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

Update csv headers, and SystemExit codes

Browse Source 
Signed-off-by: JoseSantosAMD <Jose.Santos@amd.com>
This commit is contained in:
JoseSantosAMD
2024-02-02 20:04:11 -06:00
committed by Karl W. Schulz
parent 7556316f02
commit 717a21cf84
2842 changed files with 5653 additions and 3927 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/workloads/invdev/mi100/SQ_IFETCH_LEVEL.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBuffer.kd",0,0,0,935069,935074,33554432,256,0,0,8,32,6464,0x0,0x7fb1e8e04180,504417,504417,524288,6291456,791548,101536232,12076606776836515,12076607020949616,12076607021273775,12076607021386317
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,935069,935074,32768,256,0,0,24,24,12480,0x0,0x7fb1e8e35100,27457,27457,512,8192,8659,1112808,12076607036018836,12076607036328752,12076607036334672,12076607036343359
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,4,935069,935074,4194304,256,0,0,24,24,12928,0x7fb2f49b3900,0x7fb1e8e35140,221451,221451,65536,917504,137700,17627108,12076607036410164,12076607036634350,12076607036769870,12076607036773860
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj GRBM_COUNT GRBM_GUI_ACTIVE SQ_WAVES SQ_IFETCH SQ_IFETCH_LEVEL SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBuffer.kd 0 0 0 935069 935074 33554432 256 0 0 8 32 6464 0x0 0x7fb1e8e04180 504417 504417 524288 6291456 791548 101536232 12076606776836515 12076607020949616 12076607021273775 12076607021386317
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 935069 935074 32768 256 0 0 24 24 12480 0x0 0x7fb1e8e35100 27457 27457 512 8192 8659 1112808 12076607036018836 12076607036328752 12076607036334672 12076607036343359
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 4 935069 935074 4194304 256 0 0 24 24 12928 0x7fb2f49b3900 0x7fb1e8e35140 221451 221451 65536 917504 137700 17627108 12076607036410164 12076607036634350 12076607036769870 12076607036773860
tests/workloads/invdev/mi100/SQ_INST_LEVEL_LDS.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBuffer.kd",0,0,0,936874,936879,33554432,256,0,0,8,32,6464,0x0,0x7f608a404180,0,0,0,12076632988700397,12076633233142956,12076633233467755,12076633233577518
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,936874,936879,32768,256,0,0,24,24,12480,0x0,0x7f608a435100,0,0,0,12076633248282761,12076633248580944,12076633248587504,12076633248592868
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,4,936874,936879,4194304,256,0,0,24,24,12928,0x7f61960b3900,0x7f608a435140,0,0,0,12076633248655084,12076633248875663,12076633249007503,12076633249011156
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj SQ_INSTS_LDS SQ_INST_LEVEL_LDS SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBuffer.kd 0 0 0 936874 936879 33554432 256 0 0 8 32 6464 0x0 0x7f608a404180 0 0 0 12076632988700397 12076633233142956 12076633233467755 12076633233577518
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 936874 936879 32768 256 0 0 24 24 12480 0x0 0x7f608a435100 0 0 0 12076633248282761 12076633248580944 12076633248587504 12076633248592868
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 4 936874 936879 4194304 256 0 0 24 24 12928 0x7f61960b3900 0x7f608a435140 0 0 0 12076633248655084 12076633248875663 12076633249007503 12076633249011156
tests/workloads/invdev/mi100/SQ_INST_LEVEL_SMEM.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBuffer.kd",0,0,0,936496,936501,33554432,256,0,0,8,32,6464,0x0,0x7f17e8404180,4194304,3122666,398904176,12076630480743351,12076630718390515,12076630718714354,12076630718822006
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,936496,936501,32768,256,0,0,24,24,12480,0x0,0x7f17e8435100,512,21524,2754256,12076630733442442,12076630733749898,12076630733756458,12076630733761836
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,4,936496,936501,4194304,256,0,0,24,24,12928,0x7f18f4030900,0x7f17e8435140,65536,146806,18775720,12076630733826656,12076630734045577,12076630734179017,12076630734183330
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj SQ_INSTS_SMEM SQ_INST_LEVEL_SMEM SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBuffer.kd 0 0 0 936496 936501 33554432 256 0 0 8 32 6464 0x0 0x7f17e8404180 4194304 3122666 398904176 12076630480743351 12076630718390515 12076630718714354 12076630718822006
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 936496 936501 32768 256 0 0 24 24 12480 0x0 0x7f17e8435100 512 21524 2754256 12076630733442442 12076630733749898 12076630733756458 12076630733761836
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 4 936496 936501 4194304 256 0 0 24 24 12928 0x7f18f4030900 0x7f17e8435140 65536 146806 18775720 12076630733826656 12076630734045577 12076630734179017 12076630734183330
tests/workloads/invdev/mi100/SQ_INST_LEVEL_VMEM.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBuffer.kd",0,0,0,936686,936691,33554432,256,0,0,8,32,6464,0x0,0x7fa3b8c04180,1048576,11096456,1419911252,12076631726864889,12076631974434106,12076631974755065,12076631974864438
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,936686,936691,32768,256,0,0,24,24,12480,0x0,0x7fa3b8c35100,4096,107784,13788356,12076631989518086,12076631989806097,12076631989812497,12076631989817693
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,4,936686,936691,4194304,256,0,0,24,24,12928,0x7fa4c491a900,0x7fa3b8c35140,524288,10667896,1365491840,12076631989877063,12076631990100976,12076631990238896,12076631990242964
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj SQ_INSTS_VMEM SQ_INST_LEVEL_VMEM SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBuffer.kd 0 0 0 936686 936691 33554432 256 0 0 8 32 6464 0x0 0x7fa3b8c04180 1048576 11096456 1419911252 12076631726864889 12076631974434106 12076631974755065 12076631974864438
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 936686 936691 32768 256 0 0 24 24 12480 0x0 0x7fa3b8c35100 4096 107784 13788356 12076631989518086 12076631989806097 12076631989812497 12076631989817693
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 4 936686 936691 4194304 256 0 0 24 24 12928 0x7fa4c491a900 0x7fa3b8c35140 524288 10667896 1365491840 12076631989877063 12076631990100976 12076631990238896 12076631990242964
tests/workloads/invdev/mi100/SQ_LEVEL_WAVES.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,GRBM_COUNT,GRBM_GUI_ACTIVE,CPC_ME1_BUSY_FOR_PACKET_DECODE,SQ_CYCLES,SQ_WAVES,SQ_WAVE_CYCLES,SQ_BUSY_CYCLES,SQ_LEVEL_WAVES,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,GRBM_COUNT,GRBM_GUI_ACTIVE,CPC_ME1_BUSY_FOR_PACKET_DECODE,SQ_CYCLES,SQ_WAVES,SQ_WAVE_CYCLES,SQ_BUSY_CYCLES,SQ_LEVEL_WAVES,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBuffer.kd",0,0,0,937062,937067,33554432,256,0,0,8,32,6464,0x0,0x7f30a5604180,499674,499674,16219,3997400,524288,371768834,3801967,0,1501859028,12076634246773666,12076634492293789,12076634492615709,12076634492722711
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,937062,937067,32768,256,0,0,24,24,12480,0x0,0x7f30a5635100,27467,27467,20214,219744,512,1126759,75366,0,4521000,12076634507281913,12076634507609674,12076634507615914,12076634507624380
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,4,937062,937067,4194304,256,0,0,24,24,12928,0x7f31b1191900,0x7f30a5635140,217853,217853,21633,1742832,65536,133671007,1576099,0,536498696,12076634507695903,12076634507934314,12076634508067113,12076634508071402
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj GRBM_COUNT GRBM_GUI_ACTIVE CPC_ME1_BUSY_FOR_PACKET_DECODE SQ_CYCLES SQ_WAVES SQ_WAVE_CYCLES SQ_BUSY_CYCLES SQ_LEVEL_WAVES SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBuffer.kd 0 0 0 937062 937067 33554432 256 0 0 8 32 6464 0x0 0x7f30a5604180 499674 499674 16219 3997400 524288 371768834 3801967 0 1501859028 12076634246773666 12076634492293789 12076634492615709 12076634492722711
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 937062 937067 32768 256 0 0 24 24 12480 0x0 0x7f30a5635100 27467 27467 20214 219744 512 1126759 75366 0 4521000 12076634507281913 12076634507609674 12076634507615914 12076634507624380
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 4 937062 937067 4194304 256 0 0 24 24 12928 0x7f31b1191900 0x7f30a5635140 217853 217853 21633 1742832 65536 133671007 1576099 0 536498696 12076634507695903 12076634507934314 12076634508067113 12076634508071402
tests/workloads/invdev/mi100/pmc_dispatch_info.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id
Dispatch_ID,Kernel_Name,GPU_ID
0,__amd_rocclr_fillBuffer.kd,0
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID
2 0 __amd_rocclr_fillBuffer.kd 0
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0
tests/workloads/invdev/mi100/pmc_perf.csv
+1 -1
View File
File diff suppressed because one or more lines are too long
tests/workloads/invdev/mi100/prev_analysis/1.csv
+1 -1
View File
@@ -1,4 +1,4 @@
KernelName,Count,Sum(ns),Mean(ns),Median(ns),Pct
Kernel_Name,Count,Sum(ns),Mean(ns),Median(ns),Pct
"void benchmark_func<HIP_vector_type<float, 2u>, 256, 8u, 512u>(HIP_vector_type<float, 2u>, HIP_vector_type<float, 2u>*) [clone .kd]",1,6059957.0,6059957.0,6059957.0,9.169486064060209
"void benchmark_func<int, 256, 8u, 512u>(int, int*) [clone .kd]",1,4525889.0,4525889.0,4525889.0,6.848245971544582
"void benchmark_func<HIP_vector_type<float, 2u>, 256, 8u, 256u>(HIP_vector_type<float, 2u>, HIP_vector_type<float, 2u>*) [clone .kd]",1,3056459.0,3056459.0,3056459.0,4.624811398145466
1 KernelName Kernel_Name Count Sum(ns) Mean(ns) Median(ns) Pct
2 void benchmark_func<HIP_vector_type<float, 2u>, 256, 8u, 512u>(HIP_vector_type<float, 2u>, HIP_vector_type<float, 2u>*) [clone .kd] 1 6059957.0 6059957.0 6059957.0 9.169486064060209
3 void benchmark_func<int, 256, 8u, 512u>(int, int*) [clone .kd] 1 4525889.0 4525889.0 4525889.0 6.848245971544582
4 void benchmark_func<HIP_vector_type<float, 2u>, 256, 8u, 256u>(HIP_vector_type<float, 2u>, HIP_vector_type<float, 2u>*) [clone .kd] 1 3056459.0 3056459.0 3056459.0 4.624811398145466
tests/workloads/invdev/mi100/prev_analysis/1001.csv
+1 -1
View File
@@ -1,7 +1,7 @@
Metric,Count,Unit
VALU - Vector,,Instr per wave
VMEM,,Instr per wave
LDS,0.0,Instr per wave
LDS_Per_Workgroup,0.0,Instr per wave
VALU - MFMA,,Instr per wave
SALU,46.862275449101794,Instr per wave
SMEM,1.8083832335329342,Instr per wave
1 Metric Count Unit
2 VALU - Vector Instr per wave
3 VMEM Instr per wave
4 LDS LDS_Per_Workgroup 0.0 Instr per wave
5 VALU - MFMA Instr per wave
6 SALU 46.862275449101794 Instr per wave
7 SMEM 1.8083832335329342 Instr per wave
tests/workloads/invdev/mi100/prev_analysis/101.csv
+1 -1
View File
@@ -23,4 +23,4 @@ gfx908
mi100
48
1228.8
SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
SQ|LDS_Per_Workgroup|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
1 Info
23 mi100
24 48
25 1228.8
26 SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF SQ|LDS_Per_Workgroup|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
tests/workloads/invdev/mi100/prev_analysis/1202.csv
+3 -3
View File
@@ -1,12 +1,12 @@
Metric,Avg,Min,Max,Unit
Wave Cycles,33617.72709436474,2860.667678833008,547916.0441894531,Cycles/wave
LDS Instrs,0.0,0.0,0.0,Instr per wave
LDS_Per_Workgroup Instrs,0.0,0.0,0.0,Instr per wave
Bandwidth,0.0,0.0,0.0,Bytes per wave
Bank Conficts/Access,,,,Conflicts/access
Index Accesses,0.0,0.0,0.0,Cycles per wave
Dispatch_ID Accesses,0.0,0.0,0.0,Cycles per wave
Atomic Cycles,0.0,0.0,0.0,Cycles per wave
Bank Conflict,0.0,0.0,0.0,Cycles per wave
Addr Conflict,0.0,0.0,0.0,Cycles per wave
Unaligned Stall,0.0,0.0,0.0,Cycles per wave
Mem Violations,0.0,0.0,0.0, per wave
LDS Latency,,,,Cycles
LDS_Per_Workgroup Latency,,,,Cycles
1 Metric Avg Min Max Unit
2 Wave Cycles 33617.72709436474 2860.667678833008 547916.0441894531 Cycles/wave
3 LDS Instrs LDS_Per_Workgroup Instrs 0.0 0.0 0.0 Instr per wave
4 Bandwidth 0.0 0.0 0.0 Bytes per wave
5 Bank Conficts/Access Conflicts/access
6 Index Accesses Dispatch_ID Accesses 0.0 0.0 0.0 Cycles per wave
7 Atomic Cycles 0.0 0.0 0.0 Cycles per wave
8 Bank Conflict 0.0 0.0 0.0 Cycles per wave
9 Addr Conflict 0.0 0.0 0.0 Cycles per wave
10 Unaligned Stall 0.0 0.0 0.0 Cycles per wave
11 Mem Violations 0.0 0.0 0.0 per wave
12 LDS Latency LDS_Per_Workgroup Latency Cycles
tests/workloads/invdev/mi100/prev_analysis/1901.csv
+5 -5
View File
@@ -6,16 +6,16 @@ SMEM,2.0,smem_
VALU,735.0,valu_
MFMA,,mfma_
VMEM,8.0,vmem_
LDS,0.0,lds_
LDS_Per_Workgroup,0.0,lds_
GWS,0.0,gws_
BR,22.0,br_
VGPR,24.0,vgpr_
SGPR,24.0,sgpr_
LDS Allocation,0.0,lds_alloc_
LDS_Per_Workgroup Allocation,0.0,lds_alloc_
Scratch Allocation,0.0,scratch_alloc_
Wavefronts,67894.0,wavefronts_
Workgroups,16973.0,workgroups_
LDS Req,0.0,lds_req_
LDS_Per_Workgroup Req,0.0,lds_req_
IL1 Fetch,163.0,il1_fetch_
IL1 Hit,100.0,il1_hit_
IL1_L2 Rd,0.0,il1_l2_req_
@@ -46,10 +46,10 @@ Fabric_L2 Wr,0.0,l2_fabric_wr_
Fabric_l2 Atomic,0.0,l2_fabric_atom_
HBM Rd,44.0,hbm_rd_
HBM Wr,0.0,hbm_wr_
LDS Util,0.0,lds_util_
LDS_Per_Workgroup Util,0.0,lds_util_
VL1 Coalesce,70.0,vl1_coales_
VL1 Stall,31.0,vl1_stall_
LDS Lat,,lds_lat_
LDS_Per_Workgroup Lat,,lds_lat_
vL1D Lat,,sl1_lat_
IL1 Lat,,il1_lat_
Wave Occupancy,34.0,wave_occ_
1 Metric Value Alias
6 VALU 735.0 valu_
7 MFMA mfma_
8 VMEM 8.0 vmem_
9 LDS LDS_Per_Workgroup 0.0 lds_
10 GWS 0.0 gws_
11 BR 22.0 br_
12 VGPR 24.0 vgpr_
13 SGPR 24.0 sgpr_
14 LDS Allocation LDS_Per_Workgroup Allocation 0.0 lds_alloc_
15 Scratch Allocation 0.0 scratch_alloc_
16 Wavefronts 67894.0 wavefronts_
17 Workgroups 16973.0 workgroups_
18 LDS Req LDS_Per_Workgroup Req 0.0 lds_req_
19 IL1 Fetch 163.0 il1_fetch_
20 IL1 Hit 100.0 il1_hit_
21 IL1_L2 Rd 0.0 il1_l2_req_
46 Fabric_l2 Atomic 0.0 l2_fabric_atom_
47 HBM Rd 44.0 hbm_rd_
48 HBM Wr 0.0 hbm_wr_
49 LDS Util LDS_Per_Workgroup Util 0.0 lds_util_
50 VL1 Coalesce 70.0 vl1_coales_
51 VL1 Stall 31.0 vl1_stall_
52 LDS Lat LDS_Per_Workgroup Lat lds_lat_
53 vL1D Lat sl1_lat_
54 IL1 Lat il1_lat_
55 Wave Occupancy 34.0 wave_occ_
tests/workloads/invdev/mi100/prev_analysis/2001.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id
Dispatch_ID,Kernel_Name,GPU_ID
0,__amd_rocclr_fillBuffer.kd,0
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID
2 0 __amd_rocclr_fillBuffer.kd 0
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0
tests/workloads/invdev/mi100/prev_analysis/201.csv
+2 -2
View File
@@ -12,8 +12,8 @@ VALU Util,59.71163511528944,Pct,100,59.71163511528944
MFMA Util,,Pct,100,
VALU Active Threads/Wave,63.967840400497316,Threads,64,99.94975062577706
IPC - Issue,0.8437262750969097,Instr/cycle,5,16.874525501938194
LDS BW,0.0,Gb/sec,23070.72,0.0
LDS Bank Conflict,,Conflicts/access,32,
LDS_Per_Workgroup BW,0.0,Gb/sec,23070.72,0.0
LDS_Per_Workgroup Bank Conflict,,Conflicts/access,32,
Instr Cache Hit Rate,99.99318745374036,Pct,100,99.99318745374036
Instr Cache BW,1406.6059705869952,Gb/s,4614.144,30.484656971845595
Scalar L1D Cache Hit Rate,99.35620448519533,Pct,100,99.35620448519533
1 Metric Value Unit Peak PoP
12 MFMA Util Pct 100
13 VALU Active Threads/Wave 63.967840400497316 Threads 64 99.94975062577706
14 IPC - Issue 0.8437262750969097 Instr/cycle 5 16.874525501938194
15 LDS BW LDS_Per_Workgroup BW 0.0 Gb/sec 23070.72 0.0
16 LDS Bank Conflict LDS_Per_Workgroup Bank Conflict Conflicts/access 32
17 Instr Cache Hit Rate 99.99318745374036 Pct 100 99.99318745374036
18 Instr Cache BW 1406.6059705869952 Gb/s 4614.144 30.484656971845595
19 Scalar L1D Cache Hit Rate 99.35620448519533 Pct 100 99.35620448519533
tests/workloads/invdev/mi100/prev_analysis/602.csv
+1 -1
View File
@@ -6,7 +6,7 @@ Scratch Stall,0.0,0,0,Cycles
Insufficient SIMD Waveslots,64892.16766467066,0,774244,Simd
Insufficient SIMD VGPRs,527589.6886227545,0,30743007,Simd
Insufficient SIMD SGPRs,0.0,0,0,Simd
Insufficient CU LDS,0.0,0,0,Cu
Insufficient CU LDS_Per_Workgroup,0.0,0,0,Cu
Insufficient CU Barries,0.0,0,0,Cu
Insufficient Bulky Resource,0.0,0,0,Cu
Reach CU Threadgroups Limit,0.0,0,0,Cycles
1 Metric Avg Min Max Unit
6 Insufficient SIMD Waveslots 64892.16766467066 0 774244 Simd
7 Insufficient SIMD VGPRs 527589.6886227545 0 30743007 Simd
8 Insufficient SIMD SGPRs 0.0 0 0 Simd
9 Insufficient CU LDS Insufficient CU LDS_Per_Workgroup 0.0 0 0 Cu
10 Insufficient CU Barries 0.0 0 0 Cu
11 Insufficient Bulky Resource 0.0 0 0 Cu
12 Reach CU Threadgroups Limit 0.0 0 0 Cycles
tests/workloads/invdev/mi100/prev_analysis/701.csv
+1 -1
View File
@@ -6,5 +6,5 @@ Saved Wavefronts,0.0,0,0,Wavefronts
Restored Wavefronts,0.0,0,0,Wavefronts
VGPRs,23.976047904191617,8,36,Registers
SGPRs,24.047904191616766,24,32,Registers
LDS Allocation,0.0,0,0,Bytes
LDS_Per_Workgroup Allocation,0.0,0,0,Bytes
Scratch Allocation,0.0,0,0,Bytes
1 Metric Avg Min Max Unit
6 Restored Wavefronts 0.0 0 0 Wavefronts
7 VGPRs 23.976047904191617 8 36 Registers
8 SGPRs 24.047904191616766 24 32 Registers
9 LDS Allocation LDS_Per_Workgroup Allocation 0.0 0 0 Bytes
10 Scratch Allocation 0.0 0 0 Bytes
tests/workloads/invdev/mi100/sysinfo.csv
+1 -1
View File
@@ -1,2 +1,2 @@
workload_name,host_name,host_cpu,host_distro,host_kernel,host_rocmver,date,gpu_soc,numSE,numCU,numSIMD,waveSize,maxWavesPerCU,maxWorkgroupSize,L1,L2,sclk,mclk,cur_sclk,cur_mclk,L2Banks,name,numSQC,hbmBW,ip_blocks
invdev,vesuvius,AMD EPYC 7542 32-Core Processor,,4.18.0-240.15.1.el8_3.x86_64,4.2.0-21,Thu Sep 22 11:36:21 2022 (CDT),gfx908,8,120,4,64,40,1024,16,,1502,1200,300,1200,32,mi100,48,1228.8,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
invdev,vesuvius,AMD EPYC 7542 32-Core Processor,,4.18.0-240.15.1.el8_3.x86_64,4.2.0-21,Thu Sep 22 11:36:21 2022 (CDT),gfx908,8,120,4,64,40,1024,16,,1502,1200,300,1200,32,mi100,48,1228.8,SQ|LDS_Per_Workgroup|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
1 workload_name host_name host_cpu host_distro host_kernel host_rocmver date gpu_soc numSE numCU numSIMD waveSize maxWavesPerCU maxWorkgroupSize L1 L2 sclk mclk cur_sclk cur_mclk L2Banks name numSQC hbmBW ip_blocks
2 invdev vesuvius AMD EPYC 7542 32-Core Processor 4.18.0-240.15.1.el8_3.x86_64 4.2.0-21 Thu Sep 22 11:36:21 2022 (CDT) gfx908 8 120 4 64 40 1024 16 1502 1200 300 1200 32 mi100 48 1228.8 SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF SQ|LDS_Per_Workgroup|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
tests/workloads/invdev/mi100/timestamps.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBuffer.kd",0,0,0,937113,937118,33554432,256,0,0,8,32,6464,0x0,0x7f12eac04180,12076635381227990,12076635381273678,12076635381598795,12076635381689318
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,937113,937118,32768,256,0,0,24,24,12480,0x0,0x7f12eac35100,12076635396787522,12076635396887551,12076635396894271,12076635396915901
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,4,937113,937118,4194304,256,0,0,24,24,12928,0x7f141aca3900,0x7f12eac35140,12076635396946107,12076635396958910,12076635397089629,12076635397093471
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBuffer.kd 0 0 0 937113 937118 33554432 256 0 0 8 32 6464 0x0 0x7f12eac04180 12076635381227990 12076635381273678 12076635381598795 12076635381689318
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 937113 937118 32768 256 0 0 24 24 12480 0x0 0x7f12eac35100 12076635396787522 12076635396887551 12076635396894271 12076635396915901
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 4 937113 937118 4194304 256 0 0 24 24 12928 0x7f141aca3900 0x7f12eac35140 12076635396946107 12076635396958910 12076635397089629 12076635397093471
tests/workloads/invdev/mi200/SQ_IFETCH_LEVEL.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBufferAligned.kd",0,0,0,236503,236503,33554432,256,0,0,4,32,4160,0x0,0x7fe83c204280,385832,385832,524288,4718592,682094,76386864,17833173257611,17832460487290,17833324063793,17833324177862
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,236503,236503,32768,256,0,0,12,24,13888,0x0,0x7fe83c223f80,34124,34124,512,8192,5540,624020,17833329373303,17833324063793,17833329503317,17833329508120
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,5,236503,236503,4194304,256,0,0,12,24,14336,0x7fe83f16b380,0x7fe83c223fc0,166451,166451,65536,917504,140662,15737800,17833329547519,17833329503317,17833329887317,17833329890091
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj GRBM_COUNT GRBM_GUI_ACTIVE SQ_WAVES SQ_IFETCH SQ_IFETCH_LEVEL SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBufferAligned.kd 0 0 0 236503 236503 33554432 256 0 0 4 32 4160 0x0 0x7fe83c204280 385832 385832 524288 4718592 682094 76386864 17833173257611 17832460487290 17833324063793 17833324177862
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 236503 236503 32768 256 0 0 12 24 13888 0x0 0x7fe83c223f80 34124 34124 512 8192 5540 624020 17833329373303 17833324063793 17833329503317 17833329508120
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 5 236503 236503 4194304 256 0 0 12 24 14336 0x7fe83f16b380 0x7fe83c223fc0 166451 166451 65536 917504 140662 15737800 17833329547519 17833329503317 17833329887317 17833329890091
tests/workloads/invdev/mi200/SQ_INST_LEVEL_LDS.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBufferAligned.kd",0,0,0,238351,238351,33554432,256,0,0,4,32,4160,0x0,0x7ff888c04280,0,0,0,17852547584485,17851838027568,17852691164011,17852691281181
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,238351,238351,32768,256,0,0,12,24,13888,0x0,0x7ff888c23f80,0,0,0,17852696444493,17852691164011,17852696572335,17852696576630
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,5,238351,238351,4194304,256,0,0,12,24,14336,0x7ff88bbf2380,0x7ff888c23fc0,0,0,0,17852696612159,17852696572335,17852696936336,17852696938901
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj SQ_INSTS_LDS SQ_INST_LEVEL_LDS SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBufferAligned.kd 0 0 0 238351 238351 33554432 256 0 0 4 32 4160 0x0 0x7ff888c04280 0 0 0 17852547584485 17851838027568 17852691164011 17852691281181
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 238351 238351 32768 256 0 0 12 24 13888 0x0 0x7ff888c23f80 0 0 0 17852696444493 17852691164011 17852696572335 17852696576630
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 5 238351 238351 4194304 256 0 0 12 24 14336 0x7ff88bbf2380 0x7ff888c23fc0 0 0 0 17852696612159 17852696572335 17852696936336 17852696938901
tests/workloads/invdev/mi200/SQ_INST_LEVEL_SMEM.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBufferAligned.kd",0,0,0,236281,236281,33554432,256,0,0,4,32,4160,0x0,0x7f45e2204280,3670016,2905706,325081704,17832236843728,17803198665120,17832381629870,17832381744199
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,236281,236281,32768,256,0,0,12,24,13888,0x0,0x7f45e2223f80,512,96664,10851288,17832386900391,17832381629870,17832387026519,17832387030968
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,5,236281,236281,4194304,256,0,0,12,24,14336,0x7f45e517a380,0x7f45e2223fc0,65536,639926,71704992,17832387064037,17832387026519,17832387402360,17832387404639
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj SQ_INSTS_SMEM SQ_INST_LEVEL_SMEM SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBufferAligned.kd 0 0 0 236281 236281 33554432 256 0 0 4 32 4160 0x0 0x7f45e2204280 3670016 2905706 325081704 17832236843728 17803198665120 17832381629870 17832381744199
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 236281 236281 32768 256 0 0 12 24 13888 0x0 0x7f45e2223f80 512 96664 10851288 17832386900391 17832381629870 17832387026519 17832387030968
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 5 236281 236281 4194304 256 0 0 12 24 14336 0x7f45e517a380 0x7f45e2223fc0 65536 639926 71704992 17832387064037 17832387026519 17832387402360 17832387404639
tests/workloads/invdev/mi200/SQ_INST_LEVEL_VMEM.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBufferAligned.kd",0,0,0,238129,238129,33554432,256,0,0,4,32,4160,0x0,0x7fe2c7e04280,524288,5497472,615656480,17851610507915,17849318958956,17851757718801,17851757807931
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,238129,238129,32768,256,0,0,12,24,13888,0x0,0x7fe2c7e23f80,4096,55920,6257708,17851762985423,17851757718801,17851763112570,17851763117229
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,5,238129,238129,4194304,256,0,0,12,24,14336,0x7fe2d67ec380,0x7fe2c7e23fc0,524288,10960198,1227561992,17851763151509,17851763112570,17851763483290,17851763485710
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj SQ_INSTS_VMEM SQ_INST_LEVEL_VMEM SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBufferAligned.kd 0 0 0 238129 238129 33554432 256 0 0 4 32 4160 0x0 0x7fe2c7e04280 524288 5497472 615656480 17851610507915 17849318958956 17851757718801 17851757807931
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 238129 238129 32768 256 0 0 12 24 13888 0x0 0x7fe2c7e23f80 4096 55920 6257708 17851762985423 17851757718801 17851763112570 17851763117229
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 5 238129 238129 4194304 256 0 0 12 24 14336 0x7fe2d67ec380 0x7fe2c7e23fc0 524288 10960198 1227561992 17851763151509 17851763112570 17851763483290 17851763485710
tests/workloads/invdev/mi200/SQ_LEVEL_WAVES.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,GRBM_COUNT,GRBM_GUI_ACTIVE,CPC_ME1_BUSY_FOR_PACKET_DECODE,SQ_CYCLES,SQ_WAVES,SQ_WAVE_CYCLES,SQ_BUSY_CYCLES,SQ_LEVEL_WAVES,SQ_ACCUM_PREV_HIRES,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,GRBM_COUNT,GRBM_GUI_ACTIVE,CPC_ME1_BUSY_FOR_PACKET_DECODE,SQ_CYCLES,SQ_WAVES,SQ_WAVE_CYCLES,SQ_BUSY_CYCLES,SQ_LEVEL_WAVES,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBufferAligned.kd",0,0,0,236725,236725,33554432,256,0,0,4,32,4160,0x0,0x7fc8d4204280,388118,388118,8813,3104952,524288,245379700,3015215,0,997822932,17834114824366,17833406340767,17834261594719,17834261709089
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,236725,236725,32768,256,0,0,12,24,13888,0x0,0x7fc8d4223f80,34226,34226,30380,273816,512,1732870,162674,0,6944808,17834266862991,17834261594719,17834267004326,17834267009117
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,5,236725,236725,4194304,256,0,0,12,24,14336,0x7fc8d71bf380,0x7fc8d4223fc0,166068,166068,13553,1328552,65536,78338107,1221027,0,315083700,17834267052826,17834267004326,17834267413606,17834267416207
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj GRBM_COUNT GRBM_GUI_ACTIVE CPC_ME1_BUSY_FOR_PACKET_DECODE SQ_CYCLES SQ_WAVES SQ_WAVE_CYCLES SQ_BUSY_CYCLES SQ_LEVEL_WAVES SQ_ACCUM_PREV_HIRES DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBufferAligned.kd 0 0 0 236725 236725 33554432 256 0 0 4 32 4160 0x0 0x7fc8d4204280 388118 388118 8813 3104952 524288 245379700 3015215 0 997822932 17834114824366 17833406340767 17834261594719 17834261709089
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 236725 236725 32768 256 0 0 12 24 13888 0x0 0x7fc8d4223f80 34226 34226 30380 273816 512 1732870 162674 0 6944808 17834266862991 17834261594719 17834267004326 17834267009117
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 5 236725 236725 4194304 256 0 0 12 24 14336 0x7fc8d71bf380 0x7fc8d4223fc0 166068 166068 13553 1328552 65536 78338107 1221027 0 315083700 17834267052826 17834267004326 17834267413606 17834267416207
tests/workloads/invdev/mi200/pmc_dispatch_info.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id
Dispatch_ID,Kernel_Name,GPU_ID
0,__amd_rocclr_fillBufferAligned.kd,0
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID
2 0 __amd_rocclr_fillBufferAligned.kd 0
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0
tests/workloads/invdev/mi200/pmc_perf.csv
+1 -1
View File
File diff suppressed because one or more lines are too long
tests/workloads/invdev/mi200/prev_analysis/1.csv
+1 -1
View File
@@ -1,4 +1,4 @@
KernelName,Count,Sum(ns),Mean(ns),Median(ns),Pct
Kernel_Name,Count,Sum(ns),Mean(ns),Median(ns),Pct
"void benchmark_func<int, 256, 8u, 512u>(int, int*) [clone .kd]",1,3357122.0,3357122.0,3357122.0,7.840630919671416
"void benchmark_func<HIP_vector_type<float, 2u>, 256, 8u, 512u>(HIP_vector_type<float, 2u>, HIP_vector_type<float, 2u>*) [clone .kd]",1,1724001.0,1724001.0,1724001.0,4.026441560999106
"void benchmark_func<double, 256, 8u, 512u>(double, double*) [clone .kd]",1,1714881.0,1714881.0,1714881.0,4.005141604075466
1 KernelName Kernel_Name Count Sum(ns) Mean(ns) Median(ns) Pct
2 void benchmark_func<int, 256, 8u, 512u>(int, int*) [clone .kd] 1 3357122.0 3357122.0 3357122.0 7.840630919671416
3 void benchmark_func<HIP_vector_type<float, 2u>, 256, 8u, 512u>(HIP_vector_type<float, 2u>, HIP_vector_type<float, 2u>*) [clone .kd] 1 1724001.0 1724001.0 1724001.0 4.026441560999106
4 void benchmark_func<double, 256, 8u, 512u>(double, double*) [clone .kd] 1 1714881.0 1714881.0 1714881.0 4.005141604075466
tests/workloads/invdev/mi200/prev_analysis/1001.csv
+1 -1
View File
@@ -1,7 +1,7 @@
Metric,Count,Unit
VALU - Vector,494.8622754491018,Instr per wave
VMEM,7.958083832335329,Instr per wave
LDS,0.0,Instr per wave
LDS_Per_Workgroup,0.0,Instr per wave
VALU - MFMA,0.0,Instr per wave
SALU,24.520958083832337,Instr per wave
SMEM,1.8023952095808384,Instr per wave
1 Metric Count Unit
2 VALU - Vector 494.8622754491018 Instr per wave
3 VMEM 7.958083832335329 Instr per wave
4 LDS LDS_Per_Workgroup 0.0 Instr per wave
5 VALU - MFMA 0.0 Instr per wave
6 SALU 24.520958083832337 Instr per wave
7 SMEM 1.8023952095808384 Instr per wave
tests/workloads/invdev/mi200/prev_analysis/101.csv
+1 -1
View File
@@ -23,4 +23,4 @@ gfx90a
mi200
56
1638.4
roofline|SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
roofline|SQ|LDS_Per_Workgroup|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
1 Info
23 mi200
24 56
25 1638.4
26 roofline|SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF roofline|SQ|LDS_Per_Workgroup|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
tests/workloads/invdev/mi200/prev_analysis/1202.csv
+3 -3
View File
@@ -1,12 +1,12 @@
Metric,Avg,Min,Max,Unit
Wave Cycles,18296.269474052384,1894.2776184082031,258229.23516845703,Cycles/wave
LDS Instrs,0.0,0.0,0.0,Instr per wave
LDS_Per_Workgroup Instrs,0.0,0.0,0.0,Instr per wave
Bandwidth,0.0,0.0,0.0,Bytes per wave
Bank Conficts/Access,,,,Conflicts/access
Index Accesses,0.0,0.0,0.0,Cycles per wave
Dispatch_ID Accesses,0.0,0.0,0.0,Cycles per wave
Atomic Cycles,0.0,0.0,0.0,Cycles per wave
Bank Conflict,0.0,0.0,0.0,Cycles per wave
Addr Conflict,0.0,0.0,0.0,Cycles per wave
Unaligned Stall,0.0,0.0,0.0,Cycles per wave
Mem Violations,0.0,0.0,0.0, per wave
LDS Latency,,,,Cycles
LDS_Per_Workgroup Latency,,,,Cycles
1 Metric Avg Min Max Unit
2 Wave Cycles 18296.269474052384 1894.2776184082031 258229.23516845703 Cycles/wave
3 LDS Instrs LDS_Per_Workgroup Instrs 0.0 0.0 0.0 Instr per wave
4 Bandwidth 0.0 0.0 0.0 Bytes per wave
5 Bank Conficts/Access Conflicts/access
6 Index Accesses Dispatch_ID Accesses 0.0 0.0 0.0 Cycles per wave
7 Atomic Cycles 0.0 0.0 0.0 Cycles per wave
8 Bank Conflict 0.0 0.0 0.0 Cycles per wave
9 Addr Conflict 0.0 0.0 0.0 Cycles per wave
10 Unaligned Stall 0.0 0.0 0.0 Cycles per wave
11 Mem Violations 0.0 0.0 0.0 per wave
12 LDS Latency LDS_Per_Workgroup Latency Cycles
tests/workloads/invdev/mi200/prev_analysis/1901.csv
+5 -5
View File
@@ -6,16 +6,16 @@ SMEM,2.0,smem_
VALU,495.0,valu_
MFMA,0.0,mfma_
VMEM,8.0,vmem_
LDS,0.0,lds_
LDS_Per_Workgroup,0.0,lds_
GWS,0.0,gws_
BR,10.0,br_
VGPR,12.0,vgpr_
SGPR,24.0,sgpr_
LDS Allocation,0.0,lds_alloc_
LDS_Per_Workgroup Allocation,0.0,lds_alloc_
Scratch Allocation,0.0,scratch_alloc_
Wavefronts,67894.0,wavefronts_
Workgroups,16973.0,workgroups_
LDS Req,0.0,lds_req_
LDS_Per_Workgroup Req,0.0,lds_req_
IL1 Fetch,128.0,il1_fetch_
IL1 Hit,100.0,il1_hit_
IL1_L2 Rd,0.0,il1_l2_req_
@@ -46,10 +46,10 @@ Fabric_L2 Wr,0.0,l2_fabric_wr_
Fabric_l2 Atomic,0.0,l2_fabric_atom_
HBM Rd,44.0,hbm_rd_
HBM Wr,0.0,hbm_wr_
LDS Util,0.0,lds_util_
LDS_Per_Workgroup Util,0.0,lds_util_
VL1 Coalesce,70.0,vl1_coales_
VL1 Stall,2.0,vl1_stall_
LDS Lat,,lds_lat_
LDS_Per_Workgroup Lat,,lds_lat_
vL1D Lat,,sl1_lat_
IL1 Lat,,il1_lat_
Wave Occupancy,25.0,wave_occ_
1 Metric Value Alias
6 VALU 495.0 valu_
7 MFMA 0.0 mfma_
8 VMEM 8.0 vmem_
9 LDS LDS_Per_Workgroup 0.0 lds_
10 GWS 0.0 gws_
11 BR 10.0 br_
12 VGPR 12.0 vgpr_
13 SGPR 24.0 sgpr_
14 LDS Allocation LDS_Per_Workgroup Allocation 0.0 lds_alloc_
15 Scratch Allocation 0.0 scratch_alloc_
16 Wavefronts 67894.0 wavefronts_
17 Workgroups 16973.0 workgroups_
18 LDS Req LDS_Per_Workgroup Req 0.0 lds_req_
19 IL1 Fetch 128.0 il1_fetch_
20 IL1 Hit 100.0 il1_hit_
21 IL1_L2 Rd 0.0 il1_l2_req_
46 Fabric_l2 Atomic 0.0 l2_fabric_atom_
47 HBM Rd 44.0 hbm_rd_
48 HBM Wr 0.0 hbm_wr_
49 LDS Util LDS_Per_Workgroup Util 0.0 lds_util_
50 VL1 Coalesce 70.0 vl1_coales_
51 VL1 Stall 2.0 vl1_stall_
52 LDS Lat LDS_Per_Workgroup Lat lds_lat_
53 vL1D Lat sl1_lat_
54 IL1 Lat il1_lat_
55 Wave Occupancy 25.0 wave_occ_
tests/workloads/invdev/mi200/prev_analysis/2001.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id
Dispatch_ID,Kernel_Name,GPU_ID
0,__amd_rocclr_fillBufferAligned.kd,0
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID
2 0 __amd_rocclr_fillBufferAligned.kd 0
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0
tests/workloads/invdev/mi200/prev_analysis/201.csv
+2 -2
View File
@@ -12,8 +12,8 @@ VALU Util,58.3677767471036,Pct,100,58.3677767471036
MFMA Util,0.0,Pct,100,0.0
VALU Active Threads/Wave,64.0,Threads,64,100.0
IPC - Issue,1.0,Instr/cycle,5,20.0
LDS BW,0.0,Gb/sec,22630.4,0.0
LDS Bank Conflict,,Conflicts/access,32,
LDS_Per_Workgroup BW,0.0,Gb/sec,22630.4,0.0
LDS_Per_Workgroup Bank Conflict,,Conflicts/access,32,
Instr Cache Hit Rate,99.99259859947313,Pct,100,99.99259859947313
Instr Cache BW,1671.978257589768,Gb/s,6092.8,27.441870036596768
Scalar L1D Cache Hit Rate,99.3485588612334,Pct,100,99.3485588612334
1 Metric Value Unit Peak PoP
12 MFMA Util 0.0 Pct 100 0.0
13 VALU Active Threads/Wave 64.0 Threads 64 100.0
14 IPC - Issue 1.0 Instr/cycle 5 20.0
15 LDS BW LDS_Per_Workgroup BW 0.0 Gb/sec 22630.4 0.0
16 LDS Bank Conflict LDS_Per_Workgroup Bank Conflict Conflicts/access 32
17 Instr Cache Hit Rate 99.99259859947313 Pct 100 99.99259859947313
18 Instr Cache BW 1671.978257589768 Gb/s 6092.8 27.441870036596768
19 Scalar L1D Cache Hit Rate 99.3485588612334 Pct 100 99.3485588612334
tests/workloads/invdev/mi200/prev_analysis/602.csv
+1 -1
View File
@@ -6,7 +6,7 @@ Scratch Stall,0.0,0,0,Cycles
Insufficient SIMD Waveslots,12526180.185628742,0,370304750,Simd
Insufficient SIMD VGPRs,0.0,0,0,Simd
Insufficient SIMD SGPRs,0.0,0,0,Simd
Insufficient CU LDS,0.0,0,0,Cu
Insufficient CU LDS_Per_Workgroup,0.0,0,0,Cu
Insufficient CU Barries,0.0,0,0,Cu
Insufficient Bulky Resource,0.0,0,0,Cu
Reach CU Threadgroups Limit,0.0,0,0,Cycles
1 Metric Avg Min Max Unit
6 Insufficient SIMD Waveslots 12526180.185628742 0 370304750 Simd
7 Insufficient SIMD VGPRs 0.0 0 0 Simd
8 Insufficient SIMD SGPRs 0.0 0 0 Simd
9 Insufficient CU LDS Insufficient CU LDS_Per_Workgroup 0.0 0 0 Cu
10 Insufficient CU Barries 0.0 0 0 Cu
11 Insufficient Bulky Resource 0.0 0 0 Cu
12 Reach CU Threadgroups Limit 0.0 0 0 Cycles
tests/workloads/invdev/mi200/prev_analysis/701.csv
+1 -1
View File
@@ -6,5 +6,5 @@ Saved Wavefronts,0.0,0,0,Wavefronts
Restored Wavefronts,0.0,0,0,Wavefronts
VGPRs,11.784431137724551,4,16,Registers
SGPRs,24.047904191616766,24,32,Registers
LDS Allocation,0.0,0,0,Bytes
LDS_Per_Workgroup Allocation,0.0,0,0,Bytes
Scratch Allocation,0.0,0,0,Bytes
1 Metric Avg Min Max Unit
6 Restored Wavefronts 0.0 0 0 Wavefronts
7 VGPRs 11.784431137724551 4 16 Registers
8 SGPRs 24.047904191616766 24 32 Registers
9 LDS Allocation LDS_Per_Workgroup Allocation 0.0 0 0 Bytes
10 Scratch Allocation 0.0 0 0 Bytes
tests/workloads/invdev/mi200/sysinfo.csv
+1 -1
View File
@@ -1,2 +1,2 @@
workload_name,host_name,host_cpu,host_distro,host_kernel,host_rocmver,date,gpu_soc,numSE,numCU,numSIMD,waveSize,maxWavesPerCU,maxWorkgroupSize,L1,L2,sclk,mclk,cur_sclk,cur_mclk,L2Banks,name,numSQC,hbmBW,ip_blocks
invdev,06fa5f860366,AMD EPYC 7282 16-Core Processor,Ubuntu 20.04.5 LTS,5.11.0-27-generic,5.1.3-66,Wed Sep 28 20:50:32 2022 (),gfx90a,8,104,4,64,32,1024,16,8192,1700,1600,800,1600,32,mi200,56,1638.4,roofline|SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
invdev,06fa5f860366,AMD EPYC 7282 16-Core Processor,Ubuntu 20.04.5 LTS,5.11.0-27-generic,5.1.3-66,Wed Sep 28 20:50:32 2022 (),gfx90a,8,104,4,64,32,1024,16,8192,1700,1600,800,1600,32,mi200,56,1638.4,roofline|SQ|LDS_Per_Workgroup|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
1 workload_name host_name host_cpu host_distro host_kernel host_rocmver date gpu_soc numSE numCU numSIMD waveSize maxWavesPerCU maxWorkgroupSize L1 L2 sclk mclk cur_sclk cur_mclk L2Banks name numSQC hbmBW ip_blocks
2 invdev 06fa5f860366 AMD EPYC 7282 16-Core Processor Ubuntu 20.04.5 LTS 5.11.0-27-generic 5.1.3-66 Wed Sep 28 20:50:32 2022 () gfx90a 8 104 4 64 32 1024 16 8192 1700 1600 800 1600 32 mi200 56 1638.4 roofline|SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF roofline|SQ|LDS_Per_Workgroup|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
tests/workloads/invdev/mi200/timestamps.csv
+1 -1
View File
@@ -1,4 +1,4 @@
Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,DispatchNs,BeginNs,EndNs,CompleteNs
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,grd,Workgroup_Size,LDS_Per_Workgroup,scr,vgpr,SGPR,fbar,sig,obj,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBufferAligned.kd",0,0,0,238434,238434,33554432,256,0,0,4,32,4160,0x0,0x7fd647c04280,17853365225848,17853365250419,17853365490099,17853365576439
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,238434,238434,32768,256,0,0,12,24,13888,0x0,0x7fd647c23f80,17853370260633,17853370276021,17853370289461,17853370306822
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,5,238434,238434,4194304,256,0,0,12,24,14336,0x7fd66e582380,0x7fd647c23fc0,17853370311482,17853370360501,17853370452981,17853370455238
1 Index Dispatch_ID KernelName Kernel_Name gpu-id GPU_ID queue-id queue-index pid tid grd wgr Workgroup_Size lds LDS_Per_Workgroup scr vgpr sgpr SGPR fbar sig obj DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBufferAligned.kd 0 0 0 238434 238434 33554432 256 0 0 4 32 4160 0x0 0x7fd647c04280 17853365225848 17853365250419 17853365490099 17853365576439
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 238434 238434 32768 256 0 0 12 24 13888 0x0 0x7fd647c23f80 17853370260633 17853370276021 17853370289461 17853370306822
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 5 238434 238434 4194304 256 0 0 12 24 14336 0x7fd66e582380 0x7fd647c23fc0 17853370311482 17853370360501 17853370452981 17853370455238