Implement benchmarks for vector operations

[sbaldu]

This PR implements benchmarks for the heterogeneous vector operations using Google Benchmarks.

[mdessole]

Thanks for your PR Simone, it is much appreciated!

I found a few typos regarding your benchmarking code. Moreover, I could not make it compile for any SYCL backend. Did you run tests with OneAPI or AdaptiveCPP? Maybe I am missing something here.

[mdessole]

Also, I would require google benchmark here instead of relying on git modules
find_package(benchmark REQUIRED)

[sbaldu]

Right now I am using it as a submodule because I thought that it would make it easier to use on remote machine where maybe it isn't installed, but if you think it's better I can remove the submodule and just use the find package.

[sbaldu]

Thanks for your PR Simone, it is much appreciated!

I found a few typos regarding your benchmarking code. Moreover, I could not make it compile for any SYCL backend. Did you run tests with OneAPI or AdaptiveCPP? Maybe I am missing something here.

Actually I am having problems with sycl even when compiling the tests already present, so maybe it's a problem with my machine. I'm trying to sort it out, I'll let you know.

[mdessole]

Thanks for your PR Simone, it is much appreciated!
I found a few typos regarding your benchmarking code. Moreover, I could not make it compile for any SYCL backend. Did you run tests with OneAPI or AdaptiveCPP? Maybe I am missing something here.

Actually I am having problems with sycl even when compiling the tests already present, so maybe it's a problem with my machine. I'm trying to sort it out, I'll let you know.

Maybe you'll find this preprint helpful, it contains compiling instruction for SYCL backends (see appendix A.2)

[sbaldu]

Since you've merged #4 I've added benchmarks for the streamed versions of CUDA as well

[sbaldu]

Thanks for your PR Simone, it is much appreciated!
I found a few typos regarding your benchmarking code. Moreover, I could not make it compile for any SYCL backend. Did you run tests with OneAPI or AdaptiveCPP? Maybe I am missing something here.

Actually I am having problems with sycl even when compiling the tests already present, so maybe it's a problem with my machine. I'm trying to sort it out, I'll let you know.

Maybe you'll find this preprint helpful, it contains compiling instruction for SYCL backends (see appendix A.2)

I was compiling in the way described in the paper and in the README but it stil gives error:

/opt/intel/oneapi/compiler/2025.0/bin/compiler/clang-offload-bundler: error: '/home/sbaldu/Documents/Codes/genvectorx/build/CMakeFiles/SYCLMath.dir/src/MathUtil.cxx.o': No such file or directory
icpx: error: clang-offload-bundler command failed with exit code 1 (use -v to see invocation)
gmake[2]: *** [CMakeFiles/SYCLMath.dir/build.make:74: CMakeFiles/SYCLMath.dir/src/MathUtil.cxx.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:117: CMakeFiles/SYCLMath.dir/all] Error 2
gmake: *** [Makefile:136: all] Error 2

On the other hand, I've tested the AdaptiveCPP backend and the benchmarks work correctly.

[mdessole]

Thanks for your PR Simone, it is much appreciated!
I found a few typos regarding your benchmarking code. Moreover, I could not make it compile for any SYCL backend. Did you run tests with OneAPI or AdaptiveCPP? Maybe I am missing something here.

Actually I am having problems with sycl even when compiling the tests already present, so maybe it's a problem with my machine. I'm trying to sort it out, I'll let you know.

Maybe you'll find this preprint helpful, it contains compiling instruction for SYCL backends (see appendix A.2)

I was compiling in the way described in the paper and in the README but it stil gives error:

/opt/intel/oneapi/compiler/2025.0/bin/compiler/clang-offload-bundler: error: '/home/sbaldu/Documents/Codes/genvectorx/build/CMakeFiles/SYCLMath.dir/src/MathUtil.cxx.o': No such file or directory
icpx: error: clang-offload-bundler command failed with exit code 1 (use -v to see invocation)
gmake[2]: *** [CMakeFiles/SYCLMath.dir/build.make:74: CMakeFiles/SYCLMath.dir/src/MathUtil.cxx.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:117: CMakeFiles/SYCLMath.dir/all] Error 2
gmake: *** [Makefile:136: all] Error 2

On the other hand, I've tested the AdaptiveCPP backend and the benchmarks work correctly.

I've also tested successfully with AdaptiveCpp. I think some changes to the cmake file are necessary to make it work with OneAPI, I am trying to figure it out.

[sbaldu]

Thanks for your PR Simone, it is much appreciated!
I found a few typos regarding your benchmarking code. Moreover, I could not make it compile for any SYCL backend. Did you run tests with OneAPI or AdaptiveCPP? Maybe I am missing something here.

Actually I am having problems with sycl even when compiling the tests already present, so maybe it's a problem with my machine. I'm trying to sort it out, I'll let you know.

Maybe you'll find this preprint helpful, it contains compiling instruction for SYCL backends (see appendix A.2)

I was compiling in the way described in the paper and in the README but it stil gives error:

/opt/intel/oneapi/compiler/2025.0/bin/compiler/clang-offload-bundler: error: '/home/sbaldu/Documents/Codes/genvectorx/build/CMakeFiles/SYCLMath.dir/src/MathUtil.cxx.o': No such file or directory
icpx: error: clang-offload-bundler command failed with exit code 1 (use -v to see invocation)
gmake[2]: *** [CMakeFiles/SYCLMath.dir/build.make:74: CMakeFiles/SYCLMath.dir/src/MathUtil.cxx.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:117: CMakeFiles/SYCLMath.dir/all] Error 2
gmake: *** [Makefile:136: all] Error 2

On the other hand, I've tested the AdaptiveCPP backend and the benchmarks work correctly.

I've also tested successfully with AdaptiveCpp. I think some changes to the cmake file are necessary to make it work with OneAPI, I am trying to figure it out.

I will write a small CI workflow that automatically compiles and tests the code, in order to detect this kind of problems.

[sbaldu]

Hello. As a follow-up regarding this PR, I'm currently trying to figure out what is causing the problem with the oneapi backend, but since it's not caused on the changes made here, maybe this can be merged and then we can fix the backend in a separate PR?