[Chapel Merge] Gpu radix sort performance testing update

Branch: refs/heads/main
Revision: 9397070969abf88d2776cebcb53d36ffbedfc9a9
Author: ShreyasKhandekar
Link: Gpu radix sort performance testing update by ShreyasKhandekar · Pull Request #23902 · chapel-lang/chapel · GitHub
Log Message:
Gpu radix sort performance testing update (#23902)

The PR does the following:

  • Since #23823 is fixed , we no longer need to use the --devel flag in
    our sort and scan tests.
  • updating the radix sort performance testing by changing the algorithms
    used. Earlier we had three separate scan algorithms all being tracked
    individually to see which performed best in radix sort. The radix sort
    algorithm itself was using atomics and wasn't fully GPU based.
  • Added new graph to track performance of gpuScan
  • Added new graph to retain the historical radix Sort data

[Reviewed by @e-kayrakli ]

Compare: Comparing f045784d181be75a69af7fa47d20f833a93ab306...646837dbb30b013a96284639cb03e14962a00705 · chapel-lang/chapel · GitHub

Diff:
M modules/standard/GPU.chpl
M modules/standard/GpuSort.chpl
M test/GPU-GRAPHFILES
M test/gpu/native/studies/sort/multiDimScanArray.good
M test/gpu/native/studies/sort/nonGpuScanTest.good
M test/gpu/native/studies/sort/radixSortGpu.chpl
M test/gpu/native/studies/sort/radixSortGpu.execopts
M test/gpu/native/studies/sort/radixSortGpu.gpu-execopts
M test/gpu/native/studies/sort/radixSortGpu.graph
A test/gpu/native/studies/sort/radixSortGpuHistorical.graph
M test/gpu/native/studies/sort/scanTest.chpl
M test/gpu/native/studies/sort/scanTest.compopts
A test/gpu/native/studies/sort/scanTest.gpu-execopts
A test/gpu/native/studies/sort/scanTest.gpu-keys
A test/gpu/native/studies/sort/scanTest.graph
D test/gpu/native/studies/sort/scanTest.prediff
D test/gpu/native/studies/sort/scanTestMultiDim.chpl
D test/gpu/native/studies/sort/scanTestMultiDim.good
M test/gpu/native/studies/sort/sortTest.chpl
D test/gpu/native/studies/sort/sortTest.compopts
https://github.com/chapel-lang/chapel/pull/23902.diff