chore: fix cleatext -> cleartext typo

- base_n has a different meaning in the tfhe-fft code, to make it less
confusing when working on such code, renamed base_n

.github/workflows/aws_data_tests.yml

@@ -54,7 +54,7 @@ jobs:
 
       - name: Retrieve data from cache
         id: retrieve-data-cache
-        uses: actions/cache/restore@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/restore@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         with:
           path: |
             utils/tfhe-backward-compat-data/**/*.cbor
@@ -89,7 +89,7 @@ jobs:
       - name: Store data in cache
         if: steps.retrieve-data-cache.outputs.cache-hit != 'true'
         continue-on-error: true
-        uses: actions/cache/save@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/save@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         with:
           path: |
             utils/tfhe-backward-compat-data/**/*.cbor

.github/workflows/aws_tfhe_fast_tests.yml

@@ -69,7 +69,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             dependencies:
@@ -200,7 +200,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/restore@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         with:
           path: |
             ~/.nvm
@@ -213,7 +213,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/save@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |

.github/workflows/aws_tfhe_integer_tests.yml

@@ -56,7 +56,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             integer:

.github/workflows/aws_tfhe_signed_integer_tests.yml

@@ -57,7 +57,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             integer:

.github/workflows/aws_tfhe_tests.yml

@@ -78,7 +78,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             dependencies:

.github/workflows/aws_tfhe_wasm_tests.yml

@@ -45,7 +45,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             wasm:
@@ -92,7 +92,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/restore@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         with:
           path: |
             ~/.nvm
@@ -105,7 +105,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/save@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |

.github/workflows/backward_compat_pr_change_report.yml

@@ -34,7 +34,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             backward:
@@ -79,19 +79,11 @@ jobs:
             exit 1
           fi
 
-      - name: Find existing comment
+      - name: Post/refresh backward-compat report
         if: steps.report.outputs.has_report == 'true'
-        id: find-comment
-        uses: peter-evans/find-comment@b30e6a3c0ed37e7c023ccd3f1db5c6c0b0c23aad # v4.0.0
+        uses: marocchino/sticky-pull-request-comment@0ea0beb66eb9baf113663a64ec522f60e49231c0
         with:
-          issue-number: ${{ github.event.pull_request.number }}
-          body-includes: '**Backward-compat snapshot:'
-
-      - name: Comment on PR
-        if: steps.report.outputs.has_report == 'true'
-        uses: peter-evans/create-or-update-comment@e8674b075228eee787fea43ef493e45ece1004c9 # v5.0.0
-        with:
-          comment-id: ${{ steps.find-comment.outputs.comment-id }}
-          issue-number: ${{ github.event.pull_request.number }}
-          body-path: report.md
-          edit-mode: replace
+          header: backward-compat-snapshot
+          hide_and_recreate: true
+          hide_classify: OUTDATED
+          path: report.md

.github/workflows/benchmark_cpu.yml

@@ -19,7 +19,7 @@ on:
           - shortint_oprf
           - hlapi_unsigned
           - hlapi_signed
-          - hlapi_erc20
+          - hlapi_erc7984
           - hlapi_dex
           - hlapi_noise_squash
           - hlapi_kvstore
@@ -93,8 +93,8 @@ jobs:
 
           if inputs_command == "integer_zk":
             files_to_parse.append("pke_zk_crs_sizes.csv")
-          elif inputs_command == "hlapi_erc20":
-            files_to_parse.append("erc20_pbs_count.csv")
+          elif inputs_command == "hlapi_erc7984":
+            files_to_parse.append("erc7984_pbs_count.csv")
           elif inputs_command == "hlapi_dex":
             files_to_parse.extend(
               [

.github/workflows/benchmark_cpu_common.yml

@@ -223,7 +223,7 @@ jobs:
           results_type: ${{ inputs.additional_results_type }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}_${{ matrix.bench_type }}_${{ matrix.params_type }}
           path: ${{ env.RESULTS_FILENAME }}

.github/workflows/benchmark_cpu_weekly.yml

@@ -108,14 +108,14 @@ jobs:
       SLAB_URL: ${{ secrets.SLAB_URL }}
       SLAB_BASE_URL: ${{ secrets.SLAB_BASE_URL }}
 
-  run-benchmarks-hlapi-erc20:
-    name: benchmark_cpu_weekly/run-benchmarks-hlapi-erc20
+  run-benchmarks-hlapi-erc7984:
+    name: benchmark_cpu_weekly/run-benchmarks-hlapi-erc7984
     if: needs.prepare-inputs.outputs.is_weekly_bench_group_2 == 'true'
     needs: prepare-inputs
     uses: ./.github/workflows/benchmark_cpu_common.yml
     with:
-      command: hlapi_erc20
-      additional_file_to_parse: erc20_pbs_count.csv
+      command: hlapi_erc7984
+      additional_file_to_parse: erc7984_pbs_count.csv
     secrets:
       BOT_USERNAME: ${{ secrets.BOT_USERNAME }}
       SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }}

.github/workflows/benchmark_ct_key_sizes.yml

@@ -99,7 +99,7 @@ jobs:
           --append-results
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_ct_key_sizes
           path: ${{ env.RESULTS_FILENAME }}

.github/workflows/benchmark_documentation.yml

@@ -17,6 +17,10 @@ on:
         description: "Run GPU core-crypto benchmarks"
         type: boolean
         default: true
+      run-gpu-zk-benchmarks:
+        description: "Run GPU ZK benchmarks"
+        type: boolean
+        default: true
       run-hpu-benchmarks:
         description: "Run HPU benchmarks"
         type: boolean
@@ -36,7 +40,7 @@ jobs:
     uses: ./.github/workflows/benchmark_cpu_common.yml
     if: inputs.run-cpu-benchmarks
     with:
-      command: integer,hlapi_erc20
+      command: integer,hlapi_erc7984
       op_flavor: fast_default
       bench_type: both
       precisions_set: documentation
@@ -91,7 +95,7 @@ jobs:
     with:
       profile: multi-h100-sxm5
       hardware_name: n3-H100-SXM5x8
-      command: integer_multi_bit,hlapi_erc20
+      command: integer_multi_bit,hlapi_erc7984
       op_flavor: fast_default
       bench_type: both
       precisions_set: documentation
@@ -110,7 +114,7 @@ jobs:
     uses: ./.github/workflows/benchmark_hpu_common.yml
     if: inputs.run-hpu-benchmarks
     with:
-      command: integer,hlapi_erc20
+      command: integer,hlapi_erc7984
       op_flavor: default
       bench_type: both
       precisions_set: documentation
@@ -165,21 +169,42 @@ jobs:
       SLAB_URL: ${{ secrets.SLAB_URL }}
       SLAB_BASE_URL: ${{ secrets.SLAB_BASE_URL }}
 
+  run-benchmarks-gpu-zk-server:
+    name: benchmark_documentation/run-benchmarks-gpu-zk-server
+    uses: ./.github/workflows/benchmark_gpu_common.yml
+    if: inputs.run-gpu-zk-benchmarks
+    with:
+      profile: multi-h100-sxm5
+      hardware_name: n3-H100-SXM5x8
+      command: integer_zk
+      op_flavor: default
+      bench_type: both
+    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 }}
+
   generate-svgs-with-benchmarks-run:
     name: benchmark-documentation/generate-svgs-with-benchmarks-run
     if: ${{ always() &&
-      (inputs.run-cpu-benchmarks || inputs.run-gpu-integer-benchmarks || inputs.run-gpu-core-crypto-benchmarks ||inputs.run-hpu-benchmarks) &&
+      (inputs.run-cpu-benchmarks || inputs.run-gpu-integer-benchmarks || inputs.run-gpu-core-crypto-benchmarks || inputs.run-gpu-zk-benchmarks || inputs.run-hpu-benchmarks) &&
       inputs.generate-svgs }}
     needs: [
       run-benchmarks-cpu-integer, run-benchmarks-gpu-integer, run-benchmarks-hpu-integer,
       run-benchmarks-cpu-zk-server, run-benchmarks-cpu-zk-client,
-      run-benchmarks-cpu-core-crypto, run-benchmarks-gpu-core-crypto
+      run-benchmarks-cpu-core-crypto, run-benchmarks-gpu-core-crypto,
+      run-benchmarks-gpu-zk-server
     ]
     uses: ./.github/workflows/generate_svgs.yml
     with:
       time_span_days: 5
       generate-cpu-svgs: ${{ inputs.run-cpu-benchmarks }}
-      generate-gpu-svgs: ${{ inputs.run-gpu-integer-benchmarks || inputs.run-gpu-core-crypto-benchmarks }}
+      generate-gpu-svgs: ${{ inputs.run-gpu-integer-benchmarks || inputs.run-gpu-core-crypto-benchmarks || inputs.run-gpu-zk-benchmarks }}
       generate-hpu-svgs: ${{ inputs.run-hpu-benchmarks }}
     secrets:
       DATA_EXTRACTOR_DATABASE_USER: ${{ secrets.DATA_EXTRACTOR_DATABASE_USER }}
@@ -188,7 +213,7 @@ jobs:
 
   generate-svgs-without-benchmarks-run:
     name: benchmark-documentation/generate-svgs-without-benchmarks-run
-    if: ${{ !(inputs.run-cpu-benchmarks || inputs.run-gpu-integer-benchmarks || inputs.run-gpu-core-crypto-benchmarks || inputs.run-hpu-benchmarks) &&
+    if: ${{ !(inputs.run-cpu-benchmarks || inputs.run-gpu-integer-benchmarks || inputs.run-gpu-core-crypto-benchmarks || inputs.run-gpu-zk-benchmarks || inputs.run-hpu-benchmarks) &&
       inputs.generate-svgs }}
     uses: ./.github/workflows/generate_svgs.yml
     with:

.github/workflows/benchmark_gpu.yml

@@ -37,7 +37,7 @@ on:
           - integer_zk_experimental
           - integer_aes
           - integer_aes256
-          - hlapi_erc20
+          - hlapi_erc7984
           - hlapi_dex
           - hlapi_noise_squash
       op_flavor:
@@ -123,8 +123,8 @@ jobs:
 
           if inputs_command == "integer_zk":
             files_to_parse.append("pke_zk_crs_sizes.csv")
-          elif inputs_command == "hlapi_erc20":
-            files_to_parse.append("erc20_pbs_count.csv")
+          elif inputs_command == "hlapi_erc7984":
+            files_to_parse.append("erc7984_pbs_count.csv")
           elif inputs_command == "hlapi_dex":
             files_to_parse.extend(
               [

.github/workflows/benchmark_gpu_4090.yml

@@ -89,7 +89,7 @@ jobs:
           REF_NAME: ${{ github.ref_name }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_integer_multi_bit_gpu_default
           path: ${{ env.RESULTS_FILENAME }}
@@ -173,7 +173,7 @@ jobs:
           REF_NAME: ${{ github.ref_name }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_core_crypto
           path: ${{ env.RESULTS_FILENAME }}

.github/workflows/benchmark_gpu_common.yml

@@ -126,16 +126,10 @@ jobs:
     needs: prepare-matrix
     runs-on: ubuntu-latest
     outputs:
-      # Use permanent remote instance label first as on-demand remote instance label output is set before the end of start-remote-instance step.
-      # If the latter fails due to a failed GitHub action runner set up, we have to fallback on the permanent instance.
-      # Since the on-demand remote label is set before failure, we have to do the logical OR in this order,
-      # otherwise we'll try to run the next job on a non-existing on-demand instance.
-      runner-name: ${{ steps.use-permanent-instance.outputs.runner_group || steps.start-remote-instance.outputs.label }}
-      remote-instance-outcome: ${{ steps.start-remote-instance.outcome }}
+      runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
-      - name: Start remote instance
-        id: start-remote-instance
-        continue-on-error: true
+      - name: Start instance
+        id: start-instance
         uses: zama-ai/slab-github-runner@5aee5d157f4a0201e5eaefc9cc648e5f9f5472a5 # v1.6.0
         with:
           mode: start
@@ -145,25 +139,6 @@ jobs:
           backend: ${{ inputs.backend }}
           profile: ${{ inputs.profile }}
 
-      - name: Acknowledge remote instance failure
-        if: steps.start-remote-instance.outcome == 'failure' &&
-          inputs.profile != 'single-h100'
-        run: |
-          echo "Remote instance instance has failed to start (profile provided: '${INPUTS_PROFILE}')"
-          echo "Permanent instance instance cannot be used as a substitute (profile needed: 'single-h100')"
-          exit 1
-        env:
-          INPUTS_PROFILE: ${{ inputs.profile }}
-
-      # This will allow to fallback on permanent instances running on Hyperstack.
-      - name: Use permanent remote instance
-        id: use-permanent-instance
-        if: env.SECRETS_AVAILABLE == 'true' &&
-          steps.start-remote-instance.outcome == 'failure' &&
-          inputs.profile == 'single-h100'
-        run: |
-          echo "runner_group=h100x1" >> "$GITHUB_OUTPUT"
-
   # Install dependencies only once since cuda-benchmarks uses a matrix strategy, thus running multiple times.
   install-dependencies:
     name: benchmark_gpu_common/install-dependencies
@@ -184,7 +159,6 @@ jobs:
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
       - name: Setup Hyperstack dependencies
-        if: needs.setup-instance.outputs.remote-instance-outcome == 'success'
         uses: ./.github/actions/gpu_setup
         with:
           cuda-version: ${{ matrix.cuda }}
@@ -296,7 +270,7 @@ jobs:
           filenames: ${{ inputs.additional_file_to_parse }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_${{ matrix.command }}_${{ matrix.op_flavor }}_${{ inputs.profile }}_${{ matrix.bench_type }}_${{ matrix.params_type }}
           path: ${{ env.RESULTS_FILENAME }}
@@ -333,7 +307,7 @@ jobs:
 
   teardown-instance:
     name: benchmark_gpu_common/teardown-instance
-    if: ${{ always() && needs.setup-instance.outputs.remote-instance-outcome == 'success' }}
+    if: ${{ always() && needs.setup-instance.result == 'success' }}
     needs: [ setup-instance, cuda-benchmarks, slack-notify ]
     runs-on: ubuntu-latest
     steps:

.github/workflows/benchmark_gpu_coprocessor.yml

@@ -42,7 +42,7 @@ env:
   OPTIMIZATION_TARGET: "throughput"
   BATCH_SIZE: "5000"
   SCHEDULING_POLICY: "MAX_PARALLELISM"
-  BENCHMARKS: "erc20"
+  BENCHMARKS: "erc7984"
   BRANCH_NAME: ${{ github.ref_name }}
   COMMIT_SHA: ${{ github.sha }}
   SLAB_SECRET: ${{ secrets.JOB_SECRET }}
@@ -204,7 +204,7 @@ jobs:
         uses: foundry-rs/foundry-toolchain@8789b3e21e6c11b2697f5eb56eddae542f746c10
 
       - name: Cache cargo
-        uses: actions/cache@668228422ae6a00e4ad889ee87cd7109ec5666a7 # v5.0.4
+        uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
         with:
           path: |
             ~/.cargo/registry
@@ -214,14 +214,14 @@ jobs:
           restore-keys: ${{ runner.os }}-cargo-
 
       - name: Login to GitHub Container Registry
-        uses: docker/login-action@b45d80f862d83dbcd57f89517bcf500b2ab88fb2 # v4.0.0
+        uses: docker/login-action@4907a6ddec9925e35a0a9e82d7399ccc52663121 # v4.1.0
         with:
           registry: ghcr.io
           username: ${{ github.actor }}
           password: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Login to Chainguard Registry
-        uses: docker/login-action@b45d80f862d83dbcd57f89517bcf500b2ab88fb2 # v4.0.0
+        uses: docker/login-action@4907a6ddec9925e35a0a9e82d7399ccc52663121 # v4.1.0
         with:
           registry: cgr.dev
           username: ${{ secrets.CGR_USERNAME }}
@@ -232,7 +232,7 @@ jobs:
         working-directory: fhevm/coprocessor/fhevm-engine/tfhe-worker
 
       - name: Use Node.js
-        uses: actions/setup-node@53b83947a5a98c8d113130e565377fae1a50d02f # v6.3.0
+        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
         with:
           node-version: 20.x
 
@@ -248,13 +248,13 @@ jobs:
           npm install && npm run deploy:emptyProxies && npx hardhat compile
         working-directory: fhevm/
 
-      - name: Profile erc20 no-cmux benchmark on GPU
+      - name: Profile erc7984 no-cmux benchmark on GPU
         run: |
           BENCHMARK_BATCH_SIZE="${BATCH_SIZE}" \
           FHEVM_DF_SCHEDULE="${SCHEDULING_POLICY}" \
           BENCHMARK_TYPE="THROUGHPUT_200" \
           OPTIMIZATION_TARGET="${OPTIMIZATION_TARGET}" \
-          make -e "profile_erc20_gpu"
+          make -e "profile_erc7984_gpu"
         working-directory: fhevm/coprocessor/fhevm-engine/tfhe-worker
 
       - name: Get nsys profile name
@@ -271,7 +271,7 @@ jobs:
       - name: Upload profile artifact
         env:
           REPORT_NAME: ${{ steps.nsys_profile_name.outputs.profile }}
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ env.REPORT_NAME }}
           path: fhevm/coprocessor/fhevm-engine/tfhe-worker/${{ env.REPORT_NAME }}
@@ -302,7 +302,7 @@ jobs:
         working-directory: fhevm/
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${COMMIT_SHA}_${BENCHMARKS}_${{ needs.parse-inputs.outputs.profile }}
           path: fhevm/$${{ env.RESULTS_FILENAME }}

.github/workflows/benchmark_hpu.yml

@@ -14,7 +14,7 @@ on:
           - integer
           - hlapi_unsigned
           - hlapi_signed
-          - hlapi_erc20
+          - hlapi_erc7984
       op_flavor:
         description: "Operations set to run"
         type: choice

.github/workflows/benchmark_hpu_common.yml

@@ -185,7 +185,7 @@ jobs:
           BENCH_TYPE: ${{ matrix.bench_type }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_${{ matrix.bench_type }}_${{ matrix.command }}_benchmarks
           path: ${{ env.RESULTS_FILENAME }}

.github/workflows/benchmark_perf_regression.yml

@@ -280,7 +280,7 @@ jobs:
           BENCH_TYPE: ${{ env.__TFHE_RS_BENCH_TYPE }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_regression_${{ env.RESULTS_FILE_SHA }} # RESULT_FILE_SHA is needed to avoid collision between matrix.command runs
           path: ${{ env.RESULTS_FILENAME }}

.github/workflows/benchmark_tfhe_fft.yml

@@ -99,7 +99,7 @@ jobs:
           REF_NAME: ${{ github.ref_name }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_fft
           path: ${{ env.RESULTS_FILENAME }}

.github/workflows/benchmark_tfhe_ntt.yml

@@ -99,7 +99,7 @@ jobs:
           REF_NAME: ${{ github.ref_name }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_ntt
           path: ${{ env.RESULTS_FILENAME }}

.github/workflows/benchmark_wasm_client.yml

@@ -46,7 +46,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             wasm_bench:

.github/workflows/benchmark_wasm_client_common.yml

@@ -124,7 +124,7 @@ jobs:
 
       - name: Node cache restoration
         id: node-cache
-        uses: actions/cache/restore@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/restore@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         with:
           path: |
             ~/.nvm
@@ -137,7 +137,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/save@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         if: steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |
@@ -180,7 +180,7 @@ jobs:
           REF_NAME: ${{ github.ref_name }}
 
       - name: Upload parsed results artifact
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_wasm_${{ matrix.browser }}
           path: ${{ env.RESULTS_FILENAME }}

.github/workflows/cargo_build_common.yml

@@ -138,7 +138,7 @@ jobs:
       - name: Node cache restoration
         if: inputs.run-pcc-cpu-batch == 'pcc_batch_2'
         id: node-cache
-        uses: actions/cache/restore@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/restore@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         with:
           path: |
             ~/.nvm
@@ -151,7 +151,7 @@ jobs:
           make install_node
 
       - name: Node cache save
-        uses: actions/cache/save@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/save@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         if: inputs.run-pcc-cpu-batch == 'pcc_batch_2' && steps.node-cache.outputs.cache-hit != 'true'
         with:
           path: |

.github/workflows/cargo_test_fft.yml

@@ -40,7 +40,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             fft:

.github/workflows/cargo_test_ntt.yml

@@ -42,7 +42,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             ntt:

.github/workflows/ci_lint.yml

@@ -43,14 +43,14 @@ jobs:
           echo "version=$(make zizmor_version)" >> "${GITHUB_OUTPUT}"
 
       - name: Check workflows security
-        uses: zizmorcore/zizmor-action@71321a20a9ded102f6e9ce5718a2fcec2c4f70d8 # v0.5.2
+        uses: zizmorcore/zizmor-action@b1d7e1fb5de872772f31590499237e7cce841e8e # v0.5.3
         with:
           advanced-security: 'false' # Print results directly in logs
           persona: pedantic
           version: ${{ steps.get_zizmor.outputs.version }}
 
       - name: Ensure SHA pinned actions
-        uses: zgosalvez/github-actions-ensure-sha-pinned-actions@70c4af2ed5282c51ba40566d026d6647852ffa3e # v5.0.1
+        uses: zgosalvez/github-actions-ensure-sha-pinned-actions@ca46236c6ce584ae24bc6283ba8dcf4b3ec8a066 # v5.0.4
         with:
           allowlist: |
             slsa-framework/slsa-github-generator

.github/workflows/code_coverage.yml

@@ -44,7 +44,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             tfhe:
@@ -74,7 +74,7 @@ jobs:
           make test_shortint_cov
 
       - name: Upload tfhe coverage to Codecov
-        uses: codecov/codecov-action@671740ac38dd9b0130fbe1cec585b89eea48d3de
+        uses: codecov/codecov-action@57e3a136b779b570ffcdbf80b3bdc90e7fab3de2
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
@@ -88,7 +88,7 @@ jobs:
           make test_integer_cov
 
       - name: Upload tfhe coverage to Codecov
-        uses: codecov/codecov-action@671740ac38dd9b0130fbe1cec585b89eea48d3de
+        uses: codecov/codecov-action@57e3a136b779b570ffcdbf80b3bdc90e7fab3de2
         if: steps.changed-files.outputs.tfhe_any_changed == 'true'
         with:
           token: ${{ secrets.CODECOV_TOKEN }}

.github/workflows/csprng_randomness_tests.yml

@@ -46,7 +46,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             csprng:

.github/workflows/generate_svg_common.yml

@@ -87,7 +87,7 @@ jobs:
 
       - name: Upload tables
         if: inputs.backend_comparison == false
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_${{ inputs.backend }}_${{ inputs.layer }}_subset_${{inputs.bench_subset}}_${{ inputs.pbs_kind }}_${{ inputs.bench_type }}_tables
           # This will upload all the file generated
@@ -111,7 +111,7 @@ jobs:
 
       - name: Upload comparison tables
         if: inputs.backend_comparison == true
-        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
         with:
           name: ${{ github.sha }}_backends_comparison_tables
           # This will upload all the file generated

.github/workflows/generate_svgs.yml

@@ -209,60 +209,98 @@ jobs:
       DATA_EXTRACTOR_DATABASE_HOST: ${{ secrets.DATA_EXTRACTOR_DATABASE_HOST }}
       DATA_EXTRACTOR_DATABASE_PASSWORD: ${{ secrets.DATA_EXTRACTOR_DATABASE_PASSWORD }}
 
+  gpu-zk-server-latency-table:
+    name: generate_documentation_svgs/gpu-zk-server-latency-table
+    uses: ./.github/workflows/generate_svg_common.yml
+    if: inputs.generate-gpu-svgs
+    with:
+      backend: gpu
+      hardware_name: n3-H100-SXM5x8
+      layer: integer
+      bench_subset: zk
+      pbs_kind: multi_bit
+      grouping_factor: 4
+      bench_type: latency
+      time_span_days: ${{ inputs.time_span_days }}
+      output_filename: gpu-zk-benchmark-latency
+    secrets:
+      DATA_EXTRACTOR_DATABASE_USER: ${{ secrets.DATA_EXTRACTOR_DATABASE_USER }}
+      DATA_EXTRACTOR_DATABASE_HOST: ${{ secrets.DATA_EXTRACTOR_DATABASE_HOST }}
+      DATA_EXTRACTOR_DATABASE_PASSWORD: ${{ secrets.DATA_EXTRACTOR_DATABASE_PASSWORD }}
+
+  gpu-zk-server-throughput-table:
+    name: generate_documentation_svgs/gpu-zk-server-throughput-table
+    uses: ./.github/workflows/generate_svg_common.yml
+    if: inputs.generate-gpu-svgs
+    with:
+      backend: gpu
+      hardware_name: n3-H100-SXM5x8
+      layer: integer
+      bench_subset: zk
+      pbs_kind: multi_bit
+      grouping_factor: 4
+      bench_type: throughput
+      time_span_days: ${{ inputs.time_span_days }}
+      output_filename: gpu-zk-benchmark-throughput
+    secrets:
+      DATA_EXTRACTOR_DATABASE_USER: ${{ secrets.DATA_EXTRACTOR_DATABASE_USER }}
+      DATA_EXTRACTOR_DATABASE_HOST: ${{ secrets.DATA_EXTRACTOR_DATABASE_HOST }}
+      DATA_EXTRACTOR_DATABASE_PASSWORD: ${{ secrets.DATA_EXTRACTOR_DATABASE_PASSWORD }}
+
   # -----------------------------------------------------------
-  # ERC20 benchmarks tables
+  # ERC7984 benchmarks tables
   # -----------------------------------------------------------
 
-  cpu-erc20-latency-throughput-table:
-    name: generate_documentation_svgs/cpu-erc20-latency-throughput-table
+  cpu-erc7984-latency-throughput-table:
+    name: generate_documentation_svgs/cpu-erc7984-latency-throughput-table
     uses: ./.github/workflows/generate_svg_common.yml
     if: inputs.generate-cpu-svgs
     with:
       backend: cpu
       hardware_name: hpc7a.96xlarge
       layer: hlapi
-      bench_subset: erc20
+      bench_subset: erc7984
       pbs_kind: classical
       bench_type: both
       time_span_days: ${{ inputs.time_span_days }}
-      output_filename: cpu-hlapi-erc20-benchmark-latency-throughput
+      output_filename: cpu-hlapi-erc7984-benchmark-latency-throughput
     secrets:
       DATA_EXTRACTOR_DATABASE_USER: ${{ secrets.DATA_EXTRACTOR_DATABASE_USER }}
       DATA_EXTRACTOR_DATABASE_HOST: ${{ secrets.DATA_EXTRACTOR_DATABASE_HOST }}
       DATA_EXTRACTOR_DATABASE_PASSWORD: ${{ secrets.DATA_EXTRACTOR_DATABASE_PASSWORD }}
 
-  gpu-erc20-latency-throughput-table:
-    name: generate_documentation_svgs/gpu-erc20-latency-throughput-table
+  gpu-erc7984-latency-throughput-table:
+    name: generate_documentation_svgs/gpu-erc7984-latency-throughput-table
     uses: ./.github/workflows/generate_svg_common.yml
     if: inputs.generate-gpu-svgs
     with:
       backend: gpu
       hardware_name: n3-H100-SXM5x8
       layer: hlapi
-      bench_subset: erc20
+      bench_subset: erc7984
       pbs_kind: multi_bit
       grouping_factor: 4
       bench_type: both
       time_span_days: ${{ inputs.time_span_days }}
-      output_filename: gpu-hlapi-erc20-benchmark-h100x8-sxm5-latency-throughput
+      output_filename: gpu-hlapi-erc7984-benchmark-h100x8-sxm5-latency-throughput
     secrets:
       DATA_EXTRACTOR_DATABASE_USER: ${{ secrets.DATA_EXTRACTOR_DATABASE_USER }}
       DATA_EXTRACTOR_DATABASE_HOST: ${{ secrets.DATA_EXTRACTOR_DATABASE_HOST }}
       DATA_EXTRACTOR_DATABASE_PASSWORD: ${{ secrets.DATA_EXTRACTOR_DATABASE_PASSWORD }}
 
-  hpu-erc20-latency-throughput-table:
-    name: generate_documentation_svgs/hpu-erc20-latency-throughput-table
+  hpu-erc7984-latency-throughput-table:
+    name: generate_documentation_svgs/hpu-erc7984-latency-throughput-table
     uses: ./.github/workflows/generate_svg_common.yml
     if: inputs.generate-hpu-svgs
     with:
       backend: hpu
       hardware_name: hpu_x1
       layer: hlapi
-      bench_subset: erc20
+      bench_subset: erc7984
       pbs_kind: classical
       bench_type: both
       time_span_days: ${{ inputs.time_span_days }}
-      output_filename: hpu-hlapi-erc20-benchmark-hpux1-latency-throughput.svg
+      output_filename: hpu-hlapi-erc7984-benchmark-hpux1-latency-throughput.svg
     secrets:
       DATA_EXTRACTOR_DATABASE_USER: ${{ secrets.DATA_EXTRACTOR_DATABASE_USER }}
       DATA_EXTRACTOR_DATABASE_HOST: ${{ secrets.DATA_EXTRACTOR_DATABASE_HOST }}

.github/workflows/gpu_core_h100_tests.yml

@@ -23,7 +23,7 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   pull_request:
-    types: [ labeled ]
+    types: [ labeled, opened, synchronize ]
 
 permissions:
   contents: read
@@ -38,6 +38,7 @@ jobs:
       pull-requests: read  # Needed to check for file change
     outputs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
+      core_crypto_changed: ${{ steps.changed-files.outputs.core_crypto_any_changed }}
     steps:
       - name: Checkout tfhe-rs
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd
@@ -48,7 +49,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -62,28 +63,23 @@ jobs:
               - tfhe/src/integer/server_key/radix_parallel/tests_cases_unsigned.rs
               - tfhe/src/shortint/parameters/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_core_h100_tests.yml'
+            core_crypto:
+              - tfhe/src/core_crypto/gpu/**
 
   setup-instance:
     name: gpu_core_h100_tests/setup-instance
     needs: should-run
     if: github.event_name != 'pull_request' ||
-      (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true') ||
-      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true')
+      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true') ||
+      (github.event.action != 'labeled' && needs.should-run.outputs.core_crypto_changed == 'true')
     runs-on: ubuntu-latest
     outputs:
-      # Use permanent remote instance label first as on-demand remote instance label output is set before the end of start-remote-instance step.
-      # If the latter fails due to a failed GitHub action runner set up, we have to fallback on the permanent instance.
-      # Since the on-demand remote label is set before failure, we have to do the logical OR in this order,
-      # otherwise we'll try to run the next job on a non-existing on-demand instance.
-      runner-name: ${{ steps.use-permanent-instance.outputs.runner_group || steps.start-remote-instance.outputs.label || steps.start-github-instance.outputs.runner_group }}
-      remote-instance-outcome: ${{ steps.start-remote-instance.outcome }}
+      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'
-        continue-on-error: true
         uses: zama-ai/slab-github-runner@5aee5d157f4a0201e5eaefc9cc648e5f9f5472a5 # v1.6.0
         with:
           mode: start
@@ -93,13 +89,6 @@ jobs:
           backend: hyperstack
           profile: single-h100
 
-      # This will allow to fallback on permanent instances running on Hyperstack.
-      - name: Use permanent remote instance
-        id: use-permanent-instance
-        if: env.SECRETS_AVAILABLE == 'true' && steps.start-remote-instance.outcome == 'failure'
-        run: |
-          echo "runner_group=h100x1" >> "$GITHUB_OUTPUT"
-
       # This instance will be spawned especially for pull-request from forked repository
       - name: Start GitHub instance
         id: start-github-instance
@@ -132,7 +121,6 @@ jobs:
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Setup Hyperstack dependencies
-        if: needs.setup-instance.outputs.remote-instance-outcome == 'success'
         uses: ./.github/actions/gpu_setup
         with:
           cuda-version: ${{ matrix.cuda }}
@@ -176,7 +164,7 @@ jobs:
 
   teardown-instance:
     name: gpu_core_h100_tests/teardown-instance
-    if: ${{ always() && needs.setup-instance.outputs.remote-instance-outcome == 'success' }}
+    if: ${{ always() && needs.setup-instance.result == 'success' }}
     needs: [ setup-instance, cuda-tests-linux ]
     runs-on: ubuntu-latest
     steps:

.github/workflows/gpu_fast_tests.yml

@@ -47,7 +47,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:

.github/workflows/gpu_full_h100_tests.yml

@@ -25,16 +25,10 @@ jobs:
     name: gpu_full_h100_tests/setup-instance
     runs-on: ubuntu-latest
     outputs:
-      # Use permanent remote instance label first as on-demand remote instance label output is set before the end of start-remote-instance step.
-      # If the latter fails due to a failed GitHub action runner set up, we have to fallback on the permanent instance.
-      # Since the on-demand remote label is set before failure, we have to do the logical OR in this order,
-      # otherwise we'll try to run the next job on a non-existing on-demand instance.
-      runner-name: ${{ steps.use-permanent-instance.outputs.runner_group || steps.start-remote-instance.outputs.label }}
-      remote-instance-outcome: ${{ steps.start-remote-instance.outcome }}
+      runner-name: ${{ steps.start-instance.outputs.label }}
     steps:
-      - name: Start remote instance
-        id: start-remote-instance
-        continue-on-error: true
+      - name: Start instance
+        id: start-instance
         uses: zama-ai/slab-github-runner@5aee5d157f4a0201e5eaefc9cc648e5f9f5472a5 # v1.6.0
         with:
           mode: start
@@ -44,13 +38,6 @@ jobs:
           backend: hyperstack
           profile: single-h100
 
-      # This will allow to fallback on permanent instances running on Hyperstack.
-      - name: Use permanent remote instance
-        id: use-permanent-instance
-        if: env.SECRETS_AVAILABLE == 'true' && steps.start-remote-instance.outcome == 'failure'
-        run: |
-          echo "runner_group=h100x1" >> "$GITHUB_OUTPUT"
-
   cuda-tests-linux:
     name: gpu_full_h100_tests/cuda-tests-linux
     needs: [ setup-instance ]
@@ -74,7 +61,6 @@ jobs:
           token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
 
       - name: Setup Hyperstack dependencies
-        if: needs.setup-instance.outputs.remote-instance-outcome == 'success'
         uses: ./.github/actions/gpu_setup
         with:
           cuda-version: ${{ matrix.cuda }}
@@ -118,7 +104,7 @@ jobs:
 
   teardown-instance:
     name: gpu_full_h100_tests/teardown-instance
-    if: ${{ always() && needs.setup-instance.outputs.remote-instance-outcome == 'success' }}
+    if: ${{ always() && needs.setup-instance.result == 'success' }}
     needs: [ setup-instance, cuda-tests-linux ]
     runs-on: ubuntu-latest
     steps:

.github/workflows/gpu_full_multi_gpu_tests.yml

@@ -48,7 +48,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:

.github/workflows/gpu_hlapi_h100_tests.yml

@@ -23,7 +23,7 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   pull_request:
-    types: [ labeled ]
+    types: [ labeled, opened, synchronize ]
 
 permissions:
   contents: read
@@ -38,6 +38,7 @@ jobs:
       pull-requests: read  # Needed to check for file change
     outputs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
+      core_crypto_changed: ${{ steps.changed-files.outputs.core_crypto_any_changed }}
     steps:
       - name: Checkout tfhe-rs
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd
@@ -48,7 +49,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -65,26 +66,22 @@ jobs:
               - tfhe/src/c_api/**
               - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_hlapi_h100_tests.yml'
+            core_crypto:
+              - tfhe/src/core_crypto/gpu/**
 
   setup-instance:
     name: gpu_hlapi_h100_tests/setup-instance
     needs: should-run
     if: github.event_name != 'pull_request' ||
-      (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true') ||
-      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true')
+      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true') ||
+      (github.event.action != 'labeled' && needs.should-run.outputs.core_crypto_changed == 'true')
     runs-on: ubuntu-latest
     outputs:
-      # Use permanent remote instance label first as on-demand remote instance label output is set before the end of start-remote-instance step.
-      # If the latter fails due to a failed GitHub action runner set up, we have to fallback on the permanent instance.
-      # Since the on-demand remote label is set before failure, we have to do the logical OR in this order,
-      # otherwise we'll try to run the next job on a non-existing on-demand instance.
-      runner-name: ${{ steps.use-permanent-instance.outputs.runner_group || steps.start-remote-instance.outputs.label || steps.start-github-instance.outputs.runner_group }}
-      remote-instance-outcome: ${{ steps.start-remote-instance.outcome }}
+      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'
-        continue-on-error: true
         uses: zama-ai/slab-github-runner@5aee5d157f4a0201e5eaefc9cc648e5f9f5472a5 # v1.6.0
         with:
           mode: start
@@ -94,13 +91,6 @@ jobs:
           backend: hyperstack
           profile: single-h100
 
-      # This will allow to fallback on permanent instances running on Hyperstack.
-      - name: Use permanent remote instance
-        id: use-permanent-instance
-        if: env.SECRETS_AVAILABLE == 'true' && steps.start-remote-instance.outcome == 'failure'
-        run: |
-          echo "runner_group=h100x1" >> "$GITHUB_OUTPUT"
-
       # This instance will be spawned especially for pull-request from forked repository
       - name: Start GitHub instance
         id: start-github-instance
@@ -133,7 +123,6 @@ jobs:
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Setup Hyperstack dependencies
-        if: needs.setup-instance.outputs.remote-instance-outcome == 'success'
         uses: ./.github/actions/gpu_setup
         with:
           cuda-version: ${{ matrix.cuda }}
@@ -184,7 +173,7 @@ jobs:
 
   teardown-instance:
     name: gpu_hlapi_h100_tests/teardown-instance
-    if: ${{ always() && needs.setup-instance.outputs.remote-instance-outcome == 'success' }}
+    if: ${{ always() && needs.setup-instance.result == 'success' }}
     needs: [ setup-instance, cuda-tests-linux ]
     runs-on: ubuntu-latest
     steps:

.github/workflows/gpu_integer_long_run_tests.yml

@@ -17,8 +17,8 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   schedule:
-    # Nightly tests will be triggered each evening 8p.m.
-    - cron: "0 20 * * *"
+    # Weekly tests will be triggered every Monday at 8p.m.
+    - cron: "0 20 * * 1"
   pull_request:
 
 
@@ -28,10 +28,41 @@ permissions:
 # zizmor: ignore[concurrency-limits] concurrency is managed after instance setup to ensure safe provisioning
 
 jobs:
+  should-run:
+    name: gpu_integer_long_run_tests/should-run
+    runs-on: ubuntu-latest
+    permissions:
+      pull-requests: read  # Needed to check for file change
+    outputs:
+      is_needed_in_gpu_ci: ${{ env.IS_PR == 'false' || steps.changed-files.outputs.gpu_any_changed }}
+    steps:
+      - name: Checkout tfhe-rs
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd
+        with:
+          fetch-depth: 0
+          persist-credentials: 'false'
+          token: ${{ env.CHECKOUT_TOKEN }}
+
+      - name: Check for file changes
+        id: changed-files
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
+        with:
+          files_yaml: |
+            gpu:
+              - tfhe/Cargo.toml
+              - tfhe/build.rs
+              - backends/tfhe-cuda-backend/**
+              - tfhe/src/core_crypto/gpu/**
+              - tfhe/src/integer/gpu/**
+              - tfhe/src/shortint/parameters/**
+              - '.github/workflows/gpu_integer_long_run_tests.yml'
+
   setup-instance:
     name: gpu_integer_long_run_tests/setup-instance
-    if: github.event_name != 'schedule' ||
-      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs')
+    needs: [should-run]
+    if: github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'schedule' && github.repository == 'zama-ai/tfhe-rs') ||
+      needs.should-run.outputs.is_needed_in_gpu_ci == 'true'
     runs-on: ubuntu-latest
     outputs:
       runner-name: ${{ steps.start-instance.outputs.label }}

.github/workflows/gpu_memory_sanitizer.yml

@@ -48,7 +48,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:

.github/workflows/gpu_memory_sanitizer_h100.yml

@@ -48,7 +48,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:

.github/workflows/gpu_pcc.yml

@@ -131,6 +131,10 @@ jobs:
         env:
           GCC_VERSION: ${{ matrix.gcc }}
 
+      - name: Run semgrep and lint checks on CUDA code
+        run: |
+          make semgrep_and_lint_gpu_code
+
       - name: Run fmt checks
         run: |
           make check_fmt_gpu
@@ -139,10 +143,6 @@ jobs:
         run: |
           make pcc_gpu
 
-      - name: Run semgrep and lint checks on CUDA code
-        run: |
-          make semgrep_and_lint_gpu_code
-
       - name: Run semver checks on tfhe-cuda-backend
         run: |
           make semver_check_cuda_backend

.github/workflows/gpu_signed_integer_classic_tests.yml

@@ -48,7 +48,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -63,7 +63,6 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_signed_integer_classic_tests.yml'
               - scripts/integer-tests.sh
 

.github/workflows/gpu_signed_integer_h100_tests.yml

@@ -23,7 +23,7 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   pull_request:
-    types: [ labeled ]
+    types: [ labeled, opened, synchronize ]
 
 permissions:
   contents: read
@@ -38,6 +38,7 @@ jobs:
       pull-requests: read  # Needed to check for file change
     outputs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
+      core_crypto_changed: ${{ steps.changed-files.outputs.core_crypto_any_changed }}
     steps:
       - name: Checkout tfhe-rs
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd
@@ -48,7 +49,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -63,29 +64,24 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_signed_integer_h100_tests.yml'
               - scripts/integer-tests.sh
+            core_crypto:
+              - tfhe/src/core_crypto/gpu/**
 
   setup-instance:
     name: gpu_signed_integer_h100_tests/setup-instance
     needs: should-run
     if: github.event_name != 'pull_request' ||
-      (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true') ||
-      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true')
+      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true') ||
+      (github.event.action != 'labeled' && needs.should-run.outputs.core_crypto_changed == 'true')
     runs-on: ubuntu-latest
     outputs:
-      # Use permanent remote instance label first as on-demand remote instance label output is set before the end of start-remote-instance step.
-      # If the latter fails due to a failed GitHub action runner set up, we have to fallback on the permanent instance.
-      # Since the on-demand remote label is set before failure, we have to do the logical OR in this order,
-      # otherwise we'll try to run the next job on a non-existing on-demand instance.
-      runner-name: ${{ steps.use-permanent-instance.outputs.runner_group || steps.start-remote-instance.outputs.label || steps.start-github-instance.outputs.runner_group }}
-      remote-instance-outcome: ${{ steps.start-remote-instance.outcome }}
+      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'
-        continue-on-error: true
         uses: zama-ai/slab-github-runner@5aee5d157f4a0201e5eaefc9cc648e5f9f5472a5 # v1.6.0
         with:
           mode: start
@@ -95,13 +91,6 @@ jobs:
           backend: hyperstack
           profile: single-h100
 
-      # This will allow to fallback on permanent instances running on Hyperstack.
-      - name: Use permanent remote instance
-        id: use-permanent-instance
-        if: env.SECRETS_AVAILABLE == 'true' && steps.start-remote-instance.outcome == 'failure'
-        run: |
-          echo "runner_group=h100x1" >> "$GITHUB_OUTPUT"
-
       # This instance will be spawned especially for pull-request from forked repository
       - name: Start GitHub instance
         id: start-github-instance
@@ -134,7 +123,6 @@ jobs:
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Setup Hyperstack dependencies
-        if: needs.setup-instance.outputs.remote-instance-outcome == 'success'
         uses: ./.github/actions/gpu_setup
         with:
           cuda-version: ${{ matrix.cuda }}
@@ -176,7 +164,7 @@ jobs:
 
   teardown-instance:
     name: gpu_signed_integer_h100_tests/teardown-instance
-    if: ${{ always() && needs.setup-instance.outputs.remote-instance-outcome == 'success' }}
+    if: ${{ always() && needs.setup-instance.result == 'success' }}
     needs: [ setup-instance, cuda-tests-linux ]
     runs-on: ubuntu-latest
     steps:

.github/workflows/gpu_signed_integer_tests.yml

@@ -49,7 +49,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -64,7 +64,6 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_signed_integer_tests.yml'
               - scripts/integer-tests.sh
 

.github/workflows/gpu_unsigned_integer_classic_tests.yml

@@ -48,7 +48,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -63,7 +63,6 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_unsigned_integer_classic_tests.yml'
               - scripts/integer-tests.sh
 

.github/workflows/gpu_unsigned_integer_h100_tests.yml

@@ -23,7 +23,7 @@ on:
   # Allows you to run this workflow manually from the Actions tab as an alternative.
   workflow_dispatch:
   pull_request:
-    types: [ labeled ]
+    types: [ labeled, opened, synchronize ]
 
 permissions:
   contents: read
@@ -38,6 +38,7 @@ jobs:
       pull-requests: read  # Needed to check for file change
     outputs:
       gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }}
+      core_crypto_changed: ${{ steps.changed-files.outputs.core_crypto_any_changed }}
     steps:
       - name: Checkout tfhe-rs
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd
@@ -48,7 +49,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -63,29 +64,24 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_unsigned_integer_h100_tests.yml'
               - scripts/integer-tests.sh
+            core_crypto:
+              - tfhe/src/core_crypto/gpu/**
 
   setup-instance:
     name: gpu_unsigned_integer_h100_tests/setup-instance
     needs: should-run
     if: github.event_name == 'workflow_dispatch' ||
-      (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true') ||
-      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true')
+      (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true') ||
+      (github.event.action != 'labeled' && needs.should-run.outputs.core_crypto_changed == 'true')
     runs-on: ubuntu-latest
     outputs:
-      # Use permanent remote instance label first as on-demand remote instance label output is set before the end of start-remote-instance step.
-      # If the latter fails due to a failed GitHub action runner set up, we have to fallback on the permanent instance.
-      # Since the on-demand remote label is set before failure, we have to do the logical OR in this order,
-      # otherwise we'll try to run the next job on a non-existing on-demand instance.
-      runner-name: ${{ steps.use-permanent-instance.outputs.runner_group || steps.start-remote-instance.outputs.label || steps.start-github-instance.outputs.runner_group }}
-      remote-instance-outcome: ${{ steps.start-remote-instance.outcome }}
+      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'
-        continue-on-error: true
         uses: zama-ai/slab-github-runner@5aee5d157f4a0201e5eaefc9cc648e5f9f5472a5 # v1.6.0
         with:
           mode: start
@@ -95,13 +91,6 @@ jobs:
           backend: hyperstack
           profile: single-h100
 
-      # This will allow to fallback on permanent instances running on Hyperstack.
-      - name: Use permanent remote instance
-        id: use-permanent-instance
-        if: env.SECRETS_AVAILABLE == 'true' && steps.start-remote-instance.outcome == 'failure'
-        run: |
-          echo "runner_group=h100x1" >> "$GITHUB_OUTPUT"
-
       # This instance will be spawned especially for pull-request from forked repository
       - name: Start GitHub instance
         id: start-github-instance
@@ -134,7 +123,6 @@ jobs:
           token: ${{ env.CHECKOUT_TOKEN }}
 
       - name: Setup Hyperstack dependencies
-        if: needs.setup-instance.outputs.remote-instance-outcome == 'success'
         uses: ./.github/actions/gpu_setup
         with:
           cuda-version: ${{ matrix.cuda }}
@@ -176,7 +164,7 @@ jobs:
 
   teardown-instance:
     name: gpu_unsigned_integer_h100_tests/teardown-instance
-    if: ${{ always() && needs.setup-instance.outputs.remote-instance-outcome == 'success' }}
+    if: ${{ always() && needs.setup-instance.result == 'success' }}
     needs: [ setup-instance, cuda-tests-linux ]
     runs-on: ubuntu-latest
     steps:

.github/workflows/gpu_unsigned_integer_tests.yml

@@ -49,7 +49,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -64,7 +64,6 @@ jobs:
               - tfhe/src/shortint/parameters/**
               - tfhe/src/high_level_api/**
               - tfhe/src/c_api/**
-              - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_unsigned_integer_tests.yml'
               - scripts/integer-tests.sh
 

.github/workflows/gpu_zk_tests.yml

@@ -47,7 +47,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             gpu:
@@ -55,12 +55,9 @@ jobs:
               - tfhe/build.rs
               - backends/tfhe-cuda-backend/**
               - backends/zk-cuda-backend/**
-              - tfhe/src/core_crypto/gpu/**
-              - tfhe/src/integer/gpu/**
               - tfhe/src/shortint/parameters/**
               - tfhe/src/zk/**
               - tfhe-zk-pok/**
-              - 'tfhe/docs/**/**.md'
               - '.github/workflows/gpu_zk_tests.yml'
               - ci/slab.toml
 

.github/workflows/hpu_hlapi_tests.yml

@@ -41,7 +41,7 @@ jobs:
 
       - name: Check for file changes
         id: changed-files
-        uses: tj-actions/changed-files@22103cc46bda19c2b464ffe86db46df6922fd323 # v47.0.5
+        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
         with:
           files_yaml: |
             hpu:

.github/workflows/make_release_common.yml

@@ -62,7 +62,7 @@ jobs:
           PACKAGE: ${{ inputs.package-name }}
         run: |
           cargo package -p "${PACKAGE}"
-      - uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f # v7.0.0
+      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
         with:
           name: crate-${{ inputs.package-name }}
           path: target/package/*.crate
@@ -107,7 +107,7 @@ jobs:
           path: target/package
 
       - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@b7e9a28eded4986ec6b1fa40eeee8f8f165559ec # v1.0.3
+        uses: rust-lang/crates-io-auth-action@bbd81622f20ce9e2dd9622e3218b975523e45bbe # v1.0.4
         id: auth
 
       - name: Publish crate.io package

.github/workflows/make_release_common_cuda.yml

@@ -128,7 +128,7 @@ jobs:
         run: |
           cargo package -p "${PACKAGE}"
 
-      - uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f # v7.0.0
+      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
         with:
           name: crate-${{ inputs.package-name }}
           path: target/package/*.crate
@@ -196,6 +196,13 @@ jobs:
         env:
           GCC_VERSION: ${{ matrix.gcc }}
 
+      - name: Checkout
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          fetch-depth: 0
+          persist-credentials: "false"
+          token: ${{ secrets.REPO_CHECKOUT_TOKEN }}
+
       - name: Download artifact
         uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
         with:
@@ -203,19 +210,19 @@ jobs:
           path: target/package
 
       - name: Authenticate on registry
-        uses: rust-lang/crates-io-auth-action@b7e9a28eded4986ec6b1fa40eeee8f8f165559ec # v1.0.3
+        uses: rust-lang/crates-io-auth-action@bbd81622f20ce9e2dd9622e3218b975523e45bbe # v1.0.4
         id: auth
 
       - name: Publish crate.io package
         env:
           CARGO_REGISTRY_TOKEN: ${{ steps.auth.outputs.token }}
           PACKAGE: ${{ inputs.package-name }}
-          DRY-RUN: ${{ inputs.dry-run && '--dry-run' || '' }}
+          DRY_RUN: ${{ inputs.dry-run && '--dry-run' || '' }}
         run: |
-          # dry-run expansion cannot be double quoted when variable contains empty string otherwise cargo publish
-          # would fail. This is safe since dry-run is handled in the env section above.
+          # DRY_RUN expansion cannot be double quoted when variable contains empty string otherwise cargo publish
+          # would fail. This is safe since DRY_RUN is handled in the env section above.
           # shellcheck disable=SC2086
-          cargo publish -p "${PACKAGE}" ${DRY-RUN}
+          cargo publish -p "${PACKAGE}" ${DRY_RUN}
 
       - name: Generate hash
         id: published_hash
@@ -255,7 +262,7 @@ jobs:
 
       - name: Slack Notification
         if: ${{ failure() }}
-        uses: rtCamp/action-slack-notify@e31e87e03dd19038e411e38ae27cbad084a90661
+        uses: rtCamp/action-slack-notify@e31e87e03dd19038e411e38ae27cbad084a90661 # v2.3.3
         env:
           SLACK_COLOR: ${{ job.status }}
           SLACK_MESSAGE: "Instance teardown (${{ inputs.package-name }} release) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})"

.github/workflows/make_release_tfhe.yml

@@ -89,7 +89,7 @@ jobs:
           make build_web_js_api_parallel
 
       - name: Authenticate on NPM
-        uses: actions/setup-node@53b83947a5a98c8d113130e565377fae1a50d02f # v6.3.0
+        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
         with:
           node-version: '24'
           registry-url: 'https://registry.npmjs.org'

.github/workflows/parameters_check.yml

@@ -53,7 +53,7 @@ jobs:
 
       - name: Restore Sagemath image from cache
         id: docker-cache
-        uses: actions/cache/restore@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/restore@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         with:
           path: /tmp/sagemath_image
           key: sagemath-image-${{ env.SAGEMATH_VERSION }}-${{ github.sha }}
@@ -76,7 +76,7 @@ jobs:
       - name: Store Sagemath image in cache
         if: steps.docker-cache.outputs.cache-hit != 'true'
         continue-on-error: true
-        uses: actions/cache/save@668228422ae6a00e4ad889ee87cd7109ec5666a7 #v5.0.4
+        uses: actions/cache/save@27d5ce7f107fe9357f9df03efb73ab90386fccae #v5.0.5
         with:
           path: /tmp/sagemath_image
           key: sagemath-image-${{ env.SAGEMATH_VERSION }}-${{ github.sha }}

Makefile

@@ -312,7 +312,7 @@ semgrep_and_lint_gpu_code: semgrep_lint_setup_venv
 	find "$(TFHECUDA_SRC)" -name '*.h' -o -name '*.cuh' -o -name '*.cu' \
 		| grep -v '/cmake-build-debug/' \
 		| grep -v '/build/' \
-		| xargs venv/bin/semgrep --config "$(TFHECUDA_SRC)/.semgrep/release-ordering.yaml" --scan-unknown-extensions
+		| xargs venv/bin/semgrep --error --config "$(TFHECUDA_SRC)/.semgrep/release-ordering.yaml" --scan-unknown-extensions
 	venv/bin/python3 "scripts/check_scratch_cleanup.py"
 
 .PHONY: semver_check_cuda_backend # Run semver checks on tfhe-cuda-backend
@@ -360,7 +360,7 @@ check_fmt_toml: install_taplo
 
 .PHONY: check_typos # Check for typos in codebase
 check_typos: install_typos_checker
-	@git ls-files ":!*.png" ":!*.cbor" ":!*.bcode" ":!*.ico" ":!*/twiddles.cu" | typos --file-list - && echo "No typos found"
+	@git ls-files ":!*.png" ":!*.cbor" ":!*.bcode" ":!*.ico" ":!*/twiddles.cu" ":!*.hpu" | typos --file-list - && echo "No typos found"
 
 .PHONY: clippy_gpu # Run clippy lints on tfhe with "gpu" enabled
 clippy_gpu: install_rs_check_toolchain
@@ -587,6 +587,17 @@ clippy_backward_compat_data: install_rs_check_toolchain # the toolchain is selec
 		echo "Cannot run clippy for backward compat crate on non x86 platform for now."; \
 	fi
 
+.PHONY: check_backward_compat_locks_did_not_change # Check backward compat Cargo.lock files are up to date
+check_backward_compat_locks_did_not_change: install_rs_check_toolchain
+	@for crate in `ls -1 $(BACKWARD_COMPAT_DATA_DIR)/crates/ | grep generate_`; do \
+		echo "checking Cargo.lock for $$crate"; \
+		cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" -Z unstable-options \
+			-C $(BACKWARD_COMPAT_DATA_DIR)/crates/$$crate metadata --locked --format-version 1 > /dev/null || \
+		( echo "Cargo.lock for $$crate is out of date. Update it with:" && \
+		  echo "  cd $(BACKWARD_COMPAT_DATA_DIR)/crates/$$crate && cargo metadata --format-version 1 > /dev/null" && \
+		  echo "then commit the updated Cargo.lock." && exit 1 ); \
+	done
+
 .PHONY: clippy_test_vectors # Run clippy lints on the test vectors app
 clippy_test_vectors: install_rs_check_toolchain
 	cd apps/test-vectors; RUSTFLAGS="$(RUSTFLAGS)" cargo "$(CARGO_RS_CHECK_TOOLCHAIN)" clippy --all-targets \
@@ -1934,25 +1945,25 @@ bench_hlapi_hpu: install_rs_check_toolchain
 	--bench hlapi \
 	--features=integer,internal-keycache,hpu,hpu-v80,pbs-stats -p tfhe-benchmark --
 
-.PHONY: bench_hlapi_erc20 # Run benchmarks for ERC20 operations
-bench_hlapi_erc20: install_rs_check_toolchain
+.PHONY: bench_hlapi_erc7984 # Run benchmarks for ERC7984 operations
+bench_hlapi_erc7984: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi-erc20 \
+	--bench hlapi-erc7984 \
 	--features=integer,internal-keycache,pbs-stats -p tfhe-benchmark --
 
-.PHONY: bench_hlapi_erc20_gpu # Run benchmarks for ERC20 operations on GPU
-bench_hlapi_erc20_gpu: install_rs_check_toolchain
+.PHONY: bench_hlapi_erc7984_gpu # Run benchmarks for ERC7984 operations on GPU
+bench_hlapi_erc7984_gpu: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) __TFHE_RS_PARAM_TYPE=$(BENCH_PARAM_TYPE) \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi-erc20 \
+	--bench hlapi-erc7984 \
 	--features=integer,gpu,internal-keycache,pbs-stats -p tfhe-benchmark --profile release_lto_off --
 
-.PHONY: bench_hlapi_erc20_gpu_classical # Run benchmarks for ERC20 operations on GPU with classical parameters
-bench_hlapi_erc20_gpu_classical: install_rs_check_toolchain
+.PHONY: bench_hlapi_erc7984_gpu_classical # Run benchmarks for ERC7984 operations on GPU with classical parameters
+bench_hlapi_erc7984_gpu_classical: install_rs_check_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) __TFHE_RS_PARAM_TYPE=classical \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi-erc20 \
+	--bench hlapi-erc7984 \
 	--features=integer,gpu,internal-keycache,pbs-stats -p tfhe-benchmark --profile release_lto_off --
 
 .PHONY: bench_hlapi_dex # Run benchmarks for DEX operations
@@ -1976,13 +1987,13 @@ bench_hlapi_dex_gpu_classical: install_rs_check_toolchain
 	--bench hlapi-dex \
 	--features=integer,gpu,internal-keycache,pbs-stats -p tfhe-benchmark --profile release_lto_off --
 
-.PHONY: bench_hlapi_erc20_hpu # Run benchmarks for ECR20 operations on HPU
-bench_hlapi_erc20_hpu: install_rs_check_toolchain
+.PHONY: bench_hlapi_erc7984_hpu # Run benchmarks for ECR20 operations on HPU
+bench_hlapi_erc7984_hpu: install_rs_check_toolchain
 	source ./setup_hpu.sh --config $(HPU_CONFIG); \
 	export V80_PCIE_DEV=${V80_PCIE_DEV}; \
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi-erc20 \
+	--bench hlapi-erc7984 \
 	--features=integer,internal-keycache,hpu,hpu-v80,pbs-stats -p tfhe-benchmark --
 
 .PHONY: bench_tfhe_zk_pok # Run benchmarks for the tfhe_zk_pok crate
@@ -2038,10 +2049,10 @@ bench_summary: install_rs_check_toolchain
 	--bench hlapi-noise-squash \
 	--features=integer,internal-keycache,pbs-stats -p tfhe-benchmark -- '::decomp_noise_squash_comp::'
 
-	# ERC20
+	# ERC7984
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) __TFHE_RS_PARAM_TYPE=$(BENCH_PARAM_TYPE) \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi-erc20 \
+	--bench hlapi-erc7984 \
 	--features=integer,internal-keycache -p tfhe-benchmark -- '::transfer::overflow'
 
 	# DEX
@@ -2083,10 +2094,10 @@ bench_summary_gpu: install_rs_check_toolchain
 	--bench hlapi-noise-squash \
 	--features=integer,gpu,internal-keycache,pbs-stats -p tfhe-benchmark --profile release_lto_off -- '::decomp_noise_squash_comp::'
 
-	# ERC20
+	# ERC7984
 	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) __TFHE_RS_PARAM_TYPE=$(BENCH_PARAM_TYPE) \
 	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
-	--bench hlapi-erc20 \
+	--bench hlapi-erc7984 \
 	--features=integer,gpu,internal-keycache -p tfhe-benchmark --profile release_lto_off -- '::transfer::overflow'
 
 	# DEX
@@ -2265,6 +2276,7 @@ pcc_batch_5:
 	$(call run_recipe_with_details,clippy_tfhe_lints)
 	$(call run_recipe_with_details,check_compile_tests)
 	$(call run_recipe_with_details,clippy_backward_compat_data)
+	$(call run_recipe_with_details,check_backward_compat_locks_did_not_change)
 
 .PHONY: pcc_batch_6  # duration: 6'32''
 pcc_batch_6:

backends/tfhe-cuda-backend/build.rs

@@ -1,5 +1,14 @@
 use std::path::PathBuf;
-use std::process::Command;
+
+fn get_linux_distribution_name() -> Option<String> {
+    let content = std::fs::read_to_string("/etc/os-release").ok()?;
+    for line in content.lines() {
+        if let Some(value) = line.strip_prefix("NAME=") {
+            return Some(value.trim_matches('"').to_string());
+        }
+    }
+    None
+}
 
 fn main() {
     if let Ok(val) = std::env::var("DOCS_RS") {
@@ -28,9 +37,7 @@ fn main() {
     println!("cargo::rerun-if-changed=src");
 
     if std::env::consts::OS == "linux" {
-        let output = Command::new("./get_os_name.sh").output().unwrap();
-        let distribution = String::from_utf8(output.stdout).unwrap();
-        if distribution != "Ubuntu\n" {
+        if get_linux_distribution_name().as_deref() != Some("Ubuntu") {
             println!(
                 "cargo:warning=This Linux distribution is not officially supported. \
                 Only Ubuntu is supported by tfhe-cuda-backend at this time. Build may fail\n"

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

@@ -721,7 +721,7 @@ void cuda_integer_grouped_oprf_custom_range_64_async(
     uint32_t num_blocks_intermediate, const void *seeded_lwe_input,
     const uint64_t *decomposed_scalar, const uint64_t *has_at_least_one_set,
     uint32_t num_scalars, uint32_t shift, int8_t *mem, void *const *bsks,
-    void *const *ksks);
+    void *const *compute_bsks, void *const *ksks);
 
 void cleanup_cuda_integer_grouped_oprf_custom_range_64(CudaStreamsFFI streams,
                                                        int8_t **mem_ptr_void);

backends/tfhe-cuda-backend/cuda/src/aes/aes.cuh

@@ -390,7 +390,7 @@ __host__ void vectorized_sbox_n_bytes(CudaStreams streams,
   XOR(&wires_a[6], &wires_a[15], &input_bits[7]);
   XOR(&wires_a[10], &wires_a[15], &wires_b[0]);
   XOR(&wires_a[11], &wires_a[20], &wires_a[9]);
-  FLUSH(&wires_a[6], &wires_a[10]);
+  FLUSH(&wires_a[6], &wires_a[10], &wires_a[11]);
   XOR(&wires_a[7], &input_bits[7], &wires_a[11]);
   FLUSH(&wires_a[7]);
   XOR(&wires_a[17], &wires_a[10], &wires_a[11]);
@@ -426,7 +426,7 @@ __host__ void vectorized_sbox_n_bytes(CudaStreams streams,
   XOR(&wires_b[22], &wires_b[18], &wires_a[19]);
   XOR(&wires_b[23], &wires_b[19], &wires_a[21]);
   XOR(&wires_b[24], &wires_b[20], &wires_a[18]);
-  FLUSH(&wires_b[21], &wires_b[23], &wires_b[24]);
+  FLUSH(&wires_b[21], &wires_b[22], &wires_b[23], &wires_b[24]);
   XOR(&wires_b[25], &wires_b[21], &wires_b[22]);
   FLUSH(&wires_b[25]);
 
@@ -468,7 +468,7 @@ __host__ void vectorized_sbox_n_bytes(CudaStreams streams,
 
   XOR(&wires_b[37], &wires_b[36], &wires_b[34]);
   XOR(&wires_b[38], &wires_b[27], &wires_b[36]);
-  FLUSH(&wires_b[38]);
+  FLUSH(&wires_b[38], &wires_b[37]);
   XOR(&wires_b[44], &wires_b[33], &wires_b[37]);
 
   CudaRadixCiphertextFFI *and_outs_6[] = {&wires_b[39]};
@@ -479,7 +479,7 @@ __host__ void vectorized_sbox_n_bytes(CudaStreams streams,
   XOR(&wires_b[40], &wires_b[25], &wires_b[39]);
   XOR(&wires_b[41], &wires_b[40], &wires_b[37]);
   XOR(&wires_b[43], &wires_b[29], &wires_b[40]);
-  FLUSH(&wires_b[41]);
+  FLUSH(&wires_b[41], &wires_b[40], &wires_b[43], &wires_b[44]);
   XOR(&wires_b[45], &wires_b[42], &wires_b[41]);
   FLUSH(&wires_b[45]);
 
@@ -514,6 +514,7 @@ __host__ void vectorized_sbox_n_bytes(CudaStreams streams,
   XOR(&wires_b[57], &wires_b[50], &wires_b[53]);
   XOR(&wires_b[58], &wires_c[4], &wires_b[46]);
   XOR(&wires_b[59], &wires_c[3], &wires_b[54]);
+  FLUSH(&wires_b[57], &wires_b[58]);
   XOR(&wires_b[60], &wires_b[46], &wires_b[57]);
   XOR(&wires_b[61], &wires_c[14], &wires_b[57]);
   XOR(&wires_b[62], &wires_b[52], &wires_b[58]);
@@ -589,6 +590,7 @@ __host__ void vectorized_sbox_n_bytes(CudaStreams streams,
 #undef FLUSH
 #undef AND
 #undef ADD_ONE_FLUSH
+#undef ADD_ONE
 }
 
 /**

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

@@ -489,7 +489,7 @@ template <typename Torus>
 __host__ void host_modulus_switch_multi_bit(
     cudaStream_t stream, uint32_t gpu_index, Torus *array_out, Torus *array_in,
     int size, uint32_t log_modulus, uint32_t degree, uint32_t grouping_factor) {
-  cudaSetDevice(gpu_index);
+  check_cuda_error(cudaSetDevice(gpu_index));
   int multibit_size = size / grouping_factor;
   int num_threads = 0, num_blocks = 0;
   getNumBlocksAndThreads(multibit_size, 1024, num_blocks, num_threads);

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

@@ -72,13 +72,13 @@ void cuda_integer_grouped_oprf_custom_range_64_async(
     uint32_t num_blocks_intermediate, const void *seeded_lwe_input,
     const uint64_t *decomposed_scalar, const uint64_t *has_at_least_one_set,
     uint32_t num_scalars, uint32_t shift, int8_t *mem, void *const *bsks,
-    void *const *ksks) {
+    void *const *compute_bsks, void *const *ksks) {
 
   host_integer_grouped_oprf_custom_range<uint64_t>(
       CudaStreams(streams), radix_lwe_out, num_blocks_intermediate,
       (const uint64_t *)seeded_lwe_input, decomposed_scalar,
       has_at_least_one_set, num_scalars, shift,
-      (int_grouped_oprf_custom_range_memory<uint64_t> *)mem, bsks,
+      (int_grouped_oprf_custom_range_memory<uint64_t> *)mem, bsks, compute_bsks,
       (uint64_t *const *)ksks);
 }
 

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

@@ -114,7 +114,7 @@ void host_integer_grouped_oprf_custom_range(
     const Torus *decomposed_scalar, const Torus *has_at_least_one_set,
     uint32_t num_scalars, uint32_t shift,
     int_grouped_oprf_custom_range_memory<Torus> *mem_ptr, void *const *bsks,
-    Torus *const *ksks) {
+    void *const *compute_bsks, Torus *const *ksks) {
 
   CudaRadixCiphertextFFI *computation_buffer = mem_ptr->tmp_oprf_output;
   set_zero_radix_ciphertext_slice_async<Torus>(
@@ -127,12 +127,12 @@ void host_integer_grouped_oprf_custom_range(
 
   host_integer_scalar_mul_radix<Torus>(
       streams, computation_buffer, decomposed_scalar, has_at_least_one_set,
-      mem_ptr->scalar_mul_buffer, bsks, ksks, mem_ptr->params.message_modulus,
-      num_scalars);
+      mem_ptr->scalar_mul_buffer, compute_bsks, ksks,
+      mem_ptr->params.message_modulus, num_scalars);
 
-  host_logical_scalar_shift_inplace<Torus>(streams, computation_buffer, shift,
-                                           mem_ptr->logical_scalar_shift_buffer,
-                                           bsks, ksks, num_blocks_intermediate);
+  host_logical_scalar_shift_inplace<Torus>(
+      streams, computation_buffer, shift, mem_ptr->logical_scalar_shift_buffer,
+      compute_bsks, ksks, num_blocks_intermediate);
 
   uint32_t num_blocks_output = radix_lwe_out->num_radix_blocks;
   uint32_t blocks_to_copy =

backends/tfhe-cuda-backend/cuda/src/pbs/programmable_bootstrap_multibit_128.cu

@@ -308,6 +308,7 @@ void cleanup_cuda_multi_bit_programmable_bootstrap_noise_tests_128(
     void *stream, uint32_t gpu_index, int8_t **pbs_buffer) {
   cleanup_cuda_multi_bit_programmable_bootstrap_128(stream, gpu_index,
                                                     pbs_buffer);
+  cuda_synchronize_stream(static_cast<cudaStream_t>(stream), gpu_index);
 }
 
 // Noise tests variant of the 128-bit multi-bit PBS, restricted to

backends/tfhe-cuda-backend/get_os_name.sh

@@ -1,3 +0,0 @@
-#!/usr/bin/env bash
-
-cat /etc/os-release | grep "\<NAME\>" | sed "s/NAME=\"//g" | sed "s/\"//g"

backends/tfhe-cuda-backend/src/bindings.rs

@@ -1647,6 +1647,7 @@ unsafe extern "C" {
         shift: u32,
         mem: *mut i8,
         bsks: *const *mut ffi::c_void,
+        compute_bsks: *const *mut ffi::c_void,
         ksks: *const *mut ffi::c_void,
     );
 }

backends/tfhe-hpu-backend/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tfhe-hpu-backend"
-version = "0.4.0"
+version = "0.5.0"
 edition = "2021"
 license = "BSD-3-Clause-Clear"
 description = "HPU implementation on FPGA of TFHE-rs primitives."
@@ -36,7 +36,7 @@ thiserror = "1.0.61"
 bytemuck = { workspace = true }
 anyhow = "1.0.82"
 lazy_static = "1.4.0"
-rand = "0.8.5"
+rand = "0.10.1"
 regex = "1.10.4"
 bitflags = { version = "2.5.0", features = ["serde"] }
 itertools = "0.11.0"

backends/tfhe-hpu-backend/README.md

@@ -297,8 +297,8 @@ source setup_hpu.sh --config v80 -p
 # Run hlapi benches
 make test_high_level_api_hpu
 
-# Run hlapi erc20 benches
-make bench_hlapi_erc20_hpu 
+# Run hlapi erc7984 benches
+make bench_hlapi_erc7984_hpu 
 
 # Run integer level benches
 make bench_integer_hpu

backends/tfhe-hpu-backend/config_store/sim/hpu_config.toml

@@ -109,7 +109,7 @@
   flush_behaviour = "Patient"
   flush = true
 
-[firmware.op_cfg.by_op.ERC_20]
+[firmware.op_cfg.by_op.ERC_7984]
   fill_batch_fifo = true
   min_batch_size = false
   use_tiers = true

backends/tfhe-hpu-backend/config_store/v80/hpu_config.toml

@@ -121,7 +121,7 @@
   flush_behaviour = "Patient"
   flush = true
 
-[firmware.op_cfg.by_op.ERC_20]
+[firmware.op_cfg.by_op.ERC_7984]
   fill_batch_fifo = true
   min_batch_size = false
   use_tiers = true

backends/tfhe-hpu-backend/src/asm/iop/mod.rs

@@ -230,7 +230,7 @@ iop!(
     [IOP_CMP -> "CMP_NEQ", opcode::CMP_NEQ],
     [IOP_CT_F_CT_BOOL -> "IF_THEN_ZERO", opcode::IF_THEN_ZERO],
     [IOP_CT_F_2CT_BOOL -> "IF_THEN_ELSE", opcode::IF_THEN_ELSE],
-    [IOP_2CT_F_3CT -> "ERC_20", opcode::ERC_20],
+    [IOP_2CT_F_3CT -> "ERC_7984", opcode::ERC_7984],
     [IOP_CT_F_CT -> "MEMCPY", opcode::MEMCPY],
     [IOP_CT_F_CT -> "ILOG2", opcode::ILOG2],
     [IOP_CT_F_CT -> "COUNT0", opcode::COUNT0],
@@ -240,5 +240,5 @@ iop!(
     [IOP_CT_F_CT -> "TRAIL0", opcode::TRAIL0],
     [IOP_CT_F_CT -> "TRAIL1", opcode::TRAIL1],
     [IOP_NCT_F_2NCT -> "ADD_SIMD", opcode::ADD_SIMD],
-    [IOP_2NCT_F_3NCT -> "ERC_20_SIMD", opcode::ERC_20_SIMD],
+    [IOP_2NCT_F_3NCT -> "ERC_7984_SIMD", opcode::ERC_7984_SIMD],
 );

backends/tfhe-hpu-backend/src/asm/iop/opcode.rs

@@ -74,9 +74,9 @@ pub const IF_THEN_ZERO: u8 = 0xCA;
 pub const IF_THEN_ELSE: u8 = 0xCB;
 
 // Custom algorithm
-// ERC20 -> Found xfer algorithm
+// ERC7984 -> Found xfer algorithm
 // 2Ct <- func(3Ct)
-pub const ERC_20: u8 = 0x80;
+pub const ERC_7984: u8 = 0x80;
 
 // Count bits
 pub const COUNT0: u8 = 0x81;
@@ -89,7 +89,7 @@ pub const TRAIL1: u8 = 0x87;
 
 // SIMD for maximum throughput
 pub const ADD_SIMD: u8 = 0xF0;
-pub const ERC_20_SIMD: u8 = 0xF1;
+pub const ERC_7984_SIMD: u8 = 0xF1;
 //
 // Utility operations
 // Used to handle real clone of ciphertext already uploaded in the Hpu memory

backends/tfhe-hpu-backend/src/ffi/v80/mod.rs

@@ -24,7 +24,7 @@ use mem_alloc::{MemAlloc, MemChunk};
 
 mod qdma;
 use qdma::QdmaDriver;
-use rand::Rng;
+use rand::RngExt;
 
 const DMA_XFER_ALIGN: usize = 4096_usize;
 
@@ -148,8 +148,8 @@ impl HpuHw {
         tracing::debug!("Load stage1 through JTAG");
         let pdi_stg1_tmp = format!(
             "hpu_stg1_{}.pdi",
-            rand::thread_rng()
-                .sample_iter(rand::distributions::Alphanumeric)
+            rand::rng()
+                .sample_iter(rand::distr::Alphanumeric)
                 .take(5)
                 .map(char::from)
                 .collect::<String>()

backends/tfhe-hpu-backend/src/fw/fw_impl/demo.rs

@@ -31,7 +31,7 @@ crate::impl_fw!("Demo" [
     IF_THEN_ZERO => fw_impl::ilp::iop_if_then_zero;
     IF_THEN_ELSE => fw_impl::ilp::iop_if_then_else;
 
-    ERC_20 => fw_impl::ilp::iop_erc_20;
+    ERC_7984 => fw_impl::ilp::iop_erc_7984;
 
     CMP_GT  => cmp_gt;
     CMP_GTE => cmp_gte;

backends/tfhe-hpu-backend/src/fw/fw_impl/ilp.rs

@@ -61,7 +61,7 @@ crate::impl_fw!("Ilp" [
     IF_THEN_ZERO => fw_impl::ilp::iop_if_then_zero;
     IF_THEN_ELSE => fw_impl::ilp::iop_if_then_else;
 
-    ERC_20 => fw_impl::ilp::iop_erc_20;
+    ERC_7984 => fw_impl::ilp::iop_erc_7984;
 
     MEMCPY => fw_impl::ilp::iop_memcpy;
 
@@ -74,7 +74,7 @@ crate::impl_fw!("Ilp" [
     TRAIL1 => fw_impl::ilp_log::iop_trail1;
     // SIMD Implementations
     ADD_SIMD     => fw_impl::llt::iop_add_simd;
-    ERC_20_SIMD  => fw_impl::llt::iop_erc_20_simd;
+    ERC_7984_SIMD  => fw_impl::llt::iop_erc_7984_simd;
 ]);
 
 #[instrument(level = "trace", skip(prog))]
@@ -1296,13 +1296,13 @@ pub fn iop_if_then_else(prog: &mut Program) {
         });
 }
 
-/// Implement erc_20 fund xfer
+/// Implement erc_7984 fund xfer
 /// Targeted algorithm is as follow:
 /// 1. Check that from has enough funds
 /// 2. Compute real_amount to xfer (i.e. amount or 0)
 /// 3. Compute new amount (from - new_amount, to + new_amount)
 #[instrument(level = "info", skip(prog))]
-pub fn iop_erc_20(prog: &mut Program) {
+pub fn iop_erc_7984(prog: &mut Program) {
     // Allocate metavariables:
     // Dest -> Operand
     let mut dst_from = prog.iop_template_var(OperandKind::Dst, 0);
@@ -1314,7 +1314,7 @@ pub fn iop_erc_20(prog: &mut Program) {
     let src_amount = prog.iop_template_var(OperandKind::Src, 2);
 
     // Add Comment header
-    prog.push_comment("ERC_20 (new_from, new_to) <- (from, to, amount)".to_string());
+    prog.push_comment("ERC_7984 (new_from, new_to) <- (from, to, amount)".to_string());
 
     let props = prog.params();
     let tfhe_params: asm::DigitParameters = props.clone().into();

backends/tfhe-hpu-backend/src/fw/fw_impl/llt/mod.rs

@@ -70,7 +70,7 @@ crate::impl_fw!("Llt" [
     IF_THEN_ZERO => fw_impl::ilp::iop_if_then_zero;
     IF_THEN_ELSE => fw_impl::ilp::iop_if_then_else;
 
-    ERC_20 => fw_impl::llt::iop_erc_20;
+    ERC_7984 => fw_impl::llt::iop_erc_7984;
     MEMCPY => fw_impl::ilp::iop_memcpy;
 
     COUNT0 => fw_impl::ilp_log::iop_count0;
@@ -83,7 +83,7 @@ crate::impl_fw!("Llt" [
 
     // SIMD Implementations
     ADD_SIMD     => fw_impl::llt::iop_add_simd;
-    ERC_20_SIMD  => fw_impl::llt::iop_erc_20_simd;
+    ERC_7984_SIMD  => fw_impl::llt::iop_erc_7984_simd;
 ]);
 
 // ----------------------------------------------------------------------------
@@ -225,24 +225,24 @@ pub fn iop_muls(prog: &mut Program) {
 }
 
 #[instrument(level = "trace", skip(prog))]
-pub fn iop_erc_20(prog: &mut Program) {
+pub fn iop_erc_7984(prog: &mut Program) {
     // Add Comment header
-    prog.push_comment("ERC_20 (new_from, new_to) <- (from, to, amount)".to_string());
+    prog.push_comment("ERC_7984 (new_from, new_to) <- (from, to, amount)".to_string());
     // TODO: Make sweep of kogge_blk_w
     // All these little parameters would be very handy to write an
     // exploration/compilation program which would try to minimize latency by
     // playing with these.
-    iop_erc_20_rtl(prog, 0, Some(10)).add_to_prog(prog);
+    iop_erc_7984_rtl(prog, 0, Some(10)).add_to_prog(prog);
 }
 
 #[instrument(level = "trace", skip(prog))]
-pub fn iop_erc_20_simd(prog: &mut Program) {
+pub fn iop_erc_7984_simd(prog: &mut Program) {
     // Add Comment header
-    prog.push_comment("ERC_20_SIMD (new_from, new_to) <- (from, to, amount)".to_string());
+    prog.push_comment("ERC_7984_SIMD (new_from, new_to) <- (from, to, amount)".to_string());
     simd(
         prog,
         crate::asm::iop::SIMD_N,
-        fw_impl::llt::iop_erc_20_rtl,
+        fw_impl::llt::iop_erc_7984_rtl,
         None,
     );
 }
@@ -379,7 +379,7 @@ pub fn iop_rotate_scalar_left(prog: &mut Program) {
 // Helper Functions
 // ----------------------------------------------------------------------------
 
-/// Implement erc_20 fund xfer
+/// Implement erc_7984 fund xfer
 /// Targeted algorithm is as follow:
 /// 1. Check that from has enough funds
 /// 2. Compute real_amount to xfer (i.e. amount or 0)
@@ -391,7 +391,7 @@ pub fn iop_rotate_scalar_left(prog: &mut Program) {
 ///     (dst_from[0], dst_to[0], ..., dst_from[N-1], dst_to[N-1])
 /// Where N is the batch size
 #[instrument(level = "trace", skip(prog))]
-pub fn iop_erc_20_rtl(prog: &mut Program, batch_index: u8, kogge_blk_w: Option<usize>) -> Rtl {
+pub fn iop_erc_7984_rtl(prog: &mut Program, batch_index: u8, kogge_blk_w: Option<usize>) -> Rtl {
     // Allocate metavariables:
     // Dest -> Operand
     let dst_from = prog.iop_template_var(OperandKind::Dst, 2 * batch_index);

backends/tfhe-hpu-backend/src/interface/variable.rs

@@ -156,7 +156,7 @@ impl HpuVarWrapped {
         {
             let mut inner = var.inner.lock().unwrap();
 
-            for (slot, ct) in std::iter::zip(inner.bundle.iter_mut(), ct.into_iter()) {
+            for (slot, ct) in std::iter::zip(inner.bundle.iter_mut(), ct) {
                 #[cfg(feature = "io-dump")]
                 let params = ct.params().clone();
                 for (id, cut) in ct.into_container().iter().enumerate() {

backends/zk-cuda-backend/build.rs

@@ -1,5 +1,14 @@
 use std::path::PathBuf;
-use std::process::Command;
+
+fn get_linux_distribution_name() -> Option<String> {
+    let content = std::fs::read_to_string("/etc/os-release").ok()?;
+    for line in content.lines() {
+        if let Some(value) = line.strip_prefix("NAME=") {
+            return Some(value.trim_matches('"').to_string());
+        }
+    }
+    None
+}
 
 fn main() {
     // Handle docs.rs builds (no CUDA available)
@@ -29,16 +38,10 @@ fn main() {
         println!("cargo:rustc-link-arg=-Wl,--allow-multiple-definition");
         println!("cargo:rustc-link-arg=-Wl,--no-as-needed");
 
-        // Check Linux distribution (reuse script from tfhe-cuda-backend)
         let manifest_dir = std::env::var("CARGO_MANIFEST_DIR")
             .expect("CARGO_MANIFEST_DIR must be set by cargo during build");
-        let script_path = PathBuf::from(&manifest_dir).join("../tfhe-cuda-backend/get_os_name.sh");
-        let output = Command::new(&script_path)
-            .output()
-            .expect("Failed to run get_os_name.sh — is tfhe-cuda-backend present?");
-        let distribution =
-            String::from_utf8(output.stdout).expect("get_os_name.sh output must be valid UTF-8");
-        if distribution != "Ubuntu\n" {
+
+        if get_linux_distribution_name().as_deref() != Some("Ubuntu") {
             println!(
                 "cargo:warning=This Linux distribution is not officially supported. \
                 Only Ubuntu is supported by zk-cuda-backend at this time. Build may fail\n"

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

@@ -71,11 +71,6 @@ set(CMAKE_CUDA_FLAGS_DEBUG "-g -O0 -G")
 # Additional CUDA flags (aligned with tfhe-cuda-backend)
 set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcompiler -Wall -Xcompiler -Wextra --use_fast_math --expt-relaxed-constexpr")
 
-# =============================================================================
-# Path to tfhe-cuda-backend for device utilities
-# =============================================================================
-set(TFHE_CUDA_BACKEND_DIR ${CMAKE_CURRENT_SOURCE_DIR}/../../tfhe-cuda-backend/cuda)
-
 # Core source files (without device utilities) Device utilities come from tfhe-cuda-backend.
 set(FP_CORE_SOURCES src/primitives/fp.cu src/primitives/fp2.cu src/curve.cu src/msm/pippenger/msm_pippenger.cu
                     src/msm/msm.cu)
@@ -112,7 +107,7 @@ endif()
 target_link_libraries(zk_cuda_backend PUBLIC cudart)
 
 # Include both local headers and tfhe-cuda-backend headers (for device.h)
-target_include_directories(zk_cuda_backend PUBLIC include ../src/include ${TFHE_CUDA_BACKEND_DIR}/include)
+target_include_directories(zk_cuda_backend PUBLIC include ../src/include)
 
 # =============================================================================
 # Tests and Benchmarks (optional, controlled by ZK_CUDA_BACKEND_BUILD_TESTS/BENCHMARKS)
@@ -135,4 +130,3 @@ message(STATUS "Build type: ${CMAKE_BUILD_TYPE}")
 message(STATUS "CUDA architectures: ${CMAKE_CUDA_ARCHITECTURES}")
 message(STATUS "C++ standard: ${CMAKE_CXX_STANDARD}")
 message(STATUS "CUDA standard: ${CMAKE_CUDA_STANDARD}")
-message(STATUS "tfhe-cuda-backend path: ${TFHE_CUDA_BACKEND_DIR}")

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

@@ -0,0 +1,35 @@
+#pragma once
+
+#include <cstddef>
+#include <cstdio>
+
+#include "device.h"
+
+// Variadic checked multiplication of size_t values.
+// Folds left-to-right using __builtin_mul_overflow, returning true on overflow.
+// On overflow the value written to *out is unspecified.
+template <typename... Args>
+inline bool checked_mul(size_t *out, size_t first, Args... rest) {
+  size_t result = first;
+  for (size_t value : {static_cast<size_t>(rest)...}) {
+    if (__builtin_mul_overflow(result, value, &result))
+      return true;
+  }
+  *out = result;
+  return false;
+}
+
+// Variadic safe multiplication: computes the product and panics on overflow.
+template <typename... Args> inline size_t safe_mul(size_t first, Args... rest) {
+  size_t result;
+  bool overflow = checked_mul(&result, first, rest...);
+  PANIC_IF_FALSE(!overflow, "multiplication overflow wraps size_t");
+  return result;
+}
+
+// Variadic safe multiplication with an appended sizeof(T) factor.
+// Computes (args... * sizeof(T)) with overflow checking.
+template <typename T, typename... Args>
+inline size_t safe_mul_sizeof(Args... args) {
+  return safe_mul(args..., sizeof(T));
+}

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

@@ -0,0 +1,145 @@
+#ifndef DEVICE_H
+#define DEVICE_H
+
+#include <cstdint>
+#include <cstdio>
+#include <cstdlib>
+#include <cuda_runtime.h>
+
+extern "C" {
+
+#define check_cuda_error(ans)                                                  \
+  { cuda_error((ans), __FILE__, __LINE__); }
+inline void cuda_error(cudaError_t code, const char *file, int line) {
+  if (code != cudaSuccess) {
+    std::fprintf(stderr, "Cuda error: %s %s %d\n", cudaGetErrorString(code),
+                 file, 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__,         \
+                 __LINE__, __func__, ##__VA_ARGS__);                           \
+    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);
+
+void cuda_event_record(cudaEvent_t event, cudaStream_t stream,
+                       uint32_t gpu_index);
+void cuda_stream_wait_event(cudaStream_t stream, cudaEvent_t event,
+                            uint32_t gpu_index);
+
+void cuda_event_destroy(cudaEvent_t event, uint32_t gpu_index);
+
+cudaStream_t cuda_create_stream(uint32_t gpu_index);
+
+void cuda_destroy_stream(cudaStream_t stream, uint32_t gpu_index);
+
+void cuda_synchronize_stream(cudaStream_t stream, uint32_t gpu_index);
+
+uint32_t cuda_is_available();
+
+void *cuda_malloc(uint64_t size, uint32_t gpu_index);
+
+void *cuda_malloc_with_size_tracking_async(uint64_t size, cudaStream_t stream,
+                                           uint32_t gpu_index,
+                                           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,
+                                                 cudaStream_t stream,
+                                                 uint32_t gpu_index,
+                                                 bool gpu_memory_allocated);
+
+void cuda_memcpy_async_to_gpu(void *dest, const void *src, uint64_t size,
+                              cudaStream_t stream, uint32_t gpu_index);
+
+void cuda_memcpy_with_size_tracking_async_gpu_to_gpu(
+    void *dest, void const *src, uint64_t size, cudaStream_t stream,
+    uint32_t gpu_index, bool gpu_memory_allocated);
+
+void cuda_memcpy_async_gpu_to_gpu(void *dest, void const *src, uint64_t size,
+                                  cudaStream_t stream, uint32_t gpu_index);
+
+void cuda_memcpy_gpu_to_gpu(void *dest, void const *src, uint64_t size,
+                            uint32_t gpu_index);
+
+void cuda_memcpy_async_to_cpu(void *dest, const void *src, uint64_t size,
+                              cudaStream_t stream, uint32_t gpu_index);
+
+void cuda_memset_with_size_tracking_async(void *dest, uint64_t val,
+                                          uint64_t size, cudaStream_t stream,
+                                          uint32_t gpu_index,
+                                          bool gpu_memory_allocated);
+
+void cuda_memset_async(void *dest, uint64_t val, uint64_t size,
+                       cudaStream_t stream, uint32_t gpu_index);
+
+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);
+
+void cuda_drop_with_size_tracking_async(void *ptr, cudaStream_t stream,
+                                        uint32_t gpu_index,
+                                        bool gpu_memory_allocated);
+
+void cuda_drop_async(void *ptr, cudaStream_t stream, uint32_t gpu_index);
+}
+
+uint32_t cuda_get_max_shared_memory(uint32_t gpu_index);
+
+uint32_t cuda_get_max_shared_memory_per_block(uint32_t gpu_index);
+
+bool cuda_check_support_cooperative_groups();
+
+bool cuda_check_support_thread_block_clusters();
+
+template <typename Torus>
+void cuda_set_value_async(cudaStream_t stream, uint32_t gpu_index,
+                          Torus *d_array, Torus value, Torus n);
+
+#endif

backends/zk-cuda-backend/cuda/include/helper_profile.cuh

@@ -0,0 +1,16 @@
+#ifndef HELPER_PROFILE
+#define HELPER_PROFILE
+
+#ifdef USE_NVTOOLS
+#include <nvtx3/nvToolsExt.h>
+#endif
+
+void cuda_nvtx_label_with_color(const char *name);
+void cuda_nvtx_pop();
+
+#define PUSH_RANGE(name)                                                       \
+  { cuda_nvtx_label_with_color(name); }
+#define POP_RANGE()                                                            \
+  { cuda_nvtx_pop(); }
+
+#endif

backends/zk-cuda-backend/cuda/src/helper_profile.cu

@@ -0,0 +1,43 @@
+#include "helper_profile.cuh"
+#include <stdint.h>
+
+uint32_t adler32(const unsigned char *data) {
+  const uint32_t MOD_ADLER = 65521;
+  uint32_t a = 1, b = 0;
+  size_t index;
+  for (index = 0; data[index] != 0; ++index) {
+    a = (a + data[index] * 2) % MOD_ADLER;
+    b = (b + a) % MOD_ADLER;
+  }
+  return (b << 16) | a;
+}
+
+void cuda_nvtx_label_with_color(const char *name) {
+#ifdef USE_NVTOOLS
+  int color_id = adler32((const unsigned char *)name);
+  int r, g, b;
+  r = color_id & 0x000000ff;
+  g = (color_id & 0x000ff000) >> 12;
+  b = (color_id & 0x0ff00000) >> 20;
+  if (r < 64 & g < 64 & b < 64) {
+    r = r * 3;
+    g = g * 3 + 64;
+    b = b * 4;
+  }
+
+  color_id = 0xff000000 | (r << 16) | (g << 8) | (b);
+  nvtxEventAttributes_t eventAttrib = {0};
+  eventAttrib.version = NVTX_VERSION;
+  eventAttrib.size = NVTX_EVENT_ATTRIB_STRUCT_SIZE;
+  eventAttrib.colorType = NVTX_COLOR_ARGB;
+  eventAttrib.color = color_id;
+  eventAttrib.messageType = NVTX_MESSAGE_TYPE_ASCII;
+  eventAttrib.message.ascii = name;
+  nvtxRangePushEx(&eventAttrib);
+#endif
+}
+void cuda_nvtx_pop() {
+#ifdef USE_NVTOOLS
+  nvtxRangePop();
+#endif
+}

backends/zk-cuda-backend/src/c_wrapper.cu

@@ -11,7 +11,7 @@
 #include <stddef.h>
 #include <cstring>
 
-#include "../../tfhe-cuda-backend/cuda/src/utils/helper_profile.cuh"
+#include "helper_profile.cuh"
 
 // C++ helper functions (not exported, used internally)
 // These can call template functions since they have C++ linkage

ci/data_extractor/src/benchmark_specs.py

@@ -302,7 +302,7 @@ class BenchType(enum.StrEnum):
 
 class BenchSubset(enum.StrEnum):
     All = "all"
-    Erc20 = "erc20"
+    Erc7984 = "erc7984"
     Zk = "zk"
 
     @staticmethod
@@ -310,8 +310,8 @@ class BenchSubset(enum.StrEnum):
         match bench_subset.lower():
             case "all":
                 return BenchSubset.All
-            case "erc20":
-                return BenchSubset.Erc20
+            case "erc7984":
+                return BenchSubset.Erc7984
             case "zk":
                 return BenchSubset.Zk
             case _:
@@ -611,14 +611,14 @@ class BenchDetails:
                         # Case for arithmetic operations (add, sub, mul,...)
                         self.operation_name = "::".join(parts[2:-2])
                     else:
-                        # Case for higher-level operation (erc20 transfer, dex,...)
+                        # Case for higher-level operation (erc7984 transfer, dex,...)
                         self.operation_name = "::".join(parts[2:-1])
                 else:
                     if "_PARAM_" in parts[-2]:
                         # Case for arithmetic operations (add, sub, mul,...)
                         self.operation_name = "::".join(parts[1:-2])
                     else:
-                        # Case for higher-level operation (erc20 transfer, dex,...)
+                        # Case for higher-level operation (erc7984 transfer, dex,...)
                         self.operation_name = "::".join(parts[1:-1])
                 self.rust_type = parts[-1].partition("_mean")[0]
             case Layer.Shortint:

ci/data_extractor/src/data_extractor.py

@@ -137,7 +137,7 @@ parser.add_argument(
 parser.add_argument(
     "--bench-subset",
     dest="bench_subset",
-    choices=["all", "erc20", "zk"],
+    choices=["all", "erc7984", "zk"],
     default="all",
     help="Subset of benchmarks to filter against, dedicated formatting will be applied",
 )
@@ -285,8 +285,8 @@ def perform_hardware_comparison(
 
 def get_formatter(layer: Layer, bench_subset: BenchSubset):
     match bench_subset:
-        case BenchSubset.Erc20:
-            return formatters.hlapi.Erc20Formatter
+        case BenchSubset.Erc7984:
+            return formatters.hlapi.Erc7984Formatter
         case BenchSubset.Zk:
             if layer == Layer.Wasm:
                 return formatters.wasm.ZKFormatter
@@ -442,7 +442,7 @@ def get_operands_types(layer: Layer, bench_subset: BenchSubset = None):
         return ciphertext_only
     elif bench_subset:
         match bench_subset:
-            case BenchSubset.Zk | BenchSubset.Erc20:
+            case BenchSubset.Zk | BenchSubset.Erc7984:
                 return ciphertext_only
             case BenchSubset.All:
                 return ciphertext_and_plaintext

ci/data_extractor/src/formatters/hlapi/hlapi.py

@@ -39,9 +39,9 @@ class HlApiFormatter(GenericFormatter):
 TRANSFER_IMPLEM_COLUMN_HEADER = "Transfer implementation"
 
 
-class Erc20Formatter(HlApiFormatter):
+class Erc7984Formatter(HlApiFormatter):
     """
-    Formatter for ERC20 benchmarks.
+    Formatter for ERC7984 benchmarks.
     """
 
     @staticmethod
@@ -63,7 +63,7 @@ class Erc20Formatter(HlApiFormatter):
                 bench_type = BenchType.Latency
                 conversion_func = utils.convert_latency_value_to_readable_text
 
-            # For now ERC20 benchmarks are only made on 64-bit ciphertexts.
+            # For now ERC7984 benchmarks are only made on 64-bit ciphertexts.
             value = conversion_func(timings[-1])
             formatted[test_name][bench_type] = value
 

ci/regression.toml

@@ -66,7 +66,7 @@ target.hlapi-dex = [
     "swap_claim::whitepaper",
     "swap_claim::no_cmux"
 ]
-target.hlapi-erc20 = ["transfer::whitepaper", "transfer::no_cmux"]
+target.hlapi-erc7984 = ["transfer::whitepaper", "transfer::no_cmux"]
 target.core_crypto-ks = ["keyswitch"]
 target.core_crypto-pbs = ["multi_bit_pbs"]
 parameters_filter = "PARAM_GPU_MULTI_BIT_GROUP_4_MESSAGE_2_CARRY_2_KS_PBS_TUNIFORM_2M128"
@@ -99,7 +99,7 @@ target.hlapi-dex = [
     "swap_claim::whitepaper",
     "swap_claim::no_cmux"
 ]
-target.hlapi-erc20 = ["transfer::whitepaper", "transfer::no_cmux"]
+target.hlapi-erc7984 = ["transfer::whitepaper", "transfer::no_cmux"]
 target.shortint = ["bitand"]
 target.core_crypto-ks = ["keyswitch"]
 target.core_crypto-pbs = [ "pbs_mem_optimized"]

mockups/tfhe-hpu-mockup/Cargo.toml

@@ -16,19 +16,14 @@ tfhe = { path = "../../tfhe", features = ["hpu", "hpu-debug"] }
 
 ipc-channel = "0.18.3"
 
-strum = { version = "0.26.2", features = ["derive"] }
-strum_macros = "0.26.2"
 bytemuck = { workspace = true }
 
 clap = { version = "4.4.4", features = ["derive"] }
 clap-num = "*"
-anyhow = "1.0.82"
 tracing = "0.1.40"
 tracing-subscriber = { version = "0.3.18", features = ["env-filter", "json"] }
 serde_json = "1.0"
-rand = "0.8.5"
 serde = { version = "1", features = ["derive"] }
-bitflags = "2.6.0"
 
 [[bin]]
 name = "hpu_mockup"

mockups/tfhe-hpu-mockup/README.md

@@ -164,7 +164,7 @@ cargo run --release --features="hpu" --test hpu -- u8
 # * bitwise: for ct x ct bitwise IOps
 # * cmp: for comparison IOps
 # * ternary: for if-then-else and like IOps
-# * algo: for IOps dedicated to offload sub-algorithm like ERC_20
+# * algo: for IOps dedicated to offload sub-algorithm like ERC_7984
 # Command below only run comparison IOps, for convenience, `just hpu_test "cmp"` could be also used
 cargo run --release --features="hpu" --test hpu -- cmp
 ```

nightly-toolchain.txt

@@ -1 +1 @@
-nightly-2026-01-14
+nightly-2026-04-22

tests/backward_compatibility/high_level_api.rs

@@ -25,7 +25,7 @@ use tfhe::{
     CompressedKVStore, CompressedPublicKey, CompressedServerKey,
     CompressedSquashedNoiseCiphertextList, CompressedSquashedNoiseCiphertextListBuilder, FheBool,
     FheInt8, FheUint32, FheUint64, FheUint8, ReRandomizationContext, ReRandomizationMode,
-    ReRandomizationSupport, ServerKey, SquashedNoiseFheBool, SquashedNoiseFheInt,
+    ReRandomizationSupport, Seed, ServerKey, SquashedNoiseFheBool, SquashedNoiseFheInt,
     SquashedNoiseFheUint,
 };
 use tfhe_backward_compat_data::load::{
@@ -748,6 +748,22 @@ fn test_hl_key_features(
         }
     }
 
+    // OPRF: check that oblivious pseudo-random generation works with the dedicated key.
+    // The decrypted values only need to be within range; the seed is deterministic but we
+    // don't compare to specific bit values (those are validated in the unit tests).
+    if server_key.supports_oprf() {
+        let seed = Seed(42u128);
+
+        let rand_bool = FheBool::generate_oblivious_pseudo_random(seed);
+        let _: bool = rand_bool.decrypt(client_key);
+
+        let rand_uint = FheUint8::generate_oblivious_pseudo_random(seed);
+        let _: u8 = rand_uint.decrypt(client_key);
+
+        let rand_int = FheInt8::generate_oblivious_pseudo_random(seed);
+        let _: i8 = rand_int.decrypt(client_key);
+    }
+
     Ok(())
 }
 

tfhe-benchmark/Cargo.toml

@@ -92,8 +92,8 @@ harness = false
 required-features = ["integer", "internal-keycache", "pbs-stats"]
 
 [[bench]]
-name = "hlapi-erc20"
-path = "benches/high_level_api/erc20.rs"
+name = "hlapi-erc7984"
+path = "benches/high_level_api/erc7984.rs"
 harness = false
 required-features = ["integer", "internal-keycache"]
 

tfhe-benchmark/benches/high_level_api/dex.rs

@@ -253,7 +253,7 @@ mod pbs_stats {
         let _ = swap_request_update_dex_balance_func(&from_balance, &current_dex_balance, &amount);
         let count = tfhe::get_pbs_count() * 2;
 
-        println!("ERC20 swap request update dex balance/::{type_name}: {count} PBS");
+        println!("ERC7984 swap request update dex balance/::{type_name}: {count} PBS");
 
         let params = client_key.computation_parameters();
         let params_name = params.name();
@@ -308,7 +308,7 @@ mod pbs_stats {
         let (_, _) = swap_request_finalize_func(&to_balance_0, &total_dex_token_0_in, &sent_0);
         let count = tfhe::get_pbs_count() * 2;
 
-        println!("ERC20 swap request finalize/::{type_name}: {count} PBS");
+        println!("ERC7984 swap request finalize/::{type_name}: {count} PBS");
 
         let params = client_key.computation_parameters();
         let params_name = params.name();
@@ -373,7 +373,7 @@ mod pbs_stats {
         );
         let count = tfhe::get_pbs_count();
 
-        println!("ERC20 swap claim prepare/::{type_name}: {count} PBS");
+        println!("ERC7984 swap claim prepare/::{type_name}: {count} PBS");
 
         let params = client_key.computation_parameters();
         let params_name = params.name();
@@ -433,7 +433,7 @@ mod pbs_stats {
         );
         let count = tfhe::get_pbs_count() * 2;
 
-        println!("ERC20 swap claim update dex balance/::{type_name}: {count} PBS");
+        println!("ERC7984 swap claim update dex balance/::{type_name}: {count} PBS");
 
         let params = client_key.computation_parameters();
         let params_name = params.name();

tfhe-benchmark/benches/high_level_api/erc7984.rs

@@ -252,8 +252,8 @@ where
         boolean: vec![],
         imm: vec![],
     };
-    let mut res_handle = FheHpu::iop_exec(&hpu_asm::iop::IOP_ERC_20, src);
-    // Iop erc_20 return new_from, new_to
+    let mut res_handle = FheHpu::iop_exec(&hpu_asm::iop::IOP_ERC_7984, src);
+    // Iop erc_7984 return new_from, new_to
     let new_to = res_handle.native.pop().unwrap();
     let new_from = res_handle.native.pop().unwrap();
     (new_from, new_to)
@@ -278,8 +278,8 @@ where
         boolean: vec![],
         imm: vec![],
     };
-    let res_handle = FheHpu::iop_exec(&hpu_asm::iop::IOP_ERC_20_SIMD, src);
-    // Iop erc_20 return new_from, new_to
+    let res_handle = FheHpu::iop_exec(&hpu_asm::iop::IOP_ERC_7984_SIMD, src);
+    // Iop erc_7984 return new_from, new_to
     let res = res_handle.native;
     res
 }
@@ -319,18 +319,18 @@ mod pbs_stats {
         let (_, _) = transfer_func(&from_amount, &to_amount, &amount);
         let count = tfhe::get_pbs_count();
 
-        println!("ERC20 transfer/{fn_name}::{type_name}: {count} PBS");
+        println!("ERC7984 transfer/{fn_name}::{type_name}: {count} PBS");
 
         let params = client_key.computation_parameters();
         let params_name = params.name();
 
         let test_name = if cfg!(feature = "gpu") {
-            format!("hlapi::cuda::erc20::pbs_count::{fn_name}::{params_name}::{type_name}")
+            format!("hlapi::cuda::erc7984::pbs_count::{fn_name}::{params_name}::{type_name}")
         } else {
-            format!("hlapi::erc20::pbs_count::{fn_name}::{params_name}::{type_name}")
+            format!("hlapi::erc7984::pbs_count::{fn_name}::{params_name}::{type_name}")
         };
 
-        let results_file = Path::new("erc20_pbs_count.csv");
+        let results_file = Path::new("erc7984_pbs_count.csv");
         if !results_file.exists() {
             File::create(results_file).expect("create results file failed");
         }
@@ -392,7 +392,7 @@ fn bench_transfer_latency<FheType, F>(
         &bench_id,
         params,
         params_name,
-        "erc20-transfer",
+        "erc7984-transfer",
         &OperatorType::Atomic,
         64,
         vec![],
@@ -413,7 +413,7 @@ fn bench_transfer_latency_simd<FheType, F>(
     F: for<'a> Fn(&'a Vec<FheType>, &'a Vec<FheType>, &'a Vec<FheType>) -> Vec<FheType>,
 {
     use tfhe::tfhe_hpu_backend::prelude::hpu_asm;
-    let hpu_simd_n = hpu_asm::iop::IOP_ERC_20_SIMD
+    let hpu_simd_n = hpu_asm::iop::IOP_ERC_7984_SIMD
         .format()
         .unwrap()
         .proto
@@ -453,7 +453,7 @@ fn bench_transfer_latency_simd<FheType, F>(
         &bench_id,
         params,
         params_name,
-        "erc20-simd-transfer",
+        "erc7984-simd-transfer",
         &OperatorType::Atomic,
         64,
         vec![],
@@ -507,7 +507,7 @@ fn bench_transfer_throughput<FheType, F>(
             &bench_id,
             params,
             &params_name,
-            "erc20-transfer",
+            "erc7984-transfer",
             &OperatorType::Atomic,
             64,
             vec![],
@@ -597,7 +597,7 @@ fn cuda_bench_transfer_throughput<FheType, F>(
         &bench_id,
         params,
         &params_name,
-        "erc20-transfer",
+        "erc7984-transfer",
         &OperatorType::Atomic,
         64,
         vec![],
@@ -661,7 +661,7 @@ fn hpu_bench_transfer_throughput<FheType, F>(
             &bench_id,
             params,
             &params_name,
-            "erc20-transfer",
+            "erc7984-transfer",
             &OperatorType::Atomic,
             64,
             vec![],
@@ -683,7 +683,7 @@ fn hpu_bench_transfer_throughput_simd<FheType, F>(
     F: for<'a> Fn(&'a Vec<FheType>, &'a Vec<FheType>, &'a Vec<FheType>) -> Vec<FheType> + Sync,
 {
     use tfhe::tfhe_hpu_backend::prelude::hpu_asm;
-    let hpu_simd_n = hpu_asm::iop::IOP_ERC_20_SIMD
+    let hpu_simd_n = hpu_asm::iop::IOP_ERC_7984_SIMD
         .format()
         .unwrap()
         .proto
@@ -746,7 +746,7 @@ fn hpu_bench_transfer_throughput_simd<FheType, F>(
             &bench_id,
             params,
             &params_name,
-            "erc20-simd-ransfer",
+            "erc7984-simd-ransfer",
             &OperatorType::Atomic,
             64,
             vec![],
@@ -769,7 +769,7 @@ fn main() {
 
     let mut c = Criterion::default().sample_size(10).configure_from_args();
 
-    let bench_name = "hlapi::erc20";
+    let bench_name = "hlapi::erc7984";
 
     // FheUint64 PBS counts
     // We don't run multiple times since every input is encrypted
@@ -896,7 +896,7 @@ fn main() {
 
     let mut c = Criterion::default().sample_size(10).configure_from_args();
 
-    let bench_name = "hlapi::cuda::erc20";
+    let bench_name = "hlapi::cuda::erc7984";
 
     // FheUint64 PBS counts
     // We don't run multiple times since every input is encrypted
@@ -1027,7 +1027,7 @@ fn main() {
 
     let mut c = Criterion::default().sample_size(10).configure_from_args();
 
-    let bench_name = "hlapi::hpu::erc20";
+    let bench_name = "hlapi::hpu::erc7984";
 
     match get_bench_type() {
         BenchmarkType::Latency => {
@@ -1040,7 +1040,7 @@ fn main() {
                 "transfer::whitepaper",
                 transfer_whitepaper::<FheUint64>,
             );
-            // Erc20 optimized instruction only available on Hpu
+            // Erc7984 optimized instruction only available on Hpu
             bench_transfer_latency(
                 &mut group,
                 &cks,
@@ -1049,7 +1049,7 @@ fn main() {
                 "transfer::hpu_optim",
                 transfer_hpu::<FheUint64>,
             );
-            // Erc20 SIMD instruction only available on Hpu
+            // Erc7984 SIMD instruction only available on Hpu
             bench_transfer_latency_simd(
                 &mut group,
                 &cks,
@@ -1071,7 +1071,7 @@ fn main() {
                 "transfer::whitepaper",
                 transfer_whitepaper::<FheUint64>,
             );
-            // Erc20 optimized instruction only available on Hpu
+            // Erc7984 optimized instruction only available on Hpu
             hpu_bench_transfer_throughput(
                 &mut group,
                 &cks,
@@ -1080,7 +1080,7 @@ fn main() {
                 "transfer::hpu_optim",
                 transfer_hpu::<FheUint64>,
             );
-            // Erc20 SIMD instruction only available on Hpu
+            // Erc7984 SIMD instruction only available on Hpu
             hpu_bench_transfer_throughput_simd(
                 &mut group,
                 &cks,

tfhe-benchmark/benches/integer/oprf.rs

@@ -8,6 +8,7 @@ use rayon::prelude::*;
 #[cfg(any(feature = "gpu", feature = "hpu"))]
 use std::cmp::max;
 use tfhe::integer::keycache::KEY_CACHE;
+use tfhe::integer::oprf::{OprfPrivateKey, OprfServerKey};
 use tfhe::integer::IntegerKeyKind;
 use tfhe::keycache::NamedParam;
 #[cfg(any(feature = "gpu", feature = "hpu"))]
@@ -35,32 +36,42 @@ pub fn unsigned_oprf(c: &mut Criterion) {
                     format!("{bench_name}_bounded::{param_name}::{bit_size}_bits");
 
                 bench_group.bench_function(&bench_id_oprf, |b| {
-                    let (_, sk) = KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
+                    let (cks, sks) = KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
+                    let oprf_pk = OprfPrivateKey::new(&cks);
+                    let oprf_sk = OprfServerKey::new(&oprf_pk, &cks).unwrap();
 
                     b.iter(|| {
-                        _ = black_box(sk.par_generate_oblivious_pseudo_random_unsigned_integer(
-                            Seed(0),
-                            num_block as u64,
-                        ));
+                        _ = black_box(
+                            oprf_sk.par_generate_oblivious_pseudo_random_unsigned_integer(
+                                Seed(0),
+                                num_block as u64,
+                                &sks,
+                            ),
+                        );
                     })
                 });
 
                 bench_group.bench_function(&bench_id_oprf_bounded, |b| {
-                    let (_, sk) = KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
+                    let (cks, sks) = KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
+                    let oprf_pk = OprfPrivateKey::new(&cks);
+                    let oprf_sk = OprfServerKey::new(&oprf_pk, &cks).unwrap();
 
                     b.iter(|| {
                         _ = black_box(
-                            sk.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
+                            oprf_sk.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
                                 Seed(0),
                                 bit_size as u64,
                                 num_block as u64,
+                                &sks,
                             ),
                         );
                     })
                 });
             }
             BenchmarkType::Throughput => {
-                let (_, sk) = KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
+                let (cks, sks) = KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
+                let oprf_pk = OprfPrivateKey::new(&cks);
+                let oprf_sk = OprfServerKey::new(&oprf_pk, &cks).unwrap();
 
                 bench_id_oprf = format!("{bench_name}::throughput::{param_name}::{bit_size}_bits");
                 bench_id_oprf_bounded =
@@ -71,10 +82,11 @@ pub fn unsigned_oprf(c: &mut Criterion) {
                     {
                         // Execute the operation once to know its cost.
                         reset_pbs_count();
-                        sk.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
+                        oprf_sk.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
                             Seed(0),
                             bit_size as u64,
                             num_block as u64,
+                            &sks,
                         );
                         let pbs_count = max(get_pbs_count(), 1);
                         throughput_num_threads(num_block, pbs_count)
@@ -85,11 +97,13 @@ pub fn unsigned_oprf(c: &mut Criterion) {
                         let setup = |_batch_size: usize| ();
                         let run = |_: &mut (), batch_size: usize| {
                             (0..batch_size).into_par_iter().for_each(|_| {
-                                sk.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
-                                    Seed(0),
-                                    bit_size as u64,
-                                    num_block as u64,
-                                );
+                                oprf_sk
+                                    .par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
+                                        Seed(0),
+                                        bit_size as u64,
+                                        num_block as u64,
+                                        &sks,
+                                    );
                             });
                         };
                         find_optimal_batch(run, setup) as u64
@@ -100,9 +114,10 @@ pub fn unsigned_oprf(c: &mut Criterion) {
                 bench_group.bench_function(&bench_id_oprf, |b| {
                     b.iter(|| {
                         (0..elements).into_par_iter().for_each(|_| {
-                            sk.par_generate_oblivious_pseudo_random_unsigned_integer(
+                            oprf_sk.par_generate_oblivious_pseudo_random_unsigned_integer(
                                 Seed(0),
                                 num_block as u64,
+                                &sks,
                             );
                         })
                     })
@@ -111,10 +126,11 @@ pub fn unsigned_oprf(c: &mut Criterion) {
                 bench_group.bench_function(&bench_id_oprf_bounded, |b| {
                     b.iter(|| {
                         (0..elements).into_par_iter().for_each(|_| {
-                            sk.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
+                            oprf_sk.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
                                 Seed(0),
                                 bit_size as u64,
                                 num_block as u64,
+                                &sks,
                             );
                         })
                     })
@@ -148,6 +164,8 @@ pub mod cuda {
     use criterion::black_box;
     use tfhe::core_crypto::gpu::{get_number_of_gpus, CudaStreams};
     use tfhe::integer::gpu::server_key::CudaServerKey;
+    use tfhe::integer::gpu::CudaOprfServerKey;
+    use tfhe::integer::oprf::{CompressedOprfServerKey, OprfPrivateKey};
     use tfhe::GpuIndex;
     use tfhe_csprng::seeders::Seed;
 
@@ -177,12 +195,18 @@ pub mod cuda {
                         let (cks, _cpu_sks) =
                             KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
                         let gpu_sks = CudaServerKey::new(&cks, &streams);
+                        let oprf_pk = OprfPrivateKey::new(&cks);
+                        let compressed_oprf_sk =
+                            CompressedOprfServerKey::new(&oprf_pk, &cks).unwrap();
+                        let cuda_oprf_sk =
+                            CudaOprfServerKey::decompress_from_cpu(&compressed_oprf_sk, &streams);
 
                         b.iter(|| {
                             _ = black_box(
-                                gpu_sks.par_generate_oblivious_pseudo_random_unsigned_integer(
+                                cuda_oprf_sk.par_generate_oblivious_pseudo_random_unsigned_integer(
                                     Seed(0),
                                     num_block as u64,
+                                    &gpu_sks,
                                     &streams,
                                 ),
                             );
@@ -193,14 +217,20 @@ pub mod cuda {
                         let (cks, _cpu_sks) =
                             KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
                         let gpu_sks = CudaServerKey::new(&cks, &streams);
+                        let oprf_pk = OprfPrivateKey::new(&cks);
+                        let compressed_oprf_sk =
+                            CompressedOprfServerKey::new(&oprf_pk, &cks).unwrap();
+                        let cuda_oprf_sk =
+                            CudaOprfServerKey::decompress_from_cpu(&compressed_oprf_sk, &streams);
 
                         b.iter(|| {
                             _ = black_box(
-                                gpu_sks
+                                cuda_oprf_sk
                                     .par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
                                         Seed(0),
                                         bit_size as u64,
                                         num_block as u64,
+                                        &gpu_sks,
                                         &streams,
                                     ),
                             );
@@ -210,13 +240,25 @@ pub mod cuda {
                 BenchmarkType::Throughput => {
                     let (cks, cpu_sks) = KEY_CACHE.get_from_params(param, IntegerKeyKind::Radix);
                     let gpu_sks_vec = cuda_local_keys(&cks);
+                    let cpu_oprf_pk = OprfPrivateKey::new(&cks);
+                    let cpu_oprf_sk = OprfServerKey::new(&cpu_oprf_pk, &cks).unwrap();
+                    let compressed_oprf_sk =
+                        CompressedOprfServerKey::new(&cpu_oprf_pk, &cks).unwrap();
+                    // One CudaOprfServerKey per GPU, matching `gpu_sks_vec`.
+                    let cuda_oprf_sks_vec: Vec<CudaOprfServerKey> = (0..get_number_of_gpus())
+                        .map(|gpu_index| {
+                            let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
+                            CudaOprfServerKey::decompress_from_cpu(&compressed_oprf_sk, &stream)
+                        })
+                        .collect();
 
                     // Execute the operation once to know its cost.
                     reset_pbs_count();
-                    cpu_sks.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
+                    cpu_oprf_sk.par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
                         Seed(0),
                         bit_size as u64,
                         num_block as u64,
+                        &cpu_sks,
                     );
                     let pbs_count = max(get_pbs_count(), 1); // Operation might not perform any PBS, so we take 1 as default
 
@@ -232,10 +274,11 @@ pub mod cuda {
                             (0..elements).into_par_iter().for_each(|i| {
                                 let gpu_index: u32 = i as u32 % get_number_of_gpus();
                                 let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
-                                gpu_sks_vec[gpu_index as usize]
+                                cuda_oprf_sks_vec[gpu_index as usize]
                                     .par_generate_oblivious_pseudo_random_unsigned_integer(
                                         Seed(0),
                                         num_block as u64,
+                                        &gpu_sks_vec[gpu_index as usize],
                                         &stream,
                                     );
                             })
@@ -247,11 +290,12 @@ pub mod cuda {
                             (0..elements).into_par_iter().for_each(|i| {
                                 let gpu_index: u32 = i as u32 % get_number_of_gpus();
                                 let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
-                                gpu_sks_vec[gpu_index as usize]
+                                cuda_oprf_sks_vec[gpu_index as usize]
                                     .par_generate_oblivious_pseudo_random_unsigned_integer_bounded(
                                         Seed(0),
                                         bit_size as u64,
                                         num_block as u64,
+                                        &gpu_sks_vec[gpu_index as usize],
                                         &stream,
                                     );
                             })

tfhe-benchmark/benches/shortint/oprf.rs

@@ -2,6 +2,7 @@ use benchmark::params_aliases::*;
 use criterion::{black_box, criterion_group, Criterion};
 use tfhe::keycache::NamedParam;
 use tfhe::shortint::keycache::KEY_CACHE;
+use tfhe::shortint::oprf::{OprfPrivateKey, OprfServerKey};
 use tfhe_csprng::seeders::Seed;
 
 fn oprf(c: &mut Criterion) {
@@ -12,11 +13,15 @@ fn oprf(c: &mut Criterion) {
     let param = BENCH_PARAM_MESSAGE_2_CARRY_2_KS_PBS;
 
     let keys = KEY_CACHE.get_from_param(param);
+    let cks = keys.client_key();
     let sks = keys.server_key();
 
+    let oprf_pk = OprfPrivateKey::new(cks);
+    let oprf_sk = OprfServerKey::new(&oprf_pk, cks).unwrap();
+
     bench_group.bench_function(format!("2-bits-oprf::{}", param.name()), |b| {
         b.iter(|| {
-            _ = black_box(sks.generate_oblivious_pseudo_random(Seed(0), 2));
+            _ = black_box(oprf_sk.generate_oblivious_pseudo_random(Seed(0), 2, sks));
         })
     });
 }

tfhe-benchmark/src/bin/wasm_benchmarks_parser.rs

@@ -59,7 +59,7 @@ pub fn parse_wasm_benchmarks(results_file: &Path, raw_results_file: &Path) {
         let bench_name = name_parts[0];
         let params: PBSParameters = params_from_name(name_parts[1]).into();
         println!("{name_parts:?}");
-        if bench_name.contains("_size") {
+        if full_name.contains("_size") {
             write_result(&mut file, &prefixed_full_name, *val as usize);
         } else {
             let value_in_ns = (val * 1_000_000_f32) as usize;

tfhe-csprng/src/seeders/mod.rs

@@ -168,7 +168,7 @@ mod generic_tests {
     fn test_xof_seed_getters() {
         let seed_bytes = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16];
         let bits = u128::from_le_bytes(seed_bytes);
-        let dsep = [b't', b'f', b'h', b'e', b'k', b's', b'p', b's'];
+        let dsep = *b"tfheksps";
         let seed = XofSeed::new_u128(bits, dsep);
 
         let s = u128::from_le_bytes(seed.seed().try_into().unwrap());

tfhe-zk-pok/src/proofs/pke/mod.rs

@@ -342,6 +342,28 @@ impl<G: Curve> Proof<G> {
             None => ComputeLoad::Verify,
         }
     }
+
+    pub fn to_le_bytes(&self) -> Vec<u8> {
+        let mut bytes = Vec::new();
+
+        let Self {
+            c_hat,
+            c_y,
+            pi,
+            compute_load_proof_fields,
+        } = self;
+
+        bytes.extend_from_slice(c_hat.to_le_bytes().as_ref());
+        bytes.extend_from_slice(c_y.to_le_bytes().as_ref());
+        bytes.extend_from_slice(pi.to_le_bytes().as_ref());
+        let (c_hat_t_bytes, c_h_bytes, pi_kzg_bytes) =
+            ComputeLoadProofFields::to_le_bytes(compute_load_proof_fields);
+        bytes.extend_from_slice(&c_hat_t_bytes);
+        bytes.extend_from_slice(&c_h_bytes);
+        bytes.extend_from_slice(&pi_kzg_bytes);
+
+        bytes
+    }
 }
 
 impl<G: Curve> ParameterSetConformant for Proof<G> {
@@ -404,6 +426,26 @@ pub(crate) struct ComputeLoadProofFields<G: Curve> {
     pub(crate) pi_kzg: G::G1,
 }
 
+impl<G: Curve> ComputeLoadProofFields<G> {
+    #[allow(clippy::type_complexity)]
+    fn to_le_bytes(fields: &Option<Self>) -> (Box<[u8]>, Box<[u8]>, Box<[u8]>) {
+        if let Some(ComputeLoadProofFields {
+            c_hat_t,
+            c_h,
+            pi_kzg,
+        }) = fields.as_ref()
+        {
+            (
+                Box::from(G::G2::to_le_bytes(*c_hat_t).as_ref()),
+                Box::from(G::G1::to_le_bytes(*c_h).as_ref()),
+                Box::from(G::G1::to_le_bytes(*pi_kzg).as_ref()),
+            )
+        } else {
+            (Box::from([]), Box::from([]), Box::from([]))
+        }
+    }
+}
+
 type CompressedG2<G> = <<G as Curve>::G2 as Compressible>::Compressed;
 type CompressedG1<G> = <<G as Curve>::G1 as Compressible>::Compressed;
 

tfhe-zk-pok/src/proofs/pke_v2/mod.rs

@@ -440,6 +440,44 @@ impl<G: Curve> Proof<G> {
     pub fn hash_config(&self) -> PkeV2SupportedHashConfig {
         self.hash_config
     }
+
+    pub fn to_le_bytes(&self) -> Vec<u8> {
+        let mut bytes = Vec::new();
+
+        let Self {
+            C_hat_e,
+            C_e,
+            C_r_tilde,
+            C_R,
+            C_hat_bin,
+            C_y,
+            C_h1,
+            C_h2,
+            C_hat_t,
+            pi,
+            pi_kzg,
+            compute_load_proof_fields,
+            hash_config: _,
+        } = self;
+
+        bytes.extend_from_slice(C_hat_e.to_le_bytes().as_ref());
+        bytes.extend_from_slice(C_e.to_le_bytes().as_ref());
+        bytes.extend_from_slice(C_r_tilde.to_le_bytes().as_ref());
+        bytes.extend_from_slice(C_R.to_le_bytes().as_ref());
+        bytes.extend_from_slice(C_hat_bin.to_le_bytes().as_ref());
+        bytes.extend_from_slice(C_y.to_le_bytes().as_ref());
+        bytes.extend_from_slice(C_h1.to_le_bytes().as_ref());
+        bytes.extend_from_slice(C_h2.to_le_bytes().as_ref());
+        bytes.extend_from_slice(C_hat_t.to_le_bytes().as_ref());
+        bytes.extend_from_slice(pi.to_le_bytes().as_ref());
+        bytes.extend_from_slice(pi_kzg.to_le_bytes().as_ref());
+        let (C_hat_h3_bytes, C_hat_w_bytes) =
+            ComputeLoadProofFields::to_le_bytes(compute_load_proof_fields);
+        bytes.extend_from_slice(&C_hat_h3_bytes);
+        bytes.extend_from_slice(&C_hat_w_bytes);
+
+        bytes
+    }
 }
 
 /// These fields can be pre-computed on the prover side in the faster Verifier scheme. If that's the

tfhe/Cargo.toml

@@ -64,7 +64,7 @@ tfhe-fft = { version = "0.10.1", path = "../tfhe-fft", features = [
     "serde",
     "fft128",
 ] }
-tfhe-ntt = { version = "0.7.0", path = "../tfhe-ntt" }
+tfhe-ntt = { version = "0.7.1", path = "../tfhe-ntt" }
 pulp = { workspace = true, features = ["default"] }
 tfhe-cuda-backend = { version = "0.14.0", path = "../backends/tfhe-cuda-backend", optional = true }
 aligned-vec = { workspace = true, features = ["default", "serde"] }
@@ -99,7 +99,7 @@ serde-wasm-bindgen = { workspace = true, optional = true }
 getrandom = { workspace = true, optional = true }
 bytemuck = { workspace = true }
 
-tfhe-hpu-backend = { version = "0.4", path = "../backends/tfhe-hpu-backend", optional = true }
+tfhe-hpu-backend = { version = "0.5", path = "../backends/tfhe-hpu-backend", optional = true }
 
 [features]
 default = ["avx512"]

tfhe/docs/.gitbook/assets/cpu-gpu-hpu-integer-benchmark-fheuint64-tuniform-2m128-ciphertext.svg

@@ -8,23 +8,23 @@
 	<rect x="0" y="40" width="300" height="520" fill="#fbbc04"/>
 	<rect x="300" y="40" width="420" height="520" fill="#f3f3f3"/>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="60.0">Negation (-)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="60.0">71.5 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="60.0">77.5 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="60.0">9.08 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="60.0">8.4 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="100.0">Add / Sub (+,-)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="100.0">93.2 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="100.0">91.7 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="100.0">9.07 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="100.0">8.35 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="140.0">Mul (x)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="140.0">352 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="140.0">357 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="140.0">32.8 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="140.0">122 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="180.0">Equal / Not Equal (eq, ne)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="180.0">70.1 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="180.0">72.0 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="180.0">7.03 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="180.0">6.77 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="220.0">Comparisons (ge, gt, le, lt)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="220.0">87.6 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="220.0">89.5 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="220.0">10.6 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="220.0">6.81 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="260.0">Max / Min (max, min)</text>
@@ -32,31 +32,31 @@
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="260.0">15.0 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="260.0">11.7 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="300.0">Bitwise operations (&amp;, |, ^)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="300.0">19.1 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="300.0">19.0 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="300.0">1.99 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="300.0">2.95 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="340.0">Div / Rem  (/, %)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="340.0">5.04 s</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="340.0">4.88 s</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="340.0">514 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="340.0">912 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="380.0">Left / Right Shifts (&lt;&lt;, &gt;&gt;)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="380.0">119 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="380.0">121 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="380.0">18.0 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="380.0">25.8 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="420.0">Left / Right Rotations (left_rotate, right_rotate)</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="420.0">119 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="420.0">121 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="420.0">18.0 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="420.0">27.9 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="460.0">Leading / Trailing zeros/ones</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="460.0">223 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="460.0">222 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="460.0">20.2 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="460.0">14.7 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="500.0">Log2</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="500.0">244 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="500.0">246 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="500.0">21.9 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="500.0">14.8 ms</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="540.0">Select</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="540.0">39.3 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="370.0" y="540.0">40.2 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="510.0" y="540.0">4.7 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="650.0" y="540.0">5.53 ms</text>
 	<line stroke="white" stroke-width="2" x1="0" y1="0" x2="720" y2="0"/>

tfhe/docs/.gitbook/assets/cpu-hlapi-erc7984-benchmark-latency-throughput.svg

@@ -7,13 +7,13 @@
 	<rect x="0" y="40" width="300" height="120" fill="#fbbc04"/>
 	<rect x="300" y="40" width="420" height="120" fill="#f3f3f3"/>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="60.0">whitepaper</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="405.0" y="60.0">276 ms</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="615.0" y="60.0">23.0 ops/s</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="405.0" y="60.0">253 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="615.0" y="60.0">25.2 ops/s</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="100.0">no_cmux</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="405.0" y="100.0">238 ms</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="615.0" y="100.0">24.0 ops/s</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="405.0" y="100.0">256 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="615.0" y="100.0">25.2 ops/s</text>
 	<text dominant-baseline="middle" text-anchor="start" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="6" y="140.0">overflow</text>
-	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="405.0" y="140.0">225 ms</text>
+	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="405.0" y="140.0">238 ms</text>
 	<text dominant-baseline="middle" text-anchor="middle" font-family="Arial" font-size="14" font-weight="normal" fill="black" x="615.0" y="140.0">21.3 ops/s</text>
 	<line stroke="white" stroke-width="2" x1="0" y1="0" x2="720" y2="0"/>
 	<line stroke="white" stroke-width="2" x1="0" y1="40" x2="720" y2="40"/>