ROCm support

autonomys · Sep 27, 2024 · 97a3313 · 97a3313
1 parent 84ea07b
commit 97a3313
Show file tree

Hide file tree

Showing 15 changed files with 846 additions and 14 deletions.
diff --git a/.github/workflows/rust.yml b/.github/workflows/rust.yml
@@ -54,7 +54,7 @@ jobs:
   cargo-clippy:
     strategy:
       matrix:
-        os: ${{ fromJson(github.repository_owner == 'autonomys' && '[["self-hosted", "ubuntu-20.04-x86-64"], ["self-hosted", "macos-14-arm64"], ["self-hosted", "windows-server-2022-x86-64"]]' || '["ubuntu-22.04", "macos-14", "windows-2022"]') }}
+        os: ${{ fromJson(github.repository_owner == 'autonomys' && '[["self-hosted", "ubuntu-20.04-x86-64"], ["self-hosted", "macos-14-arm64"], ["self-hosted", "windows-server-2022-x86-64"]]' || '["ubuntu-20.04", "macos-14", "windows-2022"]') }}
 
     runs-on: ${{ matrix.os }}
 
@@ -104,6 +104,42 @@ jobs:
           sub-packages: '["nvcc", "cudart"]'
         if: runner.os == 'Linux' || runner.os == 'Windows'
 
+      # TODO: ROCm compilation doesn't work in CI right now, good luck fixing it
+      # - name: Configure ROCm cache (Windows)
+      #   uses: actions/cache@0c45773b623bea8c8e75f6c82b208c3cf94ea4f9 # v4.0.2
+      #   id: rocm-cache
+      #   with:
+      #     path: C:\Program Files\AMD\ROCm
+      #     key: ${{ runner.os }}-rocm
+      #   if: runner.os == 'Windows'
+
+      - name: ROCm toolchain
+        run: |
+          ROCM_VERSION=6.2.1
+          sudo mkdir -p --mode=0755 /etc/apt/keyrings
+          curl -L https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor | sudo tee /etc/apt/keyrings/rocm.gpg > /dev/null
+          echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/$ROCM_VERSION focal main" | sudo tee /etc/apt/sources.list.d/rocm.list > /dev/null
+          echo "Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600" | sudo tee /etc/apt/preferences.d/rocm-pin-600 > /dev/null
+          sudo apt-get update
+          DEBIAN_FRONTEND=noninteractive sudo apt-get install -y --no-install-recommends rocm-hip-runtime-dev
+          echo "/opt/rocm/lib" | sudo tee /etc/ld.so.conf.d/rocm.conf > /dev/null
+          sudo ldconfig
+        if: runner.os == 'Linux'
+
+      # TODO: ROCm compilation doesn't work in CI right now, good luck fixing it
+      # - name: ROCm toolchain
+      #   run: |
+      #     $ErrorActionPreference = "Stop"
+      #     Invoke-WebRequest -Uri https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-24.Q3-WinSvr2022-For-HIP.exe -OutFile "${env:RUNNER_TEMP}\HIP-SDK-Installer.exe"
+      #     Start-Process "${env:RUNNER_TEMP}\HIP-SDK-Installer.exe" -ArgumentList '-install' -NoNewWindow -Wait
+      #     Remove-Item "${env:RUNNER_TEMP}\HIP-SDK-Installer.exe"
+      #   if: runner.os == 'Windows' && steps.rocm-cache.outputs.cache-hit != 'true'
+      #
+      # - name: ROCm toolchain environment (Windows)
+      #   run: |
+      #     Add-Content $env:GITHUB_PATH "C:\Program Files\AMD\ROCm\6.1\bin"
+      #   if: runner.os == 'Windows'
+
       - name: Configure cache
         uses: actions/cache@0c45773b623bea8c8e75f6c82b208c3cf94ea4f9 # v4.0.2
         with:
@@ -124,6 +160,22 @@ jobs:
           cargo -Zgitoxide -Zgit clippy --locked --all-targets --features runtime-benchmarks,cuda -- -D warnings
         if: runner.os == 'Linux' || runner.os == 'Windows'
 
+      - name: cargo clippy (ROCm)
+        env:
+          NVCC: off
+        run: |
+          cargo -Zgitoxide -Zgit clippy --locked --all-targets --features rocm -- -D warnings
+        if: runner.os == 'Linux'
+
+      # TODO: ROCm compilation doesn't work in CI right now, good luck fixing it
+      # - name: cargo clippy (ROCm)
+      #   env:
+      #     NVCC: off
+      #     HIPCC: hipcc.bin.exe
+      #   run: |
+      #     cargo -Zgitoxide -Zgit clippy --locked --all-targets --features rocm -- -D warnings
+      #   if: runner.os == 'Windows'
+
   cargo-docs:
     runs-on: ${{ fromJson(github.repository_owner == 'autonomys' && '["self-hosted", "ubuntu-20.04-x86-64"]' || '"ubuntu-22.04"') }}
     steps:

diff --git a/.github/workflows/snapshot-build.yml b/.github/workflows/snapshot-build.yml
@@ -197,6 +197,43 @@ jobs:
           sub-packages: '["nvcc", "cudart"]'
         if: runner.os == 'Linux' || runner.os == 'Windows'
 
+      # TODO: ROCm compilation doesn't work in CI right now, good luck fixing it
+      # - name: Configure ROCm cache (Windows)
+      #   uses: actions/cache@0c45773b623bea8c8e75f6c82b208c3cf94ea4f9 # v4.0.2
+      #   id: rocm-cache
+      #   with:
+      #     path: C:\Program Files\AMD\ROCm
+      #     key: ${{ runner.os }}-rocm
+      #   if: runner.os == 'Windows'
+
+      - name: ROCm toolchain
+        run: |
+          ROCM_VERSION=6.2.1
+          sudo mkdir -p --mode=0755 /etc/apt/keyrings
+          curl -L https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor | sudo tee /etc/apt/keyrings/rocm.gpg > /dev/null
+          echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/$ROCM_VERSION focal main" | sudo tee /etc/apt/sources.list.d/rocm.list > /dev/null
+          echo "Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600" | sudo tee /etc/apt/preferences.d/rocm-pin-600 > /dev/null
+          sudo apt-get update
+          DEBIAN_FRONTEND=noninteractive sudo apt-get install -y --no-install-recommends rocm-hip-runtime-dev
+          echo "/opt/rocm/lib" | sudo tee /etc/ld.so.conf.d/rocm.conf > /dev/null
+          sudo ldconfig
+        # TODO: ROCm packages are only available for x86-64 for now
+        if: runner.os == 'Linux' && startsWith(matrix.build.target, 'x86_64')
+
+      # TODO: ROCm compilation doesn't work in CI right now, good luck fixing it
+      # - name: ROCm toolchain
+      #   run: |
+      #     $ErrorActionPreference = "Stop"
+      #     Invoke-WebRequest -Uri https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-24.Q3-WinSvr2022-For-HIP.exe -OutFile "${env:RUNNER_TEMP}\HIP-SDK-Installer.exe"
+      #     Start-Process "${env:RUNNER_TEMP}\HIP-SDK-Installer.exe" -ArgumentList '-install' -NoNewWindow -Wait
+      #     Remove-Item "${env:RUNNER_TEMP}\HIP-SDK-Installer.exe"
+      #   if: runner.os == 'Windows' && steps.rocm-cache.outputs.cache-hit != 'true'
+      #
+      # - name: ROCm toolchain environment (Windows)
+      #   run: |
+      #     Add-Content $env:GITHUB_PATH "C:\Program Files\AMD\ROCm\6.1\bin"
+      #   if: runner.os == 'Windows'
+
       - name: Configure cache
         uses: actions/cache@0c45773b623bea8c8e75f6c82b208c3cf94ea4f9 # v4.0.2
         with:
@@ -212,6 +249,28 @@ jobs:
           cargo -Zgitoxide -Zgit build --locked -Z build-std --target ${{ matrix.build.target }} --profile production --bin subspace-farmer
         if: runner.os == 'macOS' || !startsWith(matrix.build.target, 'x86_64')
 
+      # TODO: ROCm compilation doesn't work in CI right now, good luck fixing it
+      # # ROCm can't be enabled together with CUDA for now
+      # - name: Build farmer (ROCm, Windows)
+      #   env:
+      #     NVCC: off
+      #     HIPCC: hipcc.bin.exe
+      #   run: |
+      #     cargo -Zgitoxide -Zgit build --locked -Z build-std --target ${{ matrix.build.target }} --profile production --bin subspace-farmer --features rocm
+      #     move ${{ env.PRODUCTION_TARGET }}/subspace-farmer.exe ${{ env.PRODUCTION_TARGET }}/subspace-farmer-rocm.exe
+      #   # TODO: ROCm packages are only available for x86-64 for now
+      #   if: runner.os == 'Windows' && startsWith(matrix.build.target, 'x86_64')
+
+      # ROCm can't be enabled together with CUDA for now
+      - name: Build farmer (ROCm, Ubuntu)
+        env:
+          NVCC: off
+        run: |
+          cargo -Zgitoxide -Zgit build --locked -Z build-std --target ${{ matrix.build.target }} --profile production --bin subspace-farmer --features rocm
+          mv ${{ env.PRODUCTION_TARGET }}/subspace-farmer ${{ env.PRODUCTION_TARGET }}/subspace-farmer-rocm
+        # TODO: ROCm packages are only available for x86-64 for now
+        if: runner.os == 'Linux' && startsWith(matrix.build.target, 'x86_64')
+
       - name: Build farmer
         run: |
           cargo -Zgitoxide -Zgit build --locked -Z build-std --target ${{ matrix.build.target }} --profile production --bin subspace-farmer --features cuda
@@ -270,6 +329,8 @@ jobs:
       - name: Sign Application (Windows)
         run: |
           AzureSignTool sign --azure-key-vault-url "${{ secrets.AZURE_KEY_VAULT_URI }}" --azure-key-vault-client-id "${{ secrets.AZURE_CLIENT_ID }}" --azure-key-vault-client-secret "${{ secrets.AZURE_CLIENT_SECRET }}" --azure-key-vault-tenant-id "${{ secrets.AZURE_TENANT_ID }}" --azure-key-vault-certificate "${{ secrets.AZURE_CERT_NAME }}" --file-digest sha512 --timestamp-rfc3161 http://timestamp.digicert.com -v "${{ env.PRODUCTION_TARGET }}/subspace-farmer.exe"
+          # TODO: ROCm compilation doesn't work in CI right now, good luck fixing it
+          # AzureSignTool sign --azure-key-vault-url "${{ secrets.AZURE_KEY_VAULT_URI }}" --azure-key-vault-client-id "${{ secrets.AZURE_CLIENT_ID }}" --azure-key-vault-client-secret "${{ secrets.AZURE_CLIENT_SECRET }}" --azure-key-vault-tenant-id "${{ secrets.AZURE_TENANT_ID }}" --azure-key-vault-certificate "${{ secrets.AZURE_CERT_NAME }}" --file-digest sha512 --timestamp-rfc3161 http://timestamp.digicert.com -v "${{ env.PRODUCTION_TARGET }}/subspace-farmer-rocm.exe"
           AzureSignTool sign --azure-key-vault-url "${{ secrets.AZURE_KEY_VAULT_URI }}" --azure-key-vault-client-id "${{ secrets.AZURE_CLIENT_ID }}" --azure-key-vault-client-secret "${{ secrets.AZURE_CLIENT_SECRET }}" --azure-key-vault-tenant-id "${{ secrets.AZURE_TENANT_ID }}" --azure-key-vault-certificate "${{ secrets.AZURE_CERT_NAME }}" --file-digest sha512 --timestamp-rfc3161 http://timestamp.digicert.com -v "${{ env.PRODUCTION_TARGET }}/subspace-node.exe"
         # Allow code signing to fail on non-release builds and in non-subspace repos (forks)
         continue-on-error: ${{ github.repository_owner != 'autonomys' || github.event_name != 'push' || github.ref_type != 'tag' }}
@@ -282,6 +343,12 @@ jobs:
           mv ${{ env.PRODUCTION_TARGET }}/subspace-node executables/subspace-node-${{ matrix.build.suffix }}
         if: runner.os == 'Linux'
 
+      - name: Prepare executables for uploading (Ubuntu, ROCm)
+        run: |
+          mv ${{ env.PRODUCTION_TARGET }}/subspace-farmer-rocm executables/subspace-farmer-rocm-${{ matrix.build.suffix }}
+        # TODO: ROCm packages are only available for x86-64 for now
+        if: runner.os == 'Linux' && startsWith(matrix.build.target, 'x86_64')
+
       - name: Prepare executables for uploading (macOS)
         run: |
           mkdir executables
@@ -298,6 +365,8 @@ jobs:
         run: |
           mkdir executables
           move ${{ env.PRODUCTION_TARGET }}/subspace-farmer.exe executables/subspace-farmer-${{ matrix.build.suffix }}.exe
+          # TODO: ROCm compilation doesn't work in CI right now, good luck fixing it
+          # move ${{ env.PRODUCTION_TARGET }}/subspace-farmer-rocm.exe executables/subspace-farmer-rocm-${{ matrix.build.suffix }}.exe
           move ${{ env.PRODUCTION_TARGET }}/subspace-node.exe executables/subspace-node-${{ matrix.build.suffix }}.exe
         if: runner.os == 'Windows'
 

diff --git a/Cargo.lock b/Cargo.lock
diff --git a/Dockerfile-farmer b/Dockerfile-farmer
@@ -46,11 +46,39 @@ RUN \
     curl -OL https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/$CUDA_ARCH/cuda-ubuntu2004.pin && \
     mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600 && \
     apt-get update && \
-    DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends cuda-minimal-build-12-4
+    DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends cuda-minimal-build-12-4 && \
+    echo "/usr/local/cuda/lib64" > /etc/ld.so.conf.d/cuda.conf && \
+    ldconfig
 
+# ROCm is only used on x86-64 since they don't have other packages
+ARG ROCM_VERSION=6.2.1
+RUN \
+    if [ $(uname -p) = "x86_64" ]; then \
+      DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends gpg && \
+      mkdir -p --mode=0755 /etc/apt/keyrings && \
+      curl -L https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor > /etc/apt/keyrings/rocm.gpg && \
+      echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/$ROCM_VERSION focal main" > /etc/apt/sources.list.d/rocm.list && \
+      echo "Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600" > /etc/apt/preferences.d/rocm-pin-600 && \
+      apt-get update && \
+      DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends rocm-hip-runtime-dev && \
+      echo "/opt/rocm/lib" > /etc/ld.so.conf.d/rocm.conf && \
+      ldconfig \
+    ; fi
+
+# TODO: Remove `NVCC=off` hack once `sppark` has proper features for CUDA and ROCm
+# ROCm is only used on x86-64 since they don't have other packages
 RUN \
     export PATH=/usr/local/cuda/bin${PATH:+:${PATH}} && \
-    export LD_LIBRARY_PATH=/usr/local/cuda/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}} && \
+    if [ $(uname -p) = "x86_64" ]; then \
+      NVCC=off /root/.cargo/bin/cargo -Zgitoxide -Zgit build \
+          --locked \
+          -Z build-std \
+          --profile $PROFILE \
+          --bin subspace-farmer \
+          --features rocm \
+          --target $(uname -p)-unknown-linux-gnu && \
+      mv target/*/*/subspace-farmer subspace-farmer-rocm \
+    ; fi && \
     /root/.cargo/bin/cargo -Zgitoxide -Zgit build \
         --locked \
         -Z build-std \
@@ -63,7 +91,26 @@ RUN \
 
 FROM ubuntu:20.04
 
-COPY --from=0 /code/subspace-farmer /subspace-farmer
+# Next block is for ROCm support
+# ROCm is only used on x86-64 since they don't have other packages
+ARG ROCM_VERSION=6.2.1
+RUN \
+    if [ $(uname -p) = "x86_64" ]; then \
+      apt-get update && \
+      DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends curl ca-certificates gpg && \
+      mkdir -p --mode=0755 /etc/apt/keyrings && \
+      curl -L https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor > /etc/apt/keyrings/rocm.gpg && \
+      echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/$ROCM_VERSION focal main" > /etc/apt/sources.list.d/rocm.list && \
+      echo "Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600" > /etc/apt/preferences.d/rocm-pin-600 && \
+      apt-get update && \
+      DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends hip-runtime-amd && \
+      DEBIAN_FRONTEND=noninteractive apt-get remove -y --purge --autoremove curl ca-certificates gpg && \
+      rm -rf /var/lib/apt/lists/* && \
+      echo "/opt/rocm/lib" > /etc/ld.so.conf.d/rocm.conf && \
+      ldconfig \
+    ; fi
+
+COPY --from=0 /code/subspace-farmer* /
 
 RUN mkdir /var/subspace && chown nobody:nogroup /var/subspace
 

diff --git a/crates/subspace-farmer/Cargo.toml b/crates/subspace-farmer/Cargo.toml
@@ -78,6 +78,7 @@ cluster = ["dep:async-nats"]
 numa = ["dep:hwlocality"]
 # Only Volta+ architectures are supported (GeForce RTX 20xx consumer GPUs and newer)
 cuda = ["_gpu", "subspace-proof-of-space-gpu/cuda"]
+rocm = ["_gpu", "subspace-proof-of-space-gpu/rocm"]
 # Internal feature, shouldn't be used directly
 _gpu = []