### Install CUDA 
#
# sudo sh cuda_*_linux.run
export CUDA_HOME=/usr/local/cuda
export PATH=$PATH:$CUDA_HOME/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_HOME/lib64
###init
GPU_COUNT=$(nvidia-smi --list-gpus | wc -l)
#echo $GPU_COUNT
dir=$(pwd)

### Install cuda-samples-12.8
#
# unzip cuda-samples-12.8.zip
# cd cuda-samples-12.8
# mkdir build && cd build
# cmake ..
# make -j$(nproc)
#
### Install OpenCL-Benchmark
# unzip OpenCL-Benchmark-master.zip
# 
### 设置软链接
cd $dir
#ln -s cuda-samples-12.8/build/Samples/ $dir/Samples
ln -s OpenCL-Benchmark-master OpenCL-Benchmark
ln -s nccl-tests-2.16.4/build nccl

date | tee $dir/gpu.log
hostnamectl | tee -a $dir/gpu.log
#fastfetch  | tee -a $dir/gpu.log
### 硬件信息查询
#lspci
echo -e "###lspci -d 10de:" | tee -a $dir/gpu.log
lspci -d 10de: | tee -a $dir/gpu.log
sleep 5
#
#nvidia-smi
echo -e "\n###nvidia-smi\n" | tee -a $dir/gpu.log
nvidia-smi | tee -a $dir/gpu.log
sleep 5
#
#topo
echo -e "\n###nvidia-smi topo -m\n" | tee -a $dir/gpu.log
nvidia-smi topo -m | tee -a $dir/gpu.log
sleep 5
#
#query
echo -e "\n###nvidia-smi -q -u\n" | tee -a $dir/gpu.log
nvidia-smi -q -u | tee -a $dir/gpu.log
sleep 5
#
#query
echo -e "\n###nvidia-smi -q -i 0\n" | tee -a $dir/gpu.log
nvidia-smi -q -i 0 | tee -a $dir/gpu.log
sleep 5
#
#Configured
echo -e "\n###nvidia-smi --format=csv --query-gpu=gpu_name,serial,gpu_bus_id,vbios_version,memory.total,pcie.link.gen.max,pcie.link.gen.current,pcie.link.width.max,pcie.link.width.current\n"  | tee -a $dir/gpu.log
nvidia-smi --format=csv --query-gpu=gpu_name,serial,gpu_bus_id,vbios_version,memory.total,pcie.link.gen.max,pcie.link.gen.current,pcie.link.width.max,pcie.link.width.current | tee -a $dir/gpu.log
sleep 5
#
### GPU测试
#nccl
cd $dir/nccl
NCCL_NTHREADS=$(nproc) NCCL_MIN_NCHANNELS=$GPU_COUNT ./all_reduce_perf -b 1m -e 1g -f 2 -g $GPU_COUNT  | tee -a $dir/gpu.log
sleep 10
#
#p2pBandwidthLatencyTest
echo -e "\n###p2pBandwidthLatencyTest\n" | tee -a $dir/gpu.log
cd $dir/Samples/5_Domain_Specific/p2pBandwidthLatencyTest
./p2pBandwidthLatencyTest | tee -a $dir/gpu.log
sleep 10
#
#bandwidthTest
echo -e "###bandwidthtest" | tee -a $dir/gpu.log
for (( i=0; i<GPU_COUNT; i++ )); do
    {
    cd $dir/Samples/1_Utilities/bandwidthTest
    ./bandwidthTest --device=$i >>  $dir/gpu.log 
    }
done
./bandwidthTest --device=all >>  $dir/gpu.log
cat $dir/gpu.log | grep 3200000 
sleep 10
#
#matrixMul
echo -e "\n###matrixMul\n" | tee -a $dir/gpu.log
cd $dir/Samples/0_Introduction/matrixMul
sh matrixMul-4-small.sh | tee -a $dir/gpu.log
sleep 10
#
#alignedTypes
echo -e "\n###alignedTypes\n" | tee -a $dir/gpu.log
cd $dir/Samples/6_Performance
./alignedTypes/alignedTypes | tee -a $dir/gpu.log
sleep 5
#
#LargeKernelParameter
echo -e "\n###LargeKernelParameter\n" | tee -a $dir/gpu.log
cd $dir/Samples/6_Performance
./LargeKernelParameter/LargeKernelParameter | tee -a $dir/gpu.log
sleep 5
#
#transpose
echo -e "\n###transpose\n" | tee -a $dir/gpu.log
cd $dir/Samples/6_Performance
./transpose/transpose | tee -a $dir/gpu.log
sleep 5
#
#UnifiedMemoryPerf
echo -e "\n###UnifiedMemoryPerf\n" | tee -a $dir/gpu.log
cd $dir/Samples/6_Performance
./UnifiedMemoryPerf/UnifiedMemoryPerf | tee -a $dir/gpu.log
sleep 5
#
#cudaGraphsPerfScaling
echo -e "\n###cudaGraphsPerfScaling\n" | tee -a $dir/gpu.log
cd $dir/Samples/6_Performance
./cudaGraphsPerfScaling/cudaGraphsPerfScaling | tee -a $dir/gpu.log
sleep 5
#
#OpenCL-Benchmark
echo -e "\n###OpenCL-Benchmark\n" | tee -a $dir/gpu.log
cd $dir/OpenCL-Benchmark
./make.sh | tee -a $dir/gpu.log
sleep 5

#lspci tv
lspci -tv >> $dir/gpu.log
#lspci vvv
lspci -d 10de: -vvv >> $dir/gpu.log