Enable roofline for MI350 series (#677)

Rework of roofline binaries generated from rocm-amdgpu-bench
- removed arch identifier in bin name
- removed rocm5 bins altogether

Updated required distros for roofline
- updated distro checks and bin naming
- moved up ubuntu20.04->22.04 and sles15.3->15.6 per rocm support

Enabled ctests for mi350 for test_roof_*
- removed mi350 series check to skip these specific tests

---------

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>
This commit is contained in:
cfallows-amd
2025-04-28 16:08:23 -04:00
committed by GitHub
parent ad17c4d587
commit 41e73650d5
19 changed files with 35 additions and 31 deletions
@@ -0,0 +1,12 @@
pmc: SQ_INSTS_VALU_ADD_F16 SQ_INSTS_VALU_MUL_F16 SQ_INSTS_VALU_FMA_F16 SQ_INSTS_VALU_TRANS_F16 SQ_INSTS_VALU_ADD_F32 SQ_INSTS_VALU_MUL_F32 SQ_INSTS_VALU_FMA_F32 SQ_INSTS_VALU_TRANS_F32
pmc: SQ_INSTS_VALU_ADD_F64 SQ_INSTS_VALU_MUL_F64 SQ_INSTS_VALU_FMA_F64 SQ_INSTS_VALU_TRANS_F64 SQ_INSTS_VALU_MFMA_MOPS_F16 SQ_INSTS_VALU_MFMA_MOPS_BF16 SQ_INSTS_VALU_MFMA_MOPS_F32 SQ_INSTS_VALU_MFMA_MOPS_F64 TCP_TCC_READ_REQ_sum TCC_EA0_RDREQ_32B_sum TCC_EA0_RDREQ_sum TCC_EA0_WRREQ_64B_sum TCC_EA0_WRREQ_sum TCC_BUBBLE_sum
pmc: SQ_LDS_IDX_ACTIVE SQ_LDS_BANK_CONFLICT TCP_TOTAL_CACHE_ACCESSES_sum TCP_TCC_WRITE_REQ_sum TCP_TCC_ATOMIC_WITH_RET_REQ_sum TCP_TCC_ATOMIC_WITHOUT_RET_REQ_sum
#ROOF:MI300 series and above
pmc: SQ_INSTS_VALU_MFMA_MOPS_F8
range:
gpu:
kernels:
-11
View File
@@ -46,17 +46,6 @@ class gfx950_soc(OmniSoC_Base):
)
)
)
else:
# NB: We're using generalized Mi300 perfmon configs
self.set_perfmon_dir(
str(
Path(str(config.rocprof_compute_home)).joinpath(
"rocprof_compute_soc",
"profile_configs",
"gfx950",
)
)
)
self.set_compatible_profilers(["rocprofv3"])
# Per IP block max number of simultaneous counters. GFX IP Blocks
self.set_perfmon_config(