| commit | c8096cb80eab8801a6a669c69a15f1f6b3a6c167 | |
| author | Szilárd Páll <pall.szilard@gmail.com> | |
| Mon, 30 Jul 2018 16:18:37 +0000 (30 18:18 +0200) | ||
| committer | Szilárd Páll <pall.szilard@gmail.com> | |
| Tue, 21 Aug 2018 14:54:42 +0000 (21 16:54 +0200) | ||
| tree | 59775bd949549a58e6e9ed2e67381bf17e425cac | treesnapshot (tar.gz zip) |
| parent | cd41b0fefbc65a682c6f05ebeaa7c139d45206d7 | commitdiff |
Request flushing denorms to zero in OpenCL
This change adds by default the -cl-denorms-are-zero to the flags used
for kernel compilation. This is done to:
- avoid a large performance penalty on AMD Vega with ROCm (which by
default handles denorms on GFX9 or later).
- make the defaults uniform across CUDA and OpenCL.
Fixes #2593
Change-Id: I9e6183c4367b5960e0e21f1dd342d7695acfbc44
This change adds by default the -cl-denorms-are-zero to the flags used
for kernel compilation. This is done to:
- avoid a large performance penalty on AMD Vega with ROCm (which by
default handles denorms on GFX9 or later).
- make the defaults uniform across CUDA and OpenCL.
Fixes #2593
Change-Id: I9e6183c4367b5960e0e21f1dd342d7695acfbc44
| docs/release-notes/2018/2018.3.rst | diffblobblamehistory | |
| src/gromacs/gpu_utils/ocl_compiler.cpp | diffblobblamehistory |