devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Remove LGPL Code and references.

Browse Source
This commit is contained in:
Antonio Sánchez
2023-02-08 01:25:06 +00:00
committed by Rasmus Munk Larsen
parent e71f88abce
commit e256ad1823
13 changed files with 6 additions and 904 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
bench/tensors/README
View File

@@ -17,10 +17,10 @@ following commands:
2. bash eigen_sycl_bench.sh
To compile the floating point GPU benchmarks using Intel DPCPP compiler
/path/to/dpcpp/bin/clang+ -DSYCL_COMPILER_IS_DPCPP -DNDEBUG -DEIGEN_MPL2_ONLY -DEIGEN_USE_SYCL=1 -I ../../ -O3 -DNDEBUG -fsycl -fsycl-targets="supported backend in DPCPP. i.e. spir64 or nvptx64-nvidia-cuda" -std=c++17 tensor_benchmarks_sycl.cc benchmark_main.cc -lpthread -o eigen_dpcpp_sycl
/path/to/dpcpp/bin/clang+ -DSYCL_COMPILER_IS_DPCPP -DNDEBUG -DEIGEN_USE_SYCL=1 -I ../../ -O3 -DNDEBUG -fsycl -fsycl-targets="supported backend in DPCPP. i.e. spir64 or nvptx64-nvidia-cuda" -std=c++17 tensor_benchmarks_sycl.cc benchmark_main.cc -lpthread -o eigen_dpcpp_sycl
Last but not least, we also provide a suite of benchmarks to measure the scalability of the contraction code on CPU. To compile these benchmarks, call
g++ contraction_benchmarks_cpu.cc benchmark_main.cc -I ../../ -std=c++11 -O3 -DNDEBUG -pthread -mavx -o benchmarks_cpu
To compile the contraction with DPCPP:
/path/to/dpcpp/bin/clang++ -DSYCL_COMPILER_IS_DPCPP -DNDEBUG -DEIGEN_MPL2_ONLY -DEIGEN_USE_SYCL=1 -I ../../ -O3 -DNDEBUG -fsycl -fsycl-targets="supported backend in DPCPP. i.e. spir64 or nvptx64-nvidia-cuda" -std=c++17 tensor_contract_sycl_bench.cc -lpthread -o eigen_dpcpp_contract
/path/to/dpcpp/bin/clang++ -DSYCL_COMPILER_IS_DPCPP -DNDEBUG -DEIGEN_USE_SYCL=1 -I ../../ -O3 -DNDEBUG -fsycl -fsycl-targets="supported backend in DPCPP. i.e. spir64 or nvptx64-nvidia-cuda" -std=c++17 tensor_contract_sycl_bench.cc -lpthread -o eigen_dpcpp_contract

1
bench/tensors/eigen_sycl_bench.sh
View File

@@ -10,7 +10,6 @@ benchmark_main.cc \
-march=native \
-O3 \
-DNDEBUG \
-DEIGEN_MPL2_ONLY \
-DEIGEN_USE_SYCL=1 \
-DEIGEN_SYCL_LOCAL_MEM=1 \
-no-serial-memop \

2
bench/tensors/eigen_sycl_bench_contract.sh
View File

@@ -1,7 +1,7 @@
rm -f tensor_contract_sycl_bench
: "${COMPUTECPP_PACKAGE_ROOT_DIR:?Need to set COMPUTECPP_PACKAGE_ROOT_DIR}"
echo "COMPUTECPP_PACKAGE_ROOT_DIR is set to: "$COMPUTECPP_PACKAGE_ROOT_DIR
${COMPUTECPP_PACKAGE_ROOT_DIR}/bin/compute++ tensor_contract_sycl_bench.cc -I ../../ -I ${COMPUTECPP_PACKAGE_ROOT_DIR}/include/ -std=c++11 -O3 -DNDEBUG -DEIGEN_MPL2_ONLY -DEIGEN_USE_SYCL=1 -no-serial-memop -mllvm -inline-threshold=10000 -fsycl-ih-last -sycl-driver -Xclang -cl-mad-enable -lOpenCL -lComputeCpp -lpthread -o tensor_contract_sycl_bench ${@:1}
${COMPUTECPP_PACKAGE_ROOT_DIR}/bin/compute++ tensor_contract_sycl_bench.cc -I ../../ -I ${COMPUTECPP_PACKAGE_ROOT_DIR}/include/ -std=c++11 -O3 -DNDEBUG -DEIGEN_USE_SYCL=1 -no-serial-memop -mllvm -inline-threshold=10000 -fsycl-ih-last -sycl-driver -Xclang -cl-mad-enable -lOpenCL -lComputeCpp -lpthread -o tensor_contract_sycl_bench ${@:1}
export LD_LIBRARY_PATH=${COMPUTECPP_PACKAGE_ROOT_DIR}/lib:$LD_LIBRARY_PATH
./tensor_contract_sycl_bench