[TRTLLM-7263][fix] Prevent recreation of cublas handles in lora_grouped_gemm every call #6968
Conversation
📝 WalkthroughWalkthroughIntroduce a thread-local, lazily-initialized Changes
Sequence Diagram(s)sequenceDiagram
participant Caller
participant loraOp
participant ThreadLocalWrapper as thread_local cublasWrapper
participant CUBLAS as cuBLAS/cuBLASLt
participant LoraImpl
Caller->>loraOp: lora_grouped_gemm(...)
loraOp->>ThreadLocalWrapper: check initialized?
alt not initialized
loraOp->>CUBLAS: getCublasHandle()/getCublasLtHandle()
CUBLAS-->>loraOp: handles
loraOp->>ThreadLocalWrapper: create CublasMMWrapper(handles)
else initialized
ThreadLocalWrapper-->>loraOp: reuse wrapper
end
loraOp->>LoraImpl: invoke with cublasWrapper
LoraImpl-->>Caller: return result
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Tip 🔌 Remote MCP (Model Context Protocol) integration is now available!Pro plan users can now connect to remote MCP servers from the Integrations page. Connect with popular remote MCPs such as Notion and Linear to add more context to your reviews and chats. 📜 Recent review detailsConfiguration used: .coderabbit.yaml 💡 Knowledge Base configuration:
You can enable these sources in your CodeRabbit configuration. 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
⏰ 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)
✨ Finishing Touches
🧪 Generate unit tests
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. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. CodeRabbit Commands (Invoked using PR/Issue comments)Type Other keywords and placeholders
Status, Documentation and Community
|
|
/bot run |
|
PR_Github #15534 [ run ] triggered by Bot |
![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
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (1)
cpp/tensorrt_llm/thop/loraOp.cpp(2 hunks)
🧰 Additional context used
📓 Path-based instructions (4)
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh}: In C++, close namespaces with a comment naming the namespace (e.g., } // namespace foo)
Prefer const/constexpr variables over #define for constants
Declare variables const if not modified after initialization
Use Allman brace style in C++
C++ filenames use lowerCamelCase and must be case-insensitively unique within a build target
C++ type names use UpperCamelCase
Local variables, methods, and namespaces use lowerCamelCase
Global non-static variables not in anonymous namespace use gPrefix lowerCamelCase (e.g., gExample)
Static globals or globals in anonymous namespaces use sPrefix lowerCamelCase
Locally visible static variables start with 's' (e.g., static std::once_flag sFlag;)
Member variables use mPrefix lowerCamelCase; public members may omit but are encouraged to use 'm'
Constants (enums, global/static/function-scope magic numbers) use kPREFIXED_UPPER_SNAKE (e.g., kDIGIT_NUM)
If macros are unavoidable, use UPPER_SNAKE_CASE (prefer constants over #define)
Constructor parameter that conflicts with a public member name gets trailing underscore (foo_)
Literal suffixes should be uppercase (e.g., 1234L not 1234l)
C++: use spaces only; indent 4 spaces
Run clang-format (LLVM style) before submitting; wrap lines at 120 characters
If formatting must be bypassed, use // clang-format off/on around the section
Prefer smart pointers; use unique_ptr for sole ownership, shared_ptr for shared; weak_ptr only in exceptional cases
Do not use deprecated pre-C++11 smart pointers
Use C++ style comments; avoid C comments except special inline cases; prefer // single-line
Capitalize and punctuate full-sentence comments
Follow Doxygen rules: use //! for comments and //!< for members in C++
Disable code with #if/#endif and mnemonic conditions; avoid commented-out code; avoid dead code
Do not throw exceptions across library boundaries
Use least-forceful casts; avoid removing const/volatile; avoid C-style and functional casts (except constructors); p...
Files:
cpp/tensorrt_llm/thop/loraOp.cpp
**/*.{cpp,cxx,cc,cu}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.{cpp,cxx,cc,cu}: Avoid literal values except for 0, nullptr, true, false; use named constexpr for other literals
Place semicolon of empty for/while loop on a new line
Always use brace-delimited bodies for switch/while/do-for/if/else
Use inline C comments in argument lists when parameter meaning is unclear (e.g., /* checkForErrors = */ false)
Do not use assignment in subexpressions (e.g., if (x = y) ... is forbidden)
Switch on enums should enumerate all values and omit default to catch new values at compile time
Structure switch statements; prohibit fallthrough except between empty cases; each case ends with break or throw; return at end of case not allowed; put break inside braces for compound case
Prefer anonymous namespaces over static for internal linkage of functions
Every defined function must be called at least once (no unused methods)
Files:
cpp/tensorrt_llm/thop/loraOp.cpp
**/*.{h,hpp,hxx,hh,cuh,cpp,cxx,cc,cu}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Parameter names must be consistent between declarations and definitions
Files:
cpp/tensorrt_llm/thop/loraOp.cpp
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Prepend NVIDIA copyright header (current year) to all source files
Files:
cpp/tensorrt_llm/thop/loraOp.cpp
⏰ 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)
cpp/tensorrt_llm/thop/loraOp.cpp (1)
158-158: Good: pass the persistent wrapper into LoraImplReusing a thread-local wrapper amortizes handle initialization and should remove per-call cublas{Lt} handle churn.
Confirm that LoraImpl doesn’t hold onto any per-call stream state without refreshing it (your previous comment’s setStream per call covers this). If LoraImpl internally sets wrapper’s stream in run(), keeping the explicit setStream is still safe and clearer.
|
PR_Github #15534 [ run ] completed with state |
|
/bot run |
|
PR_Github #15538 [ run ] triggered by Bot |
amitz-nv
changed the title
|
PR_Github #15538 [ run ] completed with state |
amitz-nv
force-pushed
the
dev-fix-lora-grouped-gemm-cublas-handle-lifetime
branch
from
86a7f05 to
618f582
Compare
|
/bot run |
There was a problem hiding this comment.
Actionable comments posted: 0
♻️ Duplicate comments (1)
cpp/tensorrt_llm/thop/loraOp.cpp (1)
157-158: Passing the persistent wrapper into LoraImpl is correct; no extra setStream() needed hereGiven LoraImpl::run() sets the stream internally in _runGemm() (and also handles workspace), this keeps stream management in the execution layer and avoids redundant calls in the op wrapper.
🧹 Nitpick comments (2)
cpp/tensorrt_llm/thop/loraOp.cpp (2)
133-140: Optional: guard handle creation with the input’s device for future-proofingIf this thread ever switches devices (now or in future refactors), creating the handles under the correct CUDA context is safer. Cheap to add and harmless in single-device-per-process setups.
Apply this diff around handle creation:
thread_local std::shared_ptr<tensorrt_llm::common::CublasMMWrapper> cublasWrapper; if (cublasWrapper == nullptr) { + // Ensure handles are created against the input's device/context. + at::cuda::CUDAGuard deviceGuard{input.device()}; auto cublasHandle = getCublasHandle(); auto cublasLtHandle = getCublasLtHandle(); cublasWrapper = std::make_shared<tensorrt_llm::common::CublasMMWrapper>(cublasHandle, cublasLtHandle, nullptr, nullptr); }Add the missing header (outside this hunk):
#include <ATen/cuda/CUDAContext.h>
133-140: Nit: prefix the function-scope thread_local with 's' and use consistentlyPer the coding guidelines, locally visible statics (including thread_local with static storage duration) should use an sPrefix.
Apply this diff:
-thread_local std::shared_ptr<tensorrt_llm::common::CublasMMWrapper> cublasWrapper; -if (cublasWrapper == nullptr) +thread_local std::shared_ptr<tensorrt_llm::common::CublasMMWrapper> sCublasWrapper; +if (sCublasWrapper == nullptr) { auto cublasHandle = getCublasHandle(); auto cublasLtHandle = getCublasLtHandle(); - cublasWrapper + sCublasWrapper = std::make_shared<tensorrt_llm::common::CublasMMWrapper>(cublasHandle, cublasLtHandle, nullptr, nullptr); } ... -auto mLoraImpl = std::make_shared<tensorrt_llm::kernels::LoraImpl>( - inHiddenSize, outHiddenSizes, transA, transB, numLoraModules, loraRuntimeDataType, max_low_rank, cublasWrapper); +auto mLoraImpl = std::make_shared<tensorrt_llm::kernels::LoraImpl>( + inHiddenSize, outHiddenSizes, transA, transB, numLoraModules, loraRuntimeDataType, max_low_rank, sCublasWrapper);Also applies to: 157-158
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (1)
cpp/tensorrt_llm/thop/loraOp.cpp(2 hunks)
🧰 Additional context used
📓 Path-based instructions (4)
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh}: In C++, close namespaces with a comment naming the namespace (e.g., } // namespace foo)
Prefer const/constexpr variables over #define for constants
Declare variables const if not modified after initialization
Use Allman brace style in C++
C++ filenames use lowerCamelCase and must be case-insensitively unique within a build target
C++ type names use UpperCamelCase
Local variables, methods, and namespaces use lowerCamelCase
Global non-static variables not in anonymous namespace use gPrefix lowerCamelCase (e.g., gExample)
Static globals or globals in anonymous namespaces use sPrefix lowerCamelCase
Locally visible static variables start with 's' (e.g., static std::once_flag sFlag;)
Member variables use mPrefix lowerCamelCase; public members may omit but are encouraged to use 'm'
Constants (enums, global/static/function-scope magic numbers) use kPREFIXED_UPPER_SNAKE (e.g., kDIGIT_NUM)
If macros are unavoidable, use UPPER_SNAKE_CASE (prefer constants over #define)
Constructor parameter that conflicts with a public member name gets trailing underscore (foo_)
Literal suffixes should be uppercase (e.g., 1234L not 1234l)
C++: use spaces only; indent 4 spaces
Run clang-format (LLVM style) before submitting; wrap lines at 120 characters
If formatting must be bypassed, use // clang-format off/on around the section
Prefer smart pointers; use unique_ptr for sole ownership, shared_ptr for shared; weak_ptr only in exceptional cases
Do not use deprecated pre-C++11 smart pointers
Use C++ style comments; avoid C comments except special inline cases; prefer // single-line
Capitalize and punctuate full-sentence comments
Follow Doxygen rules: use //! for comments and //!< for members in C++
Disable code with #if/#endif and mnemonic conditions; avoid commented-out code; avoid dead code
Do not throw exceptions across library boundaries
Use least-forceful casts; avoid removing const/volatile; avoid C-style and functional casts (except constructors); p...
Files:
cpp/tensorrt_llm/thop/loraOp.cpp
**/*.{cpp,cxx,cc,cu}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.{cpp,cxx,cc,cu}: Avoid literal values except for 0, nullptr, true, false; use named constexpr for other literals
Place semicolon of empty for/while loop on a new line
Always use brace-delimited bodies for switch/while/do-for/if/else
Use inline C comments in argument lists when parameter meaning is unclear (e.g., /* checkForErrors = */ false)
Do not use assignment in subexpressions (e.g., if (x = y) ... is forbidden)
Switch on enums should enumerate all values and omit default to catch new values at compile time
Structure switch statements; prohibit fallthrough except between empty cases; each case ends with break or throw; return at end of case not allowed; put break inside braces for compound case
Prefer anonymous namespaces over static for internal linkage of functions
Every defined function must be called at least once (no unused methods)
Files:
cpp/tensorrt_llm/thop/loraOp.cpp
**/*.{h,hpp,hxx,hh,cuh,cpp,cxx,cc,cu}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Parameter names must be consistent between declarations and definitions
Files:
cpp/tensorrt_llm/thop/loraOp.cpp
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Prepend NVIDIA copyright header (current year) to all source files
Files:
cpp/tensorrt_llm/thop/loraOp.cpp
🧠 Learnings (2)
📚 Learning: 2025-08-17T15:07:01.380Z
Learnt from: amitz-nv
PR: NVIDIA/TensorRT-LLM#6968
File: cpp/tensorrt_llm/thop/loraOp.cpp:133-141
Timestamp: 2025-08-17T15:07:01.380Z
Learning: In TensorRT-LLM's LoRA implementation, the LoraImpl::run() method handles setStream() internally in _runGemm(), along with setWorkspace(). Both stream and workspace are passed as arguments to run(), so there's no need to call setStream() explicitly in loraOp.cpp - this avoids redundancy and follows the intended architectural separation.
Applied to files:
cpp/tensorrt_llm/thop/loraOp.cpp
📚 Learning: 2025-08-17T15:07:01.380Z
Learnt from: amitz-nv
PR: NVIDIA/TensorRT-LLM#6968
File: cpp/tensorrt_llm/thop/loraOp.cpp:133-141
Timestamp: 2025-08-17T15:07:01.380Z
Learning: In TensorRT-LLM's LoRA implementation, the LoraImpl::run() method handles setStream() internally in _runGemm() (line 51 in lora.cpp), along with setWorkspace(). The stream parameter flows from loraOp.cpp through LoraImpl::run() to _runGemm() where setStream() is called appropriately. Adding setStream() in loraOp.cpp would be redundant and goes against the intended architectural design.
Applied to files:
cpp/tensorrt_llm/thop/loraOp.cpp
⏰ 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)
cpp/tensorrt_llm/thop/loraOp.cpp (1)
133-140: Stopping per-call cuBLAS handle recreation is the right fixThe thread_local, lazily-initialized CublasMMWrapper cleanly removes handle churn per invocation and should address the perf regression as intended. Lifetime is bound to the thread, and destruction will happen at thread teardown.
|
PR_Github #15613 [ run ] triggered by Bot |
amitz-nv
force-pushed
the
dev-fix-lora-grouped-gemm-cublas-handle-lifetime
branch
from
618f582 to
e301f90
Compare
|
/bot run |
|
PR_Github #15625 [ run ] triggered by Bot |
|
PR_Github #15613 [ run ] completed with state |
|
PR_Github #15625 [ run ] completed with state |
…ad lifetime, avoiding recreation & destruction of cublas handles every call Signed-off-by: Amit Zuker <[email protected]>
amitz-nv
force-pushed
the
dev-fix-lora-grouped-gemm-cublas-handle-lifetime
branch
from
e301f90 to
245bb0c
Compare
|
/bot run |
|
PR_Github #15745 [ run ] triggered by Bot |
|
PR_Github #15745 [ run ] completed with state |
…ed_gemm every call (NVIDIA#6968) Signed-off-by: Amit Zuker <[email protected]>
Description
In
lora_grouped_gemm- changed the lifetime of cublas wrapper that contains the cublas handles to the thread's lifetime. This saves re-creating and destroying the cublas handles on every single call, which degrades performance significantly.Implemented just like it's done in
cublas_gemm_callerfunction incpp/tensorrt_llm/thop/cublasScaledMM.cpp.Test Coverage
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.
Summary by CodeRabbit