ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-20 14:59:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add support for more Navi2x and Navi3x models. (#1152)

Browse Source 
* add support for navi2x and navi3x models

* fix syntax

* use common macro for different mi300 architectures
This commit is contained in:
Illia Silin
2024-02-02 11:35:26 -08:00
committed by GitHub
parent 171ca260b5
commit 180f16f9ac
69 changed files with 194 additions and 194 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
include/ck/ck.hpp
View File

@@ -44,16 +44,30 @@
#define CK_USE_WAVES_PER_EU 0
#endif
// define general macros for various architectures
#if defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__)
#define __gfx94__
#endif
#if defined(__gfx1010__) || defined(__gfx1011__) || defined(__gfx1012__)
#define __gfx101__
#endif
#if defined(__gfx1030__) || defined(__gfx1031__) || defined(__gfx1032__) || \
defined(__gfx1034__) || defined(__gfx1035__) || defined(__gfx1036__)
#define __gfx103__
#endif
#if defined(__gfx1100__) || defined(__gfx1101__) || defined(__gfx1102__) || defined(__gfx1103__)
#define __gfx11__
#endif
// buffer resource
#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__) || defined(__gfx940__) || defined(__gfx941__) || \
defined(__gfx942__) // for GPU code
defined(__gfx90a__) || defined(__gfx94__)
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x00020000
#elif defined(__gfx1030__) // for GPU code
#elif defined(__gfx103__)
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x31014000
#elif defined(__gfx1100__) || defined(__gfx1101__) || defined(__gfx1102__) // for GPU code
#elif defined(__gfx11__)
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x31004000
#endif
@@ -61,12 +75,12 @@
#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__) || \
defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__) // for GPU code
#elif defined(__gfx906__) || defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx103__) || \
defined(__gfx94__) // 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
#elif defined(__gfx1100__) || defined(__gfx1101__) || defined(__gfx1102__)
#elif defined(__gfx11__)
#define CK_USE_AMD_V_FMAC_F32
#define CK_USE_AMD_V_DOT2_F32_F16
#define CK_USE_AMD_V_DOT4_I32_I8_GFX11
@@ -75,23 +89,22 @@
// MFMA instruction
#ifndef __HIP_DEVICE_COMPILE__ // for host code
#define CK_USE_AMD_MFMA
#elif defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx940__) || defined(__gfx941__) || \
defined(__gfx942__) // for GPU code
#elif defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx94__) // for GPU code
#define CK_USE_AMD_MFMA
#endif
#if(defined(__gfx90a__) || defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__))
#if(defined(__gfx90a__) || defined(__gfx94__))
#define CK_USE_AMD_MFMA_BF16_1K_OP
#endif
#if defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__)
#if defined(__gfx94__)
#define CK_USE_AMD_MFMA_GFX940
#endif
// WMMA instruction
#ifndef __HIP_DEVICE_COMPILE__ // for host code
#define CK_USE_AMD_WMMA
#elif defined(__gfx1100__) || defined(__gfx1101__) || defined(__gfx1102__) // for GPU code
#elif defined(__gfx11__) // for GPU code
#define CK_USE_AMD_WMMA
#endif
@@ -107,15 +120,13 @@
// 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__) || defined(__gfx940__) || defined(__gfx941__) || \
defined(__gfx942__) // for GPU code
#elif defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx94__) // 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__) || defined(__gfx940__) || defined(__gfx941__) || \
defined(__gfx942__)) // for GPU code
#if(defined(__gfx90a__) || defined(__gfx94__)) // for GPU code
#define CK_USE_AMD_BUFFER_ATOMIC_MAX_FLOAT64 1
#else
#define CK_USE_AMD_BUFFER_ATOMIC_MAX_FLOAT64 0