AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

kfdtest: add support for gfx9.5.0 in shader store

Browse Source 
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Change-Id: I48b98ff631bd1aa1a044b60583ff256e43b17423
This commit is contained in:
Alex Sierra
2024-09-03 21:52:55 +00:00
committed by Amber Lin
parent e94ff8a36c
commit 268054cd28
1 changed files with 12 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
libhsakmt/tests/kfdtest/src/ShaderStore.cpp
+12 -12
View File
@@ -99,28 +99,28 @@ const std::vector<const char*> ShaderList = {
*/
#define SHADER_MACROS_FLAT \
" .macro FLAT_LOAD_DWORD_NSS vdst, vaddr arg0 arg1\n"\
" .if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)\n"\
" .if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)\n"\
" flat_load_dword \\vdst, \\vaddr nt sc1 sc0\n"\
" .else\n"\
" flat_load_dword \\vdst, \\vaddr \\arg0 \\arg1\n"\
" .endif\n"\
" .endm\n"\
" .macro FLAT_LOAD_DWORDX2_NSS vdst, vaddr arg0 arg1\n"\
" .if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)\n"\
" .if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)\n"\
" flat_load_dwordx2 \\vdst, \\vaddr nt sc1 sc0\n"\
" .else\n"\
" flat_load_dwordx2 \\vdst, \\vaddr \\arg0 \\arg1\n"\
" .endif\n"\
" .endm\n"\
" .macro FLAT_STORE_DWORD_NSS vaddr, vsrc arg0 arg1\n"\
" .if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)\n"\
" .if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)\n"\
" flat_store_dword \\vaddr, \\vsrc nt sc1 sc0\n"\
" .else\n"\
" flat_store_dword \\vaddr, \\vsrc \\arg0 \\arg1\n"\
" .endif\n"\
" .endm\n"\
" .macro FLAT_ATOMIC_ADD_NSS vdst, vaddr, vsrc arg0 arg1\n"\
" .if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)\n"\
" .if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)\n"\
" flat_atomic_add \\vdst, \\vaddr, \\vsrc nt sc1 sc0\n"\
" .else\n"\
" flat_atomic_add \\vdst, \\vaddr, \\vsrc \\arg0 \\arg1\n"\
@@ -363,7 +363,7 @@ const char *PollAndCopyIsa =
R"(
// Assume src buffer in s[0:1] and dst buffer in s[2:3]
// Path for Aldebaran, Aqua Vanjaram
.if (.amdgcn.gfx_generation_number == 9 && (.amdgcn.gfx_generation_minor == 4 || .amdgcn.gfx_generation_stepping == 10))
.if (.amdgcn.gfx_generation_number == 9 && (.amdgcn.gfx_generation_minor >= 4 || .amdgcn.gfx_generation_stepping == 10))
v_mov_b32 v0, s0
v_mov_b32 v1, s1
v_mov_b32 v18, 0x1
@@ -372,7 +372,7 @@ const char *PollAndCopyIsa =
s_waitcnt vmcnt(0) & lgkmcnt(0)
v_cmp_eq_i32 vcc, v16, v18
s_cbranch_vccz LOOP0
.if (.amdgcn.gfx_generation_minor == 4)
.if (.amdgcn.gfx_generation_minor >= 4)
buffer_inv sc1 sc0
.else
buffer_invl2
@@ -410,7 +410,7 @@ const char *WriteFlagAndValueIsa =
SHADER_MACROS_FLAT
R"(
// Assume two inputs buffer in s[0:1] and s[2:3]
.if (.amdgcn.gfx_generation_number == 9 && (.amdgcn.gfx_generation_minor == 4 || .amdgcn.gfx_generation_stepping == 10))
.if (.amdgcn.gfx_generation_number == 9 && (.amdgcn.gfx_generation_minor >= 4 || .amdgcn.gfx_generation_stepping == 10))
v_mov_b32 v0, s0
v_mov_b32 v1, s1
s_load_dword s18, s[2:3], 0x0 glc
@@ -480,7 +480,7 @@ const char *WriteAndSignalIsa =
const char *FlushBufferForAcquireReleaseIsa =
SHADER_START
R"(
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)
s_mov_b32 s11, 0x77
s_mov_b32 s12, 0x0
// Store some data on 5 different cache lines
@@ -510,7 +510,7 @@ const char *FlushBufferForAcquireReleaseIsa =
const char *WriteReleaseVectorIsa =
SHADER_START
R"(
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)
v_mov_b32 v11, 0x1
v_mov_b32 v12, 0x2
v_mov_b32 v13, 0x3
@@ -552,7 +552,7 @@ const char *WriteReleaseVectorIsa =
const char *WriteReleaseScalarIsa =
SHADER_START
R"(
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)
s_mov_b32 s11, 0x6
s_mov_b32 s12, 0x7
s_mov_b32 s13, 0x8
@@ -595,7 +595,7 @@ const char *WriteReleaseScalarIsa =
const char *ReadAcquireVectorIsa =
SHADER_START
R"(
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)
// Read-Acquire
s_mov_b32 s18, 0x1
LOOP:
@@ -651,7 +651,7 @@ const char *ReadAcquireVectorIsa =
const char *ReadAcquireScalarIsa =
SHADER_START
R"(
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor == 4)
.if (.amdgcn.gfx_generation_number == 9 && .amdgcn.gfx_generation_minor >= 4)
// Read-Acquire
s_mov_b32 s18, 0x1
LOOP: