Skip to content

Add ROCm 7.0 support#813

Draft
arsdragonfly wants to merge 2 commits intomicrosoft:mainfrom
arsdragonfly:arsdragonfly/rocm-7.0
Draft

Add ROCm 7.0 support#813
arsdragonfly wants to merge 2 commits intomicrosoft:mainfrom
arsdragonfly:arsdragonfly/rocm-7.0

Conversation

@arsdragonfly
Copy link
Copy Markdown

Description

Adds a ROCm 7.0 dockerfile and the cross-cutting changes that first become
relevant at this version.

Depends on #812. While #812 is unmerged, this PR's diff also includes
the ROCm 6.4 changes; once #812 merges into main, the diff here will
shrink to just the 7.0-specific commit.

This PR is the second of three stacked PRs that split #810 by ROCm version:

  1. Add ROCm 6.4 support #812Add ROCm 6.4 support
  2. This PRAdd ROCm 7.0 support
  3. Add ROCm 7.2 support (stacked on this one)

Changes (7.0-specific)

  • dockerfile/rocm7.0.x.dockerfile — new ROCm 7.0 image.
  • third_party/Makefile — inject <cassert> into hipBusBandwidth.cpp;
    newer clang in ROCm ≥ 7 rejects assert() without it.
  • superbench/benchmarks/micro_benchmarks/hipblaslt_function.py — parse
    hipblaslt-bench output by header-name lookup, so the parser handles
    both the legacy 23-column schema and the new 33+ column schema (extra
    type/scale/GB-s columns) introduced in hipBLASLt 7.x.
  • tests/benchmarks/micro_benchmarks/test_hipblaslt_function.py
    regression test for the new schema.
  • .github/workflows/build-image.yml — add rocm7.0 matrix entry.

Add dockerfile/rocm6.4.x.dockerfile and required cross-cutting changes:
- gpu_stream: register on Platform.ROCM and port the microbenchmark to HIP.
- third_party/rccl-tests: bump submodule for ROCm 6.4 compatibility.
- dockerfile/rocm6.2.x.dockerfile: bump Intel MLC to v3.12 (3.10 mirror gone).
- CI: add rocm6.4 entry to build-image workflow.
Add dockerfile/rocm7.0.x.dockerfile and required cross-cutting changes:
- third_party/Makefile: inject <cassert> into hipBusBandwidth.cpp; newer clang
  in ROCm >= 7 rejects assert() without it.
- hipblaslt_function: parse hipblaslt-bench output by header-name lookup so
  the benchmark works across both the legacy 23-column and the new 33+ column
  schemas introduced in hipblaslt 7.x. Add a regression test for the new
  schema.
- CI: add rocm7.0 entry to build-image workflow.

Stacks on top of the ROCm 6.4 PR.
Copilot AI review requested due to automatic review settings May 5, 2026 00:25
Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR adds ROCm 7.0 support to SuperBench’s container/tooling stack and updates a couple of micro-benchmarks to stay compatible with ROCm 7.x toolchain/output changes.

Changes:

  • Add new ROCm 6.4 and ROCm 7.0 Docker images and wire them into the image build workflow matrix.
  • Make hipblaslt-bench result parsing schema-robust by resolving columns via the header row (supports both older ~23-col and newer 33+ col outputs), plus a regression test.
  • Update ROCm-related native build bits (HIP bandwidth sample <cassert> include injection; GPU stream benchmark ROCm support via rocm_smi).

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file
File Description
third_party/Makefile Injects <cassert> into HIP sample source to satisfy newer clang requirements.
superbench/benchmarks/micro_benchmarks/hipblaslt_function.py Updates hipBLASLt output parsing to use header-name lookup across schema versions.
tests/benchmarks/micro_benchmarks/test_hipblaslt_function.py Adds a regression test covering the newer (33+ column) hipBLASLt output format.
superbench/benchmarks/micro_benchmarks/gpu_stream/gpu_stream.cu Adds ROCm SMI support for memory clock queries on AMD GPUs and updates docstrings accordingly.
superbench/benchmarks/micro_benchmarks/gpu_stream/gpu_stream_utils.hpp Adds ROCm SMI include gating for AMD builds.
superbench/benchmarks/micro_benchmarks/gpu_stream/CMakeLists.txt Splits CUDA vs ROCm build paths; adds hipify step and links rocm_smi.
superbench/benchmarks/micro_benchmarks/gpu_stream.py Registers gpu-stream benchmark for Platform.ROCM in addition to CUDA.
dockerfile/rocm7.0.x.dockerfile Adds a new ROCm 7.0 image definition and ROCm 7.0-specific build steps.
dockerfile/rocm6.4.x.dockerfile Adds a new ROCm 6.4 image definition (present due to stacking dependency).
dockerfile/rocm6.2.x.dockerfile Updates Intel MLC download URL/version to v3.12.
.github/workflows/build-image.yml Adds rocm6.4 and rocm7.0 to the image build matrix.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

cmake --build lapack/src/lapack-build -j$(nproc) --target install && \
cd ../.. && \
mkdir -p build/release && cd build/release && \
CMAKE_POLICY_VERSION_MINIMUM= cmake \
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants