Add ROCm 6.4, 7.0 and 7.2 support#810
Closed
arsdragonfly wants to merge 21 commits intomicrosoft:mainfrom
Closed
Conversation
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <copilot@github.com>
# Conflicts: # superbench/benchmarks/micro_benchmarks/gpu_stream/CMakeLists.txt
Co-authored-by: Copilot <copilot@github.com>
Contributor
There was a problem hiding this comment.
Pull request overview
Adds ROCm 6.4 / 7.0 / 7.2 support across SuperBench micro-benchmarks and container images, including making parsing/build steps resilient to ROCm toolchain/output changes.
Changes:
- Make
hipblaslt-benchparsing robust to evolving output schemas by using header-based column lookup, and extend unit tests to cover the newer format. - Enable
gpu-streamon ROCm by switching memory-clock queries from NVML torocm_smiand updating the CMake build flow to support HIP/hipify. - Introduce new ROCm 6.4 / 7.0 / 7.2 Dockerfiles and a standalone CMake build script for
hipblaslt-benchon ROCm 7.2.
Reviewed changes
Copilot reviewed 13 out of 13 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| third_party/Makefile | Injects <cassert> into HIP’s hipBusBandwidth sample to satisfy newer ROCm clang behavior. |
| tests/benchmarks/micro_benchmarks/test_hipblaslt_function.py | Adds a positive test for the newer hipblaslt-bench CSV schema. |
| superbench/benchmarks/micro_benchmarks/hipblaslt_function.py | Updates result parsing to map columns by header name instead of fixed indices. |
| superbench/benchmarks/micro_benchmarks/gpu_stream/gpu_stream_utils.hpp | Switches NVML include to rocm_smi when building under HIP. |
| superbench/benchmarks/micro_benchmarks/gpu_stream/gpu_stream.cu | Adds ROCm SMI-based memory clock querying for ROCm builds. |
| superbench/benchmarks/micro_benchmarks/gpu_stream/CMakeLists.txt | Adds a CUDA vs ROCm build split; hipifies sources and links rocm_smi on ROCm. |
| superbench/benchmarks/micro_benchmarks/gpu_stream.py | Registers gpu-stream for ROCm in the benchmark registry. |
| dockerfile/rocm7.2.x.dockerfile | New ROCm 7.2 image; builds hipblaslt-bench via standalone CMake to avoid upstream 7.2 build issues. |
| dockerfile/rocm7.0.x.dockerfile | New ROCm 7.0 image with updated RCCL/hipBLASLt build flow and TE install. |
| dockerfile/rocm6.4.x.dockerfile | New ROCm 6.4 image with hipBLASLt build adjustments and related environment setup. |
| dockerfile/rocm6.2.x.dockerfile | Updates Intel MLC download URL/version. |
| dockerfile/etc/hipblaslt-bench-standalone.cmake | Adds standalone CMakeLists content to build only hipblaslt-bench against system hipBLASLt. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+8
to
+13
| # Place this file at the root of an upstream hipBLASLt source tree | ||
| # (e.g. cp this to /path/to/hipBLASLt/CMakeLists-bench.txt) and invoke: | ||
| # | ||
| # cmake -B build -S /path/to/hipBLASLt -P /path/to/this/file | ||
| # | ||
| # Or use it as the top-level CMakeLists.txt by overwriting it. |
There was a problem hiding this comment.
The usage instructions are incorrect: cmake -P runs CMake in script mode and will not configure/generate a build from a project()/targets file. Either instruct users to copy this file as the top-level CMakeLists.txt and run a normal cmake -S ... -B ..., or provide a separate script-mode (-P) driver that configures a build directory via cmake -S ... -B ....
Contributor
|
@arsdragonfly Thanks for your contribution, could we break down this PR to 3 separate PR? Then each PR contain 1 ROCM version. Thanks! |
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 14 out of 14 changed files in this pull request and generated 6 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| ./bootstrap --prefix=/usr --no-system-curl --parallel=16 && \ | ||
| make -j ${NUM_MAKE_JOBS} && \ | ||
| make install && \ | ||
| rm -rf /tmp/cmake-${required_version}* \ |
|
|
||
|
|
||
| BenchmarkRegistry.register_benchmark('gpu-stream', GpuStreamBenchmark, platform=Platform.CUDA) | ||
| BenchmarkRegistry.register_benchmark('gpu-stream', GpuStreamBenchmark, platform=Platform.ROCM) |
Comment on lines
+199
to
+202
| # Install TransformerEngine — ROCm 7.0 has hip_fp4.h and gfx950 support, | ||
| # so we can use the latest dev branch with full CK fused attention. | ||
| RUN git clone --recursive https://github.com/ROCm/TransformerEngine.git && \ | ||
| cd TransformerEngine && \ |
Comment on lines
+206
to
+208
| RUN python3 -m pip install onnxscript && \ | ||
| git clone --recursive https://github.com/ROCm/TransformerEngine.git && \ | ||
| cd TransformerEngine && \ |
Comment on lines
+8
to
+13
| # Place this file at the root of an upstream hipBLASLt source tree | ||
| # (e.g. cp this to /path/to/hipBLASLt/CMakeLists-bench.txt) and invoke: | ||
| # | ||
| # cmake -B build -S /path/to/hipBLASLt -P /path/to/this/file | ||
| # | ||
| # Or use it as the top-level CMakeLists.txt by overwriting it. |
Comment on lines
+199
to
+205
| # Install TransformerEngine — ROCm 7.0 has hip_fp4.h and gfx950 support, | ||
| # so we can use the latest dev branch with full CK fused attention. | ||
| RUN git clone --recursive https://github.com/ROCm/TransformerEngine.git && \ | ||
| cd TransformerEngine && \ | ||
| NVTE_FRAMEWORK=pytorch \ | ||
| NVTE_FUSED_ATTN_CK=0 \ | ||
| NVTE_FUSED_ATTN_AOTRITON=1 \ |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.