Is there a problem with the macro HAS_WARP_MERGE_SORT()

C++
1

in native/cuda/SortUtils.cuh
“I printed the CUDA_VERSION macro in my local .cu file, and its value is 13020, which is definitely much smaller than 110600. Shouldn’t the macro HAS_WARP_MERGE_SORT() be written as CUDA_VERSION >= 11060?”

2

Looks like a valid issue. Cross-post from: Is there a problem with the macro HAS_WARP_MERGE_SORT() · Issue #183444 · pytorch/pytorch · GitHub

3

I also opened this issue, ‘Is there a problem with the macro HAS_WARP_MERGE_SORT() · Issue #183444 · pytorch/pytorch · GitHub’, but no one has responded yet.