Update peak flops for MI350 (#810)

[ROCm/rocprofiler-compute commit: e1da9355c2]
This commit is contained in:
Fei Zheng
2025-07-15 08:46:13 -06:00
committed by GitHub
parent 65673617e0
commit 1050e0201c
2 changed files with 9 additions and 9 deletions
@@ -45,7 +45,7 @@ Panel Config:
MFMA FLOPs (F8):
value: AVG(((SQ_INSTS_VALU_MFMA_MOPS_F8 * 512) / (End_Timestamp - Start_Timestamp)))
unit: GFLOP/s
peak: ((($max_sclk * $cu_per_gpu) * 4096) / 1000)
peak: ((($max_sclk * $cu_per_gpu) * 8192) / 1000)
pop: ((100 * AVG(((SQ_INSTS_VALU_MFMA_MOPS_F8 * 512) / (End_Timestamp - Start_Timestamp))))
/ ((($max_sclk * $cu_per_gpu) * 8192) / 1000))
tips:
@@ -73,16 +73,16 @@ Panel Config:
MFMA FLOPs (F64):
value: AVG(((SQ_INSTS_VALU_MFMA_MOPS_F64 * 512) / (End_Timestamp - Start_Timestamp)))
unit: GFLOP/s
peak: ((($max_sclk * $cu_per_gpu) * 256) / 1000)
peak: ((($max_sclk * $cu_per_gpu) * 128) / 1000)
pop: ((100 * AVG(((SQ_INSTS_VALU_MFMA_MOPS_F64 * 512) / (End_Timestamp - Start_Timestamp))))
/ ((($max_sclk * $cu_per_gpu) * 256) / 1000))
/ ((($max_sclk * $cu_per_gpu) * 128) / 1000))
tips:
MFMA FLOPs (F6F4):
value: AVG(((SQ_INSTS_VALU_MFMA_MOPS_F6F4 * 512) / (End_Timestamp - Start_Timestamp)))
unit: GFLOP/s
peak: ((($max_sclk * $cu_per_gpu) * 256) / 1000)
peak: ((($max_sclk * $cu_per_gpu) * 16834) / 1000)
pop: ((100 * AVG(((SQ_INSTS_VALU_MFMA_MOPS_F6F4 * 512) / (End_Timestamp - Start_Timestamp))))
/ ((($max_sclk * $cu_per_gpu) * 256) / 1000))
/ ((($max_sclk * $cu_per_gpu) * 16834) / 1000))
tips:
MFMA IOPs (Int8):
value: AVG(((SQ_INSTS_VALU_MFMA_MOPS_I8 * 512) / (End_Timestamp - Start_Timestamp)))
@@ -72,16 +72,16 @@ Panel Config:
MFMA FLOPs (F64):
value: AVG(((SQ_INSTS_VALU_MFMA_MOPS_F64 * 512) / (End_Timestamp - Start_Timestamp)))
unit: GFLOP
peak: ((($max_sclk * $cu_per_gpu) * 256) / 1000)
peak: ((($max_sclk * $cu_per_gpu) * 128) / 1000)
pop: ((100 * AVG(((SQ_INSTS_VALU_MFMA_MOPS_F64 * 512) / (End_Timestamp - Start_Timestamp))))
/ ((($max_sclk * $cu_per_gpu) * 256) / 1000))
/ ((($max_sclk * $cu_per_gpu) * 128) / 1000))
tips:
MFMA FLOPs (F6F4):
value: AVG(((SQ_INSTS_VALU_MFMA_MOPS_F6F4 * 512) / (End_Timestamp - Start_Timestamp)))
unit: GFLOP
peak: ((($max_sclk * $cu_per_gpu) * 256) / 1000)
peak: ((($max_sclk * $cu_per_gpu) * 16834) / 1000)
pop: ((100 * AVG(((SQ_INSTS_VALU_MFMA_MOPS_F6F4 * 512) / (End_Timestamp - Start_Timestamp))))
/ ((($max_sclk * $cu_per_gpu) * 256) / 1000))
/ ((($max_sclk * $cu_per_gpu) * 16834) / 1000))
tips:
MFMA IOPs (INT8):
value: AVG(((SQ_INSTS_VALU_MFMA_MOPS_I8 * 512) / (End_Timestamp - Start_Timestamp)))