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>


[ROCm/rocprofiler-compute commit: 717a21cf84]
This commit is contained in:
JoseSantosAMD
2024-02-02 20:04:11 -06:00
committed by Karl W. Schulz
parent 595ca9141a
commit 8ecc6dff7f
2842 changed files with 5653 additions and 3927 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler-compute/tests/workloads/roof_only_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
projects/rocprofiler-compute/tests/workloads/roof_only_invdev/mi200/pmc_perf.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_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,TCP_TCC_READ_REQ_sum,TCP_TOTAL_CACHE_ACCESSES_sum,TCP_TCC_WRITE_REQ_sum,TCP_TCC_ATOMIC_WITH_RET_REQ_sum,TCC_EA_RDREQ_32B_sum,TCC_EA_RDREQ_sum,TCC_EA_WRREQ_64B_sum,TCC_EA_WRREQ_sum,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_ATOMIC_WITHOUT_RET_REQ_sum,SQ_LDS_IDX_ACTIVE,SQ_LDS_BANK_CONFLICT,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_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,TCP_TCC_READ_REQ_sum,TCP_TOTAL_CACHE_ACCESSES_sum,TCP_TCC_WRITE_REQ_sum,TCP_TCC_ATOMIC_WITH_RET_REQ_sum,TCC_EA_RDREQ_32B_sum,TCC_EA_RDREQ_sum,TCC_EA_WRREQ_64B_sum,TCC_EA_WRREQ_sum,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_ATOMIC_WITHOUT_RET_REQ_sum,SQ_LDS_IDX_ACTIVE,SQ_LDS_BANK_CONFLICT,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
0,"__amd_rocclr_fillBufferAligned.kd",0,0,0,241894,241894,33554432,256,0,0,4,32,4160,0x0,0x7f6e9cc04280,0,0,0,0,0,0,0,0,0.0000000000,8388608.0000000000,4194304.0000000000,0.0000000000,0.0000000000,310.0000000000,4128768.0000000000,4128768.0000000000,0,0,0,0,0,0,0,0,0.0000000000,0,0,17931286038190,17901582538237,17931431845031,17931431958331
1,"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",0,0,2,241894,241894,32768,256,0,0,12,24,13888,0x0,0x7f6e9cc23f80,0,0,0,0,0,0,0,0,8192.0000000000,16384.0000000000,0.0000000000,0.0000000000,0.0000000000,8262.0000000000,0.0000000000,0.0000000000,0,0,0,0,0,0,0,0,0.0000000000,0,0,17931437652391,17931431845031,17931438008283,17931438016542
2,"void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd]",0,0,5,241894,241894,4194304,256,0,0,12,24,14336,0x7f6e9fb92380,0x7f6e9cc23fc0,0,0,0,0,0,0,262144,0,2097152.0000000000,2097152.0000000000,0.0000000000,0.0000000000,0.0000000000,2097343.0000000000,0.0000000000,0.0000000000,0,0,0,0,0,0,0,0,0.0000000000,0,0,17931438610067,17931438008283,17931439296934,17931439297600
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_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 TCP_TCC_READ_REQ_sum TCP_TOTAL_CACHE_ACCESSES_sum TCP_TCC_WRITE_REQ_sum TCP_TCC_ATOMIC_WITH_RET_REQ_sum TCC_EA_RDREQ_32B_sum TCC_EA_RDREQ_sum TCC_EA_WRREQ_64B_sum TCC_EA_WRREQ_sum 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_ATOMIC_WITHOUT_RET_REQ_sum SQ_LDS_IDX_ACTIVE SQ_LDS_BANK_CONFLICT DispatchNs BeginNs Start_Timestamp EndNs End_Timestamp CompleteNs
2 0 __amd_rocclr_fillBufferAligned.kd 0 0 0 241894 241894 33554432 256 0 0 4 32 4160 0x0 0x7f6e9cc04280 0 0 0 0 0 0 0 0 0.0000000000 8388608.0000000000 4194304.0000000000 0.0000000000 0.0000000000 310.0000000000 4128768.0000000000 4128768.0000000000 0 0 0 0 0 0 0 0 0.0000000000 0 0 17931286038190 17901582538237 17931431845031 17931431958331
3 1 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 0 0 2 241894 241894 32768 256 0 0 12 24 13888 0x0 0x7f6e9cc23f80 0 0 0 0 0 0 0 0 8192.0000000000 16384.0000000000 0.0000000000 0.0000000000 0.0000000000 8262.0000000000 0.0000000000 0.0000000000 0 0 0 0 0 0 0 0 0.0000000000 0 0 17931437652391 17931431845031 17931438008283 17931438016542
4 2 void benchmark_func<float, 256, 8u, 0u>(float, float*) [clone .kd] 0 0 5 241894 241894 4194304 256 0 0 12 24 14336 0x7f6e9fb92380 0x7f6e9cc23fc0 0 0 0 0 0 0 262144 0 2097152.0000000000 2097152.0000000000 0.0000000000 0.0000000000 0.0000000000 2097343.0000000000 0.0000000000 0.0000000000 0 0 0 0 0 0 0 0 0.0000000000 0 0 17931438610067 17931438008283 17931439296934 17931439297600
projects/rocprofiler-compute/tests/workloads/roof_only_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
__amd_rocclr_fillBufferAligned.kd,1,29849306794.0,29849306794.0,29849306794.0,99.23161614424411
"void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd]",1,6163252.0,6163252.0,6163252.0,0.0204892348383507
"void benchmark_func<int, 256, 8u, 512u>(int, int*) [clone .kd]",1,4454736.0,4454736.0,4454736.0,0.0148094110133506
1 KernelName Kernel_Name Count Sum(ns) Mean(ns) Median(ns) Pct
2 __amd_rocclr_fillBufferAligned.kd 1 29849306794.0 29849306794.0 29849306794.0 99.23161614424411
3 void benchmark_func<short, 256, 8u, 0u>(short, short*) [clone .kd] 1 6163252.0 6163252.0 6163252.0 0.0204892348383507
4 void benchmark_func<int, 256, 8u, 512u>(int, int*) [clone .kd] 1 4454736.0 4454736.0 4454736.0 0.0148094110133506
projects/rocprofiler-compute/tests/workloads/roof_only_invdev/mi200/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,,Instr per wave
LDS_Per_Workgroup,,Instr per wave
VALU - MFMA,,Instr per wave
SALU,,Instr per wave
SMEM,,Instr per wave
1 Metric Count Unit
2 VALU - Vector Instr per wave
3 VMEM Instr per wave
4 LDS LDS_Per_Workgroup Instr per wave
5 VALU - MFMA Instr per wave
6 SALU Instr per wave
7 SMEM Instr per wave
projects/rocprofiler-compute/tests/workloads/roof_only_invdev/mi200/prev_analysis/1202.csv
+3 -3
View File
@@ -1,12 +1,12 @@
Metric,Avg,Min,Max,Unit
Wave Cycles,,,,Cycles/wave
LDS Instrs,,,,Instr per wave
LDS_Per_Workgroup Instrs,,,,Instr per wave
Bandwidth,,,,Bytes per wave
Bank Conficts/Access,,,,Conflicts/access
Index Accesses,,,,Cycles per wave
Dispatch_ID Accesses,,,,Cycles per wave
Atomic Cycles,,,,Cycles per wave
Bank Conflict,,,,Cycles per wave
Addr Conflict,,,,Cycles per wave
Unaligned Stall,,,,Cycles per wave
Mem Violations,,,, per wave
LDS Latency,,,,Cycles
LDS_Per_Workgroup Latency,,,,Cycles
1 Metric Avg Min Max Unit
2 Wave Cycles Cycles/wave
3 LDS Instrs LDS_Per_Workgroup Instrs Instr per wave
4 Bandwidth Bytes per wave
5 Bank Conficts/Access Conflicts/access
6 Index Accesses Dispatch_ID Accesses Cycles per wave
7 Atomic Cycles Cycles per wave
8 Bank Conflict Cycles per wave
9 Addr Conflict Cycles per wave
10 Unaligned Stall Cycles per wave
11 Mem Violations per wave
12 LDS Latency LDS_Per_Workgroup Latency Cycles
projects/rocprofiler-compute/tests/workloads/roof_only_invdev/mi200/prev_analysis/1901.csv
+5 -5
View File
@@ -6,16 +6,16 @@ SMEM,,smem_
VALU,,valu_
MFMA,,mfma_
VMEM,,vmem_
LDS,,lds_
LDS_Per_Workgroup,,lds_
GWS,,gws_
BR,,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,,wavefronts_
Workgroups,,workgroups_
LDS Req,,lds_req_
LDS_Per_Workgroup Req,,lds_req_
IL1 Fetch,,il1_fetch_
IL1 Hit,,il1_hit_
IL1_L2 Rd,,il1_l2_req_
@@ -46,10 +46,10 @@ Fabric_L2 Wr,,l2_fabric_wr_
Fabric_l2 Atomic,,l2_fabric_atom_
HBM Rd,,hbm_rd_
HBM Wr,,hbm_wr_
LDS Util,,lds_util_
LDS_Per_Workgroup Util,,lds_util_
VL1 Coalesce,,vl1_coales_
VL1 Stall,,vl1_stall_
LDS Lat,,lds_lat_
LDS_Per_Workgroup Lat,,lds_lat_
vL1D Lat,,sl1_lat_
IL1 Lat,,il1_lat_
Wave Occupancy,,wave_occ_
1 Metric Value Alias
6 VALU valu_
7 MFMA mfma_
8 VMEM vmem_
9 LDS LDS_Per_Workgroup lds_
10 GWS gws_
11 BR 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 wavefronts_
17 Workgroups workgroups_
18 LDS Req LDS_Per_Workgroup Req lds_req_
19 IL1 Fetch il1_fetch_
20 IL1 Hit il1_hit_
21 IL1_L2 Rd il1_l2_req_
46 Fabric_l2 Atomic l2_fabric_atom_
47 HBM Rd hbm_rd_
48 HBM Wr hbm_wr_
49 LDS Util LDS_Per_Workgroup Util lds_util_
50 VL1 Coalesce vl1_coales_
51 VL1 Stall vl1_stall_
52 LDS Lat LDS_Per_Workgroup Lat lds_lat_
53 vL1D Lat sl1_lat_
54 IL1 Lat il1_lat_
55 Wave Occupancy wave_occ_
projects/rocprofiler-compute/tests/workloads/roof_only_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
projects/rocprofiler-compute/tests/workloads/roof_only_invdev/mi200/prev_analysis/201.csv
+2 -2
View File
@@ -12,8 +12,8 @@ VALU Util,,Pct,100,
MFMA Util,,Pct,100,
VALU Active Threads/Wave,,Threads,64,
IPC - Issue,,Instr/cycle,5,
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,,Pct,100,
Instr Cache BW,,Gb/s,6092.8,
Scalar L1D Cache Hit Rate,,Pct,100,
1 Metric Value Unit Peak PoP
12 MFMA Util Pct 100
13 VALU Active Threads/Wave Threads 64
14 IPC - Issue Instr/cycle 5
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 Pct 100
18 Instr Cache BW Gb/s 6092.8
19 Scalar L1D Cache Hit Rate Pct 100
projects/rocprofiler-compute/tests/workloads/roof_only_invdev/mi200/prev_analysis/602.csv
+1 -1
View File
@@ -6,7 +6,7 @@ Scratch Stall,,,,Cycles
Insufficient SIMD Waveslots,,,,Simd
Insufficient SIMD VGPRs,,,,Simd
Insufficient SIMD SGPRs,,,,Simd
Insufficient CU LDS,,,,Cu
Insufficient CU LDS_Per_Workgroup,,,,Cu
Insufficient CU Barries,,,,Cu
Insufficient Bulky Resource,,,,Cu
Reach CU Threadgroups Limit,,,,Cycles
1 Metric Avg Min Max Unit
6 Insufficient SIMD Waveslots Simd
7 Insufficient SIMD VGPRs Simd
8 Insufficient SIMD SGPRs Simd
9 Insufficient CU LDS Insufficient CU LDS_Per_Workgroup Cu
10 Insufficient CU Barries Cu
11 Insufficient Bulky Resource Cu
12 Reach CU Threadgroups Limit Cycles
projects/rocprofiler-compute/tests/workloads/roof_only_invdev/mi200/prev_analysis/701.csv
+1 -1
View File
@@ -6,5 +6,5 @@ Saved Wavefronts,,,,Wavefronts
Restored Wavefronts,,,,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 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