[bug] how to solve compile sycl-toolchain code occur error: '/tmp/libsycl-fallback-imf-d878ec.cubin': Permission denied #7777
Comments
|
thanks @wangzy0327 for reporting this issue. We will look into this and get back to you shortly. |
|
I've tried recreating this and I've been unsuccessful. mkdir 7777
cd 7777
git clone https://github.com/intel/llvm llvm-project
cat << 'EOD' > Vagrantfile
Vagrant.configure("2") do |config|
config.vm.box = "ubuntu/bionic64"
config.vm.synced_folder "llvm-project", "/vagrant/llvm-project"
config.vm.provider "virtualbox" do |vb|
vb.memory = "16384"
vb.cpus = "8"
end
config.vm.provision "shell", inline: <<-SHELL
apt-get update
apt-get install -qy wget gnupg2 software-properties-common
wget -q -O - https://apt.kitware.com/keys/kitware-archive-latest.asc | apt-key add - && apt-add-repository 'deb https://apt.kitware.com/ubuntu/ bionic main'
apt-get update -qy && apt-get install -qy \
wget \
git \
libncurses5-dev \
libncursesw5-dev \
vim \
build-essential \
ninja-build \
cmake \
python3-distutils \
linux-headers-`uname -r`
wget https://developer.download.nvidia.com/compute/cuda/11.2.0/local_installers/cuda_11.2.0_460.27.04_linux.run -q -O cuda.run
sh cuda.run --silent --toolkit && rm cuda.run
SHELL
end
EOD
cat << 'EOD' > example.cpp
#include <sycl/sycl.hpp>
#include <iostream>
#include <vector>
using namespace sycl;
int main(){
std::cout<<"hello sycl"<<std::endl;
const int N = 4;
const int MAGIC_NUM = 42;
std::vector<int> vec_num = {3,4,5,6,7,8,9};
std::cout<<vec_num[0]<<std::endl;
queue q;
auto dev = q.get_device();
std::cout<<"default queue select device info : "<<&dev<<std::endl;
std::cout<<"default queue select device vendor id : "<<dev.get_info<sycl::info::device::vendor_id>()<<std::endl;
std::cout<<"default queue select device vendor name : "<<dev.get_info<sycl::info::device::name>()<<std::endl;
auto devices = dev.get_devices();
for (auto &device : devices) {
std::cout << " queue get device: " << device.get_info<sycl::info::device::name>() << std::endl;
std::cout << " queue get device vendor id: " << device.get_info<sycl::info::device::vendor_id>()<< std::endl;
std::cout << " queue get device object address : " << &device << std::endl;
}
auto ctxt = q.get_context();
auto platforms = sycl::platform::get_platforms();
for (auto &platform : platforms) {
std::cout << "Platform name : " << platform.get_info<sycl::info::platform::name>()<< std::endl;
std::cout << "Platform profile : " << platform.get_info<sycl::info::platform::profile>()<<std::endl;
std::cout << "Platform version : " << platform.get_info<sycl::info::platform::version>()<<std::endl;
std::cout << "Platform vendor : " << platform.get_info<sycl::info::platform::vendor>()<<std::endl;
std::cout << "Platform object address : " << &platform<< std::endl;
auto devices = platform.get_devices();
for (auto &device : devices) {
std::cout << " Device: " << device.get_info<sycl::info::device::name>() << std::endl;
std::cout << " Device vendor id: " << device.get_info<sycl::info::device::vendor_id>()<< std::endl;
std::cout << " Device object address : " << &device << std::endl;
}
}
std::vector<kernel_id> builtinKernelIds =
dev.get_info<info::device::built_in_kernel_ids>();
std::cout<<"built in kernel size : "<<builtinKernelIds.size()<<std::endl;
auto myBundle = get_kernel_bundle<bundle_state::executable>(ctxt);
return 0;
}
EOD
cat <<'EOD' > CMakeLists.txt
cmake_minimum_required(VERSION 2.8.12)
set(DPCPP_HOME "/vagrant/llvm-project/")
set(CMAKE_C_COMPILER "${DPCPP_HOME}/build/bin/clang")
set(CMAKE_CXX_COMPILER "${DPCPP_HOME}/build/bin/clang++")
set(CMAKE_CXX_STANDARD 17)
project(hello-sycl)
set(DPCPP_SYCL_HOME "${DPCPP_HOME}/build")
include_directories("${DPCPP_SYCL_HOME}/include/sycl")
include_directories("${DPCPP_SYCL_HOME}/include")
message(STATUS "dpcpp_home : ${DPCPP_HOME}")
message(STATUS "dpcpp_cuda_sycl_home : ${DPCPP_SYCL_HOME}")
message(STATUS "find library path : ${DPCPP_SYCL_HOME}/lib")
set(CMAKE_BUILD_RPATH "${DPCPP_SYCL_HOME}/lib;${CMAKE_BUILD_RPATH}")
message(STATUS "cmake build rpath : ${CMAKE_BUILD_RPATH}")
set(CMAKE_BUILD_TYPE "Debug")
set(CMAKE_CXX_FLAGS "-fsycl -fsycl-targets=nvptx64-nvidia-cuda")
set(CMAKE_CXX_FLAGS_DEBUG "$ENV{CXXFLAGS} -O0 -Wall -g -ggdb -std=c++17 -Wno-sycl-target -Wno-linker-warnings")
# set(CMAKE_CXX_FLAGS_RELEASE "$ENV{CXXFLAGS} -O3 -Wall -std=c++17")
link_directories("${DPCPP_SYCL_HOME}/lib")
aux_source_directory(. DIR_SRCS)
add_executable(hello-sycl ${DIR_SRCS})
target_include_directories(hello-sycl PRIVATE "${DPCPP_SYCL_HOME}/include/sycl")
target_include_directories(hello-sycl PRIVATE "${DPCPP_SYCL_HOME}/include")
target_link_libraries(hello-sycl PRIVATE sycl)
EOD
cat << 'EOD' > recreate
#!/bin/bash
cd /vagrant/llvm-project
mkdir -p /vagrant/llvm-project/build
cd /vagrant/llvm-project/build
python3 ../buildbot/configure.py --cuda --cmake-opt=-DLLVM_USE_LINKER=gold --cmake-opt=-DLLVM_PARALLEL_LINK_JOBS=2 --cmake-opt=-DCUDA_CUDA_LIBRARY=/usr/local/cuda-11.2/targets/x86_64-linux/lib/stubs/libcuda.so
ninja check-sycl-deploy -k 0 # ignore some python command problems in the lit tests on 18.04
cd /vagrant
for _ in {1..100} ; do
rm -rf build
mkdir build
cmake -B build
if ! cmake --build build ; then echo FAILED ; break ; fi
done
EOD
vagrant up
vagrant ssh -c 'bash /vagrant/recreate'I had no failures to link the example with the 100 tries looped in the last fragment listed. |
|
@ldrumm I also met the problem in my scenary as following display. It quite exists the problem but is difficult to reproduce. |
|
Could you please paste plain text rather than screenshots? It's hard to extract information from a screenshot |
|
@ldrumm @Pennycook Can you try to reproduce the problem in my same environment? OS System environment OS kernel version cuda version Sycl version: 2022-12-release reproduce example: CMakeLists.txtcomplex.cctest.shI get the result which display |
|
The only difference is my kernel version is
I reran using the binary release you posted (rather than building from source as before) and still can't recreate the bug However, that's still not a permissions issue. I don't think there's anything I can do to help further, but ask the following, in order of simplicity
Without being able to reproduce locally, I'm unable to help. I suggest you follow the above steps to determine where the issue lies. |
There is enough amount of space avaliable in |
|
After I made two changes, the '/tmp/libsycl-fallback-imf-d878ec.cubin':Permission denied was tested many times and never appeared.
I am not sure which change lead to the result. Can you compare your previous environment with it ? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Describe the bug
sometimes occur the following error. how to solve it ?
clang-16: warning: linked binaries do not contain expected 'nvptx64-nvidia-cuda' target; found targets: 'nvptx64-nvidia-cuda-sm_50' [-Wsycl-target]
/home/wzy/sycl_workspace/build-cuda-2022-09-debug/bin/clang-offload-bundler: error: '/tmp/libsycl-fallback-imf-d878ec.cubin': Permission denied
clang-16: error: clang-offload-bundler command failed with exit code 1 (use -v to see invocation)
CMakeFiles/hello-sycl.dir/build.make:102: recipe for target 'hello-sycl' failed
gmake[2]: *** [hello-sycl] Error 1
CMakeFiles/Makefile2:94: recipe for target 'CMakeFiles/hello-sycl.dir/all' failed
gmake[1]: *** [CMakeFiles/hello-sycl.dir/all] Error 2
To Reproduce
any sycl-example code occurs the error somtimes.
There provides a sycl example that occurs the following error when the code compile(cmake -B build && cmake --build build).
CMakeLists.txt
cmake_minimum_required(VERSION 2.8.12)
set(CMAKE_C_COMPILER "/home/wzy/sycl_workspace/build-cuda-2022-09-debug/bin/clang")
set(CMAKE_CXX_COMPILER "/home/wzy/sycl_workspace/build-cuda-2022-09-debug/bin/clang++")
set(CMAKE_CXX_STANDARD 17)
project(hello-sycl)
set(DPCPP_HOME "/home/wzy/sycl_workspace")
set(DPCPP_SYCL_HOME "${DPCPP_HOME}/build-cuda-2022-09-debug")
include_directories("${DPCPP_SYCL_HOME}/include/sycl")
include_directories("${DPCPP_SYCL_HOME}/include")
message(STATUS "dpcpp_home : ${DPCPP_HOME}")
message(STATUS "dpcpp_cuda_sycl_home : ${DPCPP_SYCL_HOME}")
message(STATUS "find library path : ${DPCPP_SYCL_HOME}/lib")
set(CMAKE_BUILD_RPATH "${DPCPP_SYCL_HOME}/lib;${CMAKE_BUILD_RPATH}")
message(STATUS "cmake build rpath : ${CMAKE_BUILD_RPATH}")
set(CMAKE_BUILD_TYPE "Debug")
set(CMAKE_CXX_FLAGS "-fsycl -fsycl-targets=nvptx64-nvidia-cuda")
set(CMAKE_CXX_FLAGS_DEBUG "$ENV{CXXFLAGS} -O0 -Wall -g -ggdb -std=c++17 -Wno-sycl-target -Wno-linker-warnings")
set(CMAKE_CXX_FLAGS_RELEASE "$ENV{CXXFLAGS} -O3 -Wall -std=c++17")
link_directories("${DPCPP_SYCL_HOME}/lib")
aux_source_directory(. DIR_SRCS)
add_executable(hello-sycl ${DIR_SRCS})
target_include_directories(hello-sycl PRIVATE "${DPCPP_SYCL_HOME}/include/sycl")
target_include_directories(hello-sycl PRIVATE "${DPCPP_SYCL_HOME}/include")
target_link_libraries(hello-sycl PRIVATE sycl)
my-example
Environment (please complete the following information):
Additional context
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered: