intel / llvm

Intel staging area for llvm.org contribution. Home for Intel LLVM-based projects.

Other

1.21k stars 723 forks source link

NVIDIA CUDA GPU programming — processing 8 BILLION hashes/sec on V100s #6943

Open Rimatara opened 1 year ago

Rimatara commented 1 year ago

How to modify the program to run it on Windows 10 with Visual Studio 2017 and installed NVIDIA CUDA support and OneAPI?

See the program running on Linux by using NVCC compiler on Youtube link here (Github link to source files in the description below).

I would like to try to run it on my ASUS UX510UWK notebook (i7-7500U, 24GB RAM, Samsung SSD 980 PRO 1TB) with NVIDIA GeForce GTX 960M.

zjin-lcf commented 1 year ago

Do you imply that the SYCL program couldn't reach 8 billion hashes per second on the GPU ?

Rimatara commented 1 year ago

No. I am running the example above in Linux with NVIDIA CUDA support on my notebook already although I am not sure how to modify these parameters to suit my GTX 960M graphic card.

define THREADS 1500

define BLOCKS 256

define GPUS 4

zjin-lcf commented 1 year ago

The question may be posted at https://community.intel.com/ for assistance