diff --git a/example/ck_tile/01_fmha/CMakeLists.txt b/example/ck_tile/01_fmha/CMakeLists.txt index b8ca26193d..b76f08614e 100644 --- a/example/ck_tile/01_fmha/CMakeLists.txt +++ b/example/ck_tile/01_fmha/CMakeLists.txt @@ -40,7 +40,7 @@ string(REPLACE ";" "," FMHA_FWD_APIS "${FMHA_FWD_ENABLE_APIS}") set(FMHA_FWD_CODE_GEN_COMMON_ARGS ${CMAKE_CURRENT_LIST_DIR}/generate.py --api ${FMHA_FWD_APIS} - --optdim 32,64,128,256 + --optdim 32,64,128,192,256 # --filter fmha_fwd... ) set(FMHA_BWD_CODE_GEN_COMMON_ARGS diff --git a/example/ck_tile/01_fmha/codegen/ops/fmha_fwd.py b/example/ck_tile/01_fmha/codegen/ops/fmha_fwd.py index f898d5f7b2..6c8e47a336 100644 --- a/example/ck_tile/01_fmha/codegen/ops/fmha_fwd.py +++ b/example/ck_tile/01_fmha/codegen/ops/fmha_fwd.py @@ -548,7 +548,7 @@ class KernelComponentFactory: (192, 192) : [FmhaFwdTileSize( 64, 64, 32, 192, 32, 192, 4, 1, 1, 4, 1, 1, 16, 16, 16, 16, 16, 16, -1)], (256, 256) : [FmhaFwdTileSize( 64, 64, 32, 256, 32, 256, 4, 1, 1, 4, 1, 1, 16, 16, 16, 16, 16, 16, -1)], } - elif dtype == 'fp16' or dtype == 'bf16': + elif dtype == 'fp16': return { (32, 32) : [FmhaFwdTileSize(128, 64, 16, 32, 32, 32, 4, 1, 1, 4, 1, 1, 32, 32, 16, 32, 32, 16, -1)], (64, 64) : [FmhaFwdTileSize(16, 32, 64, 64, 32, 64, 1, 1, 1, 1, 1, 1, 16, 16, 32, 16, 16, 32, -1), @@ -560,7 +560,7 @@ class KernelComponentFactory: FmhaFwdTileSize(128, 64, 32, 128, 16, 128, 4, 1, 1, 4, 1, 1, 32, 32, 16, 32, 32, 16, -1), FmhaFwdTileSize(128, 128, 32, 128, 32, 128, 4, 1, 1, 4, 1, 1, 32, 32, 16, 32, 32, 16, -1)], # (160,160) : [FmhaFwdTileSize(128, 128, 32, 160, 32, 160, 4, 1, 1, 4, 1, 1, 32, 32, 16, 32, 32, 16, 1)], - (192,128) : [FmhaFwdTileSize(128, 128, 32, 128, 32, 192, 4, 1, 1, 4, 1, 1, 32, 32, 16, 32, 32, 16, -1)], + (192,128) : [FmhaFwdTileSize(128, 128, 32, 128, 32, 192, 4, 1, 1, 4, 1, 1, 32, 32, 16, 32, 32, 16, 1)], (192,192) : [FmhaFwdTileSize(128, 128, 32, 192, 32, 192, 4, 1, 1, 4, 1, 1, 32, 32, 16, 32, 32, 16, 1)], (256,256) : [FmhaFwdTileSize(128, 128, 32, 256, 32, 256, 4, 1, 1, 4, 1, 1, 32, 32, 16, 32, 32, 16, -1)], }