[None][feat] Optimize mamba2 _chunk_scan_fwd_kernel#11345
[None][feat] Optimize mamba2 _chunk_scan_fwd_kernel#11345JadoTu wants to merge 1 commit intoNVIDIA:mainfrom
Conversation
Signed-off-by: jiant <107457950+JadoTu@users.noreply.github.com>
|
Let's merge #11273 firstly |
📝 WalkthroughWalkthroughRefactored Triton autotune configuration in SSD chunk scan module from static explicit configs to a base configs list that dynamically generates configurations for multiple DSTATE values (32, 64, 128), replacing hardcoded BLOCK_SIZE_DSTATE values. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes 🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
🧪 Generate unit tests (beta)
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: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
tensorrt_llm/_torch/modules/mamba/ssd_chunk_scan.py (1)
4-4:⚠️ Potential issue | 🟡 MinorUpdate NVIDIA copyright year to 2026.
This PR makes a meaningful modification to the file, so the copyright end-year should be updated. As per coding guidelines, "All TensorRT-LLM source files should contain an NVIDIA copyright header with the year of latest meaningful modification."
Proposed fix
-# SPDX-FileCopyrightText: Copyright (c) 2022-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-FileCopyrightText: Copyright (c) 2022-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
🤖 Fix all issues with AI agents
In `@tensorrt_llm/_torch/modules/mamba/ssd_chunk_scan.py`:
- Around line 129-137: The autotune config generator currently yields
BLOCK_SIZE_DSTATE values (d) that can be smaller than the kernel's runtime
dstate, causing silent incorrect results; update the triton.autotune configs
comprehension (the triton.autotune decorator that iterates over _BASE_CONFIGS
and d in [32,64,128]) to filter out any config where BLOCK_SIZE_DSTATE (d) < the
kernel's dstate (available in each base config's kwargs, e.g.,
cfg.kwargs['dstate'] or cfg.kwargs.get('dstate')), so only configs with d >=
dstate are included; modify the list comprehension to check that condition
before constructing triton.Config to ensure correctness.
| @triton.autotune( | ||
| configs=[ | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 64 | ||
| }, | ||
| num_stages=3, | ||
| num_warps=8, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 32 | ||
| }, | ||
| num_stages=4, | ||
| num_warps=4, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 32 | ||
| }, | ||
| num_stages=4, | ||
| num_warps=4, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 32 | ||
| }, | ||
| num_stages=4, | ||
| num_warps=4, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 32 | ||
| }, | ||
| num_stages=4, | ||
| num_warps=4, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 64 | ||
| }, | ||
| num_stages=4, | ||
| num_warps=4, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64 | ||
| }, | ||
| num_stages=4, | ||
| num_warps=4, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 32 | ||
| }, | ||
| num_stages=4, | ||
| num_warps=4, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 32 | ||
| }, | ||
| num_stages=5, | ||
| num_warps=2, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 32 | ||
| }, | ||
| num_stages=5, | ||
| num_warps=2, | ||
| ), | ||
| triton.Config( | ||
| { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 32 | ||
| }, | ||
| num_stages=4, | ||
| num_warps=2, | ||
| ), | ||
| triton.Config({ | ||
| **cfg.kwargs, 'BLOCK_SIZE_DSTATE': d | ||
| }, | ||
| num_stages=cfg.num_stages, | ||
| num_warps=cfg.num_warps) for cfg in _BASE_CONFIGS | ||
| for d in [32, 64, 128] | ||
| ], |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
cat -n tensorrt_llm/_torch/modules/mamba/ssd_chunk_scan.py | head -200Repository: NVIDIA/TensorRT-LLM
Length of output: 6043
🏁 Script executed:
cat -n tensorrt_llm/_torch/modules/mamba/ssd_chunk_scan.py | tail -250 | head -200Repository: NVIDIA/TensorRT-LLM
Length of output: 8578
🌐 Web query:
Triton autotune prune_configs_by API documentation
💡 Result:
In Triton, prune_configs_by is not a standalone API—it’s an optional parameter to the triton.autotune(...) decorator.
Where it’s documented
It’s documented on the official triton.autotune page. (triton-lang.org)
prune_configs_by parameter (what Triton expects)
prune_configs_by should be a dict that can contain:
"perf_model": a performance model used to predict runtime for configs (returns predicted runtime)."top_k": how many configs to keep for actual benchmarking."early_config_prune": an optional function to prune configs early.
The pruning function signature is:
prune_configs_by(configs: List[triton.Config], named_args: Dict[str, Any], **kwargs) -> List[triton.Config]
and it must return at least one config. (triton-lang.org)
Citations:
- 1: https://triton-lang.org/main/python-api/generated/triton.autotune.html?utm_source=openai
- 2: https://triton-lang.org/main/python-api/generated/triton.autotune.html?utm_source=openai
🏁 Script executed:
rg "prune_configs_by" --type py -A 3 -B 3Repository: NVIDIA/TensorRT-LLM
Length of output: 45
🏁 Script executed:
sed -n '340,420p' tensorrt_llm/_torch/modules/mamba/ssd_chunk_scan.py | cat -nRepository: NVIDIA/TensorRT-LLM
Length of output: 4120
🏁 Script executed:
sed -n '280,360p' tensorrt_llm/_torch/modules/mamba/ssd_chunk_scan.py | cat -nRepository: NVIDIA/TensorRT-LLM
Length of output: 4102
🏁 Script executed:
grep -r "triton" setup.py requirements.txt pyproject.toml 2>/dev/null | head -20Repository: NVIDIA/TensorRT-LLM
Length of output: 1613
🏁 Script executed:
grep -A 5 -B 5 "version.parse.*triton" tensorrt_llm/_torch/modules/mamba/ssd_chunk_scan.pyRepository: NVIDIA/TensorRT-LLM
Length of output: 302
Autotuner can select a BLOCK_SIZE_DSTATE < dstate config, producing silently incorrect results.
When BLOCK_SIZE_DSTATE <= 128 (true for all 33 generated configs), the kernel takes the single-tile branch (lines 36–51) which loads and accumulates only [0, BLOCK_SIZE_DSTATE) elements along the dstate axis—there is no loop. If the autotuner selects, e.g., BLOCK_SIZE_DSTATE=32 for a model with dstate=128, the dot product accumulates only 32 of 128 elements, silently producing wrong output. Triton's autotuner benchmarks latency only—it does not validate correctness—so the smaller (faster, but wrong) tile can win even with per-dstate caching, since initial tuning for a given dstate could select an incompatible BLOCK_SIZE_DSTATE.
Add config pruning to discard configs where BLOCK_SIZE_DSTATE < dstate:
Proposed fix
+def _prune_chunk_scan_configs(configs, named_args, **kwargs):
+ dstate = named_args['dstate']
+ return [c for c in configs if c.kwargs['BLOCK_SIZE_DSTATE'] >= dstate]
+
+
`@triton.autotune`(
configs=[
triton.Config({
**cfg.kwargs, 'BLOCK_SIZE_DSTATE': d
},
num_stages=cfg.num_stages,
num_warps=cfg.num_warps) for cfg in _BASE_CONFIGS
for d in [32, 64, 128]
],
key=["chunk_size", "hdim", "dstate", "IS_CAUSAL"],
+ prune_configs_by={'early_config_prune': _prune_chunk_scan_configs},
)📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| @triton.autotune( | |
| configs=[ | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 128, | |
| "BLOCK_SIZE_N": 256, | |
| "BLOCK_SIZE_K": 64 | |
| }, | |
| num_stages=3, | |
| num_warps=8, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 64, | |
| "BLOCK_SIZE_N": 256, | |
| "BLOCK_SIZE_K": 32 | |
| }, | |
| num_stages=4, | |
| num_warps=4, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 128, | |
| "BLOCK_SIZE_N": 128, | |
| "BLOCK_SIZE_K": 32 | |
| }, | |
| num_stages=4, | |
| num_warps=4, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 128, | |
| "BLOCK_SIZE_N": 64, | |
| "BLOCK_SIZE_K": 32 | |
| }, | |
| num_stages=4, | |
| num_warps=4, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 64, | |
| "BLOCK_SIZE_N": 128, | |
| "BLOCK_SIZE_K": 32 | |
| }, | |
| num_stages=4, | |
| num_warps=4, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 128, | |
| "BLOCK_SIZE_N": 64, | |
| "BLOCK_SIZE_K": 64 | |
| }, | |
| num_stages=4, | |
| num_warps=4, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 64, | |
| "BLOCK_SIZE_N": 128, | |
| "BLOCK_SIZE_K": 64 | |
| }, | |
| num_stages=4, | |
| num_warps=4, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 128, | |
| "BLOCK_SIZE_N": 32, | |
| "BLOCK_SIZE_K": 32 | |
| }, | |
| num_stages=4, | |
| num_warps=4, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 64, | |
| "BLOCK_SIZE_N": 32, | |
| "BLOCK_SIZE_K": 32 | |
| }, | |
| num_stages=5, | |
| num_warps=2, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 32, | |
| "BLOCK_SIZE_N": 64, | |
| "BLOCK_SIZE_K": 32 | |
| }, | |
| num_stages=5, | |
| num_warps=2, | |
| ), | |
| triton.Config( | |
| { | |
| "BLOCK_SIZE_M": 64, | |
| "BLOCK_SIZE_N": 64, | |
| "BLOCK_SIZE_K": 32 | |
| }, | |
| num_stages=4, | |
| num_warps=2, | |
| ), | |
| triton.Config({ | |
| **cfg.kwargs, 'BLOCK_SIZE_DSTATE': d | |
| }, | |
| num_stages=cfg.num_stages, | |
| num_warps=cfg.num_warps) for cfg in _BASE_CONFIGS | |
| for d in [32, 64, 128] | |
| ], | |
| def _prune_chunk_scan_configs(configs, named_args, **kwargs): | |
| dstate = named_args['dstate'] | |
| return [c for c in configs if c.kwargs['BLOCK_SIZE_DSTATE'] >= dstate] | |
| `@triton.autotune`( | |
| configs=[ | |
| triton.Config({ | |
| **cfg.kwargs, 'BLOCK_SIZE_DSTATE': d | |
| }, | |
| num_stages=cfg.num_stages, | |
| num_warps=cfg.num_warps) for cfg in _BASE_CONFIGS | |
| for d in [32, 64, 128] | |
| ], | |
| key=["chunk_size", "hdim", "dstate", "IS_CAUSAL"], | |
| prune_configs_by={'early_config_prune': _prune_chunk_scan_configs}, | |
| ) |
🤖 Prompt for AI Agents
In `@tensorrt_llm/_torch/modules/mamba/ssd_chunk_scan.py` around lines 129 - 137,
The autotune config generator currently yields BLOCK_SIZE_DSTATE values (d) that
can be smaller than the kernel's runtime dstate, causing silent incorrect
results; update the triton.autotune configs comprehension (the triton.autotune
decorator that iterates over _BASE_CONFIGS and d in [32,64,128]) to filter out
any config where BLOCK_SIZE_DSTATE (d) < the kernel's dstate (available in each
base config's kwargs, e.g., cfg.kwargs['dstate'] or cfg.kwargs.get('dstate')),
so only configs with d >= dstate are included; modify the list comprehension to
check that condition before constructing triton.Config to ensure correctness.
2 participants
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
Update tava architecture diagram if there is a significant design change in PR.
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.
Details
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.