[CK_TILE] Fix flatmm on gfx11 and gfx12 (#2790)

1. Correct shuffle_b and MakeBFlatDramTileDistribution according to WMMA warp layout
2. Add FlatmmConfig16_Wmma for gfx11 and gfx12
This commit is contained in:
linqunAMD
2025-09-10 08:28:00 +08:00
committed by GitHub
parent 82890192dd
commit df4ee556d6
14 changed files with 224 additions and 67 deletions

View File

@@ -268,6 +268,9 @@ int main(int argc, char* argv[])
try
{
#if defined(CK_TILE_USE_WMMA)
return !run_flatmm_example<FlatmmConfig16_Wmma>(argc, argv);
#else
int warp_tile = arg_parser.get_int("warp_tile");
if(warp_tile == 0)
{
@@ -285,6 +288,7 @@ int main(int argc, char* argv[])
{
return !run_flatmm_example<FlatmmConfig32_950>(argc, argv);
}
#endif
}
catch(const std::runtime_error& e)
{