ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 13:11:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

topk_softmax (#1592)

Browse Source 
* topk_softmax

* remove some file

* fix atomix linear_offset

* address various comment, and change sfc get_index api to static(tuple)
This commit is contained in:
carlushuang
2024-10-26 23:52:49 +08:00
committed by GitHub
parent 31bf253aeb
commit b098b71b05
41 changed files with 5603 additions and 226 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
include/ck_tile/core/config.hpp
View File

@@ -41,6 +41,19 @@
#define CK_TILE_HOST_DEVICE_EXTERN
#endif
// implementing the "memory address space" attribute
// https://llvm.org/docs/AMDGPUUsage.html#amdgpu-address-spaces-table
#ifdef __HIPCC_
#define CK_TILE_GENERIC_ADDR __attribute__((address_space(0)))
#define CK_TILE_GLOBAL_ADDR __attribute__((address_space(1)))
#define CK_TILE_LDS_ADDR __attribute__((address_space(3)))
#define CK_TILE_BUF_RES_ADDR __attribute__((address_space(8)))
#else
#define CK_TILE_GENERIC_ADDR
#define CK_TILE_GLOBAL_ADDR
#define CK_TILE_LDS_ADDR
#define CK_TILE_BUF_RES_ADDR
#endif
#ifndef CK_TILE_USE_CUSTOM_DATA_TYPE
#define CK_TILE_USE_CUSTOM_DATA_TYPE 0 // custom data type will generate extra move/bfi code
#endif
@@ -205,3 +218,8 @@
#ifndef CK_TILE_BUFFER_LOAD_RAW_BF16_WA
#define CK_TILE_BUFFER_LOAD_RAW_BF16_WA 1
#endif
// workaround: compiler not emiting reciprocal instruction frm __frcp_rn()
#ifndef CK_TILE_WORKAROUND_SWDEV_383542
#define CK_TILE_WORKAROUND_SWDEV_383542 1
#endif