ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-11 17:00:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

Syncing up from internal repo to enable MI300. (#690)

Browse Source 
* enable gfx940

* switch between intrinsic mfma routines on mi100/200 and mi300

* fix mfma_int8 on MI300

* disable 2 int8 examples on MI300

* Update cmake-ck-dev.sh

* restore gitignore file

* modify Jenkinsfile to the internal repo

---------

Co-authored-by: Jing Zhang <jizha@amd.com>
Co-authored-by: zjing14 <zhangjing14@gmail.com>
This commit is contained in:
Illia Silin
2023-04-28 16:22:59 -07:00
committed by GitHub
parent 54c90aae13
commit 4feebedd41
47 changed files with 228 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
include/ck/ck.hpp
View File

@@ -31,7 +31,7 @@
#ifndef __HIP_DEVICE_COMPILE__ // for host code
#define CK_BUFFER_RESOURCE_3RD_DWORD -1
#elif defined(__gfx803__) || defined(__gfx900__) || defined(__gfx906__) || defined(__gfx908__) || \
defined(__gfx90a__) // for GPU code
defined(__gfx90a__) || defined(__gfx940__) // for GPU code
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x00020000
#elif defined(__gfx1030__) // for GPU code
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x31014000
@@ -43,8 +43,8 @@
#ifndef __HIP_DEVICE_COMPILE__ // for host code, define nothing
#elif defined(__gfx803__) || defined(__gfx900__) // for GPU code
#define CK_USE_AMD_V_MAC_F32
#elif defined(__gfx906__) || defined(__gfx908__) || defined(__gfx90a__) || \
defined(__gfx1030__) // for GPU code
#elif defined(__gfx906__) || defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx1030__) || \
defined(__gfx940__) // for GPU code
#define CK_USE_AMD_V_FMAC_F32
#define CK_USE_AMD_V_DOT2_F32_F16
#define CK_USE_AMD_V_DOT4_I32_I8
@@ -53,14 +53,18 @@
// MFMA instruction
#ifndef __HIP_DEVICE_COMPILE__ // for host code
#define CK_USE_AMD_MFMA
#elif defined(__gfx908__) || defined(__gfx90a__) // for GPU code
#elif defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx940__) // for GPU code
#define CK_USE_AMD_MFMA
#endif
#if defined(__gfx90a__)
#if(defined(__gfx90a__) || defined(__gfx940__))
#define CK_USE_AMD_MFMA_BF16_1K_OP
#endif
#if defined(__gfx940__)
#define CK_USE_AMD_MFMA_GFX940
#endif
// WMMA instruction
#ifndef __HIP_DEVICE_COMPILE__ // for host code
#define CK_USE_AMD_WMMA
@@ -80,13 +84,13 @@
// buffer atomic add: floating point
#ifndef __HIP_DEVICE_COMPILE__ // for host code
#define CK_USE_AMD_BUFFER_ATOMIC_ADD_FLOAT 1
#elif defined(__gfx908__) || defined(__gfx90a__) // for GPU code
#elif defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx940__) // for GPU code
#define CK_USE_AMD_BUFFER_ATOMIC_ADD_FLOAT 1
#else // for GPU code
#define CK_USE_AMD_BUFFER_ATOMIC_ADD_FLOAT 0
#endif
#if defined(__gfx90a__) // for GPU code
#if(defined(__gfx90a__) || defined(__gfx940__)) // for GPU code
#define CK_USE_AMD_BUFFER_ATOMIC_MAX_FLOAT64 1
#else
#define CK_USE_AMD_BUFFER_ATOMIC_MAX_FLOAT64 0