs[BK * BN];
-
-  // Initialize the pointers to the input blocks from A, B
-  // Tiles from these blocks are loaded to shared memory
-  A += cRow * BM * K;
-  B += cCol * BN;
-
-  // Each thread will handle multiple sub-blocks
-  const uint innerColA = threadIdx.x % BK;
-  const uint innerRowA = threadIdx.x / BK;
-  const uint innerColB = threadIdx.x % BN;
-  const uint innerRowB = threadIdx.x / BN;
-
-  // allocate thread-local cache for results in registerfile
-  Torus threadResults[TM] = {0};
-
-  auto row_A = cRow * BM + innerRowA;
-  auto col_B = cCol * BN + innerColB;
-
-  // For each thread, loop over block tiles
-  for (uint bkIdx = 0; bkIdx < K; bkIdx += BK) {
-    auto col_A = bkIdx + innerColA;
-    auto row_B = bkIdx + innerRowB;
-
-    if (row_A < M && col_A < K) {
-      As[innerRowA * BK + innerColA] = A[innerRowA * K + innerColA];
-    } else {
-      As[innerRowA * BK + innerColA] = 0;
-    }
-
-    if (col_B < N && row_B < K) {
-      Bs[innerRowB * BN + innerColB] = B[innerRowB * stride_B + innerColB];
-    } else {
-      Bs[innerRowB * BN + innerColB] = 0;
-    }
-    synchronize_threads_in_block();
-
-    // Advance blocktile for the next iteration of this loop
-    A += BK;
-    B += BK * stride_B;
-
-    // calculate per-thread results
-    for (uint dotIdx = 0; dotIdx < BK; ++dotIdx) {
-      // we make the dotproduct loop the outside loop, which facilitates
-      // reuse of the Bs entry, which we can cache in a tmp var.
-      Torus tmp = Bs[dotIdx * BN + threadCol];
-      for (uint resIdx = 0; resIdx < TM; ++resIdx) {
-        threadResults[resIdx] +=
-            As[(threadRow * TM + resIdx) * BK + dotIdx] * tmp;
-      }
-    }
-    synchronize_threads_in_block();
-  }
-
-  // Initialize the pointer to the output block of size (BLOCK_SIZE_GEMM,
-  // BLOCK_SIZE_GEMM)
-  C += cRow * BM * N + cCol * BN;
-
-  // write out the results
-  for (uint resIdx = 0; resIdx < TM; ++resIdx) {
-    int outRow = cRow * BM + threadRow * TM + resIdx;
-    int outCol = cCol * BN + threadCol;
-
-    if (outRow >= M)
-      continue;
-    if (outCol >= N)
-      continue;
-
-    C[(threadRow * TM + resIdx) * N + threadCol] += threadResults[resIdx];
-  }
-}
-
-// Finish the keyswitching operation and prepare GLWEs for accumulation.
-// 1. Finish the keyswitching computation partially performed with a GEMM:
-//  - negate the dot product between the GLWE and KSK polynomial
-//  - add the GLWE message for the N-th polynomial coeff in the message poly
-// 2. Rotate each of the GLWE . KSK poly dot products to
-//    prepare them for accumulation into a single GLWE
-template <typename Torus>
-__global__ void polynomial_accumulate_monic_monomial_mul_many_neg_and_add_C(
-    Torus *in_glwe_buffer, Torus *out_glwe_buffer, Torus const *lwe_array,
-    uint32_t lwe_dimension, uint32_t num_glwes, uint32_t polynomial_size,
-    uint32_t glwe_dimension) {
-
-  uint32_t glwe_id = blockIdx.x * blockDim.x + threadIdx.x;
-  uint32_t degree = glwe_id; // lwe 0 rotate 0, lwe 1 rotate 1, .. , lwe
-                             // poly_size-1 rotate poly_size-1
-  uint32_t coeffIdx = blockIdx.y * blockDim.y + threadIdx.y;
-
-  if (glwe_id >= num_glwes)
-    return;
-  if (coeffIdx >= polynomial_size)
-    return;
-
-  auto in_poly =
-      in_glwe_buffer + glwe_id * polynomial_size * (glwe_dimension + 1);
-  auto out_result =
-      out_glwe_buffer + glwe_id * polynomial_size * (glwe_dimension + 1);
-  if (coeffIdx == 0) {
-    // Add the message value of the input LWE (`C`) to the N-th coefficient
-    // in the GLWE . KSK dot product
-
-    // The C is added to the first position of the last polynomial in the GLWE
-    // which has (glwe_dimension+1) polynomials
-    // The C value is extracted as the last value of the LWE ct. (of index
-    // glwe_id) the LWEs have (polynomial_size + 1) values
-    in_poly[polynomial_size * glwe_dimension] =
-        lwe_array[glwe_id * (lwe_dimension + 1) + lwe_dimension] -
-        in_poly[polynomial_size * glwe_dimension];
-
-    for (int gi = 1; gi < glwe_dimension; ++gi)
-      in_poly[coeffIdx + gi * polynomial_size] =
-          -in_poly[coeffIdx + gi * polynomial_size];
-
-  } else {
-    // Otherwise simply negate the input coefficient
-    for (int gi = 1; gi < glwe_dimension + 1; ++gi)
-      in_poly[coeffIdx + gi * polynomial_size] =
-          -in_poly[coeffIdx + gi * polynomial_size];
-  }
-  // Negate all the coefficients for rotation for the first poly
-  in_poly[coeffIdx] = -in_poly[coeffIdx];
-
-  // rotate the body
-  polynomial_accumulate_monic_monomial_mul<Torus>(
-      out_result, in_poly, degree, coeffIdx, polynomial_size, 1, true);
-  // rotate the mask too
-  for (int gi = 1; gi < glwe_dimension + 1; ++gi)
-    polynomial_accumulate_monic_monomial_mul<Torus>(
-        out_result + gi * polynomial_size, in_poly + gi * polynomial_size,
-        degree, coeffIdx, polynomial_size, 1, true);
-}
-
-template <typename Torus, typename TorusVec>
-__host__ void host_fast_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, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t base_log, uint32_t level_count, uint32_t num_lwes) {
-
-  // Optimization of packing keyswitch when packing many LWEs
-
-  cudaSetDevice(gpu_index);
-  check_cuda_error(cudaGetLastError());
-
-  int glwe_accumulator_size = (glwe_dimension + 1) * polynomial_size;
-
-  // The fast path of PKS uses the scratch buffer (d_mem) differently than the
-  // old path: it needs to store the decomposed masks in the first half of this
-  // buffer and the keyswitched GLWEs in the second half of the buffer. Thus the
-  // scratch buffer for the fast path must determine the half-size of the
-  // scratch buffer as the max between the size of the GLWE and the size of the
-  // LWE-mask times two (to keep both decomposition state and decomposed
-  // intermediate value)
-  int memory_unit = glwe_accumulator_size > lwe_dimension * 2
-                        ? glwe_accumulator_size
-                        : lwe_dimension * 2;
-
-  // ping pong the buffer between successive calls
-  // split the buffer in two parts of this size
-  auto d_mem_0 = (Torus *)fp_ks_buffer;
-  auto d_mem_1 = d_mem_0 + num_lwes * memory_unit;
-
-  // Set the scratch buffer to 0 as it is used to accumulate
-  // decomposition temporary results
-  cuda_memset_async(d_mem_1, 0, num_lwes * memory_unit * sizeof(Torus), stream,
-                    gpu_index);
-  check_cuda_error(cudaGetLastError());
-
-  // decompose LWEs
-  // don't decompose LWE body - the LWE has lwe_size + 1 elements. The last
-  // element, the body is ignored by rounding down the number of blocks assuming
-  // here that the LWE dimension is a multiple of the block size
-  dim3 grid_decomp(CEIL_DIV(num_lwes, BLOCK_SIZE_DECOMP),
-                   CEIL_DIV(lwe_dimension, BLOCK_SIZE_DECOMP));
-  dim3 threads_decomp(BLOCK_SIZE_DECOMP, BLOCK_SIZE_DECOMP);
-
-  // decompose first level
-  decompose_vectorize_init<Torus, TorusVec>
-      <<<grid_decomp, threads_decomp, 0, stream>>>(lwe_array_in, d_mem_0,
-                                                   lwe_dimension, num_lwes,
-                                                   base_log, level_count);
-  check_cuda_error(cudaGetLastError());
-
-  // gemm to ks the individual LWEs to GLWEs
-  dim3 grid_gemm(CEIL_DIV(glwe_accumulator_size, BLOCK_SIZE_GEMM),
-                 CEIL_DIV(num_lwes, BLOCK_SIZE_GEMM));
-  dim3 threads_gemm(BLOCK_SIZE_GEMM * THREADS_GEMM);
-
-  auto stride_KSK_buffer = glwe_accumulator_size * level_count;
-
-  uint32_t shared_mem_size = get_shared_mem_size_tgemm<Torus>();
-  tgemm<Torus, TorusVec><<<grid_gemm, threads_gemm, shared_mem_size, stream>>>(
-      num_lwes, glwe_accumulator_size, lwe_dimension, d_mem_0, fp_ksk_array,
-      stride_KSK_buffer, d_mem_1);
-  check_cuda_error(cudaGetLastError());
-
-  auto ksk_block_size = glwe_accumulator_size;
-
-  for (int li = 1; li < level_count; ++li) {
-    decompose_vectorize_step_inplace<Torus, TorusVec>
-        <<<grid_decomp, threads_decomp, 0, stream>>>(
-            d_mem_0, lwe_dimension, num_lwes, base_log, level_count);
-    check_cuda_error(cudaGetLastError());
-
-    tgemm<Torus, TorusVec>
-        <<<grid_gemm, threads_gemm, shared_mem_size, stream>>>(
-            num_lwes, glwe_accumulator_size, lwe_dimension, d_mem_0,
-            fp_ksk_array + li * ksk_block_size, stride_KSK_buffer, d_mem_1);
-    check_cuda_error(cudaGetLastError());
-  }
-
-  // should we include the mask in the rotation ??
-  dim3 grid_rotate(CEIL_DIV(num_lwes, BLOCK_SIZE_DECOMP),
-                   CEIL_DIV(polynomial_size, BLOCK_SIZE_DECOMP));
-  dim3 threads_rotate(BLOCK_SIZE_DECOMP, BLOCK_SIZE_DECOMP);
-  // rotate the GLWEs
-  polynomial_accumulate_monic_monomial_mul_many_neg_and_add_C<Torus>
-      <<<grid_rotate, threads_rotate, 0, stream>>>(
-          d_mem_1, d_mem_0, lwe_array_in, lwe_dimension, num_lwes,
-          polynomial_size, glwe_dimension);
-  check_cuda_error(cudaGetLastError());
-
-  dim3 grid_accumulate(
-      CEIL_DIV(polynomial_size * (glwe_dimension + 1), BLOCK_SIZE_DECOMP));
-  dim3 threads_accum(BLOCK_SIZE_DECOMP);
-
-  // accumulate to a single glwe
-  accumulate_glwes<Torus><<<grid_accumulate, threads_accum, 0, stream>>>(
-      glwe_out, d_mem_0, glwe_dimension, polynomial_size, num_lwes);
-
-  check_cuda_error(cudaGetLastError());
-}
-
-#endif

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

@@ -1,8 +1,6 @@
-#include "fast_packing_keyswitch.cuh"
 #include "keyswitch.cuh"
 #include "keyswitch.h"
 #include <cstdint>
-#include <stdio.h>
 
 /* Perform keyswitch on a batch of 32 bits input LWE ciphertexts.
  * Head out to the equivalent operation on 64 bits for more details.
@@ -55,17 +53,15 @@ void cuda_keyswitch_lwe_ciphertext_vector_64(
 
 void scratch_packing_keyswitch_lwe_list_to_glwe_64(
     void *stream, uint32_t gpu_index, int8_t **fp_ks_buffer,
-    uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t num_lwes, bool allocate_gpu_memory) {
+    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, lwe_dimension,
+      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,
@@ -73,7 +69,7 @@ void cuda_packing_keyswitch_lwe_list_to_glwe_64(
     uint32_t output_polynomial_size, uint32_t base_log, uint32_t level_count,
     uint32_t num_lwes) {
 
-  host_fast_packing_keyswitch_lwe_list_to_glwe<uint64_t, ulonglong4>(
+  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),

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

@@ -158,22 +158,16 @@ void execute_keyswitch_async(cudaStream_t const *streams,
 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 lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t num_lwes, bool allocate_gpu_memory) {
+    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;
 
-  // allocate at least LWE-mask times two: to keep both decomposition state and
-  // decomposed intermediate value
-  int memory_unit = glwe_accumulator_size > lwe_dimension * 2
-                        ? glwe_accumulator_size
-                        : lwe_dimension * 2;
-
-  if (allocate_gpu_memory) {
+  if (allocate_gpu_memory)
     *fp_ks_buffer = (int8_t *)cuda_malloc_async(
-        2 * num_lwes * memory_unit * sizeof(Torus), stream, gpu_index);
-  }
+        2 * num_lwes * glwe_accumulator_size * sizeof(Torus), stream,
+        gpu_index);
 }
 
 // public functional packing keyswitch for a single LWE ciphertext
@@ -223,6 +217,43 @@ __device__ void packing_keyswitch_lwe_ciphertext_into_glwe_ciphertext(
   }
 }
 
+// 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,
@@ -240,4 +271,42 @@ __global__ void accumulate_glwes(Torus *glwe_out, Torus *glwe_array_in,
   }
 }
 
+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/device.cu

@@ -2,30 +2,6 @@
 #include <cstdint>
 #include <cuda_runtime.h>
 
-cudaEvent_t cuda_create_event(uint32_t gpu_index) {
-  check_cuda_error(cudaSetDevice(gpu_index));
-  cudaEvent_t event;
-  check_cuda_error(cudaEventCreate(&event));
-  return event;
-}
-
-void cuda_event_record(cudaEvent_t event, cudaStream_t stream,
-                       uint32_t gpu_index) {
-  check_cuda_error(cudaSetDevice(gpu_index));
-  check_cuda_error(cudaEventRecord(event, stream));
-}
-
-void cuda_stream_wait_event(cudaStream_t stream, cudaEvent_t event,
-                            uint32_t gpu_index) {
-  check_cuda_error(cudaSetDevice(gpu_index));
-  check_cuda_error(cudaStreamWaitEvent(stream, event, 0));
-}
-
-void cuda_event_destroy(cudaEvent_t event, uint32_t gpu_index) {
-  check_cuda_error(cudaSetDevice(gpu_index));
-  check_cuda_error(cudaEventDestroy(event));
-}
-
 /// Unsafe function to create a CUDA stream, must check first that GPU exists
 cudaStream_t cuda_create_stream(uint32_t gpu_index) {
   check_cuda_error(cudaSetDevice(gpu_index));
@@ -45,9 +21,6 @@ void cuda_synchronize_stream(cudaStream_t stream, uint32_t gpu_index) {
   check_cuda_error(cudaStreamSynchronize(stream));
 }
 
-// Determine if a CUDA device is available at runtime
-uint32_t cuda_is_available() { return cudaSetDevice(0) == cudaSuccess; }
-
 /// Unsafe function that will try to allocate even if gpu_index is invalid
 /// or if there's not enough memory. A safe wrapper around it must call
 /// cuda_check_valid_malloc() first

backends/tfhe-cuda-backend/cuda/src/integer/abs.cuh

@@ -58,11 +58,9 @@ host_integer_abs_kb(cudaStream_t const *streams, uint32_t const *gpu_indexes,
   host_addition<Torus>(streams[0], gpu_indexes[0], ct, mask, ct,
                        radix_params.big_lwe_dimension, num_blocks);
 
-  uint32_t requested_flag = outputFlag::FLAG_NONE;
-  uint32_t uses_carry = 0;
-  host_propagate_single_carry<Torus>(
-      streams, gpu_indexes, gpu_count, ct, nullptr, nullptr, mem_ptr->scp_mem,
-      bsks, ksks, num_blocks, requested_flag, uses_carry);
+  host_propagate_single_carry<Torus>(streams, gpu_indexes, gpu_count, ct,
+                                     nullptr, nullptr, mem_ptr->scp_mem, bsks,
+                                     ksks, num_blocks);
 
   host_integer_radix_bitop_kb(streams, gpu_indexes, gpu_count, ct, mask, ct,
                               mem_ptr->bitxor_mem, bsks, ksks, num_blocks);

backends/tfhe-cuda-backend/cuda/src/integer/addition.cu

@@ -0,0 +1,50 @@
+#include "integer/addition.cuh"
+
+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) {
+
+  SIGNED_OPERATION op = (signed_operation == 1) ? SIGNED_OPERATION::ADDITION
+                                                : SIGNED_OPERATION::SUBTRACTION;
+  int_radix_params params(pbs_type, glwe_dimension, polynomial_size,
+                          big_lwe_dimension, small_lwe_dimension, ks_level,
+                          ks_base_log, pbs_level, pbs_base_log, grouping_factor,
+                          message_modulus, carry_modulus);
+
+  scratch_cuda_integer_signed_overflowing_add_or_sub_kb<uint64_t>(
+      (cudaStream_t *)(streams), gpu_indexes, gpu_count,
+      (int_signed_overflowing_add_or_sub_memory<uint64_t> **)mem_ptr,
+      num_blocks, op, params, 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) {
+
+  auto mem = (int_signed_overflowing_add_or_sub_memory<uint64_t> *)mem_ptr;
+  SIGNED_OPERATION op = (signed_operation == 1) ? SIGNED_OPERATION::ADDITION
+                                                : SIGNED_OPERATION::SUBTRACTION;
+
+  host_integer_signed_overflowing_add_or_sub_kb<uint64_t>(
+      (cudaStream_t *)(streams), gpu_indexes, gpu_count,
+      static_cast<uint64_t *>(lhs), static_cast<uint64_t const *>(rhs),
+      static_cast<uint64_t *>(overflowed), op, bsks, (uint64_t *const *)(ksks),
+      mem, num_blocks);
+}
+
+void cleanup_signed_overflowing_add_or_sub(void *const *streams,
+                                           uint32_t const *gpu_indexes,
+                                           uint32_t gpu_count,
+                                           int8_t **mem_ptr_void) {
+  int_signed_overflowing_add_or_sub_memory<uint64_t> *mem_ptr =
+      (int_signed_overflowing_add_or_sub_memory<uint64_t> *)(*mem_ptr_void);
+
+  mem_ptr->release((cudaStream_t *)(streams), gpu_indexes, gpu_count);
+}

backends/tfhe-cuda-backend/cuda/src/integer/addition.cuh

@@ -0,0 +1,149 @@
+#ifndef TFHE_RS_ADDITION_CUH
+#define TFHE_RS_ADDITION_CUH
+
+#include "crypto/keyswitch.cuh"
+#include "device.h"
+#include "integer/comparison.cuh"
+#include "integer/integer.cuh"
+#include "integer/integer_utilities.h"
+#include "integer/negation.cuh"
+#include "integer/scalar_shifts.cuh"
+#include "linear_algebra.h"
+#include "pbs/programmable_bootstrap.h"
+#include "utils/helper.cuh"
+#include "utils/kernel_dimensions.cuh"
+#include <fstream>
+#include <iostream>
+#include <sstream>
+#include <string>
+#include <vector>
+
+template <typename Torus>
+void host_resolve_signed_overflow(
+    cudaStream_t const *streams, uint32_t const *gpu_indexes,
+    uint32_t gpu_count, Torus *result, Torus *last_block_inner_propagation,
+    Torus const *last_block_input_carry, Torus *last_block_output_carry,
+    int_resolve_signed_overflow_memory<Torus> *mem, void *const *bsks,
+    Torus *const *ksks) {
+
+  auto x = mem->x;
+
+  Torus *d_clears =
+      (Torus *)cuda_malloc_async(sizeof(Torus), streams[0], gpu_indexes[0]);
+
+  cuda_set_value_async<Torus>(streams[0], gpu_indexes[0], d_clears, 2, 1);
+
+  // replace with host function call
+  cuda_mult_lwe_ciphertext_vector_cleartext_vector_64(
+      streams[0], gpu_indexes[0], x, last_block_output_carry, d_clears,
+      mem->params.big_lwe_dimension, 1);
+
+  host_addition<Torus>(streams[0], gpu_indexes[0], last_block_inner_propagation,
+                       last_block_inner_propagation, x,
+                       mem->params.big_lwe_dimension, 1);
+  host_addition<Torus>(streams[0], gpu_indexes[0], last_block_inner_propagation,
+                       last_block_inner_propagation, last_block_input_carry,
+                       mem->params.big_lwe_dimension, 1);
+
+  host_apply_univariate_lut_kb<Torus>(streams, gpu_indexes, gpu_count, result,
+                                      last_block_inner_propagation,
+                                      mem->resolve_overflow_lut, ksks, bsks, 1);
+
+  cuda_drop_async(d_clears, streams[0], gpu_indexes[0]);
+}
+
+template <typename Torus>
+__host__ void scratch_cuda_integer_signed_overflowing_add_or_sub_kb(
+    cudaStream_t const *streams, uint32_t const *gpu_indexes,
+    uint32_t gpu_count,
+    int_signed_overflowing_add_or_sub_memory<Torus> **mem_ptr,
+    uint32_t num_blocks, SIGNED_OPERATION op, int_radix_params params,
+    bool allocate_gpu_memory) {
+
+  *mem_ptr = new int_signed_overflowing_add_or_sub_memory<Torus>(
+      streams, gpu_indexes, gpu_count, params, num_blocks, op,
+      allocate_gpu_memory);
+}
+
+/*
+ * Addition - signed_operation = 1
+ * Subtraction - signed_operation = -1
+ */
+template <typename Torus>
+__host__ void host_integer_signed_overflowing_add_or_sub_kb(
+    cudaStream_t const *streams, uint32_t const *gpu_indexes,
+    uint32_t gpu_count, Torus *lhs, Torus const *rhs, Torus *overflowed,
+    SIGNED_OPERATION op, void *const *bsks, uint64_t *const *ksks,
+    int_signed_overflowing_add_or_sub_memory<uint64_t> *mem_ptr,
+    uint32_t num_blocks) {
+
+  auto radix_params = mem_ptr->params;
+
+  uint32_t big_lwe_dimension = radix_params.big_lwe_dimension;
+  uint32_t big_lwe_size = big_lwe_dimension + 1;
+  uint32_t big_lwe_size_bytes = big_lwe_size * sizeof(Torus);
+
+  assert(radix_params.message_modulus >= 4 && radix_params.carry_modulus >= 4);
+
+  auto result = mem_ptr->result;
+  auto neg_rhs = mem_ptr->neg_rhs;
+  auto input_carries = mem_ptr->input_carries;
+  auto output_carry = mem_ptr->output_carry;
+  auto last_block_inner_propagation = mem_ptr->last_block_inner_propagation;
+
+  cuda_memcpy_async_gpu_to_gpu(result, lhs, num_blocks * big_lwe_size_bytes,
+                               streams[0], gpu_indexes[0]);
+
+  // phase 1
+  if (op == SIGNED_OPERATION::ADDITION) {
+    host_addition<Torus>(streams[0], gpu_indexes[0], result, lhs, rhs,
+                         big_lwe_dimension, num_blocks);
+  } else {
+    host_integer_radix_negation<Torus>(
+        streams, gpu_indexes, gpu_count, neg_rhs, rhs, big_lwe_dimension,
+        num_blocks, radix_params.message_modulus, radix_params.carry_modulus);
+    host_addition<Torus>(streams[0], gpu_indexes[0], result, lhs, neg_rhs,
+                         big_lwe_dimension, num_blocks);
+  }
+
+  // phase 2
+  for (uint j = 0; j < gpu_count; j++) {
+    cuda_synchronize_stream(streams[j], gpu_indexes[j]);
+  }
+
+  host_propagate_single_carry<Torus>(
+      mem_ptr->sub_streams_1, gpu_indexes, gpu_count, result, output_carry,
+      input_carries, mem_ptr->scp_mem, bsks, ksks, num_blocks);
+  host_generate_last_block_inner_propagation<Torus>(
+      mem_ptr->sub_streams_2, gpu_indexes, gpu_count,
+      last_block_inner_propagation, &lhs[(num_blocks - 1) * big_lwe_size],
+      &rhs[(num_blocks - 1) * big_lwe_size], mem_ptr->las_block_prop_mem, bsks,
+      ksks);
+
+  for (uint j = 0; j < mem_ptr->active_gpu_count; j++) {
+    cuda_synchronize_stream(mem_ptr->sub_streams_1[j], gpu_indexes[j]);
+    cuda_synchronize_stream(mem_ptr->sub_streams_2[j], gpu_indexes[j]);
+  }
+
+  // phase 3
+  auto input_carry = &input_carries[(num_blocks - 1) * big_lwe_size];
+  if (op == SIGNED_OPERATION::SUBTRACTION && num_blocks == 1) {
+    // Quick fix for the case where the subtraction is done on a single block
+    Torus *one_scalar =
+        (Torus *)cuda_malloc_async(sizeof(Torus), streams[0], gpu_indexes[0]);
+    cuda_set_value_async<Torus>(streams[0], gpu_indexes[0], one_scalar, 1, 1);
+    create_trivial_radix<Torus>(
+        streams[0], gpu_indexes[0], input_carry, one_scalar, big_lwe_dimension,
+        1, 1, radix_params.message_modulus, radix_params.carry_modulus);
+    cuda_drop_async(one_scalar, streams[0], gpu_indexes[0]);
+  }
+
+  host_resolve_signed_overflow<Torus>(
+      streams, gpu_indexes, gpu_count, overflowed, last_block_inner_propagation,
+      input_carry, output_carry, mem_ptr->resolve_overflow_mem, bsks, ksks);
+
+  cuda_memcpy_async_gpu_to_gpu(lhs, result, num_blocks * big_lwe_size_bytes,
+                               streams[0], gpu_indexes[0]);
+}
+
+#endif // TFHE_RS_ADDITION_CUH

backends/tfhe-cuda-backend/cuda/src/integer/cmux.cuh

@@ -37,32 +37,39 @@ __host__ void host_integer_radix_cmux_kb(
     uint32_t num_radix_blocks) {
 
   auto params = mem_ptr->params;
-  Torus lwe_size = params.big_lwe_dimension + 1;
-  Torus radix_lwe_size = lwe_size * num_radix_blocks;
-  cuda_memcpy_async_gpu_to_gpu(mem_ptr->buffer_in, lwe_array_true,
-                               radix_lwe_size * sizeof(Torus), streams[0],
-                               gpu_indexes[0]);
-  cuda_memcpy_async_gpu_to_gpu(mem_ptr->buffer_in + radix_lwe_size,
-                               lwe_array_false, radix_lwe_size * sizeof(Torus),
-                               streams[0], gpu_indexes[0]);
-  for (uint i = 0; i < 2 * num_radix_blocks; i++) {
-    cuda_memcpy_async_gpu_to_gpu(mem_ptr->condition_array + i * lwe_size,
-                                 lwe_condition, lwe_size * sizeof(Torus),
-                                 streams[0], gpu_indexes[0]);
+
+  // Since our CPU threads will be working on different streams we shall assert
+  // the work in the main stream is completed
+  auto true_streams = mem_ptr->zero_if_true_buffer->true_streams;
+  auto false_streams = mem_ptr->zero_if_false_buffer->false_streams;
+  for (uint j = 0; j < gpu_count; j++) {
+    cuda_synchronize_stream(streams[j], gpu_indexes[j]);
+  }
+
+  auto mem_true = mem_ptr->zero_if_true_buffer;
+  zero_out_if<Torus>(true_streams, gpu_indexes, gpu_count, mem_ptr->tmp_true_ct,
+                     lwe_array_true, lwe_condition, mem_true,
+                     mem_ptr->inverted_predicate_lut, bsks, ksks,
+                     num_radix_blocks);
+  auto mem_false = mem_ptr->zero_if_false_buffer;
+  zero_out_if<Torus>(false_streams, gpu_indexes, gpu_count,
+                     mem_ptr->tmp_false_ct, lwe_array_false, lwe_condition,
+                     mem_false, mem_ptr->predicate_lut, bsks, ksks,
+                     num_radix_blocks);
+  for (uint j = 0; j < mem_ptr->zero_if_true_buffer->active_gpu_count; j++) {
+    cuda_synchronize_stream(true_streams[j], gpu_indexes[j]);
+  }
+  for (uint j = 0; j < mem_ptr->zero_if_false_buffer->active_gpu_count; j++) {
+    cuda_synchronize_stream(false_streams[j], gpu_indexes[j]);
   }
-  integer_radix_apply_bivariate_lookup_table_kb<Torus>(
-      streams, gpu_indexes, gpu_count, mem_ptr->buffer_out, mem_ptr->buffer_in,
-      mem_ptr->condition_array, bsks, ksks, 2 * num_radix_blocks,
-      mem_ptr->predicate_lut, params.message_modulus);
 
   // If the condition was true, true_ct will have kept its value and false_ct
   // will be 0 If the condition was false, true_ct will be 0 and false_ct will
   // have kept its value
-  auto mem_true = mem_ptr->buffer_out;
-  auto mem_false = &mem_ptr->buffer_out[radix_lwe_size];
-  auto added_cts = mem_true;
-  host_addition<Torus>(streams[0], gpu_indexes[0], added_cts, mem_true,
-                       mem_false, params.big_lwe_dimension, num_radix_blocks);
+  auto added_cts = mem_ptr->tmp_true_ct;
+  host_addition<Torus>(streams[0], gpu_indexes[0], added_cts,
+                       mem_ptr->tmp_true_ct, mem_ptr->tmp_false_ct,
+                       params.big_lwe_dimension, num_radix_blocks);
 
   integer_radix_apply_univariate_lookup_table_kb<Torus>(
       streams, gpu_indexes, gpu_count, lwe_array_out, added_cts, bsks, ksks,

backends/tfhe-cuda-backend/cuda/src/integer/comparison.cu

@@ -58,9 +58,6 @@ void cuda_comparison_integer_radix_ciphertext_kb_64(
   case GE:
   case LT:
   case LE:
-    if (num_radix_blocks % 2 != 0)
-      PANIC("Cuda error (comparisons): the number of radix blocks has to be "
-            "even.")
     host_integer_radix_difference_check_kb<uint64_t>(
         (cudaStream_t *)(streams), gpu_indexes, gpu_count,
         static_cast<uint64_t *>(lwe_array_out),
@@ -71,8 +68,6 @@ void cuda_comparison_integer_radix_ciphertext_kb_64(
     break;
   case MAX:
   case MIN:
-    if (num_radix_blocks % 2 != 0)
-      PANIC("Cuda error (max/min): the number of radix blocks has to be even.")
     host_integer_radix_maxmin_kb<uint64_t>(
         (cudaStream_t *)(streams), gpu_indexes, gpu_count,
         static_cast<uint64_t *>(lwe_array_out),
@@ -94,91 +89,3 @@ void cleanup_cuda_integer_comparison(void *const *streams,
       (int_comparison_buffer<uint64_t> *)(*mem_ptr_void);
   mem_ptr->release((cudaStream_t *)(streams), gpu_indexes, gpu_count);
 }
-
-void scratch_cuda_integer_are_all_comparisons_block_true_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_radix_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory) {
-
-  int_radix_params params(pbs_type, glwe_dimension, polynomial_size,
-                          big_lwe_dimension, small_lwe_dimension, ks_level,
-                          ks_base_log, pbs_level, pbs_base_log, grouping_factor,
-                          message_modulus, carry_modulus);
-
-  scratch_cuda_integer_radix_comparison_check_kb<uint64_t>(
-      (cudaStream_t *)(streams), gpu_indexes, gpu_count,
-      (int_comparison_buffer<uint64_t> **)mem_ptr, num_radix_blocks, params, EQ,
-      false, allocate_gpu_memory);
-}
-
-void cuda_integer_are_all_comparisons_block_true_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    void *lwe_array_out, void const *lwe_array_in, int8_t *mem_ptr,
-    void *const *bsks, void *const *ksks, uint32_t num_radix_blocks) {
-
-  int_comparison_buffer<uint64_t> *buffer =
-      (int_comparison_buffer<uint64_t> *)mem_ptr;
-
-  host_integer_are_all_comparisons_block_true_kb<uint64_t>(
-      (cudaStream_t *)(streams), gpu_indexes, gpu_count,
-      static_cast<uint64_t *>(lwe_array_out),
-      static_cast<const uint64_t *>(lwe_array_in), buffer, bsks,
-      (uint64_t **)(ksks), num_radix_blocks);
-}
-
-void cleanup_cuda_integer_are_all_comparisons_block_true(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void) {
-
-  int_comparison_buffer<uint64_t> *mem_ptr =
-      (int_comparison_buffer<uint64_t> *)(*mem_ptr_void);
-  mem_ptr->release((cudaStream_t *)(streams), gpu_indexes, gpu_count);
-}
-
-void scratch_cuda_integer_is_at_least_one_comparisons_block_true_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_radix_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory) {
-
-  int_radix_params params(pbs_type, glwe_dimension, polynomial_size,
-                          big_lwe_dimension, small_lwe_dimension, ks_level,
-                          ks_base_log, pbs_level, pbs_base_log, grouping_factor,
-                          message_modulus, carry_modulus);
-
-  scratch_cuda_integer_radix_comparison_check_kb<uint64_t>(
-      (cudaStream_t *)(streams), gpu_indexes, gpu_count,
-      (int_comparison_buffer<uint64_t> **)mem_ptr, num_radix_blocks, params, EQ,
-      false, allocate_gpu_memory);
-}
-
-void cuda_integer_is_at_least_one_comparisons_block_true_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    void *lwe_array_out, void const *lwe_array_in, int8_t *mem_ptr,
-    void *const *bsks, void *const *ksks, uint32_t num_radix_blocks) {
-
-  int_comparison_buffer<uint64_t> *buffer =
-      (int_comparison_buffer<uint64_t> *)mem_ptr;
-
-  host_integer_is_at_least_one_comparisons_block_true_kb<uint64_t>(
-      (cudaStream_t *)(streams), gpu_indexes, gpu_count,
-      static_cast<uint64_t *>(lwe_array_out),
-      static_cast<const uint64_t *>(lwe_array_in), buffer, bsks,
-      (uint64_t **)(ksks), num_radix_blocks);
-}
-
-void cleanup_cuda_integer_is_at_least_one_comparisons_block_true(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void) {
-
-  int_comparison_buffer<uint64_t> *mem_ptr =
-      (int_comparison_buffer<uint64_t> *)(*mem_ptr_void);
-  mem_ptr->release((cudaStream_t *)(streams), gpu_indexes, gpu_count);
-}