fix

.gitattributes

@@ -1,3 +0,0 @@
-*.hpu filter=lfs diff=lfs merge=lfs -text
-*.bcode filter=lfs diff=lfs merge=lfs -text
-*.cbor filter=lfs diff=lfs merge=lfs -text

.github/actions/gpu_setup/action.yml

@@ -23,58 +23,38 @@ runs:
         echo "${CMAKE_SCRIPT_SHA} cmake-${CMAKE_VERSION}-linux-x86_64.sh" > checksum
         sha256sum -c checksum
         sudo bash cmake-"${CMAKE_VERSION}"-linux-x86_64.sh --skip-license --prefix=/usr/ --exclude-subdir
-        sudo apt remove -y unattended-upgrades
         sudo apt update
         sudo apt install -y cmake-format libclang-dev
       env:
         CMAKE_VERSION: 3.29.6
         CMAKE_SCRIPT_SHA: "6e4fada5cba3472ae503a11232b6580786802f0879cead2741672bf65d97488a"
 
-    - name: Install GCC
-      if: inputs.github-instance == 'true'
-      shell: bash
-      env:
-        GCC_VERSION: ${{ inputs.gcc-version }}
-      run: |
-        sudo apt-get install gcc-"{GCC_VERSION}" g++-"{GCC_VERSION}"
-        sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-"{GCC_VERSION}" 20
-        sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-"{GCC_VERSION}" 20
-
-    - name: Check GCC
-      shell: bash
-      env:
-        GCC_VERSION: ${{ inputs.gcc-version }}
-      run: |
-        which gcc-"${GCC_VERSION}"
-
     - name: Install CUDA
       if: inputs.github-instance == 'true'
       shell: bash
-      env:
-        CUDA_VERSION: ${{ inputs.cuda-version }}
-        CUDA_KEYRING_PACKAGE: cuda-keyring_1.1-1_all.deb
-        CUDA_KEYRING_SHA: "d93190d50b98ad4699ff40f4f7af50f16a76dac3bb8da1eaaf366d47898ff8df"
       run: |
         # Use Sed to extract a value from a string, this cannot be done with the ${variable//search/replace} pattern.
         # shellcheck disable=SC2001
         TOOLKIT_VERSION="$(echo "${CUDA_VERSION}" | sed 's/\(.*\)\.\(.*\)/\1-\2/')"
-        wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/${CUDA_KEYRING_PACKAGE}
+        wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/${env.CUDA_KEYRING_PACKAGE}
         echo "${CUDA_KEYRING_SHA} ${CUDA_KEYRING_PACKAGE}" > checksum
         sha256sum -c checksum
         sudo dpkg -i "${CUDA_KEYRING_PACKAGE}"
         sudo apt update
         sudo apt -y install cuda-toolkit-"${TOOLKIT_VERSION}"
+      env:
+        CUDA_VERSION: ${{ inputs.cuda-version }}
+        CUDA_KEYRING_PACKAGE: cuda-keyring_1.1-1_all.deb
+        CUDA_KEYRING_SHA: "d93190d50b98ad4699ff40f4f7af50f16a76dac3bb8da1eaaf366d47898ff8df"
 
     - name: Export CUDA variables
       shell: bash
       run: |
-        find /usr/local -executable -name "nvcc"
         CUDA_PATH=/usr/local/cuda-"${CUDA_VERSION}"
         {
           echo "CUDA_PATH=$CUDA_PATH";
           echo "LD_LIBRARY_PATH=$CUDA_PATH/lib64:$LD_LIBRARY_PATH";
           echo "CUDA_MODULE_LOADER=EAGER";
-          echo "PATH=$PATH:$CUDA_PATH/bin"; 
         } >> "${GITHUB_ENV}"
         {
           echo "PATH=$PATH:$CUDA_PATH/bin"; 
@@ -94,11 +74,6 @@ runs:
       env:
         GCC_VERSION: ${{ inputs.gcc-version }}
 
-    - name: Check setup
-      shell: bash
-      run: |
-        which nvcc
-
     - name: Check device is detected
       shell: bash
       run: nvidia-smi

.github/workflows/aws_tfhe_backward_compat_tests.yml

@@ -61,36 +61,55 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 
-      # Cache key is an aggregated hash of lfs files hashes
-      - name: Get LFS data sha
-        id: hash-lfs-data
+      - name: Use specific data branch
+        if: ${{ contains(github.event.pull_request.labels.*.name, 'data_PR') }}
+        env:
+          PR_BRANCH: ${{ github.head_ref || github.ref_name }}
         run: |
-          SHA=$(git lfs ls-files -l -I utils/tfhe-backward-compat-data | sha256sum | cut -d' ' -f1)
+          echo "BACKWARD_COMPAT_DATA_BRANCH=${PR_BRANCH}" >> "${GITHUB_ENV}"
+
+      - name: Get backward compat branch
+        id: backward_compat_branch
+        run: |
+          BRANCH="$(make backward_compat_branch)"
+          echo "branch=${BRANCH}" >> "${GITHUB_OUTPUT}"
+
+      - name: Get backward compat branch head SHA
+        id: backward_compat_sha
+        run: |
+          SHA=$(git ls-remote "${REPO_URL}" refs/heads/"${BACKWARD_COMPAT_BRANCH}" | awk '{print $1}')
           echo "sha=${SHA}" >> "${GITHUB_OUTPUT}"
+        env:
+          REPO_URL: "https://github.com/zama-ai/tfhe-backward-compat-data"
+          BACKWARD_COMPAT_BRANCH: ${{ steps.backward_compat_branch.outputs.branch }}
 
       - name: Retrieve data from cache
         id: retrieve-data-cache
-        uses: actions/cache/restore@0400d5f644dc74513175e3cd8d07132dd4860809 #v4.2.4
+        uses: actions/cache/restore@5a3ec84eff668545956fd18022155c47e93e2684 #v4.2.3
         with:
-          path: |
-            utils/tfhe-backward-compat-data/**/*.cbor
-            utils/tfhe-backward-compat-data/**/*.bcode
-          key: ${{ steps.hash-lfs-data.outputs.sha }}
+          path: tests/tfhe-backward-compat-data
+          key: ${{ steps.backward_compat_branch.outputs.branch }}_${{ steps.backward_compat_sha.outputs.sha }}
 
-      - name: Pull test data
+      - name: Clone test data
         if: steps.retrieve-data-cache.outputs.cache-hit != 'true'
-        run: |
-          make pull_backward_compat_data
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          persist-credentials: 'false'
+          token: ${{ env.CHECKOUT_TOKEN }}
+          repository: zama-ai/tfhe-backward-compat-data
+          path: tests/tfhe-backward-compat-data
+          lfs: 'true'
+          ref: ${{ steps.backward_compat_branch.outputs.branch }}
 
       - name: Run backward compatibility tests
         run: |
@@ -99,12 +118,10 @@ jobs:
       - name: Store data in cache
         if: steps.retrieve-data-cache.outputs.cache-hit != 'true'
         continue-on-error: true
-        uses: actions/cache/save@0400d5f644dc74513175e3cd8d07132dd4860809 #v4.2.4
+        uses: actions/cache/save@5a3ec84eff668545956fd18022155c47e93e2684 #v4.2.3
         with:
-          path: |
-            utils/tfhe-backward-compat-data/**/*.cbor
-            utils/tfhe-backward-compat-data/**/*.bcode
-          key: ${{ steps.hash-lfs-data.outputs.sha }}
+          path: tests/tfhe-backward-compat-data
+          key: ${{ steps.backward_compat_branch.outputs.branch }}_${{ steps.backward_compat_sha.outputs.sha }}
 
       - name: Set pull-request URL
         if: ${{ failure() && github.event_name == 'pull_request' }}

.github/workflows/aws_tfhe_fast_tests.yml

@@ -60,7 +60,7 @@ jobs:
       any_file_changed: ${{ env.IS_PULL_REQUEST == 'false' || steps.aggregated-changes.outputs.any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -168,13 +168,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 
@@ -182,11 +182,9 @@ jobs:
         if: needs.should-run.outputs.csprng_test == 'true'
         run: |
           make test_tfhe_csprng
-          make test_tfhe_csprng_big_endian
 
       - name: Run tfhe-zk-pok tests
-        # Always run it to catch non deterministic bugs earlier
-        # if: needs.should-run.outputs.zk_pok_test == 'true'
+        if: needs.should-run.outputs.zk_pok_test == 'true'
         run: |
           make test_zk_pok
 
@@ -216,7 +214,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@0400d5f644dc74513175e3cd8d07132dd4860809 #v4.2.4
+        uses: actions/cache/restore@5a3ec84eff668545956fd18022155c47e93e2684 #v4.2.3
         with:
           path: |
             ~/.nvm
@@ -229,7 +227,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@0400d5f644dc74513175e3cd8d07132dd4860809 #v4.2.4
+        uses: actions/cache/save@5a3ec84eff668545956fd18022155c47e93e2684 #v4.2.3
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |

.github/workflows/aws_tfhe_integer_tests.yml

@@ -47,7 +47,7 @@ jobs:
         steps.changed-files.outputs.integer_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -103,18 +103,18 @@ jobs:
     name: Unsigned integer tests
     needs: setup-instance
     concurrency:
-      group: ${{ github.workflow_ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
+      group: ${{ github.workflow_ref }}
       cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: "false"
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 

.github/workflows/aws_tfhe_noise_checks.yml

@@ -1,115 +0,0 @@
-name: Run noise checks 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 }}
-  SLACKIFY_MARKDOWN: true
-  PULL_REQUEST_MD_LINK: ""
-  CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN || secrets.GITHUB_TOKEN }}
-  # Secrets will be available only to zama-ai organization members
-  SECRETS_AVAILABLE: ${{ secrets.JOB_SECRET != '' }}
-
-on:
-  # Allows you to run this workflow manually from the Actions tab as an alternative.
-  workflow_dispatch:
-
-permissions:
-  contents: read
-
-jobs:
-  setup-instance:
-    name: Setup instance (noise-checks)
-    runs-on: ubuntu-latest
-    outputs:
-      runner-name: ${{ steps.start-remote-instance.outputs.label || steps.start-github-instance.outputs.runner_group }}
-    steps:
-      - name: Start remote instance
-        id: start-remote-instance
-        if: env.SECRETS_AVAILABLE == 'true'
-        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
-          # We want an hpc7a more compute, will be faster
-          profile: bench
-
-      # This instance will be spawned especially for pull-request from forked repository
-      - name: Start GitHub instance
-        id: start-github-instance
-        if: env.SECRETS_AVAILABLE == 'false'
-        run: |
-          echo "Cannot run this without secrets"
-          exit 1
-
-  noise-checks:
-    name: CPU noise checks
-    needs: setup-instance
-    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
-    timeout-minutes: 1440
-    steps:
-      - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
-        with:
-          persist-credentials: 'false'
-          token: ${{ env.CHECKOUT_TOKEN }}
-
-      - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
-        with:
-          toolchain: stable
-
-      - name: Run noise checks
-        timeout-minutes: 1440
-        run: |
-          make test_noise_check
-
-      - name: Set pull-request URL
-        if: ${{ !success() }}
-        run: |
-          echo "PULL_REQUEST_MD_LINK=[pull-request](${PR_BASE_URL}${PR_NUMBER}), "  >> "${GITHUB_ENV}"
-        env:
-          PR_BASE_URL: ${{ vars.PR_BASE_URL }}
-          PR_NUMBER: ${{ github.event.pull_request.number }}
-
-      - name: Slack Notification
-        if: ${{ !success() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@e31e87e03dd19038e411e38ae27cbad084a90661
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "Noise checks tests finished with status: ${{ job.status }}. (${{ env.PULL_REQUEST_MD_LINK }}[action run](${{ env.ACTION_RUN_URL }}))"
-
-  teardown-instance:
-    name: Teardown instance (noise-checks)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
-    needs: [ setup-instance, noise-checks ]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Stop remote instance
-        id: stop-instance
-        if: env.SECRETS_AVAILABLE == 'true'
-        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: ${{ !success() }}
-        continue-on-error: true
-        uses: rtCamp/action-slack-notify@e31e87e03dd19038e411e38ae27cbad084a90661
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "Instance teardown (noise-checks) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/aws_tfhe_signed_integer_tests.yml

@@ -48,7 +48,7 @@ jobs:
         steps.changed-files.outputs.integer_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -104,18 +104,18 @@ jobs:
     name: Signed integer tests
     needs: setup-instance
     concurrency:
-      group: ${{ github.workflow_ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
+      group: ${{ github.workflow_ref }}
       cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: "false"
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 

.github/workflows/aws_tfhe_tests.yml

@@ -69,7 +69,7 @@ jobs:
       any_file_changed: ${{ env.IS_PULL_REQUEST == 'false' || steps.aggregated-changes.outputs.any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -179,13 +179,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 

.github/workflows/aws_tfhe_wasm_tests.yml

@@ -62,13 +62,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 
@@ -78,7 +78,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@0400d5f644dc74513175e3cd8d07132dd4860809 #v4.2.4
+        uses: actions/cache/restore@5a3ec84eff668545956fd18022155c47e93e2684 #v4.2.3
         with:
           path: |
             ~/.nvm
@@ -91,7 +91,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@0400d5f644dc74513175e3cd8d07132dd4860809 #v4.2.4
+        uses: actions/cache/save@5a3ec84eff668545956fd18022155c47e93e2684 #v4.2.3
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |

.github/workflows/benchmark_boolean.yml

@@ -50,7 +50,7 @@ jobs:
       cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -68,7 +68,7 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
@@ -107,7 +107,7 @@ jobs:
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_core_crypto.yml

@@ -50,7 +50,7 @@ jobs:
       cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -68,7 +68,7 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
@@ -100,7 +100,7 @@ jobs:
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_dex.yml

@@ -50,7 +50,7 @@ jobs:
     timeout-minutes: 720  # 12 hours
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -68,12 +68,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -98,27 +98,15 @@ jobs:
         env:
           REF_NAME: ${{ github.ref_name }}
 
-      - name: Parse swap request update PBS counts
+      - name: Parse swap request PBS counts
         run: |
-          python3 ./ci/benchmark_parser.py tfhe-benchmark/dex_swap_request_update_dex_balance_pbs_count.csv "${RESULTS_FILENAME}" \
+          python3 ./ci/benchmark_parser.py tfhe-benchmark/dex_swap_request_pbs_count.csv "${RESULTS_FILENAME}" \
           --object-sizes \
           --append-results
 
-      - name: Parse swap request finalize PBS counts
+      - name: Parse swap claim PBS counts
         run: |
-          python3 ./ci/benchmark_parser.py tfhe-benchmark/dex_swap_request_finalize_pbs_count.csv "${RESULTS_FILENAME}" \
-          --object-sizes \
-          --append-results
-
-      - name: Parse swap claim prepare PBS counts
-        run: |
-          python3 ./ci/benchmark_parser.py tfhe-benchmark/dex_swap_claim_prepare_pbs_count.csv "${RESULTS_FILENAME}" \
-          --object-sizes \
-          --append-results
-
-      - name: Parse swap claim update PBS counts
-        run: |
-          python3 ./ci/benchmark_parser.py tfhe-benchmark/dex_swap_claim_update_dex_balance_pbs_count.csv "${RESULTS_FILENAME}" \
+          python3 ./ci/benchmark_parser.py tfhe-benchmark/dex_swap_claim_pbs_count.csv "${RESULTS_FILENAME}" \
           --object-sizes \
           --append-results
 

.github/workflows/benchmark_erc20.yml

@@ -51,7 +51,7 @@ jobs:
     timeout-minutes: 720  # 12 hours
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -69,12 +69,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_gpu.yml

@@ -31,7 +31,6 @@ on:
           - ks
           - ks_pbs
           - integer_zk
-          - hlapi_noise_squash
       op_flavor:
         description: "Operations set to run"
         type: choice

.github/workflows/benchmark_gpu_4090.yml

@@ -38,7 +38,7 @@ jobs:
     timeout-minutes: 1440 # 24 hours
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -57,12 +57,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -122,7 +122,7 @@ jobs:
 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -140,12 +140,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -168,8 +168,7 @@ jobs:
           --commit-date "${COMMIT_DATE}" \
           --bench-date "${BENCH_DATE}" \
           --walk-subdirs \
-        env:
-          REF_NAME: ${{ github.ref_name }}
+      
 
       - name: Upload parsed results artifact
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02

.github/workflows/benchmark_gpu_common.yml

@@ -122,23 +122,31 @@ jobs:
 
       - name: Set command output
         id: set_command
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "command=${{ toJSON(env.COMMAND) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "command=${COMMAND_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          COMMAND_OUTPUT: ${{ toJSON(env.COMMAND) }}
 
       - name: Set operation flavor output
         id: set_op_flavor
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "op_flavor=${{ toJSON(env.OP_FLAVOR) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "op_flavor=${OP_FLAVOR_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          OP_FLAVOR_OUTPUT: ${{ toJSON(env.OP_FLAVOR) }}
 
       - name: Set benchmark types output
         id: set_bench_type
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "bench_type=${BENCH_TYPE_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          BENCH_TYPE_OUTPUT: ${{ toJSON(env.BENCH_TYPE) }}
 
       - name: Set parameters types output
         id: set_params_type
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "params_type=${{ toJSON(env.PARAMS_TYPE) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "params_type=${PARAMS_TYPE_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          PARAMS_TYPE_OUTPUT: ${{ toJSON(env.PARAMS_TYPE) }}
 
   setup-instance:
     name: Setup instance (cuda-${{ inputs.profile }}-benchmarks)
@@ -192,11 +200,11 @@ jobs:
       matrix:
         # explicit include-based build matrix, of known valid options
         include:
-          - cuda: "12.8"
+          - cuda: "12.2"
             gcc: 11
     steps:
       - name: Checkout tfhe-rs repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -224,13 +232,13 @@ jobs:
         params_type: ${{ fromJSON(needs.prepare-matrix.outputs.params_type) }}
         # explicit include-based build matrix, of known valid options
         include:
-          - cuda: "12.8"
+          - cuda: "12.2"
             gcc: 11
     env:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -269,7 +277,7 @@ jobs:
           GCC_VERSION: ${{ matrix.gcc }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
@@ -312,7 +320,7 @@ jobs:
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_gpu_dex_common.yml

@@ -100,11 +100,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -129,7 +129,7 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
@@ -160,7 +160,7 @@ jobs:
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_gpu_erc20_common.yml

@@ -101,11 +101,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -130,7 +130,7 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
@@ -161,7 +161,7 @@ jobs:
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_gpu_weekly.yml

@@ -10,16 +10,37 @@ on:
 permissions: {}
 
 jobs:
-  run-benchmarks-8-h100-sxm5-integer:
-    name: Run integer benchmarks (8xH100-SXM5)
+  run-benchmarks-1-h100:
+    name: Run integer benchmarks (1xH100)
     if: github.repository == 'zama-ai/tfhe-rs'
     uses: ./.github/workflows/benchmark_gpu_common.yml
     with:
-      profile: multi-h100-sxm5
-      hardware_name: n3-H100x8-SXM5
+      profile: single-h100
+      hardware_name: n3-H100x1
+      command: integer,integer_multi_bit
+      op_flavor: default
+      bench_type: latency
+      all_precisions: true
+    secrets:
+      BOT_USERNAME: ${{ secrets.BOT_USERNAME }}
+      SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
+      SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+      REPO_CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN }}
+      JOB_SECRET: ${{ secrets.JOB_SECRET }}
+      SLAB_ACTION_TOKEN: ${{ secrets.SLAB_ACTION_TOKEN }}
+      SLAB_URL: ${{ secrets.SLAB_URL }}
+      SLAB_BASE_URL: ${{ secrets.SLAB_BASE_URL }}
+
+  run-benchmarks-2-h100:
+    name: Run integer benchmarks (2xH100)
+    if: github.repository == 'zama-ai/tfhe-rs'
+    uses: ./.github/workflows/benchmark_gpu_common.yml
+    with:
+      profile: 2-h100
+      hardware_name: n3-H100x2
       command: integer_multi_bit
       op_flavor: default
-      bench_type: both
+      bench_type: latency
       all_precisions: true
     secrets:
       BOT_USERNAME: ${{ secrets.BOT_USERNAME }}
@@ -31,16 +52,16 @@ jobs:
       SLAB_URL: ${{ secrets.SLAB_URL }}
       SLAB_BASE_URL: ${{ secrets.SLAB_BASE_URL }}
 
-  run-benchmarks-8-h100-sxm5-integer-compression:
-    name: Run integer compression benchmarks (8xH100-SXM5)
+  run-benchmarks-8-h100:
+    name: Run integer benchmarks (8xH100)
     if: github.repository == 'zama-ai/tfhe-rs'
     uses: ./.github/workflows/benchmark_gpu_common.yml
     with:
-      profile: multi-h100-sxm5
-      hardware_name: n3-H100x8-SXM5
-      command: integer_compression
+      profile: multi-h100
+      hardware_name: n3-H100x8
+      command: integer_multi_bit
       op_flavor: default
-      bench_type: both
+      bench_type: latency
       all_precisions: true
     secrets:
       BOT_USERNAME: ${{ secrets.BOT_USERNAME }}
@@ -52,37 +73,16 @@ jobs:
       SLAB_URL: ${{ secrets.SLAB_URL }}
       SLAB_BASE_URL: ${{ secrets.SLAB_BASE_URL }}
 
-  run-benchmarks-8-h100-sxm5-integer-zk:
-    name: Run integer zk benchmarks (8xH100-SXM5)
+  run-benchmarks-l40:
+    name: Run integer benchmarks (L40)
     if: github.repository == 'zama-ai/tfhe-rs'
     uses: ./.github/workflows/benchmark_gpu_common.yml
     with:
-      profile: multi-h100-sxm5
-      hardware_name: n3-H100x8-SXM5
-      command: integer_zk
+      profile: l40
+      hardware_name: n3-L40x1
+      command: integer_multi_bit,integer_compression,pbs,ks
       op_flavor: default
-      bench_type: both
-      all_precisions: true
-    secrets:
-      BOT_USERNAME: ${{ secrets.BOT_USERNAME }}
-      SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}
-      SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
-      REPO_CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN }}
-      JOB_SECRET: ${{ secrets.JOB_SECRET }}
-      SLAB_ACTION_TOKEN: ${{ secrets.SLAB_ACTION_TOKEN }}
-      SLAB_URL: ${{ secrets.SLAB_URL }}
-      SLAB_BASE_URL: ${{ secrets.SLAB_BASE_URL }}
-
-  run-benchmarks-8-h100-sxm5-noise-squash:
-    name: Run integer zk benchmarks (8xH100-SXM5)
-    if: github.repository == 'zama-ai/tfhe-rs'
-    uses: ./.github/workflows/benchmark_gpu_common.yml
-    with:
-      profile: multi-h100-sxm5
-      hardware_name: n3-H100x8-SXM5
-      command: hlapi_noise_squash
-      op_flavor: default
-      bench_type: both
+      bench_type: latency
       all_precisions: true
     secrets:
       BOT_USERNAME: ${{ secrets.BOT_USERNAME }}

.github/workflows/benchmark_hpu_hlapi.yml

@@ -1,98 +0,0 @@
-# Run all integer benchmarks on a permanent HPU instance and return parsed results to Slab CI bot.
-name: Hpu Hlapi Benchmarks
-
-on:
-  workflow_dispatch:
-
-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"
-
-permissions: {}
-
-jobs:
-  hlapi-benchmarks-hpu:
-    name: Execute HLAPI benchmarks for HPU backend
-    runs-on: v80-desktop
-    concurrency:
-      group: ${{ github.workflow }}_${{ github.ref }}
-      cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
-    timeout-minutes: 1440  # 24 hours
-    steps:
-      # Needed as long as hw_regmap repository is private
-      - name: Configure SSH
-        uses: webfactory/ssh-agent@a6f90b1f127823b31d4d4a8d96047790581349bd # v0.9.1
-        with:
-          ssh-private-key: ${{ secrets.SSH_PRIVATE_KEY }}
-
-      - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
-        with:
-          fetch-depth: 0
-          persist-credentials: 'false'
-          lfs: true
-          token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
-
-      - name: Get benchmark details
-        run: |
-          COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict "${SHA}");
-          {
-            echo "BENCH_DATE=$(date --iso-8601=seconds)";
-            echo "COMMIT_DATE=${COMMIT_DATE}";
-            echo "COMMIT_HASH=$(git describe --tags --dirty)";
-          } >> "${GITHUB_ENV}"
-        env:
-          SHA: ${{ github.sha }}
-
-      - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
-        with:
-          toolchain: nightly
-
-      - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
-        with:
-          repository: zama-ai/slab
-          path: slab
-          persist-credentials: 'false'
-          token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
-
-      - name: Run benchmarks
-        run: |
-          make pull_hpu_files
-          export V80_SERIAL_NUMBER=XFL12E4XJXWK
-          source /opt/xilinx/Vivado/2024.2/settings64.sh
-          make bench_hlapi_erc20_hpu
-          make bench_hlapi_hpu
-
-      - name: Parse results
-        run: |
-          python3 ./ci/benchmark_parser.py target/criterion "${RESULTS_FILENAME}" \
-          --database tfhe_rs \
-          --hardware "hpu_x1" \
-          --backend hpu \
-          --project-version "${COMMIT_HASH}" \
-          --branch "${REF_NAME}" \
-          --commit-date "${COMMIT_DATE}" \
-          --bench-date "${BENCH_DATE}" \
-          --walk-subdirs
-        env:
-          REF_NAME: ${{ github.ref_name }}
-
-      - name: Upload parsed results artifact
-        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
-        with:
-          name: ${{ github.sha }}_hlapi_benchmarks
-          path: ${{ env.RESULTS_FILENAME }}
-
-      - name: Send data to Slab
-        shell: bash
-        run: |
-          python3 slab/scripts/data_sender.py "${RESULTS_FILENAME}" "${JOB_SECRET}" \
-          --slab-url "${SLAB_URL}"
-        env:
-          JOB_SECRET: ${{ secrets.JOB_SECRET }}
-          SLAB_URL: ${{ secrets.SLAB_URL }}

.github/workflows/benchmark_hpu_integer.yml

@@ -3,15 +3,6 @@ name: Hpu Integer Benchmarks
 
 on:
   workflow_dispatch:
-    inputs:
-      bench_type:
-        description: "Benchmarks type"
-        type: choice
-        default: both
-        options:
-          - latency
-          - throughput
-          - both
 
 env:
   CARGO_TERM_COLOR: always
@@ -23,46 +14,13 @@ env:
 permissions: {}
 
 jobs:
-  prepare-matrix:
-    name: Prepare operations matrix
-    runs-on: v80-desktop
-    outputs:
-      bench_type: ${{ steps.set_bench_type.outputs.bench_type }}
-    steps:
-      - name: Set benchmark types
-        if: github.event_name == 'workflow_dispatch'
-        run: |
-          if [[ -z $INPUTS_BENCH_TYPE || "${INPUTS_BENCH_TYPE}" == "both" ]]; then
-            echo "BENCH_TYPE=[\"latency\", \"throughput\"]" >> "${GITHUB_ENV}"
-          else
-            echo "BENCH_TYPE=[\"${INPUTS_BENCH_TYPE}\"]" >> "${GITHUB_ENV}"
-          fi
-        env:
-          INPUTS_BENCH_TYPE: ${{ inputs.bench_type }}
-
-      - name: Default benchmark type
-        if: github.event_name != 'workflow_dispatch'
-        run: |
-          echo "BENCH_TYPE=[\"latency\"]" >> "${GITHUB_ENV}"
-
-
-      - name: Set benchmark types output
-        id: set_bench_type
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
-
   integer-benchmarks-hpu:
     name: Execute integer & erc20 benchmarks for HPU backend
-    needs: prepare-matrix
     runs-on: v80-desktop
     concurrency:
       group: ${{ github.workflow }}_${{ github.ref }}
       cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
     timeout-minutes: 1440  # 24 hours
-    strategy:
-      max-parallel: 1
-      matrix:
-        bench_type: ${{ fromJSON(needs.prepare-matrix.outputs.bench_type) }}
     steps:
       # Needed as long as hw_regmap repository is private
       - name: Configure SSH
@@ -71,11 +29,10 @@ jobs:
           ssh-private-key: ${{ secrets.SSH_PRIVATE_KEY }}
 
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
-          lfs: true
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
       - name: Get benchmark details
@@ -90,12 +47,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -104,12 +61,8 @@ jobs:
 
       - name: Run benchmarks
         run: |
-          make pull_hpu_files
-          export V80_SERIAL_NUMBER=XFL12E4XJXWK
-          source /opt/xilinx/Vivado/2024.2/settings64.sh
-          make BENCH_TYPE="${BENCH_TYPE}" bench_integer_hpu
-        env:
-          BENCH_TYPE: ${{ matrix.bench_type }}
+          make bench_integer_hpu
+          make bench_hlapi_erc20_hpu
 
       - name: Parse results
         run: |
@@ -121,16 +74,14 @@ jobs:
           --branch "${REF_NAME}" \
           --commit-date "${COMMIT_DATE}" \
           --bench-date "${BENCH_DATE}" \
-          --walk-subdirs \
-          --bench-type "${BENCH_TYPE}"
+          --walk-subdirs
         env:
           REF_NAME: ${{ github.ref_name }}
-          BENCH_TYPE: ${{ matrix.bench_type }}
 
       - name: Upload parsed results artifact
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
         with:
-          name: ${{ github.sha }}_${{ matrix.bench_type }}_integer_benchmarks
+          name: ${{ github.sha }}_integer_benchmarks
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Send data to Slab

.github/workflows/benchmark_integer.yml

@@ -78,13 +78,17 @@ jobs:
 
       - name: Set operation flavor output
         id: set_op_flavor
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "op_flavor=${{ toJSON(env.OP_FLAVOR) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "op_flavor=${OP_FLAVOR_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          OP_FLAVOR_OUTPUT: ${{ toJSON(env.OP_FLAVOR) }}
 
       - name: Set benchmark types output
         id: set_bench_type
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "bench_type=${BENCH_TYPE_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          BENCH_TYPE_OUTPUT: ${{ toJSON(env.BENCH_TYPE) }}
 
   setup-instance:
     name: Setup instance (integer-benchmarks)
@@ -120,7 +124,7 @@ jobs:
         bench_type: ${{ fromJSON(needs.prepare-matrix.outputs.bench_type) }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -138,12 +142,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_shortint.yml

@@ -47,8 +47,10 @@ jobs:
 
       - name: Set operation flavor output
         id: set_op_flavor
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "op_flavor=${{ toJSON(env.OP_FLAVOR) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "op_flavor=${OP_FLAVOR_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          OP_FLAVOR_OUTPUT: ${{ toJSON(env.OP_FLAVOR) }}
 
   setup-instance:
     name: Setup instance (shortint-benchmarks)
@@ -81,7 +83,7 @@ jobs:
         op_flavor: ${{ fromJson(needs.prepare-matrix.outputs.op_flavor) }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -99,12 +101,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_signed_integer.yml

@@ -78,13 +78,17 @@ jobs:
 
       - name: Set operation flavor output
         id: set_op_flavor
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "op_flavor=${{ toJSON(env.OP_FLAVOR) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "op_flavor=${OP_FLAVOR_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          OP_FLAVOR_OUTPUT: ${{ toJSON(env.OP_FLAVOR) }}
 
       - name: Set benchmark types output
         id: set_bench_type
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "bench_type=${BENCH_TYPE_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          BENCH_TYPE_OUTPUT: ${{ toJSON(env.BENCH_TYPE) }}
 
   setup-instance:
     name: Setup instance (signed-integer-benchmarks)
@@ -120,7 +124,7 @@ jobs:
         bench_type: ${{ fromJSON(needs.prepare-matrix.outputs.bench_type) }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -138,12 +142,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_tfhe_fft.yml

@@ -48,12 +48,12 @@ jobs:
     name: Execute FFT benchmarks in EC2
     needs: setup-ec2
     concurrency:
-      group: ${{ github.workflow_ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
+      group: ${{ github.workflow_ref }}
       cancel-in-progress: true
     runs-on: ${{ needs.setup-ec2.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -100,7 +100,7 @@ jobs:
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_tfhe_ntt.yml

@@ -48,12 +48,12 @@ jobs:
     name: Execute NTT benchmarks in EC2
     needs: setup-ec2
     concurrency:
-      group: ${{ github.workflow_ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
+      group: ${{ github.workflow_ref }}
       cancel-in-progress: true
     runs-on: ${{ needs.setup-ec2.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -100,7 +100,7 @@ jobs:
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_tfhe_zk_pok.yml

@@ -42,7 +42,7 @@ jobs:
       zk_pok_changed: ${{ steps.changed-files.outputs.zk_pok_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -90,7 +90,7 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -108,12 +108,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -144,11 +144,11 @@ jobs:
       - name: Upload parsed results artifact
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
         with:
-          name: ${{ github.sha }}_tfhe_zk_pok_${{ env.BENCH_TYPE }}
+          name: ${{ github.sha }}_tfhe_zk_pok
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_wasm_client.yml

@@ -36,7 +36,7 @@ jobs:
       wasm_bench: ${{ steps.changed-files.outputs.wasm_bench_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -88,7 +88,7 @@ jobs:
         browser: [ chrome, firefox ]
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -106,7 +106,7 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
@@ -116,7 +116,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@0400d5f644dc74513175e3cd8d07132dd4860809 #v4.2.4
+        uses: actions/cache/restore@5a3ec84eff668545956fd18022155c47e93e2684 #v4.2.3
         with:
           path: |
             ~/.nvm
@@ -129,7 +129,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@0400d5f644dc74513175e3cd8d07132dd4860809 #v4.2.4
+        uses: actions/cache/save@5a3ec84eff668545956fd18022155c47e93e2684 #v4.2.3
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |
@@ -185,7 +185,7 @@ jobs:
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/benchmark_zk_pke.yml

@@ -43,7 +43,7 @@ jobs:
       zk_pok_changed: ${{ steps.changed-files.outputs.zk_pok_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -92,8 +92,10 @@ jobs:
 
       - name: Set benchmark types output
         id: set_bench_type
-        run: | # zizmor: ignore[template-injection] this env variable is safe
-          echo "bench_type=${{ toJSON(env.BENCH_TYPE) }}" >> "${GITHUB_OUTPUT}"
+        run: |
+          echo "bench_type=${BENCH_TYPE_OUTPUT}" >> "${GITHUB_OUTPUT}"
+        env:
+          BENCH_TYPE_OUTPUT: ${{ toJSON(env.BENCH_TYPE) }}
 
   setup-instance:
     name: Setup instance (pke-zk-benchmarks)
@@ -132,7 +134,7 @@ jobs:
         bench_type: ${{ fromJSON(needs.prepare-matrix.outputs.bench_type) }}
     steps:
       - name: Checkout tfhe-rs repo with tags
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -150,12 +152,12 @@ jobs:
           SHA: ${{ github.sha }}
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: nightly
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab
@@ -194,11 +196,11 @@ jobs:
       - name: Upload parsed results artifact
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
         with:
-          name: ${{ github.sha }}_integer_zk_${{ matrix.bench_type }}
+          name: ${{ github.sha }}_integer_zk
           path: ${{ env.RESULTS_FILENAME }}
 
       - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: zama-ai/slab
           path: slab

.github/workflows/cargo_build.yml

@@ -29,13 +29,13 @@ jobs:
       fail-fast: false
 
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 
@@ -59,11 +59,6 @@ jobs:
         run: |
           make build_tfhe_csprng
 
-      - name: Build with MSRV
-        if: ${{ contains(matrix.os, 'ubuntu') }}
-        run: |
-          make build_tfhe_msrv
-
       - name: Build Release core
         if: ${{ contains(matrix.os, 'ubuntu') }}
         run: |

.github/workflows/cargo_build_tfhe_fft.yml

@@ -25,7 +25,7 @@ jobs:
       fail-fast: false
 
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}

.github/workflows/cargo_build_tfhe_ntt.yml

@@ -23,7 +23,7 @@ jobs:
         os: [ubuntu-latest, macos-latest, windows-latest]
       fail-fast: false
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}

.github/workflows/cargo_test_fft.yml

@@ -13,7 +13,7 @@ env:
   CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN || secrets.GITHUB_TOKEN }}
 
 concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
+  group: ${{ github.workflow }}-${{ github.head_ref }}
   cancel-in-progress: true
 
 permissions:
@@ -28,7 +28,7 @@ jobs:
       fft_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.fft_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -54,7 +54,7 @@ jobs:
         runner_type: [ ubuntu-latest, macos-latest, windows-latest ]
       fail-fast: false
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -84,7 +84,7 @@ jobs:
       matrix:
         runner_type: [ ubuntu-latest, macos-latest, windows-latest ]
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -108,7 +108,7 @@ jobs:
     if: needs.should-run.outputs.fft_test == 'true'
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}

.github/workflows/cargo_test_ntt.yml

@@ -11,10 +11,9 @@ env:
   CARGO_TERM_COLOR: always
   IS_PULL_REQUEST: ${{ github.event_name == 'pull_request' }}
   CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN || secrets.GITHUB_TOKEN }}
-  SECRETS_AVAILABLE: ${{ secrets.JOB_SECRET != '' }}
 
 concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
+  group: ${{ github.workflow }}-${{ github.head_ref }}
   cancel-in-progress: true
 
 permissions:
@@ -29,10 +28,10 @@ jobs:
       ntt_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.ntt_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
-          persist-credentials: "false"
+          persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Check for file changes
@@ -46,46 +45,18 @@ jobs:
               - tfhe-ntt/**
               - '.github/workflows/cargo_test_ntt.yml'
 
-  setup-instance:
-    needs: should-run
-    if: needs.should-run.outputs.ntt_test == 'true'
-    runs-on: ubuntu-latest
-    outputs:
-      matrix_os: ${{ steps.set-os-matrix.outputs.matrix_os }}
-      runner-name: ${{ steps.start-remote-instance.outputs.label }}
-    steps:
-      - name: Start remote instance
-        id: start-remote-instance
-        if: env.SECRETS_AVAILABLE == 'true'
-        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-small
-
-      - name: Set os matrix
-        id: set-os-matrix
-        env:
-          SLAB_INSTANCE: ${{ steps.start-remote-instance.outputs.label }}
-        run: |
-          INSTANCE_TO_USE="${SLAB_INSTANCE:-ubuntu-latest}"
-          echo "matrix_os=[\"${INSTANCE_TO_USE}\", \"macos-latest\", \"windows-latest\"]" >> "$GITHUB_OUTPUT"
-
   cargo-tests-ntt:
-    needs: [should-run, setup-instance]
+    needs: should-run
     if: needs.should-run.outputs.ntt_test == 'true'
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: ${{fromJson(needs.setup-instance.outputs.matrix_os)}}
+        os: [ ubuntu-latest, macos-latest, windows-latest ]
       fail-fast: false
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
-          persist-credentials: "false"
+          persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install Rust
@@ -101,16 +72,16 @@ jobs:
         run: make test_ntt_no_std
 
   cargo-tests-ntt-nightly:
-    needs: [should-run, setup-instance]
+    needs: should-run
     if: needs.should-run.outputs.ntt_test == 'true'
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: ${{fromJson(needs.setup-instance.outputs.matrix_os)}}
+        os: [ ubuntu-latest, macos-latest, windows-latest ]
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
-          persist-credentials: "false"
+          persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install Rust
@@ -126,7 +97,7 @@ jobs:
         run: make test_ntt_no_std_nightly
 
   cargo-tests-ntt-successful:
-    needs: [should-run, cargo-tests-ntt, cargo-tests-ntt-nightly]
+    needs: [ should-run, cargo-tests-ntt, cargo-tests-ntt-nightly ]
     if: ${{ always() }}
     runs-on: ubuntu-latest
     steps:
@@ -149,28 +120,3 @@ jobs:
         run: |
           echo "Some tfhe-ntt tests failed"
           exit 1
-
-  teardown-instance:
-    name: Teardown instance (cargo-tests-ntt-successful)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
-    needs: [setup-instance, cargo-tests-ntt-successful]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Stop remote instance
-        id: stop-instance
-        if: env.SECRETS_AVAILABLE == 'true'
-        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@e31e87e03dd19038e411e38ae27cbad084a90661
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "Instance teardown (cargo-tests-ntt) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/check_commit.yml

@@ -3,15 +3,14 @@ name: Check commit and PR compliance
 on:
   pull_request:
 
-permissions: {}
+permissions:
+  contents: read
+  pull-requests: read # Permission needed to scan commits in a pull-request
 
 jobs:
   check-commit-pr:
     name: Check commit and PR
     runs-on: ubuntu-latest
-    permissions:
-      contents: read
-      pull-requests: write # Permission needed to scan commits in a pull-request and write issue comment
     steps:
       - name: Check first line
         uses: gsactions/commit-message-checker@16fa2d5de096ae0d35626443bcd24f1e756cafee

.github/workflows/ci_lint.yml

@@ -18,7 +18,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}

.github/workflows/code_coverage.yml

@@ -48,13 +48,13 @@ jobs:
     timeout-minutes: 5760 # 4 days
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 
@@ -90,7 +90,7 @@ jobs:
           make test_shortint_cov
 
       - name: Upload tfhe coverage to Codecov
-        uses: codecov/codecov-action@fdcc8476540edceab3de004e990f80d881c6cc00
+        uses: codecov/codecov-action@18283e04ce6e62d37312384ff67231eb8fd56d24
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
@@ -104,7 +104,7 @@ jobs:
           make test_integer_cov
 
       - name: Upload tfhe coverage to Codecov
-        uses: codecov/codecov-action@fdcc8476540edceab3de004e990f80d881c6cc00
+        uses: codecov/codecov-action@18283e04ce6e62d37312384ff67231eb8fd56d24
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
         with:
           token: ${{ secrets.CODECOV_TOKEN }}

.github/workflows/coprocessor-benchmark-gpu.yml

@@ -1,291 +0,0 @@
-# Run all fhevm coprocessor benchmarks on a GPU instance on Hyperstack and return parsed results to Slab CI bot.
-name: Cuda Coprocessor benchmarks
-
-on:
-  workflow_dispatch:
-  schedule:
-    # Weekly tests @ 1AM
-    - cron: "0 1 * * 6"
-
-permissions:
-  contents: read
-
-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"
-  CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN || secrets.GITHUB_TOKEN }}
-  PROFILE: "multi-h100-sxm5 (n3-H100x8-SXM5)"
-  BENCHMARK_TYPE: "ALL"
-  OPTIMIZATION_TARGET: "throughput"
-  BATCH_SIZE: "5000"
-  SCHEDULING_POLICY: "MAX_PARALLELISM"
-  BENCHMARKS: "erc20"
-  BRANCH_NAME: ${{ github.ref_name }}
-  COMMIT_SHA: ${{ github.sha }}
-  SLAB_SECRET: ${{ secrets.JOB_SECRET }}
-
-jobs:
-  parse-inputs:
-    name: coprocessor-benchmark-gpu/parse-inputs
-    runs-on: ubuntu-latest
-    permissions:
-      contents: 'read'
-    outputs:
-      profile: ${{ steps.parse_profile.outputs.profile }}
-      hardware_name: ${{ steps.parse_hardware_name.outputs.name }}
-    steps:
-      - name: Parse profile
-        id: parse_profile
-        run: |
-          # shellcheck disable=SC2001
-          PROFILE_VAL=$(echo "${PROFILE}" | sed 's|\(.*\)[[:space:]](.*)|\1|')
-          echo "profile=$PROFILE_VAL" >> "${GITHUB_OUTPUT}"
-
-      - name: Parse hardware name
-        id: parse_hardware_name
-        run: |
-          # shellcheck disable=SC2001
-          PROFILE_VAL=$(echo "${PROFILE}" | sed 's|.*[[:space:]](\(.*\))|\1|')
-          echo "name=$PROFILE_VAL" >> "${GITHUB_OUTPUT}"
-
-  setup-instance:
-    name: coprocessor-benchmark-gpu/setup-instance
-    needs: parse-inputs
-    runs-on: ubuntu-latest
-    permissions:
-      contents: 'read'
-    outputs:
-      runner-name: ${{ steps.start-remote-instance.outputs.label }}
-    steps:
-      - name: Start remote instance
-        id: start-remote-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: ${{ needs.parse-inputs.outputs.profile }}
-
-  benchmark:
-    name: coprocessor-benchmark-gpu/benchmark-gpu (bpr)
-    needs: [ parse-inputs, setup-instance ]
-    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
-    continue-on-error: true
-    timeout-minutes: 720  # 12 hours
-    permissions:
-      contents: 'read'
-      packages: 'read'
-    strategy:
-      fail-fast: false
-      # explicit include-based build matrix, of known valid options
-      matrix:
-        include:
-          - os: ubuntu-22.04
-            cuda: "12.8"
-            gcc: 11
-    env:
-      HW_NAME: "${{ needs.parse-inputs.outputs.hardware_name }}"
-
-    steps:
-      - name: Install git LFS
-        run: |
-          sudo apt-get remove -y unattended-upgrades
-          sudo apt-get update
-          sudo apt-get install -y git-lfs protobuf-compiler
-          git lfs install
-
-      - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
-        with:
-          path: tfhe-rs
-          persist-credentials: false
-
-      - name: Check fhEVM and TFHE-rs repos
-        run: |
-          pwd
-          ls
-
-      - name: Checkout fhevm
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
-        with:
-          repository: zama-ai/fhevm
-          persist-credentials: 'false'
-          fetch-depth: 0
-          lfs: true
-          ref: antoniu/use-tfhe-main-benches
-          path: fhevm
-
-      - name: Get benchmark details
-        run: |
-          COMMIT_DATE_ENV=$(git --no-pager show -s --format=%cd --date=iso8601-strict "${COMMIT_SHA}")
-          {
-            echo "BENCH_DATE=$(date --iso-8601=seconds)";
-            echo "COMMIT_DATE=$COMMIT_DATE_ENV";
-            echo "COMMIT_HASH=$(git rev-parse HEAD)";
-          } >> "${GITHUB_ENV}"
-        working-directory: tfhe-rs/
-
-      - name: Check fhEVM and TFHE-rs repos
-        run: |
-          pwd
-          ls
-          mv tfhe-rs fhevm/coprocessor/
-
-      - name: Checkout LFS objects
-        run: git lfs checkout
-        working-directory: fhevm/
-
-      - name: Setup Hyperstack dependencies
-        uses: ./fhevm/.github/actions/gpu_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-          github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
-
-      - name: Install rust
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
-        with:
-          toolchain: nightly
-
-      - name: Install cargo dependencies
-        run: |
-          sudo apt-get install -y protobuf-compiler cmake pkg-config libssl-dev \
-                                  libclang-dev docker-compose-v2 docker.io acl
-          sudo usermod -aG docker "$USER"
-          newgrp docker
-          sudo setfacl --modify user:"$USER":rw /var/run/docker.sock
-          cargo install sqlx-cli
-
-      - name: Install foundry
-        uses: foundry-rs/foundry-toolchain@82dee4ba654bd2146511f85f0d013af94670c4de
-
-      - name: Cache cargo
-        uses: actions/cache@0400d5f644dc74513175e3cd8d07132dd4860809 # v4.2.4
-        with:
-          path: |
-            ~/.cargo/registry
-            ~/.cargo/git
-            target
-          key: ${{ runner.os }}-cargo-${{ hashFiles('**/Cargo.lock') }}
-          restore-keys: ${{ runner.os }}-cargo-
-
-      - name: Login to GitHub Container Registry
-        uses: docker/login-action@184bdaa0721073962dff0199f1fb9940f07167d1 # v3.5.0
-        with:
-          registry: ghcr.io
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-
-      - name: Init database
-        run: make init_db
-        working-directory: fhevm/coprocessor/fhevm-engine/coprocessor
-
-      - name: Use Node.js
-        uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4.4.0
-        with:
-          node-version: 20.x
-
-      - name: Build contracts
-        env:
-          HARDHAT_NETWORK: hardhat
-        run: |
-          ls
-          pwd
-          cp ./host-contracts/.env.example ./host-contracts/.env
-          npm --prefix ./host-contracts ci --include=optional
-          cd host-contracts && npm install && npm run deploy:emptyProxies && npx hardhat compile
-        working-directory: fhevm/
-
-      - name: Profile erc20 no-cmux benchmark on GPU
-        run: |
-          BENCHMARK_BATCH_SIZE="${BATCH_SIZE}" FHEVM_DF_SCHEDULE="${SCHEDULING_POLICY}" BENCHMARK_TYPE="LATENCY" OPTIMIZATION_TARGET="${OPTIMIZATION_TARGET}" make -e "profile_erc20_gpu"
-        working-directory: fhevm/coprocessor/fhevm-engine/coprocessor
-
-      - name: Get nsys profile name
-        id: nsys_profile_name
-        run: echo "profile=coprocessor_profile_$(date +"%Y-%m-%d-%Hh").nsys-rep" >> "$GITHUB_OUTPUT"
-
-      - name: Timestamp nsys profile # zizmor: ignore[template-injection]
-        env:
-          REPORT_NAME: ${{ steps.nsys_profile_name.outputs.profile }}
-        run: |
-          mv report1.nsys-rep ${{ env.REPORT_NAME }}
-        working-directory: fhevm/coprocessor/fhevm-engine/coprocessor
-
-      - name: Upload profile artifact
-        env:
-          REPORT_NAME: ${{ steps.nsys_profile_name.outputs.profile }}
-        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
-        with:
-          name: ${{ env.REPORT_NAME }}
-          path: fhevm/coprocessor/fhevm-engine/coprocessor/${{ env.REPORT_NAME }}
-
-      - name: Run latency benchmark on GPU
-        run: |
-          BENCHMARK_BATCH_SIZE="${BATCH_SIZE}" FHEVM_DF_SCHEDULE="${SCHEDULING_POLICY}" BENCHMARK_TYPE="LATENCY" OPTIMIZATION_TARGET="${OPTIMIZATION_TARGET}" make -e "benchmark_${BENCHMARKS}_gpu"
-        working-directory: fhevm/coprocessor/fhevm-engine/coprocessor
-
-      - name: Run throughput benchmarks on GPU
-        run: |
-          BENCHMARK_BATCH_SIZE="${BATCH_SIZE}" FHEVM_DF_SCHEDULE="${SCHEDULING_POLICY}" BENCHMARK_TYPE="THROUGHPUT_200" OPTIMIZATION_TARGET="${OPTIMIZATION_TARGET}" make -e "benchmark_${BENCHMARKS}_gpu"
-        working-directory: fhevm/coprocessor/fhevm-engine/coprocessor
-
-      - name: Parse results
-        run: |
-          python3 ./ci/benchmark_parser.py coprocessor/fhevm-engine/target/criterion "${RESULTS_FILENAME}" \
-          --database coprocessor \
-          --hardware "${HW_NAME}" \
-          --backend gpu \
-          --project-version "${COMMIT_HASH}" \
-          --branch "${BRANCH_NAME}" \
-          --commit-date "${COMMIT_DATE}" \
-          --bench-date "${BENCH_DATE}" \
-          --walk-subdirs \
-          --crate "coprocessor/fhevm-engine/coprocessor" \
-          --name-suffix "operation_batch_size_${BATCH_SIZE}-schedule_${SCHEDULING_POLICY}-optimization_target_${OPTIMIZATION_TARGET}"
-        working-directory: fhevm/
-
-      - name: Upload parsed results artifact
-        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
-        with:
-          name: ${COMMIT_SHA}_${BENCHMARKS}_${{ needs.parse-inputs.outputs.profile }}
-          path: fhevm/$${{ env.RESULTS_FILENAME }}
-
-      - name: Checkout Slab repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
-        with:
-          repository: zama-ai/slab
-          path: slab
-          persist-credentials: 'false'
-          token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
-
-      - name: Send data to Slab
-        shell: bash
-        env:
-          SLAB_URL: ${{ secrets.SLAB_URL }}
-        run: |
-          python3 slab/scripts/data_sender.py fhevm/"${RESULTS_FILENAME}" "${SLAB_SECRET}" \
-          --slab-url "${SLAB_URL}"
-
-  teardown-instance:
-    name: coprocessor-benchmark-gpu/teardown
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
-    needs: [ setup-instance, benchmark ]
-    runs-on: ubuntu-latest
-    permissions:
-      contents: 'read'
-    steps:
-      - name: Stop remote 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 }}

.github/workflows/csprng_randomness_tests.yml

@@ -60,13 +60,13 @@ jobs:
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 

.github/workflows/data_pr_close.yml

@@ -0,0 +1,142 @@
+name: Close or Merge corresponding PR on the data repo
+
+# When a PR with the data_PR tag is closed or merged, this will close the corresponding PR in the data repo.
+
+env:
+  TARGET_REPO_API_URL: ${{ github.api_url }}/repos/zama-ai/tfhe-backward-compat-data
+  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 }}
+  PR_BRANCH: ${{ github.head_ref || github.ref_name }}
+  CLOSE_TYPE: ${{ github.event.pull_request.merged && 'merge' || 'close' }}
+
+# only trigger on pull request closed events
+on:
+  pull_request:
+    types: [ closed ]
+
+# The same pattern is used for jobs that use the github api:
+# - save the result of the API call in the env var "GH_API_RES". Since the var is multiline
+# we use this trick: https://docs.github.com/en/actions/using-workflows/workflow-commands-for-github-actions#example-of-a-multiline-string
+# - "set +e" will make sure we reach the last "echo EOF" even in case of error
+# - "set -o" pipefail makes one line piped command return the error of the first failure
+# - 'RES="$?"' and 'exit $RES' are used to return the error code if a command failed. Without it, with "set +e"
+# the script will always return 0 because of the "echo EOF".
+
+
+
+permissions: {}
+
+jobs:
+  auto_close_job:
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'data_PR') }}
+    runs-on: ubuntu-latest
+    steps:
+    - name: Find corresponding Pull Request in the data repo
+      run: |
+        {
+          set +e
+          set -o pipefail
+          echo 'TARGET_REPO_PR<<EOF'
+          curl --fail-with-body --no-progress-meter -L -X GET \
+          -H "Accept: application/vnd.github+json" \
+          -H "X-GitHub-Api-Version: 2022-11-28"  \
+          "${TARGET_REPO_API_URL}"/pulls\?head="${REPO_OWNER}":"${PR_BRANCH}" | jq -e '.[0]' | sed 's/null/{ "message": "corresponding PR not found" }/'
+          RES="$?"
+          echo EOF
+        } >> "${GITHUB_ENV}"
+        exit $RES
+      env:
+        REPO_OWNER: ${{ github.repository_owner }}
+
+    - name: Comment on the PR to indicate the reason of the close
+      run: |
+        BODY="'{ \"body\": \"PR ${CLOSE_TYPE}d because the corresponding PR in main repo was ${CLOSE_TYPE}d: ${REPO}#${EVENT_NUMBER}\" }'"
+        {
+          set +e
+          set -o pipefail
+          echo 'GH_API_RES<<EOF'
+          curl --fail-with-body --no-progress-meter -L -X POST \
+          -H "Accept: application/vnd.github+json" \
+          -H "Authorization: Bearer ${TOKEN}" \
+          -H "X-GitHub-Api-Version: 2022-11-28" \
+          "${COMMENTS_URL}" \
+          -d "${BODY}"
+          RES="$?"
+          echo EOF
+        } >> "${GITHUB_ENV}"
+        exit $RES
+      env:
+        REPO: ${{ github.repository }}
+        EVENT_NUMBER: ${{ github.event.number }}
+        COMMENTS_URL: ${{ fromJson(env.TARGET_REPO_PR).comments_url }}
+        TOKEN: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+    - name: Merge the Pull Request in the data repo
+      if: ${{ github.event.pull_request.merged }}
+      run: |
+        {
+          set +e
+          set -o pipefail
+          echo 'GH_API_RES<<EOF'
+          curl --fail-with-body --no-progress-meter -L -X PUT \
+          -H "Accept: application/vnd.github+json" \
+          -H "Authorization: Bearer ${TOKEN}" \
+          -H "X-GitHub-Api-Version: 2022-11-28" \
+          "${TARGET_REPO_PR_URL}"/merge \
+          -d '{ "merge_method": "rebase" }'
+          RES="$?"
+          echo EOF
+        } >> "${GITHUB_ENV}"
+        exit $RES
+      env:
+        TARGET_REPO_PR_URL: ${{ fromJson(env.TARGET_REPO_PR).url }}
+        TOKEN: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+    - name: Close the Pull Request in the data repo
+      if: ${{ !github.event.pull_request.merged }}
+      run: |
+        {
+          set +e
+          set -o pipefail
+          echo 'GH_API_RES<<EOF'
+          curl --fail-with-body --no-progress-meter -L -X PATCH \
+          -H "Accept: application/vnd.github+json" \
+          -H "Authorization: Bearer ${TOKEN}" \
+          -H "X-GitHub-Api-Version: 2022-11-28" \
+          "${TARGET_REPO_PR_URL}" \
+          -d '{ "state": "closed" }'
+          RES="$?"
+          echo EOF
+        } >> "${GITHUB_ENV}"
+        exit $RES
+      env:
+        TARGET_REPO_PR_URL: ${{ fromJson(env.TARGET_REPO_PR).url }}
+        TOKEN: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+    - name: Delete the associated branch in the data repo
+      run: |
+        {
+          set +e
+          set -o pipefail
+          echo 'GH_API_RES<<EOF'
+          curl --fail-with-body --no-progress-meter -L -X DELETE \
+          -H "Accept: application/vnd.github+json" \
+          -H "Authorization: Bearer ${TOKEN}" \
+          -H "X-GitHub-Api-Version: 2022-11-28" \
+          "${TARGET_REPO_API_URL}"/git/refs/heads/"${PR_BRANCH}"
+          RES="$?"
+          echo EOF
+        } >> "${GITHUB_ENV}"
+        exit $RES
+      env:
+        TOKEN: ${{ secrets.FHE_ACTIONS_TOKEN }}
+
+    - name: Slack Notification
+      if: ${{ always() && job.status == 'failure' }}
+      continue-on-error: true
+      uses: rtCamp/action-slack-notify@e31e87e03dd19038e411e38ae27cbad084a90661
+      env:
+        SLACK_COLOR: ${{ job.status }}
+        SLACK_MESSAGE: "Failed to auto-${{ env.CLOSE_TYPE }} PR on data repo: ${{ fromJson(env.GH_API_RES || env.TARGET_REPO_PR).message }}"

.github/workflows/gpu_4090_tests.yml

@@ -39,13 +39,13 @@ jobs:
 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 

.github/workflows/gpu_code_validation_tests.yml

@@ -1,152 +0,0 @@
-# Compile and test tfhe-cuda-backend on an AWS instance
-name: Cuda - CPU Memory Checks
-
-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 }}
-  SLACKIFY_MARKDOWN: true
-  IS_PULL_REQUEST: ${{ github.event_name == 'pull_request' }}
-  PULL_REQUEST_MD_LINK: ""
-  CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN || secrets.GITHUB_TOKEN }}
-  # Secrets will be available only to zama-ai organization members
-  SECRETS_AVAILABLE: ${{ secrets.JOB_SECRET != '' }}
-  EXTERNAL_CONTRIBUTION_RUNNER: "gpu_ubuntu-22.04"
-
-on:
-  # Allows you to run this workflow manually from the Actions tab as an alternative.
-  workflow_dispatch:
-  schedule:
-    # every 3 months
-    - cron: "0 0 1 */3 *"
-
-permissions:
-  contents: read
-
-jobs:
-  setup-instance:
-    name: Setup instance (cuda-tests)
-    runs-on: ubuntu-latest
-    if: github.event_name != 'pull_request' ||
-      (github.event.action == 'labeled' && github.event.label.name == 'approved')
-    outputs:
-      runner-name: ${{ steps.start-remote-instance.outputs.label || steps.start-github-instance.outputs.runner_group }}
-    steps:
-      - name: Start remote instance
-        id: start-remote-instance
-        if: env.SECRETS_AVAILABLE == 'true'
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
-        with:
-          mode: start
-          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
-          slab-url: ${{ secrets.SLAB_BASE_URL }}
-          job-secret: ${{ secrets.JOB_SECRET }}
-          backend: hyperstack
-          profile: gpu-test
-
-      # This instance will be spawned especially for pull-request from forked repository
-      - name: Start GitHub instance
-        id: start-github-instance
-        if: env.SECRETS_AVAILABLE == 'false'
-        run: |
-          echo "runner_group=${EXTERNAL_CONTRIBUTION_RUNNER}" >> "$GITHUB_OUTPUT"
-
-  cuda-tests-linux:
-    name: CUDA Memory Checks tests
-    needs: [ setup-instance ]
-    if: github.event_name != 'pull_request' ||
-      (github.event_name == 'pull_request' && needs.setup-instance.result != 'skipped')
-    concurrency:
-      group: ${{ github.workflow_ref }}
-      cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
-    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
-    timeout-minutes: 5760
-    strategy:
-      fail-fast: false
-      # explicit include-based build matrix, of known valid options
-      matrix:
-        include:
-          - os: ubuntu-22.04
-            cuda: "12.8"
-            gcc: 11 
-    steps:
-      - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
-        with:
-          persist-credentials: 'false'
-          token: ${{ env.CHECKOUT_TOKEN }}
-
-      - name: Setup Hyperstack dependencies
-        uses: ./.github/actions/gpu_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-          github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
-
-      - name: Find tools
-        run: |
-          sudo apt update && sudo apt install -y valgrind 
-          find /usr -executable -name "compute-sanitizer"
-          which valgrind
-
-      - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
-        with:
-          toolchain: stable
-
-      - name: Run memory sanitizer
-        run: |
-          make test_high_level_api_gpu_valgrind
-
-  slack-notify:
-    name: Slack Notification
-    needs: [ setup-instance, cuda-tests-linux ]
-    runs-on: ubuntu-latest
-    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
-    continue-on-error: true
-    steps:
-      - name: Set pull-request URL
-        if: env.SECRETS_AVAILABLE == 'true' && github.event_name == 'pull_request'
-        run: |
-          echo "PULL_REQUEST_MD_LINK=[pull-request](${PR_BASE_URL}${PR_NUMBER}), "  >> "${GITHUB_ENV}"
-        env:
-          PR_BASE_URL: ${{ vars.PR_BASE_URL }}
-          PR_NUMBER: ${{ github.event.pull_request.number }}
-
-      - name: Send message
-        if: env.SECRETS_AVAILABLE == 'true'
-        uses: rtCamp/action-slack-notify@e31e87e03dd19038e411e38ae27cbad084a90661
-        env:
-          SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
-          SLACK_MESSAGE: "GPU Memory Checks tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.PULL_REQUEST_MD_LINK }}[action run](${{ env.ACTION_RUN_URL }}))"
-
-  teardown-instance:
-    name: Teardown instance (cuda-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
-    needs: [ setup-instance, cuda-tests-linux ]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Stop remote instance
-        id: stop-instance
-        if: env.SECRETS_AVAILABLE == 'true'
-        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@e31e87e03dd19038e411e38ae27cbad084a90661
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "Instance teardown (cuda-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_fast_h100_tests.yml

@@ -37,7 +37,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -122,11 +122,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -140,12 +140,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Run core crypto and internal CUDA backend tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_core_crypto_gpu

.github/workflows/gpu_fast_tests.yml

@@ -36,7 +36,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -107,11 +107,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -124,14 +124,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
-
       - name: Run core crypto and internal CUDA backend tests
         run: |
           make test_core_crypto_gpu

.github/workflows/gpu_full_h100_tests.yml

@@ -62,11 +62,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
@@ -79,12 +79,10 @@ jobs:
           gcc-version: ${{ matrix.gcc }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Run core crypto, integer and internal CUDA backend tests
         run: |
           make test_gpu

.github/workflows/gpu_full_multi_gpu_tests.yml

@@ -37,7 +37,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -109,11 +109,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -126,12 +126,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Run multi-bit CUDA integer compression tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_integer_compression_gpu
@@ -139,7 +137,7 @@ jobs:
       # No need to test core_crypto and classic PBS in integer since it's already tested on single GPU.
       - name: Run multi-bit CUDA integer tests
         run: |
-          BIG_TESTS_INSTANCE=TRUE NO_BIG_PARAMS_GPU=TRUE make test_integer_multi_bit_gpu_ci
+          BIG_TESTS_INSTANCE=TRUE make test_integer_multi_bit_gpu_ci
 
       - name: Run user docs tests
         run: |

.github/workflows/gpu_integer_long_run_tests.yml

@@ -11,7 +11,6 @@ env:
   SLACK_USERNAME: ${{ secrets.BOT_USERNAME }}
   SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
   CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN || secrets.GITHUB_TOKEN }}
-  IS_PR: ${{ github.event_name == 'pull_request' }}
 
 on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
@@ -19,8 +18,6 @@ on:
   schedule:
     # Nightly tests will be triggered each evening 8p.m.
     - cron: "0 20 * * *"
-  pull_request:
-
 
 permissions:
   contents: read
@@ -58,12 +55,12 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     timeout-minutes: 4320 # 72 hours
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -75,19 +72,13 @@ jobs:
           gcc-version: ${{ matrix.gcc }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Run tests
         run: |
-          if [[ "${IS_PR}" == "true" ]]; then
-            make test_integer_short_run_gpu
-          else
-            make test_integer_long_run_gpu
-          fi
+          make test_integer_long_run_gpu
 
   slack-notify:
     name: Slack Notification

.github/workflows/gpu_memory_sanitizer.yml

@@ -1,149 +0,0 @@
-# Compile and test tfhe-cuda-backend on an AWS instance
-name: Cuda - GPU Memory Checks
-
-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 }}
-  SLACKIFY_MARKDOWN: true
-  IS_PULL_REQUEST: ${{ github.event_name == 'pull_request' }}
-  PULL_REQUEST_MD_LINK: ""
-  CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN || secrets.GITHUB_TOKEN }}
-  # Secrets will be available only to zama-ai organization members
-  SECRETS_AVAILABLE: ${{ secrets.JOB_SECRET != '' }}
-  EXTERNAL_CONTRIBUTION_RUNNER: "gpu_ubuntu-22.04"
-
-on:
-  # Allows you to run this workflow manually from the Actions tab as an alternative.
-  pull_request:
-    types: [ labeled ]
-  workflow_dispatch:
-
-permissions:
-  contents: read
-
-jobs:
-  setup-instance:
-    name: Setup instance (cuda-tests)
-    runs-on: ubuntu-latest
-    if: github.event_name != 'pull_request' ||
-      (github.event.action == 'labeled' && github.event.label.name == 'approved')
-    outputs:
-      runner-name: ${{ steps.start-remote-instance.outputs.label || steps.start-github-instance.outputs.runner_group }}
-    steps:
-      - name: Start remote instance
-        id: start-remote-instance
-        if: env.SECRETS_AVAILABLE == 'true'
-        uses: zama-ai/slab-github-runner@79939325c3c429837c10d6041e4fd8589d328bac
-        with:
-          mode: start
-          github-token: ${{ secrets.SLAB_ACTION_TOKEN }}
-          slab-url: ${{ secrets.SLAB_BASE_URL }}
-          job-secret: ${{ secrets.JOB_SECRET }}
-          backend: hyperstack
-          profile: gpu-test
-
-      # This instance will be spawned especially for pull-request from forked repository
-      - name: Start GitHub instance
-        id: start-github-instance
-        if: env.SECRETS_AVAILABLE == 'false'
-        run: |
-          echo "runner_group=${EXTERNAL_CONTRIBUTION_RUNNER}" >> "$GITHUB_OUTPUT"
-
-  cuda-tests-linux:
-    name: CUDA Memory Checks tests
-    needs: [ setup-instance ]
-    if: github.event_name != 'pull_request' ||
-      (github.event_name == 'pull_request' && needs.setup-instance.result != 'skipped')
-    concurrency:
-      group: ${{ github.workflow_ref }}
-      cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
-    runs-on: ${{ needs.setup-instance.outputs.runner-name }}
-    timeout-minutes: 240
-    strategy:
-      fail-fast: false
-      # explicit include-based build matrix, of known valid options
-      matrix:
-        include:
-          - os: ubuntu-22.04
-            cuda: "12.8"
-            gcc: 11 
-    steps:
-      - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
-        with:
-          persist-credentials: 'false'
-          token: ${{ env.CHECKOUT_TOKEN }}
-
-      - name: Setup Hyperstack dependencies
-        uses: ./.github/actions/gpu_setup
-        with:
-          cuda-version: ${{ matrix.cuda }}
-          gcc-version: ${{ matrix.gcc }}
-          github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
-
-      - name: Find tools
-        run: |
-          find /usr -executable -name "compute-sanitizer"
-
-      - name: Install latest stable
-        uses: dtolnay/rust-toolchain@b3b07ba8b418998c39fb20f53e8b695cdcc8de1b # zizmor: ignore[stale-action-refs] this action doesn't create releases
-        with:
-          toolchain: stable
-
-      - name: Run memory sanitizer
-        run: |
-          make test_high_level_api_gpu_sanitizer
-
-  slack-notify:
-    name: Slack Notification
-    needs: [ setup-instance, cuda-tests-linux ]
-    runs-on: ubuntu-latest
-    if: ${{ always() && needs.cuda-tests-linux.result != 'skipped' && failure() }}
-    continue-on-error: true
-    steps:
-      - name: Set pull-request URL
-        if: env.SECRETS_AVAILABLE == 'true' && github.event_name == 'pull_request'
-        run: |
-          echo "PULL_REQUEST_MD_LINK=[pull-request](${PR_BASE_URL}${PR_NUMBER}), "  >> "${GITHUB_ENV}"
-        env:
-          PR_BASE_URL: ${{ vars.PR_BASE_URL }}
-          PR_NUMBER: ${{ github.event.pull_request.number }}
-
-      - name: Send message
-        if: env.SECRETS_AVAILABLE == 'true'
-        uses: rtCamp/action-slack-notify@e31e87e03dd19038e411e38ae27cbad084a90661
-        env:
-          SLACK_COLOR: ${{ needs.cuda-tests-linux.result }}
-          SLACK_MESSAGE: "GPU Memory Checks tests finished with status: ${{ needs.cuda-tests-linux.result }}. (${{ env.PULL_REQUEST_MD_LINK }}[action run](${{ env.ACTION_RUN_URL }}))"
-
-  teardown-instance:
-    name: Teardown instance (cuda-tests)
-    if: ${{ always() && needs.setup-instance.result == 'success' }}
-    needs: [ setup-instance, cuda-tests-linux ]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Stop remote instance
-        id: stop-instance
-        if: env.SECRETS_AVAILABLE == 'true'
-        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@e31e87e03dd19038e411e38ae27cbad084a90661
-        env:
-          SLACK_COLOR: ${{ job.status }}
-          SLACK_MESSAGE: "Instance teardown (cuda-tests) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/gpu_pcc.yml

@@ -72,7 +72,7 @@ jobs:
 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -94,7 +94,7 @@ jobs:
           CUDA_VERSION: ${{ matrix.cuda }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 

.github/workflows/gpu_signed_integer_classic_tests.yml

@@ -37,7 +37,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -109,11 +109,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -126,12 +126,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Run signed integer tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_signed_integer_gpu_ci

.github/workflows/gpu_signed_integer_h100_tests.yml

@@ -37,7 +37,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -122,11 +122,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -140,12 +140,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Run signed integer multi-bit tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_signed_integer_multi_bit_gpu_ci

.github/workflows/gpu_signed_integer_tests.yml

@@ -25,6 +25,9 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   pull_request:
+  schedule:
+    # Nightly tests @ 1AM after each work day
+    - cron: "0 1 * * MON-FRI"
 
 permissions:
   contents: read
@@ -38,7 +41,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -110,11 +113,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -127,12 +130,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Should run nightly tests
         if: github.event_name == 'schedule'
         run: |

.github/workflows/gpu_unsigned_integer_classic_tests.yml

@@ -37,7 +37,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -109,11 +109,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -126,12 +126,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Run unsigned integer tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_unsigned_integer_gpu_ci

.github/workflows/gpu_unsigned_integer_h100_tests.yml

@@ -37,7 +37,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -122,11 +122,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11 
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -140,12 +140,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Run unsigned integer multi-bit tests
         run: |
           BIG_TESTS_INSTANCE=TRUE make test_unsigned_integer_multi_bit_gpu_ci

.github/workflows/gpu_unsigned_integer_tests.yml

@@ -25,6 +25,9 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   pull_request:
+  schedule:
+    # Nightly tests @ 1AM after each work day
+    - cron: "0 1 * * MON-FRI"
 
 permissions:
   contents: read
@@ -38,7 +41,7 @@ jobs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -110,11 +113,11 @@ jobs:
       matrix:
         include:
           - os: ubuntu-22.04
-            cuda: "12.8"
+            cuda: "12.2"
             gcc: 11
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}
@@ -127,12 +130,10 @@ jobs:
           github-instance: ${{ env.SECRETS_AVAILABLE == 'false' }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
-      - name: Enable nvidia multi-process service
-        run: |
-          nvidia-cuda-mps-control -d
+
       - name: Should run nightly tests
         if: github.event_name == 'schedule'
         run: |

.github/workflows/hpu_hlapi_tests.yml

@@ -13,7 +13,7 @@ env:
   CHECKOUT_TOKEN: ${{ secrets.REPO_CHECKOUT_TOKEN || secrets.GITHUB_TOKEN }}
 
 concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref }}${{ github.ref == 'refs/heads/main' && github.sha || '' }}
+  group: ${{ github.workflow }}-${{ github.head_ref }}
   cancel-in-progress: true
 
 
@@ -28,7 +28,7 @@ jobs:
       hpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.hpu_any_changed }}
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -50,7 +50,7 @@ jobs:
     if: needs.should-run.outputs.hpu_test == 'true'
     runs-on: large_ubuntu_16
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ env.CHECKOUT_TOKEN }}

.github/workflows/integer_long_run_tests.yml

@@ -51,13 +51,13 @@ jobs:
     timeout-minutes: 4320 # 72 hours
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 

.github/workflows/m1_tests.yml

@@ -40,13 +40,13 @@ jobs:
     timeout-minutes: 720
 
     steps:
-      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: "false"
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 

.github/workflows/make_release.yml

@@ -49,7 +49,7 @@ jobs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -87,10 +87,10 @@ jobs:
     # For provenance of npmjs publish
     permissions:
       contents: read
-      id-token: write # also needed for OIDC token exchange on crates.io
+      id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -100,23 +100,20 @@ jobs:
         run: |
           echo "NPM_TAG=latest" >> "${GITHUB_ENV}"
       - name: Download artifact
-        uses: actions/download-artifact@634f93cb2916e3fdff6788551b99b062d0335ce0 # v5.0.0
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
         with:
           name: crate
           path: target/package
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
       - name: Publish crate.io package
         if: ${{ inputs.push_to_crates }}
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           # DRY_RUN expansion cannot be double quoted when variable contains empty string otherwise cargo publish 
           # would fail. This is safe since DRY_RUN is handled in the env section above.
           # shellcheck disable=SC2086
-          cargo publish -p tfhe ${DRY_RUN}
+          cargo publish -p tfhe --token "${CRATES_TOKEN}" ${DRY_RUN}
 
       - name: Generate hash
         id: published_hash

.github/workflows/make_release_cuda.yml

@@ -60,14 +60,14 @@ jobs:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: "false"
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 
@@ -93,9 +93,6 @@ jobs:
             echo "CUDAHOSTCXX=/usr/bin/g++-${GCC_VERSION}";
             echo "HOME=/home/ubuntu";
           } >> "${GITHUB_ENV}"
-        env:
-          GCC_VERSION: ${{ matrix.gcc }}
-
       - name: Prepare package
         run: |
           cargo package -p tfhe-cuda-backend
@@ -122,9 +119,6 @@ jobs:
     name: Publish CUDA Release
     needs: [setup-instance, package] # for comparing hashes
     runs-on: ${{ needs.setup-instance.outputs.runner-name }}
-    permissions:
-      # Needed for OIDC token exchange on crates.io
-      id-token: write
     strategy:
       fail-fast: false
       # explicit include-based build matrix, of known valid options
@@ -137,7 +131,7 @@ jobs:
       CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }}
     steps:
       - name: Install latest stable
-        uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # zizmor: ignore[stale-action-refs] this action doesn't create releases
+        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
         with:
           toolchain: stable
 
@@ -166,19 +160,15 @@ jobs:
         env:
           GCC_VERSION: ${{ matrix.gcc }}
 
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
-
       - name: Publish crate.io package
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           # DRY_RUN expansion cannot be double quoted when variable contains empty string otherwise cargo publish 
           # would fail. This is safe since DRY_RUN is handled in the env section above.
           # shellcheck disable=SC2086
-          cargo publish -p tfhe-cuda-backend ${DRY_RUN}
+          cargo publish -p tfhe-cuda-backend --token "${CRATES_TOKEN}" ${DRY_RUN}
 
       - name: Generate hash
         id: published_hash

.github/workflows/make_release_hpu.yml

@@ -31,7 +31,7 @@ jobs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -66,30 +66,23 @@ jobs:
     name: Publish tfhe-hpu-backend Release
     runs-on: ubuntu-latest
     needs: [verify_tag, package] # for comparing hashes
-    permissions:
-      # Needed for OIDC token exchange on crates.io
-      id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
-
       - name: Publish crate.io package
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           # DRY_RUN expansion cannot be double quoted when variable contains empty string otherwise cargo publish 
           # would fail. This is safe since DRY_RUN is handled in the env section above.
           # shellcheck disable=SC2086
-          cargo publish -p tfhe-hpu-backend ${DRY_RUN}
+          cargo publish -p tfhe-hpu-backend --token "${CRATES_TOKEN}" ${DRY_RUN}
 
       - name: Generate hash
         id: published_hash

.github/workflows/make_release_tfhe_csprng.yml

@@ -30,7 +30,7 @@ jobs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -67,33 +67,27 @@ jobs:
     name: Publish tfhe-csprng Release
     needs: [verify_tag, package]
     runs-on: ubuntu-latest
-    permissions:
-      # Needed for OIDC token exchange on crates.io
-      id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
       - name: Download artifact
-        uses: actions/download-artifact@634f93cb2916e3fdff6788551b99b062d0335ce0 # v5.0.0
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
         with:
           name: crate-tfhe-csprng
           path: target/package
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
       - name: Publish crate.io package
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           # DRY_RUN expansion cannot be double quoted when variable contains empty string otherwise cargo publish 
           # would fail. This is safe since DRY_RUN is handled in the env section above.
           # shellcheck disable=SC2086
-          cargo publish -p tfhe-csprng ${DRY_RUN}
+          cargo publish -p tfhe-csprng --token "${CRATES_TOKEN}" ${DRY_RUN}
       - name: Generate hash
         id: published_hash
         run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"

.github/workflows/make_release_tfhe_fft.yml

@@ -32,7 +32,7 @@ jobs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -67,30 +67,23 @@ jobs:
     name: Publish tfhe-fft Release
     runs-on: ubuntu-latest
     needs: [verify_tag, package] # for comparing hashes
-    permissions:
-      # Needed for OIDC token exchange on crates.io
-      id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
-
       - name: Publish crate.io package
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           # DRY_RUN expansion cannot be double quoted when variable contains empty string otherwise cargo publish 
           # would fail. This is safe since DRY_RUN is handled in the env section above.
           # shellcheck disable=SC2086
-          cargo publish -p tfhe-fft ${DRY_RUN}
+          cargo publish -p tfhe-fft --token "${CRATES_TOKEN}" ${DRY_RUN}
 
       - name: Generate hash
         id: published_hash

.github/workflows/make_release_tfhe_ntt.yml

@@ -32,7 +32,7 @@ jobs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -67,30 +67,23 @@ jobs:
     name: Publish tfhe-ntt Release
     runs-on: ubuntu-latest
     needs: [verify_tag, package] # for comparing hashes
-    permissions:
-      # Needed for OIDC token exchange on crates.io
-      id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
-
       - name: Publish crate.io package
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           # DRY_RUN expansion cannot be double quoted when variable contains empty string otherwise cargo publish 
           # would fail. This is safe since DRY_RUN is handled in the env section above.
           # shellcheck disable=SC2086
-          cargo publish -p tfhe-ntt ${DRY_RUN}
+          cargo publish -p tfhe-ntt --token "${CRATES_TOKEN}" ${DRY_RUN}
 
       - name: Generate hash
         id: published_hash

.github/workflows/make_release_tfhe_versionable.yml

@@ -26,7 +26,7 @@ jobs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -60,29 +60,23 @@ jobs:
     name: Publish tfhe-versionable-derive Release
     needs: [ verify_tag, package-derive ] # for comparing hashes
     runs-on: ubuntu-latest
-    permissions:
-      # Needed for OIDC token exchange on crates.io
-      id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
       - name: Download artifact
-        uses: actions/download-artifact@634f93cb2916e3fdff6788551b99b062d0335ce0 # v5.0.0
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
         with:
           name: crate-tfhe-versionable-derive
           path: target/package
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
       - name: Publish crate.io package
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
         run: |
-          cargo publish -p tfhe-versionable-derive
+          cargo publish -p tfhe-versionable-derive --token "${CRATES_TOKEN}"
       - name: Generate hash
         id: published_hash
         run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"
@@ -109,7 +103,7 @@ jobs:
       hash: ${{ steps.hash.outputs.hash }}
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
@@ -145,24 +139,21 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
       - name: Download artifact
-        uses: actions/download-artifact@634f93cb2916e3fdff6788551b99b062d0335ce0 # v5.0.0
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
         with:
           name: crate-tfhe-versionable
           path: target/package
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
       - name: Publish crate.io package
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
         run: |
-          cargo publish -p tfhe-versionable
+          cargo publish -p tfhe-versionable --token "${CRATES_TOKEN}"
       - name: Generate hash
         id: published_hash
         run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"

.github/workflows/make_release_zk_pok.yml

@@ -24,7 +24,7 @@ jobs:
         hash: ${{ steps.hash.outputs.hash }}
       steps:
         - name: Checkout
-          uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+          uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
           with:
             fetch-depth: 0
             persist-credentials: 'false'
@@ -64,33 +64,27 @@ jobs:
     name: Publish tfhe-zk-pok Release
     needs: [verify_tag, package] # for comparing hashes
     runs-on: ubuntu-latest
-    permissions:
-      # Needed for OIDC token exchange on crates.io
-      id-token: write
     steps:
       - name: Checkout
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           fetch-depth: 0
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
       - name: Download artifact
-        uses: actions/download-artifact@634f93cb2916e3fdff6788551b99b062d0335ce0 # v5.0.0
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
         with:
           name: crate-zk-pok
           path: target/package
-      - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@e919bc7605cde86df457cf5b93c5e103838bd879 # v1.0.1
-        id: auth
       - name: Publish crate.io package
         env:
-          CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
+          CRATES_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run && '--dry-run' || '' }}
         run: |
           # DRY_RUN expansion cannot be double quoted when variable contains empty string otherwise cargo publish 
           # would fail. This is safe since DRY_RUN is handled in the env section above.
           # shellcheck disable=SC2086
-          cargo publish -p tfhe-zk-pok ${DRY_RUN}
+          cargo publish -p tfhe-zk-pok --token "${CRATES_TOKEN}" ${DRY_RUN}
       - name: Verify hash
         id: published_hash
         run: cd target/package && echo "pub_hash=$(sha256sum ./*.crate | base64 -w0)" >> "${GITHUB_OUTPUT}"

.github/workflows/parameters_check.yml

@@ -19,17 +19,17 @@ jobs:
     runs-on: large_ubuntu_16-22.04
     steps:
       - name: Checkout tfhe-rs
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
       - name: Checkout lattice-estimator
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           repository: malb/lattice-estimator
           path: lattice_estimator
-          ref: '52f4b7a99ae7b5dfd088c5c295070bd38ff0d1e0'
+          ref: 'e80ec6bbbba212428b0e92d0467c18629cf9ed67'
           persist-credentials: 'false'
 
       - name: Install Sage

.github/workflows/sync_on_push.yml

@@ -15,13 +15,13 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout repo
-        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
           fetch-depth: 0
           persist-credentials: 'false'
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
       - name: git-sync
-        uses: valtech-sd/git-sync@e734cfe9485a92e720eac5af8a4555dde5fecf88
+        uses: wei/git-sync@55c6b63b4f21607da0e9877ca9b4d11a29fc6d83
         with:
           source_repo: "zama-ai/tfhe-rs"
           source_branch: "main"

.github/workflows/unverified_prs.yml

@@ -1,26 +0,0 @@
-name: 'Close unverified PRs'
-on:
-  schedule:
-    - cron: '30 1 * * *'
-
-permissions: {}
-
-jobs:
-  stale:
-    runs-on: ubuntu-latest
-    permissions:
-      issues: read
-      pull-requests: write
-    steps:
-      - uses: actions/stale@5bef64f19d7facfb25b37b414482c7164d639639 # v9.1.0
-        with:
-          stale-pr-message: 'This PR is unverified and has been open for 2 days, it will now be closed. If you want to contribute please sign the CLA as indicated by the bot.'
-          days-before-stale: 2
-          days-before-close: 0
-          # We are not interested in suppressing issues so have a currently non existent label
-          # if we ever accept issues to become stale/closable this label will be the signal for that
-          only-issue-labels: can-be-auto-closed
-          # Only unverified PRs are an issue
-          exempt-pr-labels: cla-signed
-          # We don't want people commenting to keep an unverified PR
-          ignore-updates: true

.gitignore

@@ -36,7 +36,7 @@ package-lock.json
 .env
 __pycache__
 
+# Dir used for backward compatibility test data
+# First directive is to ignore symlinks
+tests/tfhe-backward-compat-data
 ci/
-
-# In case someone clones the lattice-estimator locally to verify security
-/lattice-estimator

.linelint.yml

@@ -10,7 +10,6 @@ ignore:
   - keys
   - coverage
   - utils/tfhe-lints/ui/main.stderr
-  - utils/tfhe-backward-compat-data/**/*.ron # ron files are autogenerated
 
 rules:
   # checks if file ends in a newline character

CODEOWNERS

@@ -1,28 +1,18 @@
 # Specifying a path without code owners means that path won't have owners and is akin to a negation
 # i.e. the `core_crypto` dir is owned and needs owner approval/review, but not the `gpu` sub dir
 # See https://docs.github.com/en/repositories/managing-your-repositorys-settings-and-features/customizing-your-repository/about-code-owners#example-of-a-codeowners-file
-
-/backends/tfhe-cuda-backend/            @agnesLeroy
-/backends/tfhe-hpu-backend/             @zama-ai/hardware
-
-/tfhe/examples/hpu                      @zama-ai/hardware
-
 /tfhe/src/core_crypto/                  @IceTDrinker
-/tfhe/src/core_crypto/gpu               @agnesLeroy
-/tfhe/src/core_crypto/hpu               @zama-ai/hardware
+/tfhe/src/core_crypto/gpu
 
 /tfhe/src/shortint/                     @mayeul-zama
 
 /tfhe/src/integer/                      @tmontaigu
-/tfhe/src/integer/gpu                   @agnesLeroy
-/tfhe/src/integer/hpu                   @zama-ai/hardware
+/tfhe/src/integer/gpu
 
 /tfhe/src/high_level_api/               @tmontaigu
 
 /Makefile                               @IceTDrinker @soonum
 
-/mockups/tfhe-hpu-mockup                @zama-ai/hardware
-
 /.github/                               @soonum
 
 /CODEOWNERS                             @IceTDrinker

CONTRIBUTING.md

@@ -170,8 +170,6 @@ On the contrary, these changes are *not* data breaking:
  * Renaming a type (unless it implements the `Named` trait).
  * Adding a variant to the end of an enum.
 
-Historical data from previous TFHE-rs versions are stored inside `utils/tfhe-backward-compat-data`. They are used to check on every PR that backward compatibility has been preserved.
-
 ## Example: adding a field
 
 Suppose you want to add an i32 field to a type named `MyType`. The original type is defined as:

Cargo.toml

@@ -18,7 +18,7 @@ members = [
 ]
 
 exclude = [
-    "utils/tfhe-backward-compat-data",
+    "tests/backward_compatibility_tests",
     "utils/tfhe-lints",
     "apps/trivium",
 ]
@@ -30,10 +30,9 @@ itertools = "0.14"
 num-complex = "0.4"
 pulp = { version = "0.21", default-features = false }
 rand = "0.8"
-rayon = "1.11"
+rayon = "1"
 serde = { version = "1.0", default-features = false }
 wasm-bindgen = "0.2.100"
-getrandom = "0.2.8"
 
 [profile.bench]
 lto = "fat"

Makefile

@@ -22,7 +22,11 @@ BENCH_TYPE?=latency
 BENCH_PARAM_TYPE?=classical
 BENCH_PARAMS_SET?=default
 NODE_VERSION=22.6
-BACKWARD_COMPAT_DATA_DIR=utils/tfhe-backward-compat-data
+BACKWARD_COMPAT_DATA_URL=https://github.com/zama-ai/tfhe-backward-compat-data.git
+BACKWARD_COMPAT_DATA_BRANCH?=$(shell ./scripts/backward_compat_data_version.py)
+BACKWARD_COMPAT_DATA_PROJECT=tfhe-backward-compat-data
+BACKWARD_COMPAT_DATA_DIR=$(BACKWARD_COMPAT_DATA_PROJECT)
+TFHE_SPEC:=tfhe
 WASM_PACK_VERSION="0.13.1"
 # We are kind of hacking the cut here, the version cannot contain a quote '"'
 WASM_BINDGEN_VERSION:=$(shell grep '^wasm-bindgen[[:space:]]*=' Cargo.toml | cut -d '"' -f 2 | xargs)
@@ -96,12 +100,6 @@ install_rs_build_toolchain:
 	( echo "Unable to install $(RS_BUILD_TOOLCHAIN) toolchain, check your rustup installation. \
 	Rustup can be downloaded at https://rustup.rs/" && exit 1 )
 
-.PHONY: install_rs_msrv_toolchain # Install the msrv toolchain
-install_rs_msrv_toolchain:
-	@rustup toolchain install --profile default "$(MIN_RUST_VERSION)" || \
-	( echo "Unable to install $(MIN_RUST_VERSION) toolchain, check your rustup installation. \
-	Rustup can be downloaded at https://rustup.rs/" && exit 1 )
-
 .PHONY: install_build_wasm32_target # Install the wasm32 toolchain used for builds
 install_build_wasm32_target: install_rs_build_toolchain
 	rustup +$(RS_BUILD_TOOLCHAIN) target add wasm32-unknown-unknown || \
@@ -161,24 +159,20 @@ install_tarpaulin: install_rs_build_toolchain
 
 .PHONY: install_cargo_dylint # Install custom tfhe-rs lints
 install_cargo_dylint:
-	cargo install --locked cargo-dylint dylint-link
+	cargo install cargo-dylint dylint-link
 
 .PHONY: install_typos_checker # Install typos checker
 install_typos_checker: install_rs_build_toolchain
 	@typos --version > /dev/null 2>&1 || \
-	cargo $(CARGO_RS_BUILD_TOOLCHAIN) install --locked typos-cli || \
+	cargo $(CARGO_RS_BUILD_TOOLCHAIN) install typos-cli || \
 	( echo "Unable to install typos-cli, unknown error." && exit 1 )
 
 .PHONY: install_zizmor # Install zizmor workflow security checker
 install_zizmor: install_rs_build_toolchain
 	@zizmor --version > /dev/null 2>&1 || \
-	cargo $(CARGO_RS_BUILD_TOOLCHAIN) install --locked zizmor --version ~1.9 || \
+	cargo $(CARGO_RS_BUILD_TOOLCHAIN) install zizmor || \
 	( echo "Unable to install zizmor, unknown error." && exit 1 )
 
-.PHONY: install_cargo_cross # Install cross for big endian tests
-install_cargo_cross: install_rs_build_toolchain
-	cargo $(CARGO_RS_BUILD_TOOLCHAIN) install --locked cross
-
 .PHONY: setup_venv # Setup Python virtualenv for wasm tests
 setup_venv:
 	python3 -m venv venv
@@ -254,9 +248,6 @@ install_mlc: install_rs_build_toolchain
 .PHONY: fmt # Format rust code
 fmt: install_rs_check_toolchain
 	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" fmt
-	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" -Z unstable-options -C $(BACKWARD_COMPAT_DATA_DIR) fmt
-	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" -Z unstable-options -C utils/tfhe-lints fmt
-	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" -Z unstable-options -C apps/trivium fmt
 
 .PHONY: fmt_js # Format javascript code
 fmt_js: check_nvm_installed
@@ -278,9 +269,6 @@ fmt_c_tests:
 .PHONY: check_fmt # Check rust code format
 check_fmt: install_rs_check_toolchain
 	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" fmt --check
-	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" -Z unstable-options -C $(BACKWARD_COMPAT_DATA_DIR) fmt --check
-	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" -Z unstable-options -C utils/tfhe-lints fmt --check
-	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" -Z unstable-options -C apps/trivium fmt --check
 
 .PHONY: check_fmt_c_tests  # Check C tests format
 check_fmt_c_tests:
@@ -308,28 +296,28 @@ clippy_gpu: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=boolean,shortint,integer,internal-keycache,gpu,pbs-stats,extended-types,zk-pok \
 		--all-targets \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: check_gpu # Run check on tfhe with "gpu" enabled
 check_gpu: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" check \
 		--features=boolean,shortint,integer,internal-keycache,gpu,pbs-stats \
 		--all-targets \
-		-p tfhe
+		-p $(TFHE_SPEC)
 
 .PHONY: clippy_hpu # Run clippy lints on tfhe with "hpu" enabled
 clippy_hpu: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=boolean,shortint,integer,internal-keycache,hpu,pbs-stats,extended-types \
 		--all-targets \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_gpu_hpu # Run clippy lints on tfhe with "gpu" and "hpu" enabled
 clippy_gpu_hpu: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=boolean,shortint,integer,internal-keycache,gpu,hpu,pbs-stats,extended-types,zk-pok \
 		--all-targets \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: fix_newline # Fix newline at end of file issues to be UNIX compliant
 fix_newline: check_linelint_installed
@@ -350,55 +338,55 @@ check_workflow_security: install_zizmor
 .PHONY: clippy_core # Run clippy lints on core_crypto with and without experimental features
 clippy_core: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=experimental \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=nightly-avx512 \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=experimental,nightly-avx512 \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=zk-pok \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_boolean # Run clippy lints enabling the boolean features
 clippy_boolean: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=boolean \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_shortint # Run clippy lints enabling the shortint features
 clippy_shortint: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=shortint \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=shortint,experimental \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=zk-pok,shortint \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_integer # Run clippy lints enabling the integer features
 clippy_integer: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=integer \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=integer,experimental \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=integer,experimental,extended-types \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy # Run clippy lints enabling the boolean, shortint, integer
 clippy: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		--features=boolean,shortint,integer \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_rustdoc # Run clippy lints on doctests enabling the boolean, shortint, integer and zk-pok
 clippy_rustdoc: install_rs_check_toolchain
@@ -409,7 +397,7 @@ clippy_rustdoc: install_rs_check_toolchain
 	CARGO_TERM_QUIET=true CLIPPYFLAGS="-D warnings" RUSTDOCFLAGS="--no-run --nocapture --test-builder ./scripts/clippy_driver.sh -Z unstable-options" \
 		cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" test --doc \
 		--features=boolean,shortint,integer,zk-pok,pbs-stats,strings,experimental \
-		-p tfhe
+		-p $(TFHE_SPEC)
 
 .PHONY: clippy_rustdoc_gpu # Run clippy lints on doctests enabling the boolean, shortint, integer and zk-pok
 clippy_rustdoc_gpu: install_rs_check_toolchain
@@ -420,22 +408,22 @@ clippy_rustdoc_gpu: install_rs_check_toolchain
 	CARGO_TERM_QUIET=true CLIPPYFLAGS="-D warnings" RUSTDOCFLAGS="--no-run --nocapture --test-builder ./scripts/clippy_driver.sh -Z unstable-options" \
 		cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" test --doc \
 		--features=boolean,shortint,integer,zk-pok,pbs-stats,strings,experimental,gpu \
-		-p tfhe
+		-p $(TFHE_SPEC)
 
 .PHONY: clippy_c_api # Run clippy lints enabling the boolean, shortint and the C API
 clippy_c_api: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=boolean-c-api,shortint-c-api,high-level-c-api,extended-types \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_js_wasm_api # Run clippy lints enabling the boolean, shortint, integer and the js wasm API
 clippy_js_wasm_api: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=boolean-client-js-wasm-api,shortint-client-js-wasm-api,integer-client-js-wasm-api,high-level-client-js-wasm-api,zk-pok,extended-types \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
 		--features=boolean-client-js-wasm-api,shortint-client-js-wasm-api,integer-client-js-wasm-api,high-level-client-js-wasm-api,extended-types \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_tasks # Run clippy lints on helper tasks crate.
 clippy_tasks: install_rs_check_toolchain
@@ -456,24 +444,20 @@ clippy_ws_tests: install_rs_check_toolchain
 clippy_all_targets: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		--features=boolean,shortint,integer,internal-keycache,zk-pok,strings,pbs-stats,extended-types \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		--features=boolean,shortint,integer,internal-keycache,zk-pok,strings,pbs-stats,extended-types,experimental \
-		-p tfhe -- --no-deps -D warnings
+		-p $(TFHE_SPEC) -- --no-deps -D warnings
 
 .PHONY: clippy_tfhe_csprng # Run clippy lints on tfhe-csprng
 clippy_tfhe_csprng: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		--features=parallel,software-prng -p tfhe-csprng -- --no-deps -D warnings
-	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
-		--features=parallel -p tfhe-csprng -- --no-deps -D warnings
 
 .PHONY: clippy_zk_pok # Run clippy lints on tfhe-zk-pok
 clippy_zk_pok: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		-p tfhe-zk-pok -- --no-deps -D warnings
-	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
-		-p tfhe-zk-pok --features=experimental -- --no-deps -D warnings
 
 .PHONY: clippy_versionable # Run clippy lints on tfhe-versionable
 clippy_versionable: install_rs_check_toolchain
@@ -493,22 +477,10 @@ clippy_param_dedup: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		-p param_dedup -- --no-deps -D warnings
 
-.PHONY: clippy_backward_compat_data # Run clippy lints on tfhe-backward-compat-data
-clippy_backward_compat_data: install_rs_check_toolchain # the toolchain is selected with toolchain.toml
-	@# Some old crates are x86 specific, only run in that case
-	@if uname -a | grep -q x86; then \
-		RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" -Z unstable-options \
-			-C $(BACKWARD_COMPAT_DATA_DIR) clippy --all-targets \
-			-- --no-deps -D warnings; \
-	else \
-		echo "Cannot run clippy for backward compat crate on non x86 platform for now."; \
-	fi
-
 .PHONY: clippy_all # Run all clippy targets
 clippy_all: clippy_rustdoc clippy clippy_boolean clippy_shortint clippy_integer clippy_all_targets \
 clippy_c_api clippy_js_wasm_api clippy_tasks clippy_core clippy_tfhe_csprng clippy_zk_pok clippy_trivium \
-clippy_versionable clippy_tfhe_lints clippy_ws_tests clippy_bench clippy_param_dedup \
-clippy_backward_compat_data
+clippy_versionable clippy_tfhe_lints clippy_ws_tests clippy_bench clippy_param_dedup
 
 .PHONY: clippy_fast # Run main clippy targets
 clippy_fast: clippy_rustdoc clippy clippy_all_targets clippy_c_api clippy_js_wasm_api clippy_tasks \
@@ -524,12 +496,6 @@ clippy_hpu_backend: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
 		-p tfhe-hpu-backend -- --no-deps -D warnings
 
-.PHONY: clippy_hpu_mockup # Run clippy lints on tfhe-hpu-mockup
-clippy_hpu_mockup: install_rs_check_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy \
-		--all-targets \
-		-p tfhe-hpu-mockup -- --no-deps -D warnings
-
 .PHONY: check_rust_bindings_did_not_change # Check rust bindings are up to date for tfhe-cuda-backend
 check_rust_bindings_did_not_change:
 	cargo build -p tfhe-cuda-backend && "$(MAKE)" fmt_gpu && \
@@ -542,79 +508,67 @@ check_rust_bindings_did_not_change:
 tfhe_lints: install_cargo_dylint
 	RUSTFLAGS="$(RUSTFLAGS)" cargo dylint --all -p tfhe --no-deps -- \
 		--features=boolean,shortint,integer,strings,zk-pok
-	RUSTFLAGS="$(RUSTFLAGS)" cargo dylint --all -p tfhe-zk-pok --no-deps -- \
-		--features=experimental
-
 
 .PHONY: build_core # Build core_crypto without experimental features
 build_core: install_rs_build_toolchain install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		-p tfhe
+		-p $(TFHE_SPEC)
 	@if [[ "$(AVX512_SUPPORT)" == "ON" ]]; then \
 		RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-			--features=nightly-avx512 -p tfhe; \
+			--features=nightly-avx512 -p $(TFHE_SPEC); \
 	fi
 
 .PHONY: build_core_experimental # Build core_crypto with experimental features
 build_core_experimental: install_rs_build_toolchain install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		--features=experimental -p tfhe
+		--features=experimental -p $(TFHE_SPEC)
 	@if [[ "$(AVX512_SUPPORT)" == "ON" ]]; then \
 		RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-			--features=experimental,nightly-avx512 -p tfhe; \
+			--features=experimental,nightly-avx512 -p $(TFHE_SPEC); \
 	fi
 
 .PHONY: build_boolean # Build with boolean enabled
 build_boolean: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		--features=boolean -p tfhe --all-targets
+		--features=boolean -p $(TFHE_SPEC) --all-targets
 
 .PHONY: build_shortint # Build with shortint enabled
 build_shortint: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		--features=shortint -p tfhe --all-targets
+		--features=shortint -p $(TFHE_SPEC) --all-targets
 
 .PHONY: build_integer # Build with integer enabled
 build_integer: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		--features=integer -p tfhe --all-targets
+		--features=integer -p $(TFHE_SPEC) --all-targets
 
 .PHONY: build_tfhe_full # Build with boolean, shortint and integer enabled
 build_tfhe_full: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		--features=boolean,shortint,integer -p tfhe --all-targets
+		--features=boolean,shortint,integer -p $(TFHE_SPEC) --all-targets
 
 .PHONY: build_tfhe_coverage # Build with test coverage enabled
 build_tfhe_coverage: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS) --cfg tarpaulin" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
-		--features=boolean,shortint,integer,internal-keycache -p tfhe --tests
-
-# As of 05/08/2025 this is the set of features that can be easily compiled without additional
-# toolkits
-.PHONY: build_tfhe_msrv # Build with msrv compiler
-build_tfhe_msrv: install_rs_build_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" cargo +$(MIN_RUST_VERSION) build --profile dev \
-		--features=boolean,extended-types,hpu,hpu-debug \
-		--features=hpu-v80,integer,noise-asserts \
-		--features=pbs-stats,shortint,strings,zk-pok -p tfhe
+		--features=boolean,shortint,integer,internal-keycache -p $(TFHE_SPEC) --tests
 
 .PHONY: build_c_api # Build the C API for boolean, shortint and integer
 build_c_api: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
 		--features=boolean-c-api,shortint-c-api,high-level-c-api,zk-pok,extended-types \
-		-p tfhe
+		-p $(TFHE_SPEC)
 
 .PHONY: build_c_api_gpu # Build the C API for boolean, shortint and integer
 build_c_api_gpu: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
 		--features=boolean-c-api,shortint-c-api,high-level-c-api,zk-pok,extended-types,gpu \
-		-p tfhe
+		-p $(TFHE_SPEC)
 
 .PHONY: build_c_api_experimental_deterministic_fft # Build the C API for boolean, shortint and integer with experimental deterministic FFT
 build_c_api_experimental_deterministic_fft: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) build --profile $(CARGO_PROFILE) \
 		--features=boolean-c-api,shortint-c-api,high-level-c-api,zk-pok,experimental-force_fft_algo_dif4 \
-		-p tfhe
+		-p $(TFHE_SPEC)
 
 .PHONY: build_web_js_api # Build the js API targeting the web browser
 build_web_js_api: install_rs_build_toolchain install_wasm_pack
@@ -649,10 +603,10 @@ build_tfhe_csprng: install_rs_build_toolchain
 .PHONY: test_core_crypto # Run the tests of the core_crypto module including experimental ones
 test_core_crypto: install_rs_build_toolchain install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=experimental,zk-pok -p tfhe -- core_crypto::
+		--features=experimental,zk-pok -p $(TFHE_SPEC) -- core_crypto::
 	@if [[ "$(AVX512_SUPPORT)" == "ON" ]]; then \
 		RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-			--features=experimental,zk-pok,nightly-avx512 -p tfhe -- core_crypto::; \
+			--features=experimental,zk-pok,nightly-avx512 -p $(TFHE_SPEC) -- core_crypto::; \
 	fi
 
 .PHONY: test_core_crypto_cov # Run the tests of the core_crypto module with code coverage
@@ -661,13 +615,13 @@ test_core_crypto_cov: install_rs_build_toolchain install_rs_check_toolchain inst
 		--out xml --output-dir coverage/core_crypto --line --engine llvm --timeout 500 \
 		--implicit-test-threads $(COVERAGE_EXCLUDED_FILES) \
 		--features=experimental,internal-keycache \
-		-p tfhe -- core_crypto::
+		-p $(TFHE_SPEC) -- core_crypto::
 	@if [[ "$(AVX512_SUPPORT)" == "ON" ]]; then \
 		RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) tarpaulin --profile $(CARGO_PROFILE) \
 			--out xml --output-dir coverage/core_crypto_avx512 --line --engine llvm --timeout 500 \
 			--implicit-test-threads $(COVERAGE_EXCLUDED_FILES) \
 			--features=experimental,internal-keycache,nightly-avx512 \
-			-p tfhe -- -Z unstable-options --report-time core_crypto::; \
+			-p $(TFHE_SPEC) -- -Z unstable-options --report-time core_crypto::; \
 	fi
 
 .PHONY: test_cuda_backend # Run the internal tests of the CUDA backend
@@ -684,48 +638,16 @@ test_gpu: test_core_crypto_gpu test_integer_gpu test_cuda_backend
 .PHONY: test_core_crypto_gpu # Run the tests of the core_crypto module including experimental on the gpu backend
 test_core_crypto_gpu: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=gpu -p tfhe -- core_crypto::gpu::
+		--features=gpu -p $(TFHE_SPEC) -- core_crypto::gpu::
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --doc --profile $(CARGO_PROFILE) \
-		--features=gpu -p tfhe -- core_crypto::gpu::
+		--features=gpu -p $(TFHE_SPEC) -- core_crypto::gpu::
 
 .PHONY: test_integer_gpu # Run the tests of the integer module including experimental on the gpu backend
 test_integer_gpu: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=integer,gpu -p tfhe -- integer::gpu::server_key:: --test-threads=2
+		--features=integer,gpu -p $(TFHE_SPEC) -- integer::gpu::server_key:: --test-threads=4
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --doc --profile $(CARGO_PROFILE) \
-		--features=integer,gpu -p tfhe -- integer::gpu::server_key::
-
-.PHONY: test_integer_gpu_debug # Run the tests of the integer module with Debug flags for CUDA
-test_integer_gpu_debug: install_rs_build_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile release_lto_off \
-		--features=integer,gpu-debug -vv -p tfhe -- integer::gpu::server_key:: --test-threads=1 --nocapture
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --doc --profile release_lto_off \
-		--features=integer,gpu-debug -p tfhe -- integer::gpu::server_key::
-
-.PHONY: test_high_level_api_gpu_valgrind # Run the tests of the integer module with Debug flags for CUDA
-test_high_level_api_gpu_valgrind: install_rs_build_toolchain install_cargo_nextest
-	export RUSTFLAGS="-C target-cpu=x86-64" && \
-	export CARGO_RS_BUILD_TOOLCHAIN="$(CARGO_RS_BUILD_TOOLCHAIN)" && \
-	export TFHE_SPEC="tfhe" && \
-	export CARGO_PROFILE="$(CARGO_PROFILE)" &&	scripts/check_memory_errors.sh --cpu
-
-.PHONY: test_high_level_api_gpu_sanitizer # Run the tests of the integer module with Debug flags for CUDA
-test_high_level_api_gpu_sanitizer: install_rs_build_toolchain install_cargo_nextest
-	export RUSTFLAGS="-C target-cpu=x86-64" && \
-	export CARGO_RS_BUILD_TOOLCHAIN="$(CARGO_RS_BUILD_TOOLCHAIN)" && \
-	export TFHE_SPEC="tfhe" && \
-	export CARGO_PROFILE="$(CARGO_PROFILE)" &&	scripts/check_memory_errors.sh --gpu
-
-.PHONY: test_integer_hl_test_gpu_check_warnings
-test_integer_hl_test_gpu_check_warnings: install_rs_build_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build \
-		--features=integer,internal-keycache,gpu-debug,zk-pok -vv -p tfhe &> /tmp/gpu_compile_output
-	WARNINGS=$$(cat /tmp/gpu_compile_output | grep ": warning #" | grep "\[tfhe-cuda-backend" | grep -v "inline qualifier" || true) && \
-	if [[ "$${WARNINGS}" != "" ]]; then \
-	    echo "FAILING BECAUSE CUDA COMPILATION WARNINGS WERE DETECTED: " && \
-		echo "$${WARNINGS}" && exit 1; \
-	fi
-
+		--features=integer,gpu -p $(TFHE_SPEC) -- integer::gpu::server_key::
 
 .PHONY: test_integer_long_run_gpu # Run the long run integer tests on the gpu backend
 test_integer_long_run_gpu: install_rs_check_toolchain install_cargo_nextest
@@ -733,27 +655,21 @@ test_integer_long_run_gpu: install_rs_check_toolchain install_cargo_nextest
 	LONG_TESTS=TRUE \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_BUILD_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --avx512-support "$(AVX512_SUPPORT)" \
-		--tfhe-package "tfhe" --backend "gpu"
-
-.PHONY: test_integer_short_run_gpu # Run the long run integer tests on the gpu backend
-test_integer_short_run_gpu: install_rs_check_toolchain install_cargo_nextest
-	TFHE_RS_TEST_LONG_TESTS_MINIMAL=TRUE \
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=integer,gpu -p tfhe -- integer::gpu::server_key::radix::tests_long_run::test_random_op_sequence integer::gpu::server_key::radix::tests_long_run::test_signed_random_op_sequence --test-threads=1 --nocapture
+		--tfhe-package "$(TFHE_SPEC)" --backend "gpu"
 
 .PHONY: test_integer_compression
 test_integer_compression: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=integer -p tfhe -- integer::ciphertext::compressed_ciphertext_list::tests::
+		--features=integer -p $(TFHE_SPEC) -- integer::ciphertext::compressed_ciphertext_list::tests::
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --doc --profile $(CARGO_PROFILE) \
-		--features=integer -p tfhe -- integer::ciphertext::compress
+		--features=integer -p $(TFHE_SPEC) -- integer::ciphertext::compress
 
 .PHONY: test_integer_compression_gpu
 test_integer_compression_gpu: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=integer,gpu -p tfhe -- integer::gpu::ciphertext::compressed_ciphertext_list::tests::
+		--features=integer,gpu -p $(TFHE_SPEC) -- integer::gpu::ciphertext::compressed_ciphertext_list::tests::
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --doc --profile $(CARGO_PROFILE) \
-		--features=integer,gpu -p tfhe -- integer::gpu::ciphertext::compress
+		--features=integer,gpu -p $(TFHE_SPEC) -- integer::gpu::ciphertext::compress
 
 .PHONY: test_integer_gpu_ci # Run the tests for integer ci on gpu backend
 test_integer_gpu_ci: install_rs_check_toolchain install_cargo_nextest
@@ -762,7 +678,7 @@ test_integer_gpu_ci: install_rs_check_toolchain install_cargo_nextest
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --backend "gpu" \
-		--tfhe-package "tfhe"
+		--tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_unsigned_integer_gpu_ci # Run the tests for unsigned integer ci on gpu backend
 test_unsigned_integer_gpu_ci: install_rs_check_toolchain install_cargo_nextest
@@ -771,7 +687,7 @@ test_unsigned_integer_gpu_ci: install_rs_check_toolchain install_cargo_nextest
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --backend "gpu" \
-		--unsigned-only --tfhe-package "tfhe"
+		--unsigned-only --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_signed_integer_gpu_ci # Run the tests for signed integer ci on gpu backend
 test_signed_integer_gpu_ci: install_rs_check_toolchain install_cargo_nextest
@@ -780,7 +696,7 @@ test_signed_integer_gpu_ci: install_rs_check_toolchain install_cargo_nextest
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --backend "gpu" \
-		--signed-only --tfhe-package "tfhe"
+		--signed-only --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_integer_multi_bit_gpu_ci # Run the tests for integer ci on gpu backend running only multibit tests
 test_integer_multi_bit_gpu_ci: install_rs_check_toolchain install_cargo_nextest
@@ -789,7 +705,7 @@ test_integer_multi_bit_gpu_ci: install_rs_check_toolchain install_cargo_nextest
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --backend "gpu" \
-		--tfhe-package "tfhe"
+		--tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_unsigned_integer_multi_bit_gpu_ci # Run the tests for unsigned integer ci on gpu backend running only multibit tests
 test_unsigned_integer_multi_bit_gpu_ci: install_rs_check_toolchain install_cargo_nextest
@@ -798,7 +714,7 @@ test_unsigned_integer_multi_bit_gpu_ci: install_rs_check_toolchain install_cargo
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --backend "gpu" \
-		--unsigned-only --tfhe-package "tfhe"
+		--unsigned-only --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_signed_integer_multi_bit_gpu_ci # Run the tests for signed integer ci on gpu backend running only multibit tests
 test_signed_integer_multi_bit_gpu_ci: install_rs_check_toolchain install_cargo_nextest
@@ -807,34 +723,34 @@ test_signed_integer_multi_bit_gpu_ci: install_rs_check_toolchain install_cargo_n
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --backend "gpu" \
-		--signed-only --tfhe-package "tfhe"
+		--signed-only --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_integer_hpu_ci # Run the tests for integer ci on hpu backend
 test_integer_hpu_ci: install_rs_check_toolchain install_cargo_nextest
-	cargo test --release -p tfhe --features hpu-v80 --test hpu
+	cargo test --release -p $(TFHE_SPEC) --features hpu-v80 --test hpu
 
 .PHONY: test_integer_hpu_mockup_ci # Run the tests for integer ci on hpu backend and mockup
 test_integer_hpu_mockup_ci: install_rs_check_toolchain install_cargo_nextest
 	source ./setup_hpu.sh --config sim ; \
 	cargo build --release --bin hpu_mockup; \
-	coproc target/release/hpu_mockup --params mockups/tfhe-hpu-mockup/params/tuniform_64b_pfail64_psi64.toml > mockup.log; \
+    coproc target/release/hpu_mockup --params mockups/tfhe-hpu-mockup/params/tuniform_64b_pfail64_psi64.toml > mockup.log; \
 	HPU_TEST_ITER=1 \
-	cargo test --profile devo -p tfhe --features hpu --test hpu -- u32 && \
+	cargo test --profile devo -p $(TFHE_SPEC) --features hpu --test hpu -- u32 && \
 	kill %1
 
 .PHONY: test_integer_hpu_mockup_ci_fast # Run the quick tests for integer ci on hpu backend and mockup.
 test_integer_hpu_mockup_ci_fast: install_rs_check_toolchain install_cargo_nextest
 	source ./setup_hpu.sh --config sim ; \
 	cargo build --profile devo --bin hpu_mockup; \
-	coproc target/devo/hpu_mockup --params mockups/tfhe-hpu-mockup/params/tuniform_64b_fast.toml > mockup.log; \
+    coproc target/devo/hpu_mockup --params mockups/tfhe-hpu-mockup/params/tuniform_64b_fast.toml > mockup.log; \
 	HPU_TEST_ITER=1 \
-	cargo test --profile devo -p tfhe --features hpu --test hpu -- u32 && \
+	cargo test --profile devo -p $(TFHE_SPEC) --features hpu --test hpu -- u32 && \
 	kill %1
 
 .PHONY: test_boolean # Run the tests of the boolean module
 test_boolean: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=boolean -p tfhe -- boolean::
+		--features=boolean -p $(TFHE_SPEC) -- boolean::
 
 .PHONY: test_boolean_cov # Run the tests of the boolean module with code coverage
 test_boolean_cov: install_rs_check_toolchain install_tarpaulin
@@ -842,13 +758,13 @@ test_boolean_cov: install_rs_check_toolchain install_tarpaulin
 		--out xml --output-dir coverage/boolean --line --engine llvm --timeout 500 \
 		$(COVERAGE_EXCLUDED_FILES) \
 		--features=boolean,internal-keycache \
-		-p tfhe -- -Z unstable-options --report-time boolean::
+		-p $(TFHE_SPEC) -- -Z unstable-options --report-time boolean::
 
 .PHONY: test_c_api_rs # Run the rust tests for the C API
 test_c_api_rs: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
 		--features=boolean-c-api,shortint-c-api,high-level-c-api \
-		-p tfhe \
+		-p $(TFHE_SPEC) \
 		c_api
 
 .PHONY: test_c_api_c # Run the C tests for the C API
@@ -867,19 +783,19 @@ test_shortint_ci: install_rs_build_toolchain install_cargo_nextest
 	BIG_TESTS_INSTANCE="$(BIG_TESTS_INSTANCE)" \
 	FAST_TESTS="$(FAST_TESTS)" \
 		./scripts/shortint-tests.sh --rust-toolchain $(CARGO_RS_BUILD_TOOLCHAIN) \
-		--cargo-profile "$(CARGO_PROFILE)" --tfhe-package "tfhe"
+		--cargo-profile "$(CARGO_PROFILE)" --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_shortint_multi_bit_ci # Run the tests for shortint ci running only multibit tests
 test_shortint_multi_bit_ci: install_rs_build_toolchain install_cargo_nextest
 	BIG_TESTS_INSTANCE="$(BIG_TESTS_INSTANCE)" \
 	FAST_TESTS="$(FAST_TESTS)" \
 		./scripts/shortint-tests.sh --rust-toolchain $(CARGO_RS_BUILD_TOOLCHAIN) \
-		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --tfhe-package "tfhe"
+		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_shortint # Run all the tests for shortint
 test_shortint: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=shortint,internal-keycache -p tfhe -- shortint::
+		--features=shortint,internal-keycache -p $(TFHE_SPEC) -- shortint::
 
 .PHONY: test_shortint_cov # Run the tests of the shortint module with code coverage
 test_shortint_cov: install_rs_check_toolchain install_tarpaulin
@@ -887,7 +803,7 @@ test_shortint_cov: install_rs_check_toolchain install_tarpaulin
 		--out xml --output-dir coverage/shortint --line --engine llvm --timeout 500 \
 		$(COVERAGE_EXCLUDED_FILES) \
 		--features=shortint,internal-keycache \
-		-p tfhe -- -Z unstable-options --report-time shortint::
+		-p $(TFHE_SPEC) -- -Z unstable-options --report-time shortint::
 
 .PHONY: test_integer_ci # Run the tests for integer ci
 test_integer_ci: install_rs_check_toolchain install_cargo_nextest
@@ -896,7 +812,7 @@ test_integer_ci: install_rs_check_toolchain install_cargo_nextest
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --avx512-support "$(AVX512_SUPPORT)" \
-		--tfhe-package "tfhe"
+		--tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_unsigned_integer_ci # Run the tests for unsigned integer ci
 test_unsigned_integer_ci: install_rs_check_toolchain install_cargo_nextest
@@ -905,7 +821,7 @@ test_unsigned_integer_ci: install_rs_check_toolchain install_cargo_nextest
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --avx512-support "$(AVX512_SUPPORT)" \
-		--unsigned-only --tfhe-package "tfhe"
+		--unsigned-only --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_signed_integer_ci # Run the tests for signed integer ci
 test_signed_integer_ci: install_rs_check_toolchain install_cargo_nextest
@@ -914,7 +830,7 @@ test_signed_integer_ci: install_rs_check_toolchain install_cargo_nextest
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --avx512-support "$(AVX512_SUPPORT)" \
-		--signed-only --tfhe-package "tfhe"
+		--signed-only --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_integer_multi_bit_ci # Run the tests for integer ci running only multibit tests
 test_integer_multi_bit_ci: install_rs_check_toolchain install_cargo_nextest
@@ -923,7 +839,7 @@ test_integer_multi_bit_ci: install_rs_check_toolchain install_cargo_nextest
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --avx512-support "$(AVX512_SUPPORT)" \
-		--tfhe-package "tfhe"
+		--tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_unsigned_integer_multi_bit_ci # Run the tests for unsigned integer ci running only multibit tests
 test_unsigned_integer_multi_bit_ci: install_rs_check_toolchain install_cargo_nextest
@@ -932,7 +848,7 @@ test_unsigned_integer_multi_bit_ci: install_rs_check_toolchain install_cargo_nex
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --avx512-support "$(AVX512_SUPPORT)" \
-		--unsigned-only --tfhe-package "tfhe"
+		--unsigned-only --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_signed_integer_multi_bit_ci # Run the tests for signed integer ci running only multibit tests
 test_signed_integer_multi_bit_ci: install_rs_check_toolchain install_cargo_nextest
@@ -941,7 +857,7 @@ test_signed_integer_multi_bit_ci: install_rs_check_toolchain install_cargo_nexte
 	NIGHTLY_TESTS="$(NIGHTLY_TESTS)" \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_CHECK_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --multi-bit --avx512-support "$(AVX512_SUPPORT)" \
-		--signed-only --tfhe-package "tfhe"
+		--signed-only --tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_integer_long_run # Run the long run integer tests
 test_integer_long_run: install_rs_check_toolchain install_cargo_nextest
@@ -949,31 +865,22 @@ test_integer_long_run: install_rs_check_toolchain install_cargo_nextest
 	LONG_TESTS=TRUE \
 		./scripts/integer-tests.sh --rust-toolchain $(CARGO_RS_BUILD_TOOLCHAIN) \
 		--cargo-profile "$(CARGO_PROFILE)" --avx512-support "$(AVX512_SUPPORT)" \
-		--tfhe-package "tfhe"
-
-.PHONY: test_noise_check # Run dedicated noise and pfail check tests
-test_noise_check: install_rs_check_toolchain
-	@# First run the sanity checks to make sure the atomic patterns are correct
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=boolean,shortint,integer,nightly-avx512 -p tfhe -- sanity_check
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=boolean,shortint,integer,nightly-avx512 -p tfhe -- noise_check \
-		--test-threads=1 --nocapture
+		--tfhe-package "$(TFHE_SPEC)"
 
 .PHONY: test_safe_serialization # Run the tests for safe serialization
 test_safe_serialization: install_rs_build_toolchain install_cargo_nextest
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=boolean,shortint,integer,internal-keycache -p tfhe -- safe_serialization::
+		--features=boolean,shortint,integer,internal-keycache -p $(TFHE_SPEC) -- safe_serialization::
 
 .PHONY: test_zk # Run the tests for the zk module of the TFHE-rs crate
 test_zk: install_rs_build_toolchain install_cargo_nextest
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=shortint,zk-pok -p tfhe -- zk::
+		--features=shortint,zk-pok -p $(TFHE_SPEC) -- zk::
 
 .PHONY: test_integer # Run all the tests for integer
 test_integer: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=integer,internal-keycache -p tfhe -- integer::
+		--features=integer,internal-keycache -p $(TFHE_SPEC) -- integer::
 
 .PHONY: test_integer_cov # Run the tests of the integer module with code coverage
 test_integer_cov: install_rs_check_toolchain install_tarpaulin
@@ -982,17 +889,17 @@ test_integer_cov: install_rs_check_toolchain install_tarpaulin
 		--implicit-test-threads \
 		--exclude-files $(COVERAGE_EXCLUDED_FILES) \
 		--features=integer,internal-keycache \
-		-p tfhe -- -Z unstable-options --report-time integer::
+		-p $(TFHE_SPEC) -- -Z unstable-options --report-time integer::
 
 .PHONY: test_high_level_api # Run all the tests for high_level_api
 test_high_level_api: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=boolean,shortint,integer,internal-keycache,zk-pok,strings -p tfhe \
+		--features=boolean,shortint,integer,internal-keycache,zk-pok,strings -p $(TFHE_SPEC) \
 		-- high_level_api::
 
 test_high_level_api_gpu: install_rs_build_toolchain install_cargo_nextest
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) nextest run --cargo-profile $(CARGO_PROFILE) \
-		--test-threads=4 --features=integer,internal-keycache,gpu,zk-pok -p tfhe \
+		--test-threads=4 --features=integer,internal-keycache,gpu,zk-pok -p $(TFHE_SPEC) \
 		-E "test(/high_level_api::.*gpu.*/)"
 
 test_high_level_api_hpu: install_rs_build_toolchain install_cargo_nextest
@@ -1000,13 +907,13 @@ ifeq ($(HPU_CONFIG), v80)
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) nextest run --cargo-profile $(CARGO_PROFILE) \
 		--build-jobs=$(CARGO_BUILD_JOBS) \
 		--test-threads=1 \
-		--features=integer,internal-keycache,hpu,hpu-v80 -p tfhe \
+		--features=integer,internal-keycache,hpu,hpu-v80 -p $(TFHE_SPEC) \
 		-E "test(/high_level_api::.*hpu.*/)"
 else
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) nextest run --cargo-profile $(CARGO_PROFILE) \
 		--build-jobs=$(CARGO_BUILD_JOBS) \
 		--test-threads=1 \
-		--features=integer,internal-keycache,hpu -p tfhe \
+		--features=integer,internal-keycache,hpu -p $(TFHE_SPEC) \
 		-E "test(/high_level_api::.*hpu.*/)"
 endif
 
@@ -1014,7 +921,7 @@ endif
 .PHONY: test_strings # Run the tests for strings ci
 test_strings: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		--features=shortint,integer,strings -p tfhe \
+		--features=shortint,integer,strings -p $(TFHE_SPEC) \
 		-- strings::
 
 
@@ -1022,24 +929,24 @@ test_strings: install_rs_build_toolchain
 test_user_doc: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) --doc \
 		--features=boolean,shortint,integer,internal-keycache,pbs-stats,zk-pok,strings \
-		-p tfhe \
+		-p $(TFHE_SPEC) \
 		-- test_user_docs::
 
 .PHONY: test_user_doc_gpu # Run tests for GPU from the .md documentation
 test_user_doc_gpu: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) --doc \
-		--features=internal-keycache,integer,zk-pok,gpu -p tfhe \
+		--features=internal-keycache,integer,zk-pok,gpu -p $(TFHE_SPEC) \
 		-- test_user_docs::
 
 .PHONY: test_user_doc_hpu # Run tests for HPU from the .md documentation
 test_user_doc_hpu: install_rs_build_toolchain
 ifeq ($(HPU_CONFIG), v80)
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) --doc \
-		--features=internal-keycache,integer,hpu,hpu-v80 -p tfhe \
+		--features=internal-keycache,integer,hpu,hpu-v80 -p $(TFHE_SPEC) \
 		-- test_user_docs::
 else
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) --doc \
-		--features=internal-keycache,integer,hpu -p tfhe \
+		--features=internal-keycache,integer,hpu -p $(TFHE_SPEC) \
 		-- test_user_docs::
 endif
 
@@ -1073,16 +980,10 @@ test_tfhe_csprng: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
 		-p tfhe-csprng
 
-.PHONY: test_tfhe_csprng_big_endian # Run tfhe-csprng tests on an emulated big endian system
-test_tfhe_csprng_big_endian: install_rs_build_toolchain install_cargo_cross
-	RUSTFLAGS="" cross $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		-p tfhe-csprng --target=powerpc64-unknown-linux-gnu
-
-
 .PHONY: test_zk_pok # Run tfhe-zk-pok tests
 test_zk_pok: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
-		-p tfhe-zk-pok --features experimental
+		-p tfhe-zk-pok
 
 .PHONY: test_zk_wasm_x86_compat_ci
 test_zk_wasm_x86_compat_ci: check_nvm_installed
@@ -1112,11 +1013,16 @@ test_tfhe_lints: install_cargo_dylint
 # Here we use the "patch" functionality of Cargo to make sure the repo used for the data is the same as the one used for the code.
 .PHONY: test_backward_compatibility_ci
 test_backward_compatibility_ci: install_rs_build_toolchain
-	TFHE_BACKWARD_COMPAT_DATA_DIR="../$(BACKWARD_COMPAT_DATA_DIR)" RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
+	TFHE_BACKWARD_COMPAT_DATA_DIR="$(BACKWARD_COMPAT_DATA_DIR)" RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --profile $(CARGO_PROFILE) \
+		--config "patch.'$(BACKWARD_COMPAT_DATA_URL)'.$(BACKWARD_COMPAT_DATA_PROJECT).path=\"tests/$(BACKWARD_COMPAT_DATA_DIR)\"" \
 		--features=shortint,integer,zk-pok -p tests test_backward_compatibility -- --nocapture
 
 .PHONY: test_backward_compatibility # Same as test_backward_compatibility_ci but tries to clone the data repo first if needed
-test_backward_compatibility: pull_backward_compat_data test_backward_compatibility_ci
+test_backward_compatibility: tests/$(BACKWARD_COMPAT_DATA_DIR) test_backward_compatibility_ci
+
+.PHONY: backward_compat_branch # Prints the required backward compatibility branch
+backward_compat_branch:
+	@echo "$(BACKWARD_COMPAT_DATA_BRANCH)"
 
 .PHONY: doc # Build rust doc
 doc: install_rs_check_toolchain
@@ -1124,7 +1030,7 @@ doc: install_rs_check_toolchain
 	DOCS_RS=1 \
 	RUSTDOCFLAGS="--html-in-header katex-header.html" \
 	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" doc \
-		--features=boolean,shortint,integer,strings,gpu,internal-keycache,experimental,zk-pok --no-deps -p tfhe
+		--features=boolean,shortint,integer,strings,gpu,internal-keycache,experimental,zk-pok --no-deps -p $(TFHE_SPEC)
 
 .PHONY: docs # Build rust doc alias for doc
 docs: doc
@@ -1135,7 +1041,7 @@ lint_doc: install_rs_check_toolchain
 	DOCS_RS=1 \
 	RUSTDOCFLAGS="--html-in-header katex-header.html -Dwarnings" \
 	cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" doc \
-		--features=boolean,shortint,integer,strings,gpu,internal-keycache,experimental,zk-pok -p tfhe --no-deps
+		--features=boolean,shortint,integer,strings,gpu,internal-keycache,experimental,zk-pok -p $(TFHE_SPEC) --no-deps
 
 .PHONY: lint_docs # Build rust doc with linting enabled alias for lint_doc
 lint_docs: lint_doc
@@ -1161,10 +1067,6 @@ check_intra_md_links: install_mlc
 check_md_links: install_mlc
 	mlc --match-file-extension tfhe/docs
 
-.PHONY: check_doc_paths_use_dash # Check paths use "-" instead of "_" in docs for gitbook compatibility
-check_doc_paths_use_dash:
-	python3 ./scripts/check_doc_paths_use_dash.py
-
 .PHONY: check_parameter_export_ok # Checks exported "current" shortint parameter module is correct
 check_parameter_export_ok:
 	python3 ./scripts/check_current_param_export.py
@@ -1173,7 +1075,7 @@ check_parameter_export_ok:
 check_compile_tests: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --no-run \
 		--features=experimental,boolean,shortint,integer,internal-keycache \
-		-p tfhe
+		-p $(TFHE_SPEC)
 
 	@if [[ "$(OS)" == "Linux" || "$(OS)" == "Darwin" ]]; then \
 		"$(MAKE)" build_c_api && \
@@ -1184,7 +1086,7 @@ check_compile_tests: install_rs_build_toolchain
 check_compile_tests_benches_gpu: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --no-run \
 		--features=experimental,boolean,shortint,integer,internal-keycache,gpu,zk-pok \
-		-p tfhe
+		-p $(TFHE_SPEC)
 	mkdir -p "$(TFHECUDA_BUILD)" && \
 		cd "$(TFHECUDA_BUILD)" && \
 		cmake .. -DCMAKE_BUILD_TYPE=Debug -DTFHE_CUDA_BACKEND_BUILD_TESTS=ON -DTFHE_CUDA_BACKEND_BUILD_BENCHMARKS=ON && \
@@ -1312,11 +1214,11 @@ bench_signed_integer_gpu: install_rs_check_toolchain
 
 .PHONY: bench_integer_hpu # Run benchmarks for integer on HPU backend
 bench_integer_hpu: install_rs_check_toolchain
-	source ./setup_hpu.sh --config $(HPU_CONFIG) -p ; \
+	source ./setup_hpu.sh --config $(HPU_CONFIG) ; \
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_OP_FLAVOR=$(BENCH_OP_FLAVOR) __TFHE_RS_FAST_BENCH=$(FAST_BENCH) __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
 	--bench integer-bench \
-	--features=integer,internal-keycache,pbs-stats,hpu,hpu-v80 -p tfhe-benchmark --
+	--features=integer,internal-keycache,pbs-stats,hpu,hpu-v80 -p tfhe-benchmark -- --quick
 
 .PHONY: bench_integer_compression # Run benchmarks for unsigned integer compression
 bench_integer_compression: install_rs_check_toolchain
@@ -1485,26 +1387,6 @@ bench_web_js_api_parallel_firefox_ci: setup_venv
 	nvm use $(NODE_VERSION) && \
 	$(MAKE) bench_web_js_api_parallel_firefox
 
-.PHONY: bench_hlapi # Run benchmarks for integer operations
-bench_hlapi: install_rs_check_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi \
-	--features=integer,internal-keycache,nightly-avx512 -p tfhe-benchmark --
-
-.PHONY: bench_hlapi_gpu # Run benchmarks for integer operations on GPU
-bench_hlapi_gpu: install_rs_check_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi \
-	--features=integer,gpu,internal-keycache,nightly-avx512 -p tfhe-benchmark --
-
-.PHONY: bench_hlapi_hpu # Run benchmarks for HLAPI operations on HPU
-bench_hlapi_hpu: install_rs_check_toolchain
-	source ./setup_hpu.sh --config $(HPU_CONFIG) -p ; \
-	RUSTFLAGS="$(RUSTFLAGS)" \
-	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi \
-	--features=integer,internal-keycache,hpu,hpu-v80 -p tfhe-benchmark --
-
 .PHONY: bench_hlapi_erc20 # Run benchmarks for ERC20 operations
 bench_hlapi_erc20: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
@@ -1531,31 +1413,17 @@ bench_hlapi_dex_gpu: install_rs_check_toolchain
 
 .PHONY: bench_hlapi_erc20_hpu # Run benchmarks for ECR20 operations on HPU
 bench_hlapi_erc20_hpu: install_rs_check_toolchain
-	source ./setup_hpu.sh --config $(HPU_CONFIG) -p ; \
+	source ./setup_hpu.sh --config $(HPU_CONFIG) ; \
 	RUSTFLAGS="$(RUSTFLAGS)" \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
 	--bench hlapi-erc20 \
-	--features=integer,internal-keycache,hpu,hpu-v80 -p tfhe-benchmark --
+	--features=integer,internal-keycache,hpu,hpu-v80 -p tfhe-benchmark -- --quick
 
 .PHONY: bench_tfhe_zk_pok # Run benchmarks for the tfhe_zk_pok crate
 bench_tfhe_zk_pok: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench -p tfhe-zk-pok --
 
-.PHONY: bench_hlapi_noise_squash # Run benchmarks for noise squash operation
-bench_hlapi_noise_squash: install_rs_check_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) \
-	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi-noise-squash \
-	--features=integer,internal-keycache,pbs-stats,nightly-avx512 -p tfhe-benchmark --
-
-.PHONY: bench_hlapi_noise_squash_gpu # Run benchmarks for noise squash operation on GPU
-bench_hlapi_noise_squash_gpu: install_rs_check_toolchain
-	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) \
-	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi-noise-squash \
-	--features=integer,gpu,internal-keycache,pbs-stats,nightly-avx512 -p tfhe-benchmark --
-
 #
 # Utility tools
 #
@@ -1564,13 +1432,13 @@ bench_hlapi_noise_squash_gpu: install_rs_check_toolchain
 gen_key_cache: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS) --cfg tarpaulin" cargo $(CARGO_RS_BUILD_TOOLCHAIN) run --profile $(CARGO_PROFILE) \
 		--example generates_test_keys \
-		--features=boolean,shortint,experimental,internal-keycache -p tfhe \
+		--features=boolean,shortint,experimental,internal-keycache -p $(TFHE_SPEC) \
 		-- $(MULTI_BIT_ONLY) $(COVERAGE_ONLY)
 
 .PHONY: gen_key_cache_core_crypto # Run function to generate keys and cache them for core_crypto tests
 gen_key_cache_core_crypto: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) test --tests --profile $(CARGO_PROFILE) \
-		--features=experimental,internal-keycache -p tfhe -- --nocapture \
+		--features=experimental,internal-keycache -p $(TFHE_SPEC) -- --nocapture \
 		core_crypto::keycache::generate_keys
 
 .PHONY: measure_hlapi_compact_pk_ct_sizes # Measure sizes of public keys and ciphertext for high-level API
@@ -1611,15 +1479,13 @@ parse_wasm_benchmarks: install_rs_check_toolchain
 .PHONY: write_params_to_file # Gather all crypto parameters into a file with a Sage readable format.
 write_params_to_file: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_CHECK_TOOLCHAIN) run \
-	--example write_params_to_file --features=boolean,shortint,hpu,internal-keycache
+	--example write_params_to_file --features=boolean,shortint,internal-keycache
 
-.PHONY: pull_backward_compat_data # Pull the data files needed for backward compatibility tests
-pull_backward_compat_data:
-	./scripts/pull_lfs_data.sh $(BACKWARD_COMPAT_DATA_DIR)
+.PHONY: clone_backward_compat_data # Clone the data repo needed for backward compatibility tests
+clone_backward_compat_data:
+	./scripts/clone_backward_compat_data.sh $(BACKWARD_COMPAT_DATA_URL) $(BACKWARD_COMPAT_DATA_BRANCH) tests/$(BACKWARD_COMPAT_DATA_DIR)
 
-.PHONY: pull_hpu_files # Pull the hpu files
-pull_hpu_files:
-	./scripts/pull_lfs_data.sh backends/tfhe-hpu-backend/
+tests/$(BACKWARD_COMPAT_DATA_DIR): clone_backward_compat_data
 
 #
 # Real use case examples
@@ -1645,20 +1511,19 @@ sha256_bool: install_rs_check_toolchain
 
 .PHONY: pcc # pcc stands for pre commit checks (except GPU)
 pcc: no_tfhe_typo no_dbg_log check_parameter_export_ok check_fmt check_typos lint_doc \
-check_md_docs_are_tested check_intra_md_links check_doc_paths_use_dash \
-clippy_all check_compile_tests test_tfhe_lints \
+check_md_docs_are_tested check_intra_md_links clippy_all check_compile_tests test_tfhe_lints \
 tfhe_lints
 
 .PHONY: pcc_gpu # pcc stands for pre commit checks for GPU compilation
 pcc_gpu: check_rust_bindings_did_not_change clippy_rustdoc_gpu \
-clippy_gpu clippy_cuda_backend clippy_bench_gpu check_compile_tests_benches_gpu test_integer_hl_test_gpu_check_warnings
+clippy_gpu clippy_cuda_backend clippy_bench_gpu check_compile_tests_benches_gpu
 
 .PHONY: pcc_hpu # pcc stands for pre commit checks for HPU compilation
-pcc_hpu: clippy_hpu clippy_hpu_backend clippy_hpu_mockup test_integer_hpu_mockup_ci_fast
+pcc_hpu: clippy_hpu clippy_hpu_backend test_integer_hpu_mockup_ci_fast
 
 .PHONY: fpcc # pcc stands for pre commit checks, the f stands for fast
 fpcc: no_tfhe_typo no_dbg_log check_parameter_export_ok check_fmt check_typos lint_doc \
-check_md_docs_are_tested check_intra_md_links check_doc_paths_use_dash clippy_fast check_compile_tests
+check_md_docs_are_tested clippy_fast check_compile_tests
 
 .PHONY: conformance # Automatically fix problems that can be fixed
 conformance: fix_newline fmt fmt_js

README.md

@@ -18,7 +18,6 @@
   <a href="https://github.com/zama-ai/tfhe-rs/releases"><img src="https://img.shields.io/github/v/release/zama-ai/tfhe-rs?style=flat-square"></a>
   <a href="LICENSE"><img src="https://img.shields.io/badge/License-BSD--3--Clause--Clear-%23ffb243?style=flat-square"></a>
   <a href="https://github.com/zama-ai/bounty-program"><img src="https://img.shields.io/badge/Contribute-Zama%20Bounty%20Program-%23ffd208?style=flat-square"></a>
-  <a href="https://slsa.dev"><img alt="SLSA 3" src="https://slsa.dev/images/gh-badge-level3.svg" /></a>
 </p>
 
 ## About
@@ -149,7 +148,7 @@ To run this code, use the following command:
 > Note that when running code that uses `TFHE-rs`, it is highly recommended
 to run in release mode with cargo's `--release` flag to have the best performances possible.
 
-*Find an example with more explanations in [this part of the documentation](https://docs.zama.ai/tfhe-rs/get-started/quick-start)*
+*Find an example with more explanations in [this part of the documentation](https://docs.zama.ai/tfhe-rs/get-started/quick_start)*
 
 <p align="right">
   <a href="#about" > ↑ Back to top </a>
@@ -201,9 +200,11 @@ When a new update is published in the Lattice Estimator, we update parameters ac
 
 ### Security model
 
-By default, the parameter sets used in the High-Level API have a failure probability $\le 2^{-128}$ to securely work in the IND-CPA^D model using the algorithmic techniques provided in our code base [1].
+By default, the parameter sets used in the High-Level API with the x86 CPU backend have a failure probability $\le 2^{128}$ to securely work in the IND-CPA^D model using the algorithmic techniques provided in our code base [1].
 If you want to work within the IND-CPA security model, which is less strict than the IND-CPA-D model, the parameter sets can easily be changed and would have slightly better performance. More details can be found in the [TFHE-rs documentation](https://docs.zama.ai/tfhe-rs).
 
+The default parameters used in the High-Level API with the GPU backend are chosen considering the IND-CPA security model, and are selected with a bootstrapping failure probability fixed at $p_{error} \le 2^{-64}$. In particular, it is assumed that the results of decrypted computations are not shared by the secret key owner with any third parties, as such an action can lead to leakage of the secret encryption key. If you are designing an application where decryptions must be shared, you will need to craft custom encryption parameters which are chosen in consideration of the IND-CPA^D security model [2].
+
 [1] Bernard, Olivier, et al. "Drifting Towards Better Error Probabilities in Fully Homomorphic Encryption Schemes". https://eprint.iacr.org/2024/1718.pdf
 
 [2] Li, Baiyu, et al. "Securing approximate homomorphic encryption using differential privacy." Annual International Cryptology Conference. Cham: Springer Nature Switzerland, 2022. https://eprint.iacr.org/2022/816.pdf

_typos.toml

@@ -12,7 +12,7 @@ extend-ignore-identifiers-re = [
     "herlo",
     # Example in trivium
     "C9217BA0D762ACA1",
-    "0x[0-9a-fA-F]+",
+    "0x[0-9a-fA-F]+"
 ]
 
 [files]
@@ -20,6 +20,4 @@ extend-exclude = [
     "backends/tfhe-cuda-backend/cuda/src/fft128/twiddles.cu",
     "backends/tfhe-cuda-backend/cuda/src/fft/twiddles.cu",
     "backends/tfhe-hpu-backend/config_store/**/*.link_summary",
-    "*.cbor",
-    "*.bcode",
 ]

apps/trivium/README.md

@@ -129,7 +129,7 @@ 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 `V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128`).
+The same implementation is also available for generic Ciphertexts representing bits (meant to be used with parameters `V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128`).
 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.
@@ -137,10 +137,10 @@ its setup a little more intricate.
 Example code:
 ```rust
 use tfhe::shortint::prelude::*;
-use tfhe::shortint::parameters::current_params::{
-    V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+use tfhe::shortint::parameters::v1_2::{
+    V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
 };
 use tfhe::{ConfigBuilder, generate_keys, FheUint64};
 use tfhe::prelude::*;
@@ -148,17 +148,17 @@ use tfhe_trivium::TriviumStreamShortint;
 
 fn test_shortint() {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+    let (client_key, server_key): (ClientKey, ServerKey) = gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();

apps/trivium/benches/kreyvium_shortint.rs

@@ -1,9 +1,9 @@
 use criterion::Criterion;
 use tfhe::prelude::*;
-use tfhe::shortint::parameters::current_params::{
-    V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
+use tfhe::shortint::parameters::v1_2::{
+    V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
 };
 use tfhe::shortint::prelude::*;
 use tfhe::{generate_keys, ConfigBuilder, FheUint64};
@@ -11,19 +11,19 @@ use tfhe_trivium::{KreyviumStreamShortint, TransCiphering};
 
 pub fn kreyvium_shortint_warmup(c: &mut Criterion) {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+        gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB000000000000".to_string();
@@ -64,19 +64,19 @@ pub fn kreyvium_shortint_warmup(c: &mut Criterion) {
 
 pub fn kreyvium_shortint_gen(c: &mut Criterion) {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+        gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB000000000000".to_string();
@@ -112,19 +112,19 @@ pub fn kreyvium_shortint_gen(c: &mut Criterion) {
 
 pub fn kreyvium_shortint_trans(c: &mut Criterion) {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+        gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB000000000000".to_string();

apps/trivium/benches/trivium_shortint.rs

@@ -1,9 +1,9 @@
 use criterion::Criterion;
 use tfhe::prelude::*;
-use tfhe::shortint::parameters::current_params::{
-    V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
+use tfhe::shortint::parameters::v1_2::{
+    V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
 };
 use tfhe::shortint::prelude::*;
 use tfhe::{generate_keys, ConfigBuilder, FheUint64};
@@ -11,19 +11,19 @@ use tfhe_trivium::{TransCiphering, TriviumStreamShortint};
 
 pub fn trivium_shortint_warmup(c: &mut Criterion) {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+        gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();
@@ -64,19 +64,19 @@ pub fn trivium_shortint_warmup(c: &mut Criterion) {
 
 pub fn trivium_shortint_gen(c: &mut Criterion) {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+        gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();
@@ -112,19 +112,19 @@ pub fn trivium_shortint_gen(c: &mut Criterion) {
 
 pub fn trivium_shortint_trans(c: &mut Criterion) {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+        gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();

apps/trivium/src/kreyvium/test.rs

@@ -1,16 +1,16 @@
 use crate::{KreyviumStream, KreyviumStreamByte, KreyviumStreamShortint, TransCiphering};
 use tfhe::prelude::*;
-use tfhe::shortint::parameters::current_params::{
-    V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
+use tfhe::shortint::parameters::v1_2::{
+    V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
 };
 use tfhe::{generate_keys, ConfigBuilder, FheBool, FheUint64, FheUint8};
 // Values for these tests come from the github repo renaud1239/Kreyvium,
 // commit fd6828f68711276c25f55e605935028f5e843f43
 
 fn get_hexadecimal_string_from_lsb_first_stream(a: Vec<bool>) -> String {
-    assert!(a.len().is_multiple_of(8));
+    assert!(a.len() % 8 == 0);
     let mut hexadecimal: String = "".to_string();
     for test in a.chunks(8) {
         // Encoding is bytes in LSB order
@@ -63,7 +63,7 @@ fn get_hexadecimal_string_from_lsb_first_stream(a: Vec<bool>) -> String {
 }
 
 fn get_hexagonal_string_from_bytes(a: Vec<u8>) -> String {
-    assert!(a.len().is_multiple_of(8));
+    assert!(a.len() % 8 == 0);
     let mut hexadecimal: String = "".to_string();
     for test in a {
         hexadecimal.push_str(&format!("{test:02X?}"));
@@ -221,19 +221,19 @@ use tfhe::shortint::prelude::*;
 #[test]
 fn kreyvium_test_shortint_long() {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+        gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB000000000000".to_string();

apps/trivium/src/trivium/test.rs

@@ -1,16 +1,16 @@
 use crate::{TransCiphering, TriviumStream, TriviumStreamByte, TriviumStreamShortint};
 use tfhe::prelude::*;
-use tfhe::shortint::parameters::current_params::{
-    V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
-    V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
+use tfhe::shortint::parameters::v1_2::{
+    V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128,
+    V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128,
 };
 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
 
 fn get_hexadecimal_string_from_lsb_first_stream(a: Vec<bool>) -> String {
-    assert!(a.len().is_multiple_of(8));
+    assert!(a.len() % 8 == 0);
     let mut hexadecimal: String = "".to_string();
     for test in a.chunks(8) {
         // Encoding is bytes in LSB order
@@ -63,7 +63,7 @@ fn get_hexadecimal_string_from_lsb_first_stream(a: Vec<bool>) -> String {
 }
 
 fn get_hexagonal_string_from_bytes(a: Vec<u8>) -> String {
-    assert!(a.len().is_multiple_of(8));
+    assert!(a.len() % 8 == 0);
     let mut hexadecimal: String = "".to_string();
     for test in a {
         hexadecimal.push_str(&format!("{test:02X?}"));
@@ -357,19 +357,19 @@ use tfhe::shortint::prelude::*;
 #[test]
 fn trivium_test_shortint_long() {
     let config = ConfigBuilder::default()
-        .use_custom_parameters(V1_4_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
+        .use_custom_parameters(V1_2_PARAM_MESSAGE_2_CARRY_2_KS_PBS_GAUSSIAN_2M128)
         .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(V1_4_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
+        gen_keys(V1_2_PARAM_MESSAGE_1_CARRY_1_KS_PBS_GAUSSIAN_2M128);
 
     let ksk = KeySwitchingKey::new(
         (&client_key, Some(&server_key)),
         (&underlying_ck, &underlying_sk),
-        V1_4_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
+        V1_2_PARAM_KEYSWITCH_1_1_KS_PBS_TO_2_2_KS_PBS_GAUSSIAN_2M128,
     );
 
     let key_string = "0053A6F94C9FF24598EB".to_string();

backends/tfhe-cuda-backend/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tfhe-cuda-backend"
-version = "0.11.0"
+version = "0.10.0"
 edition = "2021"
 authors = ["Zama team"]
 license = "BSD-3-Clause-Clear"
@@ -19,4 +19,3 @@ bindgen = "0.71"
 [features]
 experimental-multi-arch = []
 profile = []
-debug = []

backends/tfhe-cuda-backend/build.rs

@@ -53,10 +53,6 @@ fn main() {
             cmake_config.define("USE_NVTOOLS", "OFF");
         }
 
-        if cfg!(feature = "debug") {
-            cmake_config.define("CMAKE_BUILD_TYPE", "Debug");
-        }
-
         // Build the CMake project
         let dest = cmake_config.build();
         println!("cargo:rustc-link-search=native={}", dest.display());

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

@@ -52,8 +52,6 @@ endif()
 
 if(NOT CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE Release)
-else()
-  message("Building CUDA backend in ${CMAKE_BUILD_TYPE}")
 endif()
 
 # Add OpenMP support
@@ -78,10 +76,8 @@ endif()
 
 add_compile_definitions(CUDA_ARCH=${CUDA_ARCH})
 
-string(TOLOWER "${CMAKE_BUILD_TYPE}" CMAKE_BUILD_TYPE_LOWERCASE)
-
 # Check if the DEBUG flag is defined
-if(CMAKE_BUILD_TYPE_LOWERCASE STREQUAL "debug")
+if(CMAKE_BUILD_TYPE STREQUAL "Debug")
   # Debug mode
   message("Compiling in Debug mode")
   add_definitions(-DDEBUG)

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

@@ -26,15 +26,6 @@ void cuda_modulus_switch_inplace_64(void *stream, uint32_t gpu_index,
                                     void *lwe_array_out, uint32_t size,
                                     uint32_t log_modulus);
 
-void cuda_modulus_switch_64(void *stream, uint32_t gpu_index, void *lwe_out,
-                            const void *lwe_in, uint32_t size,
-                            uint32_t log_modulus);
-
-void cuda_centered_modulus_switch_64(void *stream, uint32_t gpu_index,
-                                     void *lwe_out, const void *lwe_in,
-                                     uint32_t lwe_dimension,
-                                     uint32_t log_modulus);
-
 void cuda_improve_noise_modulus_switch_64(
     void *stream, uint32_t gpu_index, void *lwe_array_out,
     void const *lwe_array_in, void const *lwe_array_indexes,

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

@@ -19,11 +19,6 @@ inline void cuda_error(cudaError_t code, const char *file, int line) {
     std::abort();
   }
 }
-
-// The PANIC macro should be used to validate user-inputs to GPU functions
-// it will execute in all targets, including production settings
-// e.g., cudaMemCopy to the device should check that the destination pointer is
-// a device pointer
 #define PANIC(format, ...)                                                     \
   {                                                                            \
     std::fprintf(stderr, "%s::%d::%s: panic.\n" format "\n", __FILE__,         \
@@ -31,31 +26,6 @@ inline void cuda_error(cudaError_t code, const char *file, int line) {
     std::abort();                                                              \
   }
 
-// This is a generic assertion checking macro with user defined printf-style
-// message
-#define PANIC_IF_FALSE(cond, format, ...)                                      \
-  do {                                                                         \
-    if (!(cond)) {                                                             \
-      PANIC(format "\n\n %s\n", ##__VA_ARGS__, #cond);                         \
-    }                                                                          \
-  } while (0)
-
-#ifndef GPU_ASSERTS_DISABLE
-// The GPU assert should be used to validate assumptions in algorithms,
-// for example, checking that two user-provided quantities have a certain
-// relationship or that the size of the buffer  provided to a function is
-// sufficient when it is filled with some algorithm that depends on
-// user-provided inputs e.g., OPRF corrections buffer should not have a size
-// higher than the number of blocks in the datatype that is generated
-#define GPU_ASSERT(cond, format, ...)                                          \
-  PANIC_IF_FALSE(cond, format, ##__VA_ARGS__)
-#else
-#define GPU_ASSERT(cond)                                                       \
-  do {                                                                         \
-  } while (0)
-#endif
-
-uint32_t cuda_get_device();
 void cuda_set_device(uint32_t gpu_index);
 
 cudaEvent_t cuda_create_event(uint32_t gpu_index);
@@ -79,13 +49,12 @@ void *cuda_malloc(uint64_t size, uint32_t gpu_index);
 
 void *cuda_malloc_with_size_tracking_async(uint64_t size, cudaStream_t stream,
                                            uint32_t gpu_index,
-                                           uint64_t &size_tracker,
+                                           uint64_t *size_tracker,
                                            bool allocate_gpu_memory);
 
 void *cuda_malloc_async(uint64_t size, cudaStream_t stream, uint32_t gpu_index);
 
 bool cuda_check_valid_malloc(uint64_t size, uint32_t gpu_index);
-uint64_t cuda_device_total_memory(uint32_t gpu_index);
 
 void cuda_memcpy_with_size_tracking_async_to_gpu(void *dest, const void *src,
                                                  uint64_t size,
@@ -119,8 +88,6 @@ void cuda_memset_async(void *dest, uint64_t val, uint64_t size,
 
 int cuda_get_number_of_gpus();
 
-int cuda_get_number_of_sms();
-
 void cuda_synchronize_device(uint32_t gpu_index);
 
 void cuda_drop(void *ptr, uint32_t gpu_index);

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

@@ -6,7 +6,6 @@
 
 extern std::mutex m;
 extern bool p2p_enabled;
-extern const int THRESHOLD_MULTI_GPU;
 
 extern "C" {
 int32_t cuda_setup_multi_gpu(int device_0_id);
@@ -16,26 +15,15 @@ int32_t cuda_setup_multi_gpu(int device_0_id);
 template <typename Torus>
 using LweArrayVariant = std::variant<std::vector<Torus *>, Torus *>;
 
-/// get_variant_element() resolves access when the input may be either a single
-/// pointer or a vector of pointers. If the variant holds a single pointer, the
-/// index is ignored and that pointer is returned; if it holds a vector, the
-/// element at `index` is returned.
-///
-/// This function replaces the previous macro:
-/// - Easier to debug and read than a macro
-/// - Deduces the pointer type from the variant (no need to name a Torus type
-/// explicitly)
-/// - Defined in a header, so it’s eligible for inlining by the optimizer
-template <typename Torus>
-inline Torus
-get_variant_element(const std::variant<std::vector<Torus>, Torus> &variant,
-                    size_t index) {
-  if (std::holds_alternative<std::vector<Torus>>(variant)) {
-    return std::get<std::vector<Torus>>(variant)[index];
-  } else {
-    return std::get<Torus>(variant);
-  }
-}
+// Macro to define the visitor logic using std::holds_alternative for vectors
+#define GET_VARIANT_ELEMENT(variant, index)                                    \
+  [&] {                                                                        \
+    if (std::holds_alternative<std::vector<Torus *>>(variant)) {               \
+      return std::get<std::vector<Torus *>>(variant)[index];                   \
+    } else {                                                                   \
+      return std::get<Torus *>(variant);                                       \
+    }                                                                          \
+  }()
 
 int get_active_gpu_count(int num_inputs, int gpu_count);
 

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

@@ -3,26 +3,6 @@
 
 #include "../../pbs/pbs_enums.h"
 
-typedef struct {
-  void *ptr;
-  uint32_t num_radix_blocks;
-  uint32_t lwe_dimension;
-} CudaLweCiphertextListFFI;
-
-typedef struct {
-  void *ptr;
-  uint32_t storage_log_modulus;
-  uint32_t lwe_per_glwe;
-  // Input LWEs are grouped by groups of `lwe_per_glwe`(the last group may be
-  // smaller)
-  // Each group is then packed into one GLWE with `lwe_per_glwe` bodies (one for
-  // each LWE of the group). In the end the total number of bodies is equal to
-  // the number of input LWE
-  uint32_t total_lwe_bodies_count;
-  uint32_t glwe_dimension;
-  uint32_t polynomial_size;
-} CudaPackedGlweCiphertextListFFI;
-
 extern "C" {
 uint64_t scratch_cuda_integer_compress_radix_ciphertext_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -30,29 +10,28 @@ uint64_t scratch_cuda_integer_compress_radix_ciphertext_64(
     uint32_t compression_polynomial_size, uint32_t lwe_dimension,
     uint32_t ks_level, uint32_t ks_base_log, uint32_t num_radix_blocks,
     uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    uint32_t lwe_per_glwe, bool allocate_gpu_memory);
+    uint32_t lwe_per_glwe, uint32_t storage_log_modulus,
+    bool allocate_gpu_memory);
 
 uint64_t scratch_cuda_integer_decompress_radix_ciphertext_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
     int8_t **mem_ptr, uint32_t encryption_glwe_dimension,
     uint32_t encryption_polynomial_size, uint32_t compression_glwe_dimension,
     uint32_t compression_polynomial_size, uint32_t lwe_dimension,
-    uint32_t pbs_level, uint32_t pbs_base_log,
-    uint32_t num_blocks_to_decompress, uint32_t message_modulus,
-    uint32_t carry_modulus, PBS_TYPE pbs_type, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t num_radix_blocks,
+    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
+    uint32_t storage_log_modulus, uint32_t body_count, bool allocate_gpu_memory,
+    bool allocate_ms_array);
 
 void cuda_integer_compress_radix_ciphertext_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaPackedGlweCiphertextListFFI *glwe_array_out,
-    CudaLweCiphertextListFFI const *lwe_array_in, void *const *fp_ksk,
-    int8_t *mem_ptr);
+    void *glwe_array_out, void const *lwe_array_in, void *const *fp_ksk,
+    uint32_t num_nths, int8_t *mem_ptr);
 
 void cuda_integer_decompress_radix_ciphertext_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaLweCiphertextListFFI *lwe_array_out,
-    CudaPackedGlweCiphertextListFFI const *glwe_in,
-    uint32_t const *indexes_array, void *const *bsks, int8_t *mem_ptr);
+    void *lwe_array_out, void const *glwe_in, uint32_t const *indexes_array,
+    uint32_t indexes_array_size, void *const *bsks, int8_t *mem_ptr);
 
 void cleanup_cuda_integer_compress_radix_ciphertext_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -61,41 +40,6 @@ void cleanup_cuda_integer_compress_radix_ciphertext_64(
 void cleanup_cuda_integer_decompress_radix_ciphertext_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
     int8_t **mem_ptr_void);
-
-uint64_t scratch_cuda_integer_compress_radix_ciphertext_128(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t compression_glwe_dimension,
-    uint32_t compression_polynomial_size, uint32_t lwe_dimension,
-    uint32_t ks_level, uint32_t ks_base_log, uint32_t num_radix_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    uint32_t lwe_per_glwe, bool allocate_gpu_memory);
-
-uint64_t scratch_cuda_integer_decompress_radix_ciphertext_128(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t compression_glwe_dimension,
-    uint32_t compression_polynomial_size, uint32_t lwe_dimension,
-    uint32_t num_radix_blocks, uint32_t message_modulus, uint32_t carry_modulus,
-    bool allocate_gpu_memory);
-
-void cuda_integer_compress_radix_ciphertext_128(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaPackedGlweCiphertextListFFI *glwe_array_out,
-    CudaLweCiphertextListFFI const *lwe_array_in, void *const *fp_ksk,
-    int8_t *mem_ptr);
-
-void cuda_integer_decompress_radix_ciphertext_128(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaLweCiphertextListFFI *lwe_array_out,
-    CudaPackedGlweCiphertextListFFI const *glwe_in,
-    uint32_t const *indexes_array, int8_t *mem_ptr);
-
-void cleanup_cuda_integer_compress_radix_ciphertext_128(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void);
-
-void cleanup_cuda_integer_decompress_radix_ciphertext_128(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void);
 }
 
 #endif

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

@@ -5,33 +5,40 @@
 
 template <typename Torus> struct int_compression {
   int_radix_params compression_params;
+  uint32_t storage_log_modulus;
+  uint32_t lwe_per_glwe;
+
+  uint32_t body_count;
+
   // Compression
   int8_t *fp_ks_buffer;
   Torus *tmp_lwe;
   Torus *tmp_glwe_array_out;
   bool gpu_memory_allocated;
-  uint32_t lwe_per_glwe;
 
   int_compression(cudaStream_t const *streams, uint32_t const *gpu_indexes,
                   uint32_t gpu_count, int_radix_params compression_params,
                   uint32_t num_radix_blocks, uint32_t lwe_per_glwe,
-                  bool allocate_gpu_memory, uint64_t &size_tracker) {
+                  uint32_t storage_log_modulus, bool allocate_gpu_memory,
+                  uint64_t *size_tracker) {
     gpu_memory_allocated = allocate_gpu_memory;
     this->compression_params = compression_params;
+    this->lwe_per_glwe = lwe_per_glwe;
+    this->storage_log_modulus = storage_log_modulus;
+    this->body_count = num_radix_blocks;
 
-    uint64_t glwe_accumulator_size = (compression_params.glwe_dimension + 1) *
-                                     compression_params.polynomial_size;
+    Torus glwe_accumulator_size = (compression_params.glwe_dimension + 1) *
+                                  compression_params.polynomial_size;
 
-    tmp_lwe = static_cast<Torus *>(cuda_malloc_with_size_tracking_async(
+    tmp_lwe = (Torus *)cuda_malloc_with_size_tracking_async(
         num_radix_blocks * (compression_params.small_lwe_dimension + 1) *
             sizeof(Torus),
-        streams[0], gpu_indexes[0], size_tracker, allocate_gpu_memory));
-    tmp_glwe_array_out =
-        static_cast<Torus *>(cuda_malloc_with_size_tracking_async(
-            lwe_per_glwe * glwe_accumulator_size * sizeof(Torus), streams[0],
-            gpu_indexes[0], size_tracker, allocate_gpu_memory));
+        streams[0], gpu_indexes[0], size_tracker, allocate_gpu_memory);
+    tmp_glwe_array_out = (Torus *)cuda_malloc_with_size_tracking_async(
+        lwe_per_glwe * glwe_accumulator_size * sizeof(Torus), streams[0],
+        gpu_indexes[0], size_tracker, allocate_gpu_memory);
 
-    size_tracker += scratch_packing_keyswitch_lwe_list_to_glwe<Torus>(
+    *size_tracker += 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,
@@ -51,7 +58,11 @@ template <typename Torus> struct int_compression {
 template <typename Torus> struct int_decompression {
   int_radix_params encryption_params;
   int_radix_params compression_params;
-  uint32_t num_blocks_to_decompress;
+
+  uint32_t storage_log_modulus;
+
+  uint32_t num_radix_blocks;
+  uint32_t body_count;
 
   Torus *tmp_extracted_glwe;
   Torus *tmp_extracted_lwe;
@@ -63,63 +74,57 @@ template <typename Torus> struct int_decompression {
   int_decompression(cudaStream_t const *streams, uint32_t const *gpu_indexes,
                     uint32_t gpu_count, int_radix_params encryption_params,
                     int_radix_params compression_params,
-                    uint32_t num_blocks_to_decompress, bool allocate_gpu_memory,
-                    uint64_t &size_tracker) {
+                    uint32_t num_radix_blocks, uint32_t body_count,
+                    uint32_t storage_log_modulus, bool allocate_gpu_memory,
+                    uint64_t *size_tracker) {
     gpu_memory_allocated = allocate_gpu_memory;
     this->encryption_params = encryption_params;
     this->compression_params = compression_params;
-    this->num_blocks_to_decompress = num_blocks_to_decompress;
+    this->storage_log_modulus = storage_log_modulus;
+    this->num_radix_blocks = num_radix_blocks;
+    this->body_count = body_count;
 
-    uint64_t glwe_accumulator_size = (compression_params.glwe_dimension + 1) *
-                                     compression_params.polynomial_size;
-    uint64_t lwe_accumulator_size = (compression_params.glwe_dimension *
-                                         compression_params.polynomial_size +
-                                     1);
+    Torus glwe_accumulator_size = (compression_params.glwe_dimension + 1) *
+                                  compression_params.polynomial_size;
+    Torus lwe_accumulator_size = (compression_params.glwe_dimension *
+                                      compression_params.polynomial_size +
+                                  1);
+    decompression_rescale_lut = new int_radix_lut<Torus>(
+        streams, gpu_indexes, gpu_count, encryption_params, 1, num_radix_blocks,
+        allocate_gpu_memory, size_tracker);
 
     tmp_extracted_glwe = (Torus *)cuda_malloc_with_size_tracking_async(
-        num_blocks_to_decompress * glwe_accumulator_size * sizeof(Torus),
-        streams[0], gpu_indexes[0], size_tracker, allocate_gpu_memory);
+        num_radix_blocks * glwe_accumulator_size * sizeof(Torus), streams[0],
+        gpu_indexes[0], size_tracker, allocate_gpu_memory);
     tmp_indexes_array = (uint32_t *)cuda_malloc_with_size_tracking_async(
-        num_blocks_to_decompress * sizeof(uint32_t), streams[0], gpu_indexes[0],
+        num_radix_blocks * sizeof(uint32_t), streams[0], gpu_indexes[0],
         size_tracker, allocate_gpu_memory);
     tmp_extracted_lwe = (Torus *)cuda_malloc_with_size_tracking_async(
-        num_blocks_to_decompress * lwe_accumulator_size * sizeof(Torus),
-        streams[0], gpu_indexes[0], size_tracker, allocate_gpu_memory);
+        num_radix_blocks * lwe_accumulator_size * sizeof(Torus), streams[0],
+        gpu_indexes[0], size_tracker, allocate_gpu_memory);
 
-    // rescale is only needed on 64-bit decompression
-    if constexpr (std::is_same_v<Torus, uint64_t>) {
-      decompression_rescale_lut = new int_radix_lut<Torus>(
-          streams, gpu_indexes, gpu_count, encryption_params, 1,
-          num_blocks_to_decompress, allocate_gpu_memory, size_tracker);
+    // 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 = [](Torus x) -> Torus { return x; };
 
-      // 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-bit message onto a 4-bit
-      // space, we want to keep the original 2-bit value in the 4-bit space,
-      // so we apply the identity and the encoding will rescale it for us.
-      decompression_rescale_lut = new int_radix_lut<Torus>(
-          streams, gpu_indexes, gpu_count, encryption_params, 1,
-          num_blocks_to_decompress, allocate_gpu_memory, size_tracker);
-      auto decompression_rescale_f = [](Torus x) -> Torus { return x; };
+    auto effective_compression_message_modulus =
+        encryption_params.carry_modulus;
+    auto effective_compression_carry_modulus = 1;
 
-      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),
+        decompression_rescale_lut->get_degree(0),
+        decompression_rescale_lut->get_max_degree(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,
+        gpu_memory_allocated);
 
-      generate_device_accumulator_with_encoding<Torus>(
-          streams[0], gpu_indexes[0], decompression_rescale_lut->get_lut(0, 0),
-          decompression_rescale_lut->get_degree(0),
-          decompression_rescale_lut->get_max_degree(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, gpu_memory_allocated);
-      auto active_gpu_count =
-          get_active_gpu_count(num_blocks_to_decompress, gpu_count);
-      decompression_rescale_lut->broadcast_lut(streams, gpu_indexes,
-                                               active_gpu_count);
-    }
+    decompression_rescale_lut->broadcast_lut(streams, gpu_indexes, 0);
   }
   void release(cudaStream_t const *streams, uint32_t const *gpu_indexes,
                uint32_t gpu_count) {
@@ -129,11 +134,9 @@ template <typename Torus> struct int_decompression {
                                        gpu_indexes[0], gpu_memory_allocated);
     cuda_drop_with_size_tracking_async(tmp_indexes_array, streams[0],
                                        gpu_indexes[0], gpu_memory_allocated);
-    if constexpr (std::is_same_v<Torus, uint64_t>) {
-      decompression_rescale_lut->release(streams, gpu_indexes, gpu_count);
-      delete decompression_rescale_lut;
-      decompression_rescale_lut = nullptr;
-    }
+
+    decompression_rescale_lut->release(streams, gpu_indexes, gpu_count);
+    delete decompression_rescale_lut;
   }
 };
 #endif

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

@@ -37,10 +37,6 @@ enum SIGNED_OPERATION { ADDITION = 1, SUBTRACTION = -1 };
 
 enum outputFlag { FLAG_NONE = 0, FLAG_OVERFLOW = 1, FLAG_CARRY = 2 };
 
-enum Direction { Trailing = 0, Leading = 1 };
-
-enum BitValue { Zero = 0, One = 1 };
-
 extern "C" {
 
 typedef struct {
@@ -52,34 +48,6 @@ typedef struct {
   uint32_t lwe_dimension;
 } CudaRadixCiphertextFFI;
 
-typedef struct {
-  uint64_t const *chosen_multiplier_has_at_least_one_set;
-  uint64_t const *decomposed_chosen_multiplier;
-
-  uint32_t const num_scalars;
-  uint32_t const active_bits;
-  uint64_t const shift_pre;
-  uint32_t const shift_post;
-  uint32_t const ilog2_chosen_multiplier;
-  uint32_t const chosen_multiplier_num_bits;
-
-  bool const is_chosen_multiplier_zero;
-  bool const is_abs_chosen_multiplier_one;
-  bool const is_chosen_multiplier_negative;
-  bool const is_chosen_multiplier_pow2;
-  bool const chosen_multiplier_has_more_bits_than_numerator;
-  // if signed: test if chosen_multiplier >= 2^{num_bits - 1}
-  bool const is_chosen_multiplier_geq_two_pow_numerator;
-
-  uint32_t const ilog2_divisor;
-
-  bool const is_divisor_zero;
-  bool const is_abs_divisor_one;
-  bool const is_divisor_negative;
-  bool const is_divisor_pow2;
-  bool const divisor_has_more_bits_than_numerator;
-} CudaScalarDivisorFFI;
-
 uint64_t scratch_cuda_apply_univariate_lut_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
     int8_t **mem_ptr, void const *input_lut, uint32_t lwe_dimension,
@@ -87,8 +55,7 @@ uint64_t scratch_cuda_apply_univariate_lut_kb_64(
     uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
     uint32_t grouping_factor, uint32_t input_lwe_ciphertext_count,
     uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    uint64_t lut_degree, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    uint64_t lut_degree, bool allocate_gpu_memory, bool allocate_ms_array);
 uint64_t 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,
@@ -97,7 +64,7 @@ uint64_t scratch_cuda_apply_many_univariate_lut_kb_64(
     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, uint64_t lut_degree, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_ms_array);
 void cuda_apply_univariate_lut_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
     CudaRadixCiphertextFFI *output_radix_lwe,
@@ -118,8 +85,7 @@ uint64_t scratch_cuda_apply_bivariate_lut_kb_64(
     uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
     uint32_t grouping_factor, uint32_t input_lwe_ciphertext_count,
     uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    uint64_t lut_degree, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    uint64_t lut_degree, bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_apply_bivariate_lut_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -149,7 +115,7 @@ uint64_t scratch_cuda_full_propagation_64(
     uint32_t polynomial_size, uint32_t ks_level, uint32_t ks_base_log,
     uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
     uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_full_propagation_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -168,7 +134,7 @@ uint64_t scratch_cuda_integer_mult_radix_ciphertext_kb_64(
     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, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_mult_radix_ciphertext_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -202,7 +168,7 @@ uint64_t scratch_cuda_integer_radix_logical_scalar_shift_kb_64(
     uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
     uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
     uint32_t carry_modulus, PBS_TYPE pbs_type, SHIFT_OR_ROTATE_TYPE shift_type,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_radix_logical_scalar_shift_kb_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -217,7 +183,7 @@ uint64_t scratch_cuda_integer_radix_arithmetic_scalar_shift_kb_64(
     uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
     uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
     uint32_t carry_modulus, PBS_TYPE pbs_type, SHIFT_OR_ROTATE_TYPE shift_type,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_radix_arithmetic_scalar_shift_kb_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -240,8 +206,7 @@ uint64_t scratch_cuda_integer_radix_shift_and_rotate_kb_64(
     uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
     uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
     uint32_t carry_modulus, PBS_TYPE pbs_type, SHIFT_OR_ROTATE_TYPE shift_type,
-    bool is_signed, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    bool is_signed, bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_radix_shift_and_rotate_kb_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -262,7 +227,7 @@ uint64_t scratch_cuda_integer_radix_comparison_kb_64(
     uint32_t grouping_factor, uint32_t lwe_ciphertext_count,
     uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
     COMPARISON_TYPE op_type, bool is_signed, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_ms_array);
 
 void cuda_comparison_integer_radix_ciphertext_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -292,8 +257,7 @@ uint64_t scratch_cuda_integer_radix_bitop_kb_64(
     uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
     uint32_t grouping_factor, uint32_t lwe_ciphertext_count,
     uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    BITOP_TYPE op_type, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    BITOP_TYPE op_type, bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_bitop_integer_radix_ciphertext_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -322,7 +286,7 @@ uint64_t scratch_cuda_integer_radix_cmux_kb_64(
     uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
     uint32_t grouping_factor, uint32_t lwe_ciphertext_count,
     uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_cmux_integer_radix_ciphertext_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -344,7 +308,7 @@ uint64_t scratch_cuda_integer_radix_scalar_rotate_kb_64(
     uint32_t ks_base_log, uint32_t pbs_level, uint32_t pbs_base_log,
     uint32_t grouping_factor, uint32_t num_blocks, uint32_t message_modulus,
     uint32_t carry_modulus, PBS_TYPE pbs_type, SHIFT_OR_ROTATE_TYPE shift_type,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_radix_scalar_rotate_kb_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -364,8 +328,7 @@ uint64_t scratch_cuda_propagate_single_carry_kb_64_inplace(
     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,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    uint32_t uses_carry, bool allocate_gpu_memory, bool allocate_ms_array);
 
 uint64_t scratch_cuda_add_and_propagate_single_carry_kb_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -374,8 +337,7 @@ uint64_t scratch_cuda_add_and_propagate_single_carry_kb_64_inplace(
     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,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    uint32_t uses_carry, bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_propagate_single_carry_kb_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -410,7 +372,7 @@ uint64_t scratch_cuda_integer_overflowing_sub_kb_64_inplace(
     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, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_overflowing_sub_kb_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -433,8 +395,7 @@ uint64_t scratch_cuda_integer_radix_partial_sum_ciphertexts_vec_kb_64(
     uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
     uint32_t num_blocks_in_radix, uint32_t max_num_radix_in_vec,
     uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool reduce_degrees_for_single_carry_propagation, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_radix_partial_sum_ciphertexts_vec_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -453,8 +414,7 @@ uint64_t scratch_cuda_integer_scalar_mul_kb_64(
     uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
     uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
     uint32_t num_blocks, uint32_t message_modulus, uint32_t carry_modulus,
-    PBS_TYPE pbs_type, uint32_t num_scalar_bits, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    PBS_TYPE pbs_type, bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_scalar_multiplication_integer_radix_ciphertext_64_inplace(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -476,8 +436,7 @@ uint64_t scratch_cuda_integer_div_rem_radix_ciphertext_kb_64(
     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,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    PBS_TYPE pbs_type, bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_div_rem_radix_ciphertext_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -498,8 +457,7 @@ uint64_t scratch_cuda_integer_compute_prefix_sum_hillis_steele_64(
     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,
-    uint64_t lut_degree, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    uint64_t lut_degree, bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_compute_prefix_sum_hillis_steele_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -525,8 +483,7 @@ uint64_t scratch_cuda_integer_abs_inplace_radix_ciphertext_kb_64(
     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,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    PBS_TYPE pbs_type, bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_abs_inplace_radix_ciphertext_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -546,7 +503,7 @@ uint64_t scratch_cuda_integer_are_all_comparisons_block_true_kb_64(
     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, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_are_all_comparisons_block_true_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -567,7 +524,7 @@ uint64_t scratch_cuda_integer_is_at_least_one_comparisons_block_true_kb_64(
     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, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_integer_is_at_least_one_comparisons_block_true_kb_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
@@ -584,230 +541,5 @@ void cleanup_cuda_integer_is_at_least_one_comparisons_block_true(
 void extend_radix_with_trivial_zero_blocks_msb_64(
     CudaRadixCiphertextFFI *output, CudaRadixCiphertextFFI const *input,
     void *const *streams, uint32_t const *gpu_indexes);
-
-void trim_radix_blocks_lsb_64(CudaRadixCiphertextFFI *output,
-                              CudaRadixCiphertextFFI const *input,
-                              void *const *streams,
-                              uint32_t const *gpu_indexes);
-
-uint64_t scratch_cuda_apply_noise_squashing_kb(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t lwe_dimension, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t input_glwe_dimension,
-    uint32_t input_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 num_original_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_apply_noise_squashing_kb(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *output_radix_lwe,
-    CudaRadixCiphertextFFI const *input_radix_lwe, int8_t *mem_ptr,
-    void *const *ksks,
-    CudaModulusSwitchNoiseReductionKeyFFI const *ms_noise_reduction_key,
-    void *const *bsks);
-
-void cleanup_cuda_apply_noise_squashing_kb(void *const *streams,
-                                           uint32_t const *gpu_indexes,
-                                           uint32_t gpu_count,
-                                           int8_t **mem_ptr_void);
-
-uint64_t scratch_cuda_sub_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,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_sub_and_propagate_single_carry_kb_64_inplace(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *lhs_array, const CudaRadixCiphertextFFI *rhs_array,
-    CudaRadixCiphertextFFI *carry_out, const CudaRadixCiphertextFFI *carry_in,
-    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
-    CudaModulusSwitchNoiseReductionKeyFFI const *ms_noise_reduction_key,
-    uint32_t requested_flag, uint32_t uses_carry);
-
-void cleanup_cuda_sub_and_propagate_single_carry(void *const *streams,
-                                                 uint32_t const *gpu_indexes,
-                                                 uint32_t gpu_count,
-                                                 int8_t **mem_ptr_void);
-
-uint64_t scratch_cuda_integer_unsigned_scalar_div_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
-    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
-    uint32_t num_blocks, uint32_t message_modulus, uint32_t carry_modulus,
-    PBS_TYPE pbs_type, const CudaScalarDivisorFFI *scalar_divisor_ffi,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_integer_unsigned_scalar_div_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *numerator_ct, int8_t *mem_ptr, void *const *bsks,
-    void *const *ksks,
-    const CudaModulusSwitchNoiseReductionKeyFFI *ms_noise_reduction_key,
-    const CudaScalarDivisorFFI *scalar_divisor_ffi);
-
-void cleanup_cuda_integer_unsigned_scalar_div_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void);
-
-uint64_t scratch_cuda_extend_radix_with_sign_msb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
-    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
-    uint32_t num_blocks, uint32_t num_additional_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_extend_radix_with_sign_msb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *output, CudaRadixCiphertextFFI const *input,
-    int8_t *mem_ptr, uint32_t num_additional_blocks, void *const *bsks,
-    void *const *ksks,
-    CudaModulusSwitchNoiseReductionKeyFFI const *ms_noise_reduction_key);
-
-void cleanup_cuda_extend_radix_with_sign_msb_64(void *const *streams,
-                                                uint32_t const *gpu_indexes,
-                                                uint32_t gpu_count,
-                                                int8_t **mem_ptr_void);
-
-uint64_t scratch_cuda_integer_signed_scalar_div_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
-    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
-    uint32_t num_blocks, uint32_t message_modulus, uint32_t carry_modulus,
-    PBS_TYPE pbs_type, const CudaScalarDivisorFFI *scalar_divisor_ffi,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_integer_signed_scalar_div_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *numerator_ct, int8_t *mem_ptr, void *const *bsks,
-    void *const *ksks,
-    const CudaModulusSwitchNoiseReductionKeyFFI *ms_noise_reduction_key,
-    const CudaScalarDivisorFFI *scalar_divisor_ffi, uint32_t numerator_bits);
-
-void cleanup_cuda_integer_signed_scalar_div_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void);
-
-uint64_t scratch_integer_unsigned_scalar_div_rem_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
-    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
-    uint32_t num_blocks, uint32_t message_modulus, uint32_t carry_modulus,
-    PBS_TYPE pbs_type, const CudaScalarDivisorFFI *scalar_divisor_ffi,
-    uint32_t const active_bits_divisor, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_integer_unsigned_scalar_div_rem_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *quotient_ct, CudaRadixCiphertextFFI *remainder_ct,
-    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
-    const CudaModulusSwitchNoiseReductionKeyFFI *ms_noise_reduction_key,
-    const CudaScalarDivisorFFI *scalar_divisor_ffi,
-    uint64_t const *divisor_has_at_least_one_set,
-    uint64_t const *decomposed_divisor, uint32_t const num_scalars_divisor,
-    void const *clear_blocks, void const *h_clear_blocks,
-    uint32_t num_clear_blocks);
-
-void cleanup_cuda_integer_unsigned_scalar_div_rem_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void);
-
-uint64_t scratch_integer_signed_scalar_div_rem_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
-    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
-    uint32_t num_blocks, uint32_t message_modulus, uint32_t carry_modulus,
-    PBS_TYPE pbs_type, const CudaScalarDivisorFFI *scalar_divisor_ffi,
-    uint32_t const active_bits_divisor, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_integer_signed_scalar_div_rem_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *quotient_ct, CudaRadixCiphertextFFI *remainder_ct,
-    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
-    CudaModulusSwitchNoiseReductionKeyFFI const *ms_noise_reduction_key,
-    const CudaScalarDivisorFFI *scalar_divisor_ffi,
-    uint64_t const *divisor_has_at_least_one_set,
-    uint64_t const *decomposed_divisor, uint32_t const num_scalars_divisor,
-    uint32_t numerator_bits);
-
-void cleanup_cuda_integer_signed_scalar_div_rem_radix_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void);
-
-uint64_t scratch_integer_count_of_consecutive_bits_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
-    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
-    uint32_t num_blocks, uint32_t counter_num_blocks, uint32_t message_modulus,
-    uint32_t carry_modulus, PBS_TYPE pbs_type, Direction direction,
-    BitValue bit_value, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_integer_count_of_consecutive_bits_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *output_ct, CudaRadixCiphertextFFI const *input_ct,
-    int8_t *mem_ptr, void *const *bsks, void *const *ksks,
-    const CudaModulusSwitchNoiseReductionKeyFFI *ms_noise_reduction_key);
-
-void cleanup_cuda_integer_count_of_consecutive_bits_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr_void);
-
-uint64_t scratch_cuda_integer_grouped_oprf_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
-    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
-    uint32_t num_blocks_to_process, uint32_t num_blocks,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    bool allocate_gpu_memory, uint32_t message_bits_per_block,
-    uint32_t total_random_bits, PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_integer_grouped_oprf_async_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *radix_lwe_out, const void *seeded_lwe_input,
-    uint32_t num_blocks_to_process, int8_t *mem, void *const *bsks,
-    CudaModulusSwitchNoiseReductionKeyFFI const *ms_noise_reduction_key);
-
-void cleanup_cuda_integer_grouped_oprf_64(void *const *streams,
-                                          uint32_t const *gpu_indexes,
-                                          uint32_t gpu_count,
-                                          int8_t **mem_ptr_void);
-
-uint64_t scratch_integer_ilog2_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    int8_t **mem_ptr, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t lwe_dimension, uint32_t ks_level, uint32_t ks_base_log,
-    uint32_t pbs_level, uint32_t pbs_base_log, uint32_t grouping_factor,
-    uint32_t message_modulus, uint32_t carry_modulus, PBS_TYPE pbs_type,
-    uint32_t input_num_blocks, uint32_t counter_num_blocks,
-    uint32_t num_bits_in_ciphertext, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
-
-void cuda_integer_ilog2_kb_64(
-    void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,
-    CudaRadixCiphertextFFI *output_ct, CudaRadixCiphertextFFI const *input_ct,
-    CudaRadixCiphertextFFI const *trivial_ct_neg_n,
-    CudaRadixCiphertextFFI const *trivial_ct_2,
-    CudaRadixCiphertextFFI const *trivial_ct_m_minus_1_block, int8_t *mem_ptr,
-    void *const *bsks, void *const *ksks,
-    const CudaModulusSwitchNoiseReductionKeyFFI *ms_noise_reduction_key);
-
-void cleanup_cuda_integer_ilog2_kb_64(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/pbs/pbs_128_utilities.h

@@ -1,13 +0,0 @@
-#ifndef CUDA_BOOTSTRAP_128_H
-#define CUDA_BOOTSTRAP_128_H
-
-#include "pbs_enums.h"
-#include <stdint.h>
-
-uint64_t scratch_cuda_programmable_bootstrap_128_vector_64(
-    void *stream, uint32_t gpu_index, int8_t **pbs_buffer,
-    uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
-    uint32_t level_count, uint32_t input_lwe_ciphertext_count,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
-
-#endif // CUDA_BOOTSTRAP_128_H

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

@@ -3,7 +3,6 @@
 #include <stdint.h>
 enum PBS_TYPE { MULTI_BIT = 0, CLASSICAL = 1 };
 enum PBS_VARIANT { DEFAULT = 0, CG = 1, TBC = 2 };
-enum PBS_MS_REDUCTION_T { NO_REDUCTION = 0, DRIFT = 1, CENTERED = 2 };
 
 extern "C" {
 typedef struct {

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

@@ -66,9 +66,6 @@ void cuda_multi_bit_programmable_bootstrap_lwe_ciphertext_vector(
     uint32_t num_many_lut, uint32_t lut_stride);
 
 template <typename Torus>
-uint64_t get_buffer_size_full_sm_multibit_programmable_bootstrap_128_keybundle(
-    uint32_t polynomial_size);
-template <typename Torus>
 uint64_t get_buffer_size_full_sm_multibit_programmable_bootstrap_keybundle(
     uint32_t polynomial_size);
 template <typename Torus>
@@ -98,12 +95,8 @@ uint64_t get_buffer_size_full_sm_tbc_multibit_programmable_bootstrap(
 
 template <typename Torus, class params>
 uint32_t get_lwe_chunk_size(uint32_t gpu_index, uint32_t max_num_pbs,
-                            uint32_t polynomial_size,
-                            uint64_t full_sm_keybundle);
-template <typename Torus, class params>
-uint32_t get_lwe_chunk_size_128(uint32_t gpu_index, uint32_t max_num_pbs,
-                                uint32_t polynomial_size,
-                                uint64_t full_sm_keybundle);
+                            uint32_t polynomial_size);
+
 template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::MULTI_BIT> {
   int8_t *d_mem_keybundle = NULL;
   int8_t *d_mem_acc_step_one = NULL;
@@ -122,7 +115,7 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::MULTI_BIT> {
              uint32_t polynomial_size, uint32_t level_count,
              uint32_t input_lwe_ciphertext_count, uint32_t lwe_chunk_size,
              PBS_VARIANT pbs_variant, bool allocate_gpu_memory,
-             uint64_t &size_tracker) {
+             uint64_t *size_tracker) {
     gpu_memory_allocated = allocate_gpu_memory;
     cuda_set_device(gpu_index);
 
@@ -288,146 +281,4 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::MULTI_BIT> {
   }
 };
 
-template <typename InputTorus>
-struct pbs_buffer_128<InputTorus, PBS_TYPE::MULTI_BIT> {
-  int8_t *d_mem_keybundle = NULL;
-  int8_t *d_mem_acc_step_one = NULL;
-  int8_t *d_mem_acc_step_two = NULL;
-  int8_t *d_mem_acc_cg = NULL;
-  int8_t *d_mem_acc_tbc = NULL;
-  uint32_t lwe_chunk_size;
-  double *keybundle_fft;
-  __uint128_t *global_accumulator;
-  double *global_join_buffer;
-
-  PBS_VARIANT pbs_variant;
-  bool gpu_memory_allocated;
-
-  pbs_buffer_128(cudaStream_t stream, uint32_t gpu_index,
-                 uint32_t glwe_dimension, uint32_t polynomial_size,
-                 uint32_t level_count, uint32_t input_lwe_ciphertext_count,
-                 uint32_t lwe_chunk_size, PBS_VARIANT pbs_variant,
-                 bool allocate_gpu_memory, uint64_t &size_tracker) {
-    gpu_memory_allocated = allocate_gpu_memory;
-    cuda_set_device(gpu_index);
-
-    this->pbs_variant = pbs_variant;
-    this->lwe_chunk_size = lwe_chunk_size;
-    auto max_shared_memory = cuda_get_max_shared_memory(gpu_index);
-
-    // default
-    uint64_t full_sm_keybundle =
-        get_buffer_size_full_sm_multibit_programmable_bootstrap_128_keybundle<
-            __uint128_t>(polynomial_size);
-    uint64_t full_sm_accumulate_step_one =
-        get_buffer_size_full_sm_multibit_programmable_bootstrap_step_one<
-            __uint128_t>(polynomial_size);
-    uint64_t full_sm_accumulate_step_two =
-        get_buffer_size_full_sm_multibit_programmable_bootstrap_step_two<
-            __uint128_t>(polynomial_size);
-    uint64_t partial_sm_accumulate_step_one =
-        get_buffer_size_partial_sm_multibit_programmable_bootstrap_step_one<
-            __uint128_t>(polynomial_size);
-    // cg
-    uint64_t full_sm_cg_accumulate =
-        get_buffer_size_full_sm_cg_multibit_programmable_bootstrap<__uint128_t>(
-            polynomial_size);
-    uint64_t partial_sm_cg_accumulate =
-        get_buffer_size_partial_sm_cg_multibit_programmable_bootstrap<
-            __uint128_t>(polynomial_size);
-
-    auto num_blocks_keybundle = input_lwe_ciphertext_count * lwe_chunk_size *
-                                (glwe_dimension + 1) * (glwe_dimension + 1) *
-                                level_count;
-    auto num_blocks_acc_step_one =
-        level_count * (glwe_dimension + 1) * input_lwe_ciphertext_count;
-    auto num_blocks_acc_step_two =
-        input_lwe_ciphertext_count * (glwe_dimension + 1);
-    auto num_blocks_acc_cg =
-        level_count * (glwe_dimension + 1) * input_lwe_ciphertext_count;
-
-    // Keybundle
-    if (max_shared_memory < full_sm_keybundle)
-      d_mem_keybundle = (int8_t *)cuda_malloc_with_size_tracking_async(
-          num_blocks_keybundle * full_sm_keybundle, stream, gpu_index,
-          size_tracker, allocate_gpu_memory);
-
-    switch (pbs_variant) {
-    case PBS_VARIANT::CG:
-      // Accumulator CG
-      if (max_shared_memory < partial_sm_cg_accumulate)
-        d_mem_acc_cg = (int8_t *)cuda_malloc_with_size_tracking_async(
-            num_blocks_acc_cg * full_sm_cg_accumulate, stream, gpu_index,
-            size_tracker, allocate_gpu_memory);
-      else if (max_shared_memory < full_sm_cg_accumulate)
-        d_mem_acc_cg = (int8_t *)cuda_malloc_with_size_tracking_async(
-            num_blocks_acc_cg * partial_sm_cg_accumulate, stream, gpu_index,
-            size_tracker, allocate_gpu_memory);
-      break;
-    case PBS_VARIANT::DEFAULT:
-      // Accumulator step one
-      if (max_shared_memory < partial_sm_accumulate_step_one)
-        d_mem_acc_step_one = (int8_t *)cuda_malloc_with_size_tracking_async(
-            num_blocks_acc_step_one * full_sm_accumulate_step_one, stream,
-            gpu_index, size_tracker, allocate_gpu_memory);
-      else if (max_shared_memory < full_sm_accumulate_step_one)
-        d_mem_acc_step_one = (int8_t *)cuda_malloc_with_size_tracking_async(
-            num_blocks_acc_step_one * partial_sm_accumulate_step_one, stream,
-            gpu_index, size_tracker, allocate_gpu_memory);
-
-      // Accumulator step two
-      if (max_shared_memory < full_sm_accumulate_step_two)
-        d_mem_acc_step_two = (int8_t *)cuda_malloc_with_size_tracking_async(
-            num_blocks_acc_step_two * full_sm_accumulate_step_two, stream,
-            gpu_index, size_tracker, allocate_gpu_memory);
-      break;
-    default:
-      PANIC("Cuda error (PBS): unsupported implementation variant.")
-    }
-
-    keybundle_fft = (double *)cuda_malloc_with_size_tracking_async(
-        num_blocks_keybundle * (polynomial_size / 2) * 4 * sizeof(double),
-        stream, gpu_index, size_tracker, allocate_gpu_memory);
-    global_accumulator = (__uint128_t *)cuda_malloc_with_size_tracking_async(
-        input_lwe_ciphertext_count * (glwe_dimension + 1) * polynomial_size *
-            sizeof(__uint128_t),
-        stream, gpu_index, size_tracker, allocate_gpu_memory);
-    global_join_buffer = (double *)cuda_malloc_with_size_tracking_async(
-        level_count * (glwe_dimension + 1) * input_lwe_ciphertext_count *
-            (polynomial_size / 2) * 4 * sizeof(double),
-        stream, gpu_index, size_tracker, allocate_gpu_memory);
-  }
-
-  void release(cudaStream_t stream, uint32_t gpu_index) {
-
-    if (d_mem_keybundle)
-      cuda_drop_with_size_tracking_async(d_mem_keybundle, stream, gpu_index,
-                                         gpu_memory_allocated);
-    switch (pbs_variant) {
-    case DEFAULT:
-      if (d_mem_acc_step_one)
-        cuda_drop_with_size_tracking_async(d_mem_acc_step_one, stream,
-                                           gpu_index, gpu_memory_allocated);
-      if (d_mem_acc_step_two)
-        cuda_drop_with_size_tracking_async(d_mem_acc_step_two, stream,
-                                           gpu_index, gpu_memory_allocated);
-      break;
-    case CG:
-      if (d_mem_acc_cg)
-        cuda_drop_with_size_tracking_async(d_mem_acc_cg, stream, gpu_index,
-                                           gpu_memory_allocated);
-      break;
-    default:
-      PANIC("Cuda error (PBS): unsupported implementation variant.")
-    }
-
-    cuda_drop_with_size_tracking_async(keybundle_fft, stream, gpu_index,
-                                       gpu_memory_allocated);
-    cuda_drop_with_size_tracking_async(global_accumulator, stream, gpu_index,
-                                       gpu_memory_allocated);
-    cuda_drop_with_size_tracking_async(global_join_buffer, stream, gpu_index,
-                                       gpu_memory_allocated);
-  }
-};
-
 #endif // CUDA_MULTI_BIT_UTILITIES_H

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

@@ -83,24 +83,23 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
   Torus *temp_lwe_array_in;
 
   PBS_VARIANT pbs_variant;
-  PBS_MS_REDUCTION_T noise_reduction_type;
+  bool uses_noise_reduction;
   bool gpu_memory_allocated;
 
   pbs_buffer(cudaStream_t stream, uint32_t gpu_index, uint32_t lwe_dimension,
              uint32_t glwe_dimension, uint32_t polynomial_size,
              uint32_t level_count, uint32_t input_lwe_ciphertext_count,
              PBS_VARIANT pbs_variant, bool allocate_gpu_memory,
-             PBS_MS_REDUCTION_T noise_reduction_type, uint64_t &size_tracker)
-      : noise_reduction_type(noise_reduction_type) {
+             bool allocate_ms_array, uint64_t *size_tracker) {
     gpu_memory_allocated = allocate_gpu_memory;
     cuda_set_device(gpu_index);
+    this->uses_noise_reduction = allocate_ms_array;
     this->pbs_variant = pbs_variant;
 
     auto max_shared_memory = cuda_get_max_shared_memory(gpu_index);
     this->temp_lwe_array_in = (Torus *)cuda_malloc_with_size_tracking_async(
         (lwe_dimension + 1) * input_lwe_ciphertext_count * sizeof(Torus),
-        stream, gpu_index, size_tracker,
-        noise_reduction_type == PBS_MS_REDUCTION_T::DRIFT);
+        stream, gpu_index, size_tracker, allocate_ms_array);
     switch (pbs_variant) {
     case PBS_VARIANT::DEFAULT: {
       uint64_t full_sm_step_one =
@@ -235,48 +234,45 @@ template <typename Torus> struct pbs_buffer<Torus, PBS_TYPE::CLASSICAL> {
       cuda_drop_with_size_tracking_async(global_accumulator, stream, gpu_index,
                                          gpu_memory_allocated);
 
-    if (noise_reduction_type == PBS_MS_REDUCTION_T::DRIFT)
+    if (uses_noise_reduction)
       cuda_drop_with_size_tracking_async(temp_lwe_array_in, stream, gpu_index,
                                          gpu_memory_allocated);
   }
 };
 
-template <typename Torus, PBS_TYPE pbs_type> struct pbs_buffer_128;
+template <PBS_TYPE pbs_type> struct pbs_buffer_128;
 
-template <typename InputTorus>
-struct pbs_buffer_128<InputTorus, PBS_TYPE::CLASSICAL> {
+template <> struct pbs_buffer_128<PBS_TYPE::CLASSICAL> {
   int8_t *d_mem;
 
   __uint128_t *global_accumulator;
   double *global_join_buffer;
-  InputTorus *temp_lwe_array_in;
+  __uint128_t *temp_lwe_array_in;
   uint64_t *trivial_indexes;
 
   PBS_VARIANT pbs_variant;
-  PBS_MS_REDUCTION_T noise_reduction_type;
+  bool uses_noise_reduction;
   bool gpu_memory_allocated;
 
   pbs_buffer_128(cudaStream_t stream, uint32_t gpu_index,
                  uint32_t lwe_dimension, uint32_t glwe_dimension,
                  uint32_t polynomial_size, uint32_t level_count,
                  uint32_t input_lwe_ciphertext_count, PBS_VARIANT pbs_variant,
-                 bool allocate_gpu_memory,
-                 PBS_MS_REDUCTION_T noise_reduction_type,
-                 uint64_t &size_tracker)
-      : noise_reduction_type(noise_reduction_type) {
+                 bool allocate_gpu_memory, bool allocate_ms_array,
+                 uint64_t *size_tracker) {
     gpu_memory_allocated = allocate_gpu_memory;
     cuda_set_device(gpu_index);
     this->pbs_variant = pbs_variant;
-
-    if (noise_reduction_type == PBS_MS_REDUCTION_T::DRIFT) {
+    this->uses_noise_reduction = allocate_ms_array;
+    if (allocate_ms_array) {
       this->temp_lwe_array_in =
-          (InputTorus *)cuda_malloc_with_size_tracking_async(
+          (__uint128_t *)cuda_malloc_with_size_tracking_async(
               (lwe_dimension + 1) * input_lwe_ciphertext_count *
-                  sizeof(InputTorus),
-              stream, gpu_index, size_tracker, allocate_gpu_memory);
+                  sizeof(__uint128_t),
+              stream, gpu_index, size_tracker, allocate_ms_array);
       this->trivial_indexes = (uint64_t *)cuda_malloc_with_size_tracking_async(
           input_lwe_ciphertext_count * sizeof(uint64_t), stream, gpu_index,
-          size_tracker, allocate_gpu_memory);
+          size_tracker, allocate_ms_array);
       uint64_t *h_trivial_indexes = new uint64_t[input_lwe_ciphertext_count];
       for (uint32_t i = 0; i < input_lwe_ciphertext_count; i++)
         h_trivial_indexes[i] = i;
@@ -425,7 +421,7 @@ struct pbs_buffer_128<InputTorus, PBS_TYPE::CLASSICAL> {
       cuda_drop_with_size_tracking_async(global_accumulator, stream, gpu_index,
                                          gpu_memory_allocated);
 
-    if (noise_reduction_type == PBS_MS_REDUCTION_T::DRIFT) {
+    if (uses_noise_reduction) {
       cuda_drop_with_size_tracking_async(temp_lwe_array_in, stream, gpu_index,
                                          gpu_memory_allocated);
       cuda_drop_with_size_tracking_async(trivial_indexes, stream, gpu_index,
@@ -505,7 +501,7 @@ uint64_t scratch_cuda_programmable_bootstrap_tbc(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, CLASSICAL> **pbs_buffer,
     uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
     uint32_t level_count, uint32_t input_lwe_ciphertext_count,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 #endif
 
 template <typename Torus>
@@ -513,14 +509,14 @@ uint64_t scratch_cuda_programmable_bootstrap_cg(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, CLASSICAL> **pbs_buffer,
     uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
     uint32_t level_count, uint32_t input_lwe_ciphertext_count,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 template <typename Torus>
 uint64_t scratch_cuda_programmable_bootstrap(
     void *stream, uint32_t gpu_index, pbs_buffer<Torus, CLASSICAL> **buffer,
     uint32_t lwe_dimension, uint32_t glwe_dimension, uint32_t polynomial_size,
     uint32_t level_count, uint32_t input_lwe_ciphertext_count,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 template <typename Torus>
 bool has_support_to_cuda_programmable_bootstrap_tbc(uint32_t num_samples,
@@ -529,10 +525,6 @@ bool has_support_to_cuda_programmable_bootstrap_tbc(uint32_t num_samples,
                                                     uint32_t level_count,
                                                     uint32_t max_shared_memory);
 
-bool has_support_to_cuda_programmable_bootstrap_128_cg(
-    uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
-    uint32_t num_samples, uint32_t max_shared_memory);
-
 #ifdef __CUDACC__
 __device__ inline int get_start_ith_ggsw(int i, uint32_t polynomial_size,
                                          int glwe_dimension,
@@ -549,11 +541,6 @@ __device__ T *get_ith_mask_kth_block(T *ptr, int i, int k, int level,
                                      uint32_t polynomial_size,
                                      int glwe_dimension, uint32_t level_count);
 
-template <typename T, uint32_t polynomial_size, uint32_t glwe_dimension,
-          uint32_t level_count, uint32_t level_id>
-__device__ const T *get_ith_mask_kth_block_2_2_params(const T *ptr,
-                                                      int iteration, int k);
-
 template <typename T>
 __device__ T *get_ith_body_kth_block(T *ptr, int i, int k, int level,
                                      uint32_t polynomial_size,

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

@@ -61,19 +61,19 @@ uint64_t scratch_cuda_programmable_bootstrap_32(
     void *stream, uint32_t gpu_index, int8_t **buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
     uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_ms_array);
 
 uint64_t scratch_cuda_programmable_bootstrap_64(
     void *stream, uint32_t gpu_index, int8_t **buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
     uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_ms_array);
 
 uint64_t scratch_cuda_programmable_bootstrap_128(
     void *stream, uint32_t gpu_index, int8_t **buffer, uint32_t lwe_dimension,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
     uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory,
-    PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_ms_array);
 
 void cuda_programmable_bootstrap_lwe_ciphertext_vector_32(
     void *stream, uint32_t gpu_index, void *lwe_array_out,
@@ -100,7 +100,7 @@ void cuda_programmable_bootstrap_lwe_ciphertext_vector_128(
     void const *lut_vector, void const *lwe_array_in,
     void const *bootstrapping_key,
     CudaModulusSwitchNoiseReductionKeyFFI const *ms_noise_reduction_key,
-    void const *ms_noise_reduction_ptr, int8_t *buffer, uint32_t lwe_dimension,
+    void *ms_noise_reduction_ptr, 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);
 

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

@@ -15,11 +15,6 @@ void cuda_convert_lwe_multi_bit_programmable_bootstrap_key_64(
     uint32_t input_lwe_dim, uint32_t glwe_dim, uint32_t level_count,
     uint32_t polynomial_size, uint32_t grouping_factor);
 
-void cuda_convert_lwe_multi_bit_programmable_bootstrap_key_128(
-    void *stream, uint32_t gpu_index, void *dest, void const *src,
-    uint32_t input_lwe_dim, uint32_t glwe_dim, uint32_t level_count,
-    uint32_t polynomial_size, uint32_t grouping_factor);
-
 uint64_t scratch_cuda_multi_bit_programmable_bootstrap_64(
     void *stream, uint32_t gpu_index, int8_t **pbs_buffer,
     uint32_t glwe_dimension, uint32_t polynomial_size, uint32_t level_count,
@@ -38,25 +33,6 @@ void cuda_multi_bit_programmable_bootstrap_lwe_ciphertext_vector_64(
 void cleanup_cuda_multi_bit_programmable_bootstrap(void *stream,
                                                    uint32_t gpu_index,
                                                    int8_t **pbs_buffer);
-
-uint64_t scratch_cuda_multi_bit_programmable_bootstrap_128_vector_64(
-    void *stream, uint32_t gpu_index, int8_t **buffer, uint32_t glwe_dimension,
-    uint32_t polynomial_size, uint32_t level_count,
-    uint32_t input_lwe_ciphertext_count, bool allocate_gpu_memory);
-
-void cuda_multi_bit_programmable_bootstrap_lwe_ciphertext_vector_128(
-    void *stream, uint32_t gpu_index, void *lwe_array_out,
-    void const *lwe_output_indexes, void const *lut_vector,
-    void const *lut_vector_indexes, void const *lwe_array_in,
-    void const *lwe_input_indexes, void const *bootstrapping_key,
-    int8_t *mem_ptr, 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);
-
-void cleanup_cuda_multi_bit_programmable_bootstrap_128(void *stream,
-                                                       const uint32_t gpu_index,
-                                                       int8_t **buffer);
 }
 
 #endif // CUDA_MULTI_BIT_H

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

@@ -24,7 +24,7 @@ uint64_t scratch_cuda_expand_without_verification_64(
     const uint32_t *num_lwes_per_compact_list, const bool *is_boolean_array,
     uint32_t num_compact_lists, uint32_t message_modulus,
     uint32_t carry_modulus, PBS_TYPE pbs_type, KS_TYPE casting_key_type,
-    bool allocate_gpu_memory, PBS_MS_REDUCTION_T noise_reduction_type);
+    bool allocate_gpu_memory, bool allocate_ms_array);
 
 void cuda_expand_without_verification_64(
     void *const *streams, uint32_t const *gpu_indexes, uint32_t gpu_count,

backends/tfhe-cuda-backend/cuda/include/zk/zk_utilities.h

@@ -27,7 +27,7 @@ template <typename Torus> struct zk_expand_mem {
                 int_radix_params casting_params, KS_TYPE casting_key_type,
                 const uint32_t *num_lwes_per_compact_list,
                 const bool *is_boolean_array, uint32_t num_compact_lists,
-                bool allocate_gpu_memory, uint64_t &size_tracker)
+                bool allocate_gpu_memory, uint64_t *size_tracker)
       : computing_params(computing_params), casting_params(casting_params),
         num_compact_lists(num_compact_lists),
         casting_key_type(casting_key_type) {
@@ -112,15 +112,15 @@ template <typename Torus> struct zk_expand_mem {
 
     // Hint for future readers: if message_modulus == 4 then
     // packed_messages_per_lwe becomes 2
-    auto num_packed_msgs = log2_int(params.message_modulus);
+    auto packed_messages_per_lwe = log2_int(params.message_modulus);
 
     // Adjust indexes to permute the output and access the correct LUT
     auto h_indexes_in = static_cast<Torus *>(
-        malloc(num_packed_msgs * num_lwes * sizeof(Torus)));
+        malloc(packed_messages_per_lwe * num_lwes * sizeof(Torus)));
     auto h_indexes_out = static_cast<Torus *>(
-        malloc(num_packed_msgs * num_lwes * sizeof(Torus)));
+        malloc(packed_messages_per_lwe * num_lwes * sizeof(Torus)));
     auto h_lut_indexes = static_cast<Torus *>(
-        malloc(num_packed_msgs * num_lwes * sizeof(Torus)));
+        malloc(packed_messages_per_lwe * num_lwes * sizeof(Torus)));
     auto h_body_id_per_compact_list =
         static_cast<uint32_t *>(malloc(num_lwes * sizeof(uint32_t)));
     auto h_lwe_compact_input_indexes =
@@ -138,10 +138,6 @@ template <typename Torus> struct zk_expand_mem {
     auto compact_list_id = 0;
     auto idx = 0;
     auto count = 0;
-    // During flattening, all num_lwes LWEs from all compact lists are stored
-    // sequentially on a Torus array. h_lwe_compact_input_indexes stores the
-    // index of the first LWE related to the compact list that contains the i-th
-    // LWE
     for (int i = 0; i < num_lwes; i++) {
       h_lwe_compact_input_indexes[i] = idx;
       count++;
@@ -152,8 +148,6 @@ template <typename Torus> struct zk_expand_mem {
       }
     }
 
-    // Stores the index of the i-th LWE (within each compact list) related to
-    // the k-th compact list.
     auto offset = 0;
     for (int k = 0; k < num_compact_lists; k++) {
       auto num_lwes_in_kth_compact_list = num_lwes_per_compact_list[k];
@@ -165,80 +159,46 @@ template <typename Torus> struct zk_expand_mem {
       offset += num_lwes_in_kth_compact_list;
     }
 
-    /*
-     * Each LWE contains encrypted data in both carry and message spaces
-     * that needs to be extracted.
-     *
-     * The loop processes each compact list (k) and for each LWE within that
-     * list:
-     * 1. Sets input indexes to read each LWE twice (for carry and message
-     * extraction)
-     * 2. Creates output indexes to properly reorder the results
-     * 3. Selects appropriate LUT index based on whether boolean sanitization is
-     * needed
-     *
-     * We want the output to have always first the content of the message part
-     * and then the content of the carry part of each LWE.
-     *
-     * i.e. msg_extract(LWE_0), carry_extract(LWE_0), msg_extract(LWE_1),
-     * carry_extract(LWE_1), ...
-     *
-     * Aiming that behavior, with 4 LWEs we would have:
-     *
-     * // Each LWE is processed twice
-     * h_indexes_in   = {0, 1, 2, 3, 0, 1, 2, 3}
-     *
-     * // First 4 use message LUT, last 4 use carry LUT
-     * h_lut_indexes  = {0, 0, 0, 0, 1, 1, 1, 1}
-     *
-     * // Reorders output so message and carry for each LWE appear together
-     * h_indexes_out  = {0, 2, 4, 6, 1, 3, 5, 7}
-     *
-     * If an LWE contains a boolean value, its LUT index is shifted by
-     * num_packed_msgs to use the sanitization LUT (which ensures output is
-     * exactly 0 or 1).
-     */
     offset = 0;
     for (int k = 0; k < num_compact_lists; k++) {
-      auto num_lwes_in_kth = num_lwes_per_compact_list[k];
-      for (int i = 0; i < num_packed_msgs * num_lwes_in_kth; i++) {
-        auto lwe_index = i + num_packed_msgs * offset;
-        auto lwe_index_in_list = i % num_lwes_in_kth;
-        h_indexes_in[lwe_index] = lwe_index_in_list + offset;
-        h_indexes_out[lwe_index] =
-            num_packed_msgs * h_indexes_in[lwe_index] + i / num_lwes_in_kth;
+      auto num_lwes_in_kth_compact_list = num_lwes_per_compact_list[k];
+      for (int i = 0;
+           i < packed_messages_per_lwe * num_lwes_in_kth_compact_list; i++) {
+        Torus j = i % num_lwes_in_kth_compact_list;
+        h_indexes_in[i + packed_messages_per_lwe * offset] = j + offset;
+        h_indexes_out[i + packed_messages_per_lwe * offset] =
+            packed_messages_per_lwe * (j + offset) +
+            (i / num_lwes_in_kth_compact_list);
         // If the input relates to a boolean, shift the LUT so the correct one
         // with sanitization is used
-        auto boolean_offset =
-            is_boolean_array[h_indexes_out[lwe_index]] ? num_packed_msgs : 0;
-        h_lut_indexes[lwe_index] = i / num_lwes_in_kth + boolean_offset;
+        h_lut_indexes[i + packed_messages_per_lwe * offset] =
+            (is_boolean_array[h_indexes_out[i +
+                                            packed_messages_per_lwe * offset]]
+                 ? packed_messages_per_lwe
+                 : 0) +
+            i / num_lwes_in_kth_compact_list;
       }
-      offset += num_lwes_in_kth;
+      offset += num_lwes_in_kth_compact_list;
     }
 
     message_and_carry_extract_luts->set_lwe_indexes(
         streams[0], gpu_indexes[0], h_indexes_in, h_indexes_out);
     auto lut_indexes = message_and_carry_extract_luts->get_lut_indexes(0, 0);
+    message_and_carry_extract_luts->broadcast_lut(streams, gpu_indexes, 0);
 
     cuda_memcpy_with_size_tracking_async_to_gpu(
         d_lwe_compact_input_indexes, h_lwe_compact_input_indexes,
         num_lwes * sizeof(uint32_t), streams[0], gpu_indexes[0],
         allocate_gpu_memory);
     cuda_memcpy_with_size_tracking_async_to_gpu(
-        lut_indexes, h_lut_indexes, num_packed_msgs * num_lwes * sizeof(Torus),
-        streams[0], gpu_indexes[0], allocate_gpu_memory);
+        lut_indexes, h_lut_indexes,
+        packed_messages_per_lwe * num_lwes * sizeof(Torus), streams[0],
+        gpu_indexes[0], allocate_gpu_memory);
     cuda_memcpy_with_size_tracking_async_to_gpu(
         d_body_id_per_compact_list, h_body_id_per_compact_list,
         num_lwes * sizeof(uint32_t), streams[0], gpu_indexes[0],
         allocate_gpu_memory);
 
-    auto active_gpu_count = get_active_gpu_count(2 * num_lwes, gpu_count);
-    message_and_carry_extract_luts->broadcast_lut(streams, gpu_indexes,
-                                                  active_gpu_count);
-
-    message_and_carry_extract_luts->allocate_lwe_vector_for_non_trivial_indexes(
-        streams, gpu_indexes, active_gpu_count, 2 * num_lwes, size_tracker,
-        allocate_gpu_memory);
     // The expanded LWEs will always be on the casting key format
     tmp_expanded_lwes = (Torus *)cuda_malloc_with_size_tracking_async(
         num_lwes * (casting_params.big_lwe_dimension + 1) * sizeof(Torus),