[None][feat] Add a standalone buffer cache class and reuse buffers between cduagraph and no-graph flow #7669
Conversation
📝 WalkthroughWalkthroughAdds a new CUDA memory buffer manager ( Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant M as DeepGemmFusedMoE.get_workspace
participant B as Buffers (singleton)
participant GP as Graph Pool
participant RP as Runtime Pool
participant Cuda as CUDA
Note over M: Detect capture_graph = torch.cuda.is_current_stream_capturing()
M->>B: get_buffer(shape,dtype, buffer_name, create_if_miss=True, pin_memory=capture_graph)
alt pin_memory (graph capture)
B->>GP: Find suitable buffer by name
alt Found
GP-->>B: Return candidate
else Not found
B->>Cuda: Allocate torch.zeros(..., device="cuda")
B->>GP: Store buffer
end
else not capturing
B->>RP: Check runtime buffer by name
B->>GP: Check graph buffers by name
Note over B: Select buffer with more elements if both exist
alt Suitable found
B-->>M: Return reshaped view
else None suitable
B->>RP: Delete existing runtime buffer (if any)
B->>Cuda: Allocate torch.zeros(..., device="cuda")
B->>RP: Store runtime buffer
end
end
B-->>M: Return reshaped view (workspace_*)
Note over M: Repeat for workspace_0, workspace_1, workspace_sf
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Pre-merge checks (3 warnings)❌ Failed checks (3 warnings)
Pre-merge checks and finishing touches❌ Failed checks (3 warnings)
✨ Finishing touches
🧪 Generate unit tests
Tip 👮 Agentic pre-merge checks are now available in preview!Pro plan users can now enable pre-merge checks in their settings to enforce checklists before merging PRs.
Please see the documentation for more information. Example: reviews:
pre_merge_checks:
custom_checks:
- name: "Undocumented Breaking Changes"
mode: "warning"
instructions: |
Pass/fail criteria: All breaking changes to public APIs, CLI flags, environment variables, configuration keys, database schemas, or HTTP/GraphQL endpoints must be documented in the "Breaking Change" section of the PR description and in CHANGELOG.md. Exclude purely internal or private changes (e.g., code not exported from package entry points or explicitly marked as internal).Please share your feedback with us on this Discord post. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 10
🧹 Nitpick comments (3)
tensorrt_llm/_torch/memory_buffer_utils.py (1)
77-77: Improve comment grammar.The comment has grammatical issues that should be fixed for clarity.
- # Reach here, no buffer is found. Then, we will use a new buffer to replace the small one. Release the memory first. + # If we reach here, no suitable buffer was found. Release any existing small buffer before allocating a new one.tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py (2)
417-417: Consider adding documentation for CUDA graph capture detection.The use of
torch.cuda.is_current_stream_capturing()to detect graph capture context is a key design decision that should be documented.Add a comment explaining the detection logic:
+ # Detect if we're in CUDA graph capture mode by checking the current stream state. + # During graph capture, we want to pin memory to avoid it being released back to the allocator. capture_graph = torch.cuda.is_current_stream_capturing()
369-371: Promote GetMemoryBuffer to instance attribute
Move thebuffer = GetMemoryBuffer()out of the class scope into__init__asself.buffer = GetMemoryBuffer()and document that this manager is intentionally shared across all MoE layer instances.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
tensorrt_llm/_torch/memory_buffer_utils.py(1 hunks)tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py(3 hunks)
🧰 Additional context used
📓 Path-based instructions (3)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Use only spaces, no tabs; indent with 4 spaces.
Files:
tensorrt_llm/_torch/memory_buffer_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+.
Indent Python code with 4 spaces; do not use tabs.
Maintain module namespace when importing; prefer 'from package.subpackage import foo' then 'foo.SomeClass()' instead of importing the class directly.
Python filenames should be snake_case (e.g., some_file.py).
Python classes use PascalCase names.
Functions and methods use snake_case names.
Local variables use snake_case; prefix 'k' for variables that start with a number (e.g., k_99th_percentile).
Global variables use upper SNAKE_CASE prefixed with 'G' (e.g., G_MY_GLOBAL).
Constants use upper SNAKE_CASE (e.g., MY_CONSTANT).
Avoid shadowing variables from an outer scope.
Initialize all externally visible members of a class in the constructor.
Prefer docstrings for interfaces that may be used outside a file; comments for in-function or file-local interfaces.
Use Google-style docstrings for classes and functions (Sphinx-parsable).
Document attributes and variables inline so they render under the class/function docstring.
Avoid reflection when a simpler, explicit approach suffices (e.g., avoid dict(**locals()) patterns).
In try/except, catch the most specific exceptions possible.
For duck-typing try/except, keep the try body minimal and use else for the main logic.
Files:
tensorrt_llm/_torch/memory_buffer_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py
**/*.{cpp,cxx,cc,h,hpp,hh,hxx,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Prepend the NVIDIA Apache-2.0 copyright header with current year to the top of all source files (e.g., .cpp, .h, .cu, .py).
Files:
tensorrt_llm/_torch/memory_buffer_utils.pytensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py
🧠 Learnings (1)
📚 Learning: 2025-08-19T03:35:20.866Z
Learnt from: djns99
PR: NVIDIA/TensorRT-LLM#6915
File: cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu:4616-4626
Timestamp: 2025-08-19T03:35:20.866Z
Learning: In the MOE profiler TMA workspace preparation (cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_kernels.cu), the overlapping of TMA WS regions for NONE and FINALIZE variants is deliberate design to save memory space, as confirmed by djns99. The comment "reuse the same pointers to save space" reflects this intentional behavior.
Applied to files:
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py
🧬 Code graph analysis (2)
tensorrt_llm/_torch/memory_buffer_utils.py (1)
tensorrt_llm/_utils.py (1)
numel(968-969)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py (1)
tensorrt_llm/_torch/memory_buffer_utils.py (2)
GetMemoryBuffer(95-97)get_buffer(30-89)
🪛 Ruff (0.12.2)
tensorrt_llm/_torch/memory_buffer_utils.py
31-31: Unused method argument: create_if_miss
(ARG002)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Pre-commit Check
🔇 Additional comments (1)
tensorrt_llm/_torch/modules/fused_moe/fused_moe_deepgemm.py (1)
718-724: Ignore workspace reuse issue: safe per-stream buffer reuse. The code bindsworkspace_0toaux_streamandworkspace_1to the main stream, and CUDA streams guarantee that operations within a stream execute in submission order; the initial event‐record/join further synchronizes the two streams, preventing concurrent writes to the same buffer. (developer.nvidia.com)Likely an incorrect or invalid review comment.
|
/bot run |
|
PR_Github #18379 [ run ] triggered by Bot |
|
PR_Github #18379 [ run ] completed with state |
HuiGao-NV
force-pushed
the
add_buffer_cache_class
branch
from
0ba7adb to
a5574c4
Compare
|
/bot run |
|
PR_Github #18412 [ run ] triggered by Bot |
|
PR_Github #18412 [ run ] completed with state |
|
/bot run --stage-list="B200_PCIe-PyTorch-1" --add-multi-gpu-test --disable-fail-fast |
|
PR_Github #18456 [ run ] triggered by Bot |
|
PR_Github #18456 [ run ] completed with state |
HuiGao-NV
force-pushed
the
add_buffer_cache_class
branch
from
8781b72 to
051bb73
Compare
|
/bot run --disable-fail-fast |
HuiGao-NV
force-pushed
the
add_buffer_cache_class
branch
from
051bb73 to
e56a4f6
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #18472 [ run ] triggered by Bot |
|
PR_Github #18472 [ run ] completed with state |
|
/bot run --stage-list="GB200-8_GPUs-2_Nodes-PyTorch-1,GB200-8_GPUs-2_Nodes-PyTorch-4,DGX_B200-8_GPUs-PyTorch-1" |
|
PR_Github #18486 [ run ] triggered by Bot |
|
PR_Github #18486 [ run ] completed with state |
HuiGao-NV
force-pushed
the
add_buffer_cache_class
branch
from
e56a4f6 to
4c7c30e
Compare
|
/bot run --stage-list="GB200-8_GPUs-2_Nodes-PyTorch-1,GB200-8_GPUs-2_Nodes-PyTorch-4,DGX_B200-8_GPUs-PyTorch-1" |
|
PR_Github #18505 [ run ] triggered by Bot |
|
PR_Github #18505 [ run ] completed with state |
HuiGao-NV
force-pushed
the
add_buffer_cache_class
branch
from
4c7c30e to
fb10bc7
Compare
|
/bot run --stage-list="B200_PCIe-PyTorch-2" --add-multi-gpu-test |
|
PR_Github #19965 [ run ] triggered by Bot |
|
PR_Github #19965 [ run ] completed with state |
|
/bot run --only-multi-gpu-test |
|
PR_Github #19996 [ run ] triggered by Bot |
|
PR_Github #19996 [ run ] completed with state |
|
error: subprocess.CalledProcessError: Command 'nvidia-smi -L' returned non-zero exit status 255. |
|
/bot run --only-multi-gpu-test --disable-fail-fast |
|
PR_Github #20006 [ run ] triggered by Bot |
HuiGao-NV
force-pushed
the
add_buffer_cache_class
branch
from
f1acc86 to
f72369d
Compare
|
PR_Github #20006 [ run ] completed with state |
|
pipeline of SBSA timed out when setup enviroment. Rerun the stages in it. |
|
/bot run --only-multi-gpu-test --stage-list="GB200-4_GPUs-PyTorch-1" |
|
PR_Github #20048 [ run ] triggered by Bot |
HuiGao-NV
force-pushed
the
add_buffer_cache_class
branch
from
f72369d to
d53d10b
Compare
|
/bot skip --comment "All stages passed before rebasement" |
|
PR_Github #20048 [ run ] completed with state |
|
PR_Github #20098 [ skip ] triggered by Bot |
and no-graph flow Signed-off-by: Hui Gao <huig@nvidia.com>
Signed-off-by: Hui Gao <huig@nvidia.com>
Signed-off-by: Hui Gao <huig@nvidia.com>
Signed-off-by: Hui Gao <huig@nvidia.com>
Signed-off-by: Hui Gao <huig@nvidia.com>
HuiGao-NV
force-pushed
the
add_buffer_cache_class
branch
from
d53d10b to
bd30b29
Compare
|
PR_Github #20098 [ skip ] completed with state |
|
/bot skip --comment "All stages passed before rebasement" |
|
PR_Github #20103 [ skip ] triggered by Bot |
|
PR_Github #20103 [ skip ] completed with state |
4 participants
and no-graph flow
Summary by CodeRabbit
Description
Test Coverage
PR Checklist
Please review the following before submitting your PR:
PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.
PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.
Test cases are provided for new code paths (see test instructions)
Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
The reviewers assigned automatically/manually are appropriate for the PR.
Please check this after reviewing the above items as appropriate for this PR.
GitHub Bot Help
/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...Provide a user friendly way for developers to interact with a Jenkins server.
Run
/bot [-h|--help]to print this help message.See details below for each supported subcommand.
run [--reuse-test (optional)pipeline-id --disable-fail-fast --skip-test --stage-list "A10-PyTorch-1, xxx" --gpu-type "A30, H100_PCIe" --test-backend "pytorch, cpp" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" --detailed-log --debug(experimental)]Launch build/test pipelines. All previously running jobs will be killed.
--reuse-test (optional)pipeline-id(OPTIONAL) : Allow the new pipeline to reuse build artifacts and skip successful test stages from a specified pipeline or the last pipeline if no pipeline-id is indicated. If the Git commit ID has changed, this option will be always ignored. The DEFAULT behavior of the bot is to reuse build artifacts and successful test results from the last pipeline.--disable-reuse-test(OPTIONAL) : Explicitly prevent the pipeline from reusing build artifacts and skipping successful test stages from a previous pipeline. Ensure that all builds and tests are run regardless of previous successes.--disable-fail-fast(OPTIONAL) : Disable fail fast on build/tests/infra failures.--skip-test(OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.--stage-list "A10-PyTorch-1, xxx"(OPTIONAL) : Only run the specified test stages. Examples: "A10-PyTorch-1, xxx". Note: Does NOT update GitHub check status.--gpu-type "A30, H100_PCIe"(OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.--test-backend "pytorch, cpp"(OPTIONAL) : Skip test stages which don't match the specified backends. Only support [pytorch, cpp, tensorrt, triton]. Examples: "pytorch, cpp" (does not run test stages with tensorrt or triton backend). Note: Does NOT update GitHub pipeline status.--only-multi-gpu-test(OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.--disable-multi-gpu-test(OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.--add-multi-gpu-test(OPTIONAL) : Force run the multi-GPU tests in addition to running L0 pre-merge pipeline.--post-merge(OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.--extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx"(OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx".--detailed-log(OPTIONAL) : Enable flushing out all logs to the Jenkins console. This will significantly increase the log volume and may slow down the job.--debug(OPTIONAL) : Experimental feature. Enable access to the CI container for debugging purpose. Note: Specify exactly one stage in thestage-listparameter to access the appropriate container environment. Note: Does NOT update GitHub check status.For guidance on mapping tests to stage names, see
docs/source/reference/ci-overview.mdand the
scripts/test_to_stage_mapping.pyhelper.kill
killKill all running builds associated with pull request.
skip
skip --comment COMMENTSkip testing for latest commit on pull request.
--comment "Reason for skipping build/test"is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.reuse-pipeline
reuse-pipelineReuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.