NVIDIA / NVIDIA/raft

[BUG] ptxas error : Value of threads per SM for entry is out of the range.

Open
#2,003 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug
Dominant language
Cuda
Stars
1k
Forks
251
Avg merge
1d 8h
Merged PRs (30d)
31

Description

Describe the bug
I am building the C++ shared lib of raft from the source. When I used ./build.sh libraft --compile-lib to build the lib, it returned
ptxas error : Value of threads per SM for entry _ZN4raft9neighbors12experimental10nn_descent6detail17local_join_kernelIiNS3_12InternalID_tIiEEEEvPKT_S9_PK4int2S9_S9_SC_iPK6__halfiPT0_PfiPiSI_ is out of range. .minnctapersm will be ignored.

Steps/Code to reproduce bug
user@myserver:~/raft$ ./build.sh libraft --compile-lib
Building for the architecture of the GPU in the system...
-- Auto detection of gpu-archs: 89
-- CPM: Using local package Thrust@1.17.2.0
-- CPM: Using local package rmm@23.12.0
-- CPM: Adding package NvidiaCutlass@2.10.0 (v2.10.0)
-- CMake Version: 3.28.0-rc5
-- CUDART: /usr/local/cuda-11.8/lib64/libcudart.so
-- CUDA Driver: /usr/local/cuda-11.8/lib64/stubs/libcuda.so
-- NVRTC: /usr/local/cuda-11.8/lib64/libnvrtc.so
-- Default Install Location: /home/bld/raft_build
-- CUDA Compilation Architectures: 53;60;61;70;72;75;80;86
-- Enable caching of reference results in conv unit tests
-- Enable rigorous conv problem sizes in conv unit tests
-- Using NVCC flags: -DCUTLASS_TEST_LEVEL=0;-DCUTLASS_TEST_ENABLE_CACHED_RESULTS=1;-DCUTLASS_CONV_UNIT_TEST_RIGOROUS_SIZE_ENABLED=1;-DCUTLASS_DEBUG_TRACE_LEVEL=0;$<$BOOL:1:-Xcompiler=-Wconversion>;$<$BOOL:1:-Xcompiler=-fno-strict-aliasing>
-- CUTLASS Revision: 31fcbf17
-- Configuring cublas ...
-- cuBLAS Disabled.
-- Configuring cuBLAS ... done.
-- CPM: Using local package cuco@0.0.1
-- _RAPIDS_POLICY_CALLERS_VERSION: 23.12
-- _RAPIDS_POLICY_REMOVED_IN: 24.02
CMake Deprecation Warning at build/_deps/rapids-cmake-src/rapids-cmake/cmake/detail/policy.cmake:57 (message):
rapids-cmake policy [deprecated=23.12 removed=24.02]: Usage of
rapids_export_find_package_file without an explicit EXPORT_SET key has
been deprecated.
Call Stack (most recent call first):
build/_deps/rapids-cmake-src/rapids-cmake/export/find_package_file.cmake:73 (rapids_cmake_policy)
build/_deps/rapids-cmake-src/rapids-cmake/find/generate_module.cmake:216 (rapids_export_find_package_file)
CMakeLists.txt:535 (rapids_find_generate_module)

-- _RAPIDS_POLICY_CALLERS_VERSION: 23.12
-- _RAPIDS_POLICY_REMOVED_IN: 24.02
CMake Deprecation Warning at build/_deps/rapids-cmake-src/rapids-cmake/cmake/detail/policy.cmake:57 (message):
rapids-cmake policy [deprecated=23.12 removed=24.02]: Usage of
rapids_export_find_package_file without an explicit EXPORT_SET key has
been deprecated.
Call Stack (most recent call first):
build/_deps/rapids-cmake-src/rapids-cmake/export/find_package_file.cmake:73 (rapids_cmake_policy)
build/_deps/rapids-cmake-src/rapids-cmake/find/generate_module.cmake:223 (rapids_export_find_package_file)
CMakeLists.txt:535 (rapids_find_generate_module)

-- Configuring done (1.0s)
CMake Warning (dev) at CMakeLists.txt:254 (target_link_libraries):
The library that is being linked to, CUDA::nvToolsExt, is marked as being
deprecated by the owner. The message provided by the developer is:

nvToolsExt has been superseded by nvtx3 since CUDA 10.0 and CMake 3.25.
Use CUDA::nvtx3 and include <nvtx3/nvToolsExt.h> instead.

This warning is for project developers. Use -Wno-dev to suppress it.

CMake Warning (dev) at CMakeLists.txt:254 (target_link_libraries):
The library that is being linked to, CUDA::nvToolsExt, is marked as being
deprecated by the owner. The message provided by the developer is:

nvToolsExt has been superseded by nvtx3 since CUDA 10.0 and CMake 3.25.
Use CUDA::nvtx3 and include <nvtx3/nvToolsExt.h> instead.

This warning is for project developers. Use -Wno-dev to suppress it.

-- Generating done (0.0s)
-- Build files have been written to: /home/bld/raft/cpp/build
-- Compiling targets: ;raft_lib, verbose=
[1/4] Building CUDA object CMakeFiles/raft_objs.dir/src/raft_runtime/neighbors/cagra_build.cu.o
FAILED: CMakeFiles/raft_objs.dir/src/raft_runtime/neighbors/cagra_build.cu.o
/usr/local/cuda-11.8/bin/nvcc -forward-unknown-to-host-compiler -DCUTLASS_NAMESPACE=raft_cutlass -DFMT_HEADER_ONLY=1 -DNVTX_ENABLED -DRAFT_COMPILED -DRAFT_EXPLICIT_INSTANTIATE_ONLY -DRAFT_SYSTEM_LITTLE_ENDIAN=1 -DSPDLOG_FMT_EXTERNAL -DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_CUDA -DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -I/home/bld/raft/cpp/include -I/home/bld/raft/cpp/build/_deps/thrust-src -I/home/bld/raft/cpp/build/_deps/thrust-src/dependencies/cub -I/home/bld/anaconda3/include/rapids/libcudacxx -I/home/bld/raft/cpp/build/_deps/nvidiacutlass-src/include -I/home/bld/raft/cpp/build/_deps/nvidiacutlass-build/include -I/usr/local/cuda-11.8/include -isystem /home/bld/anaconda3/include -isystem /usr/local/cuda-11.8/targets/x86_64-linux/include -O3 -DNDEBUG -std=c++17 "--generate-code=arch=compute_89,code=[sm_89]" -Xcompiler=-fPIC -Xcompiler=-Wno-deprecated-declarations -Xcompiler=-Wall,-Werror,-Wno-error=deprecated-declarations -Werror=all-warnings --expt-extended-lambda --expt-relaxed-constexpr -DCUDA_API_PER_THREAD_DEFAULT_STREAM -Xfatbin=-compress-all -Xcompiler=-fopenmp -MD -MT CMakeFiles/raft_objs.dir/src/raft_runtime/neighbors/cagra_build.cu.o -MF CMakeFiles/raft_objs.dir/src/raft_runtime/neighbors/cagra_build.cu.o.d -x cu -c /home/bld/raft/cpp/src/raft_runtime/neighbors/cagra_build.cu -o CMakeFiles/raft_objs.dir/src/raft_runtime/neighbors/cagra_build.cu.o
ptxas error : Value of threads per SM for entry ZN4raft9neighbors12experimental10nn_descent6detail17local_join_kernelIiNS3_12InternalID_tIiEEEEvPKT_S9_PK4int2S9_S9_SC_iPK6__halfiPT0_PfiPiSI is out of range. .minnctapersm will be ignored
ptxas fatal : Ptx assembly aborted due to errors
ninja: build stopped: subcommand failed.

Expected behavior
Finish compiling correctly.

Environment details (please complete the following information):

  • Environment location: Bare-metal
  • Method of RAFT install: source
  • System: Ubuntu 22.04
  • GCC: 11.4.0
  • Cuda: 11.8
  • Nvidia driver: 535.86.05
  • GPU: RTX4090

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the failure with ./build.sh libraft --compile-lib in the reported Ubuntu 22.04, CUDA 11.8, RTX 4090 environment. Then inspect src/raft_runtime/neighbors/cagra_build.cu and the local_join_kernel compilation path; done means the raft shared library builds successfully without the ptxas error.

Written by the indexing model from the issue text.

Assessment

Tech stack
cmake, cpp
Domain
build-system
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.