Multi-GPU scheduling

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

@@ -10,13 +10,23 @@ on:
         type: choice
         options:
           - "erc20"
+          - "dex"
           - "synthetics"
           - "all"
       batch_size:
         description: "Batch sizes (# FHE operations executed per batch)"
         required: true
         type: string
         default: "5000"
+      scheduling_policy:
+        description: "Scheduling policy"
+        required: true
+        type: choice
+        options:
+          - "MAX_PARALLELISM"
+          - "MAX_LOCALITY"
+          - "LOOP"
+          - "FINE_GRAIN"
 
 env:
   CARGO_TERM_COLOR: always
@@ -54,6 +64,7 @@ jobs:
     timeout-minutes: 720  # 12 hours
     permissions:
       contents: 'read'
+      packages: 'read'
     strategy:
       fail-fast: false
       # explicit include-based build matrix, of known valid options
@@ -64,11 +75,12 @@ jobs:
             gcc: 11
 
     steps:
-      - name: Install git LFS
+      - name: Wait for unattended upgrades to complete
         run: |
-          sudo systemctl stop unattended-upgrades
-          sleep 10
           sudo apt-get update
+
+      - name: Install git LFS
+        run: |
           sudo apt-get install -y git-lfs
           git lfs install
 
@@ -91,13 +103,16 @@ jobs:
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
-        uses: dtolnay/rust-toolchain@888c2e1ea69ab0d4330cbf0af1ecc7b68f368cc1
+        uses: dtolnay/rust-toolchain@a54c7afa936fefeb4456b2dd8068152669aa8203
         with:
           toolchain: nightly
 
       - name: Install cargo dependencies
         run: |
           sudo systemctl stop docker
+          DEBIAN_FRONTEND=noninteractive sudo apt-get remove -y docker docker-engine docker.io containerd runc
+          DEBIAN_FRONTEND=noninteractive sudo apt-get purge -y docker-ce docker-ce-cli containerd.io docker-compose-plugin docker-compose
+          sudo rm -rf /etc/bash_completion.d/docker /usr/local/bin/docker-compose /etc/bash_completion.d/docker-compose
           DEBIAN_FRONTEND=noninteractive sudo apt-get install -y protobuf-compiler cmake \
                                                                  pkg-config libssl-dev \
                                                                  libclang-dev docker-compose-v2 \
@@ -118,6 +133,13 @@ jobs:
           key: ${{ runner.os }}-cargo-${{ hashFiles('**/Cargo.lock') }}
           restore-keys: ${{ runner.os }}-cargo-
 
+      - name: Login to GitHub Container Registry
+        uses: docker/login-action@9780b0c442fbb1117ed29e0efdff1e18412f7567 # v3.3.0
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
       - name: Init database
         run: make init_db
         working-directory: coprocessor/fhevm-engine/coprocessor
@@ -127,17 +149,13 @@ jobs:
         with:
           node-version: 20.x
 
-      - name: Build contracts
-        env:
-          HARDHAT_NETWORK: hardhat
+      - name: Start localstack
         run: |
-          cp ./host-contracts/.env.example ./host-contracts/.env
-          npm --prefix ./host-contracts ci --include=optional
-          cd host-contracts && npm install && npm run deploy:emptyProxies && npx hardhat compile
+          docker run --rm -d -p 4566:4566 --name localstack localstack/localstack:latest
 
       - name: Run benchmarks on CPU
         run: |
-          BENCHMARK_BATCH_SIZE="${{ inputs.batch_size }}" make "benchmark_${{ inputs.benchmarks }}_cpu"
+          DATABASE_URL=postgresql://postgres:postgres@localhost:5432/coprocessor TXN_SENDER_TEST_GLOBAL_LOCALSTACK=1 BENCHMARK_BATCH_SIZE="${{ inputs.batch_size }}" FHEVM_DF_SCHEDULE="${{ inputs.scheduling_policy }}" make -e "benchmark_${{ inputs.benchmarks }}_cpu"
         working-directory: coprocessor/fhevm-engine/coprocessor
 
       - name: Parse results
@@ -151,7 +169,7 @@ jobs:
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
-          --name-suffix "operation_batch_size_${{ inputs.batch_size }}"
+          --name-suffix "operation_batch_size_${{ inputs.batch_size }}-schedule_${{ inputs.scheduling_policy }}"
 
       - name: Upload parsed results artifact
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02

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

@@ -23,13 +23,40 @@ on:
         type: choice
         options:
           - "erc20"
+          - "dex"
           - "synthetics"
           - "all"
       batch_size:
         description: "Batch sizes (# FHE operations executed per batch)"
         required: true
         type: string
         default: "5000"
+      scheduling_policy:
+        description: "Scheduling policy"
+        required: true
+        type: choice
+        options:
+          - "MAX_PARALLELISM"
+          - "MAX_LOCALITY"
+          - "LOOP"
+          - "FINE_GRAIN"
+      optimization_target:
+        description: "Optimization target"
+        required: false
+        type: choice
+        options:
+          - "throughput"
+          - "latency"
+        default: "throughput"
+      benchmark_type:
+        description: "Benchmark type"
+        required: false
+        type: choice
+        options:
+          - "ALL"
+          - "THROUGHPUT"
+          - "LATENCY"
+        default: "ALL"
 
 env:
   CARGO_TERM_COLOR: always
@@ -87,6 +114,7 @@ jobs:
     timeout-minutes: 720  # 12 hours
     permissions:
       contents: 'read'
+      packages: 'read'
     strategy:
       fail-fast: false
       # explicit include-based build matrix, of known valid options
@@ -125,7 +153,7 @@ jobs:
           {
             echo "BENCH_DATE=$(date --iso-8601=seconds)";
             echo "COMMIT_DATE=$(git --no-pager show -s --format=%cd --date=iso8601-strict ${{ github.sha }})";
-            echo "COMMIT_HASH=$(git describe --tags --dirty)";
+            echo "COMMIT_HASH=$(git rev-parse HEAD)";
           } >> "${GITHUB_ENV}"
 
       - name: Install rust
@@ -155,6 +183,13 @@ jobs:
           key: ${{ runner.os }}-cargo-${{ hashFiles('**/Cargo.lock') }}
           restore-keys: ${{ runner.os }}-cargo-
 
+      - name: Login to GitHub Container Registry
+        uses: docker/login-action@9780b0c442fbb1117ed29e0efdff1e18412f7567 # v3.3.0
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
       - name: Init database
         run: make init_db
         working-directory: coprocessor/fhevm-engine/coprocessor
@@ -174,7 +209,7 @@ jobs:
 
       - name: Run benchmarks on GPU
         run: |
-          BENCHMARK_BATCH_SIZE="${{ inputs.batch_size }}" make "benchmark_${{ inputs.benchmarks }}_gpu"
+          BENCHMARK_BATCH_SIZE="${{ inputs.batch_size }}" FHEVM_DF_SCHEDULE="${{ inputs.scheduling_policy }}" BENCHMARK_TYPE="${{ inputs.benchmark_type }}" OPTIMIZATION_TARGET="${{ inputs.optimization_target }}" make -e "benchmark_${{ inputs.benchmarks }}_gpu"
         working-directory: coprocessor/fhevm-engine/coprocessor
 
       - name: Parse results
@@ -188,7 +223,8 @@ jobs:
           --commit-date "${{ env.COMMIT_DATE }}" \
           --bench-date "${{ env.BENCH_DATE }}" \
           --walk-subdirs \
-          --name-suffix "operation_batch_size_${{ inputs.batch_size }}"
+          --crate "coprocessor/fhevm-engine/coprocessor" \
+          --name-suffix "operation_batch_size_${{ inputs.batch_size }}-schedule_${{ inputs.scheduling_policy }}-optimization_target_${{ inputs.optimization_target }}"
 
       - name: Upload parsed results artifact
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02

ci/benchmark_parser.py

@@ -99,7 +99,7 @@
 parser.add_argument(
     "--crate",
     dest="crate",
-    default="fhevm-engine/coprocessor",
+    default="coprocessor/fhevm-engine/coprocessor",
     help="Crate for which benchmarks have run",
 )
 

coprocessor/fhevm-engine/coprocessor/Cargo.toml

@@ -50,6 +50,9 @@ scheduler = { path = "../scheduler" }
 [features]
 nightly-avx512 = ["tfhe/nightly-avx512"]
 gpu = ["tfhe/gpu", "scheduler/gpu", "fhevm-engine-common/gpu"]
+bench = []
+latency = ["fhevm-engine-common/latency"]
+throughput = ["fhevm-engine-common/throughput"]
 
 [dev-dependencies]
 testcontainers = { workspace = true }
@@ -81,3 +84,8 @@ harness = false
 name = "synthetics"
 path = "benches/synthetics.rs"
 harness = false
+
+[[bench]]
+name = "dex"
+path = "benches/dex.rs"
+harness = false

coprocessor/fhevm-engine/coprocessor/Makefile

@@ -1,5 +1,6 @@
 
 DB_URL ?= DATABASE_URL=postgres://postgres:[email protected]:5432/coprocessor
+OPTIMIZATION_TARGET ?= throughput
 
 .PHONY: build
 build:
@@ -41,25 +42,33 @@ clean_run:
 
 .PHONY: benchmark_erc20_gpu # Run ECR20 transfer benchmarks on GPU
 benchmark_erc20_gpu:
-	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench erc20 --features=gpu --
+	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench erc20 --features=gpu,bench,$(OPTIMIZATION_TARGET) --
+
+.PHONY: benchmark_dex_gpu # Run DEX benchmarks on GPU
+benchmark_dex_gpu:
+	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench dex --features=gpu,bench,$(OPTIMIZATION_TARGET) --
 
 .PHONY: benchmark_synthetics_gpu # Run all benchmarks on GPU
 benchmark_synthetics_gpu:
-	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench synthetics --features=gpu --
+	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench synthetics --features=gpu,bench,$(OPTIMIZATION_TARGET) --
 
 .PHONY: benchmark_all_gpu # Run all benchmarks on GPU
 benchmark_all_gpu:
-	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench erc20 --bench synthetics --features=gpu --
+	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench erc20 --bench dex --bench synthetics --features=gpu,bench,$(OPTIMIZATION_TARGET) --
 
 
 .PHONY: benchmark_erc20_cpu # Run ECR20 transfer benchmarks on CPU
 benchmark_erc20_cpu:
-	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench erc20 --
+	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench erc20 --features=bench --
+
+.PHONY: benchmark_dex_cpu # Run DEX benchmarks on CPU
+benchmark_dex_cpu:
+	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench dex --features=bench --
 
 .PHONY: benchmark_synthetics_cpu # Run all benchmarks on CPU
 benchmark_synthetics_cpu:
-	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench synthetics --
+	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench synthetics --features=bench --
 
 .PHONY: benchmark_all_cpu # Run all benchmarks on CPU
 benchmark_all_cpu:
-	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench erc20 --bench synthetics --
+	RUSTFLAGS="-C target-cpu=native" $(DB_URL) cargo +nightly bench --bench erc20 --bench dex --bench synthetics --features=bench --