Add 'projects/rocprofiler-compute/' from commit 'd2cec001161fc49761bd71a498474a447b1d6975'
git-subtree-dir: projects/rocprofiler-compute git-subtree-mainline:8a4d7262f8git-subtree-split:d2cec00116
This commit is contained in:
systems-assistant[bot]
BIN
Binary file not shown.
BIN
Binary file not shown.
Binary file not shown.
@@ -0,0 +1,132 @@
|
||||
[roofline] Generating pmc_perf.csv (roofline counters only).
|
||||
Omniperf version: 2.0.0-RC1
|
||||
Profiler choice: rocprofv1
|
||||
Path: /home1/josantos/omniperf/tests/workloads/kernel_names/MI200
|
||||
Target: MI200
|
||||
Command: ./tests/vcopy -n 1048576 -b 256 -i 3
|
||||
Kernel Selection: None
|
||||
Dispatch Selection: None
|
||||
IP Blocks: All
|
||||
|
||||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||||
Collecting Performance Counters
|
||||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||||
|
||||
[profiling] Current input file: tests/workloads/kernel_names/MI200/perfmon/pmc_perf_0.txt
|
||||
|-> [rocprof] RPL: on '240321_162548' from '/opt/rocm-6.0.2' in '/home1/josantos/omniperf'
|
||||
|-> [rocprof] RPL: profiling '""./tests/vcopy -n 1048576 -b 256 -i 3""'
|
||||
|-> [rocprof] RPL: input file 'tests/workloads/kernel_names/MI200/perfmon/pmc_perf_0.txt'
|
||||
|-> [rocprof] RPL: output dir '/tmp/rpl_data_240321_162548_4134506'
|
||||
|-> [rocprof] RPL: result dir '/tmp/rpl_data_240321_162548_4134506/input0_results_240321_162548'
|
||||
|-> [rocprof] ROCProfiler: input from "/tmp/rpl_data_240321_162548_4134506/input0.xml"
|
||||
|-> [rocprof] gpu_index =
|
||||
|-> [rocprof] kernel =
|
||||
|-> [rocprof] range =
|
||||
|-> [rocprof] 16 metrics
|
||||
|-> [rocprof] 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
|
||||
|-> [rocprof] vcopy testing on GCD 0
|
||||
|-> [rocprof] Finished allocating vectors on the CPU
|
||||
|-> [rocprof] Finished allocating vectors on the GPU
|
||||
|-> [rocprof] Finished copying vectors to the GPU
|
||||
|-> [rocprof] sw thinks it moved 1.000000 KB per wave
|
||||
|-> [rocprof] Total threads: 1048576, Grid Size: 4096 block Size:256, Wavefronts:16384:
|
||||
|-> [rocprof] Launching the kernel on the GPU
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished copying the output vector from the GPU to the CPU
|
||||
|-> [rocprof] Releasing GPU memory
|
||||
|-> [rocprof] Releasing CPU memory
|
||||
|-> [rocprof]
|
||||
|-> [rocprof] ROCPRofiler: 3 contexts collected, output directory /tmp/rpl_data_240321_162548_4134506/input0_results_240321_162548
|
||||
|-> [rocprof] File 'tests/workloads/kernel_names/MI200/pmc_perf_0.csv' is generating
|
||||
|-> [rocprof]
|
||||
[profiling] Current input file: tests/workloads/kernel_names/MI200/perfmon/pmc_perf_1.txt
|
||||
|-> [rocprof] RPL: on '240321_162548' from '/opt/rocm-6.0.2' in '/home1/josantos/omniperf'
|
||||
|-> [rocprof] RPL: profiling '""./tests/vcopy -n 1048576 -b 256 -i 3""'
|
||||
|-> [rocprof] RPL: input file 'tests/workloads/kernel_names/MI200/perfmon/pmc_perf_1.txt'
|
||||
|-> [rocprof] RPL: output dir '/tmp/rpl_data_240321_162548_4134693'
|
||||
|-> [rocprof] RPL: result dir '/tmp/rpl_data_240321_162548_4134693/input0_results_240321_162548'
|
||||
|-> [rocprof] ROCProfiler: input from "/tmp/rpl_data_240321_162548_4134693/input0.xml"
|
||||
|-> [rocprof] gpu_index =
|
||||
|-> [rocprof] kernel =
|
||||
|-> [rocprof] range =
|
||||
|-> [rocprof] 9 metrics
|
||||
|-> [rocprof] 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
|
||||
|-> [rocprof] vcopy testing on GCD 0
|
||||
|-> [rocprof] Finished allocating vectors on the CPU
|
||||
|-> [rocprof] Finished allocating vectors on the GPU
|
||||
|-> [rocprof] Finished copying vectors to the GPU
|
||||
|-> [rocprof] sw thinks it moved 1.000000 KB per wave
|
||||
|-> [rocprof] Total threads: 1048576, Grid Size: 4096 block Size:256, Wavefronts:16384:
|
||||
|-> [rocprof] Launching the kernel on the GPU
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished copying the output vector from the GPU to the CPU
|
||||
|-> [rocprof] Releasing GPU memory
|
||||
|-> [rocprof] Releasing CPU memory
|
||||
|-> [rocprof]
|
||||
|-> [rocprof] ROCPRofiler: 3 contexts collected, output directory /tmp/rpl_data_240321_162548_4134693/input0_results_240321_162548
|
||||
|-> [rocprof] File 'tests/workloads/kernel_names/MI200/pmc_perf_1.csv' is generating
|
||||
|-> [rocprof]
|
||||
[profiling] Current input file: tests/workloads/kernel_names/MI200/perfmon/pmc_perf_2.txt
|
||||
|-> [rocprof] RPL: on '240321_162549' from '/opt/rocm-6.0.2' in '/home1/josantos/omniperf'
|
||||
|-> [rocprof] RPL: profiling '""./tests/vcopy -n 1048576 -b 256 -i 3""'
|
||||
|-> [rocprof] RPL: input file 'tests/workloads/kernel_names/MI200/perfmon/pmc_perf_2.txt'
|
||||
|-> [rocprof] RPL: output dir '/tmp/rpl_data_240321_162549_4134881'
|
||||
|-> [rocprof] RPL: result dir '/tmp/rpl_data_240321_162549_4134881/input0_results_240321_162549'
|
||||
|-> [rocprof] ROCProfiler: input from "/tmp/rpl_data_240321_162549_4134881/input0.xml"
|
||||
|-> [rocprof] gpu_index =
|
||||
|-> [rocprof] kernel =
|
||||
|-> [rocprof] range =
|
||||
|-> [rocprof] 2 metrics
|
||||
|-> [rocprof] SQ_LDS_IDX_ACTIVE, SQ_LDS_BANK_CONFLICT
|
||||
|-> [rocprof] vcopy testing on GCD 0
|
||||
|-> [rocprof] Finished allocating vectors on the CPU
|
||||
|-> [rocprof] Finished allocating vectors on the GPU
|
||||
|-> [rocprof] Finished copying vectors to the GPU
|
||||
|-> [rocprof] sw thinks it moved 1.000000 KB per wave
|
||||
|-> [rocprof] Total threads: 1048576, Grid Size: 4096 block Size:256, Wavefronts:16384:
|
||||
|-> [rocprof] Launching the kernel on the GPU
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished copying the output vector from the GPU to the CPU
|
||||
|-> [rocprof] Releasing GPU memory
|
||||
|-> [rocprof] Releasing CPU memory
|
||||
|-> [rocprof]
|
||||
|-> [rocprof] ROCPRofiler: 3 contexts collected, output directory /tmp/rpl_data_240321_162549_4134881/input0_results_240321_162549
|
||||
|-> [rocprof] File 'tests/workloads/kernel_names/MI200/pmc_perf_2.csv' is generating
|
||||
|-> [rocprof]
|
||||
[profiling] Current input file: tests/workloads/kernel_names/MI200/perfmon/timestamps.txt
|
||||
|-> [rocprof] RPL: on '240321_162549' from '/opt/rocm-6.0.2' in '/home1/josantos/omniperf'
|
||||
|-> [rocprof] RPL: profiling '""./tests/vcopy -n 1048576 -b 256 -i 3""'
|
||||
|-> [rocprof] RPL: input file 'tests/workloads/kernel_names/MI200/perfmon/timestamps.txt'
|
||||
|-> [rocprof] RPL: output dir '/tmp/rpl_data_240321_162549_4135068'
|
||||
|-> [rocprof] RPL: result dir '/tmp/rpl_data_240321_162549_4135068/input0_results_240321_162549'
|
||||
|-> [rocprof] ROCProfiler: input from "/tmp/rpl_data_240321_162549_4135068/input0.xml"
|
||||
|-> [rocprof] gpu_index =
|
||||
|-> [rocprof] kernel =
|
||||
|-> [rocprof] range =
|
||||
|-> [rocprof] 0 metrics
|
||||
|-> [rocprof] vcopy testing on GCD 0
|
||||
|-> [rocprof] Finished allocating vectors on the CPU
|
||||
|-> [rocprof] Finished allocating vectors on the GPU
|
||||
|-> [rocprof] Finished copying vectors to the GPU
|
||||
|-> [rocprof] sw thinks it moved 1.000000 KB per wave
|
||||
|-> [rocprof] Total threads: 1048576, Grid Size: 4096 block Size:256, Wavefronts:16384:
|
||||
|-> [rocprof] Launching the kernel on the GPU
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished executing kernel
|
||||
|-> [rocprof] Finished copying the output vector from the GPU to the CPU
|
||||
|-> [rocprof] Releasing GPU memory
|
||||
|-> [rocprof] Releasing CPU memory
|
||||
|-> [rocprof]
|
||||
|-> [rocprof] ROCPRofiler: 3 contexts collected, output directory /tmp/rpl_data_240321_162549_4135068/input0_results_240321_162549
|
||||
|-> [rocprof] File 'tests/workloads/kernel_names/MI200/timestamps.csv' is generating
|
||||
|-> [rocprof]
|
||||
[roofline] Checking for roofline.csv in tests/workloads/kernel_names/MI200
|
||||
[roofline] No roofline data found. Generating...
|
||||
[roofline] Empirical Roofline PDFs saved!
|
||||
@@ -0,0 +1,5 @@
|
||||
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 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
|
||||
|
||||
gpu:
|
||||
range:
|
||||
kernel:
|
||||
@@ -0,0 +1,5 @@
|
||||
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_ATOMIC_WITHOUT_RET_REQ_sum
|
||||
|
||||
gpu:
|
||||
range:
|
||||
kernel:
|
||||
@@ -0,0 +1,5 @@
|
||||
pmc: SQ_LDS_IDX_ACTIVE SQ_LDS_BANK_CONFLICT
|
||||
|
||||
gpu:
|
||||
range:
|
||||
kernel:
|
||||
@@ -0,0 +1,5 @@
|
||||
pmc:
|
||||
|
||||
gpu:
|
||||
range:
|
||||
kernel:
|
||||
@@ -0,0 +1,4 @@
|
||||
Dispatch_ID,Kernel_Name,GPU_ID
|
||||
0,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2
|
||||
1,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2
|
||||
2,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2
|
||||
|
@@ -0,0 +1,4 @@
|
||||
Dispatch_ID,Kernel_Name,GPU_ID,Grid_Size,Workgroup_Size,LDS_Per_Workgroup,Scratch_Per_Workitem,Arch_VGPR,Accum_VGPR,SGPR,wave_size,obj,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,wave_size_1,obj_1,SQ_LDS_IDX_ACTIVE,SQ_LDS_BANK_CONFLICT,wave_size_2,obj_2,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,Start_Timestamp,End_Timestamp
|
||||
0,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,1048576,256,0,0,8,0,16,64,0x7f79a6568ec0,0,0,0,0,0,0,0,0,0.0,64,0x7f964ade8ec0,0,0,64,0x7f8ceb5a8ec0,0,0,0,0,0,0,0,0,131072.0,524288.0,131072.0,0.0,0.0,131080.0,131072.0,131072.0,1412081937504773,1412081937524933
|
||||
1,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,1048576,256,0,0,8,0,16,64,0x7f79a6568ec0,0,0,0,0,0,0,0,0,0.0,64,0x7f964ade8ec0,0,0,64,0x7f8ceb5a8ec0,0,0,0,0,0,0,0,0,131072.0,524288.0,131072.0,0.0,0.0,131073.0,91142.0,91142.0,1412081937544293,1412081937559973
|
||||
2,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,1048576,256,0,0,8,0,16,64,0x7f79a6568ec0,0,0,0,0,0,0,0,0,0.0,64,0x7f964ade8ec0,0,0,64,0x7f8ceb5a8ec0,0,0,0,0,0,0,0,0,131072.0,524288.0,131072.0,0.0,0.0,131081.0,91346.0,91346.0,1412081937618533,1412081937635173
|
||||
|
@@ -0,0 +1,5 @@
|
||||
device,HBMBw,HBMBwLow,hbmBwHigh,L2Bw,L2BwLow,L2BwHigh,L1Bw,L1BwLow,L1BwHigh,LDSBw,LDSBwLow,LDSBwHigh,FP32Flops,FP32FlopsLow,FP32FlopsHigh,FP64Flops,FP64FlopsLow,FP64FlopsHigh,MFMABF16Flops,MFMABF16FlopsLow,MFMABF16FlopsHigh,MFMAF16Flops,MFMAF16FlopsLow,MFMAF16FlopsHigh,MFMAF32Flops,MFMAF32FlopsLow,MFMAF32FlopsHigh,MFMAF64Flops,MFMAF64FlopsLow,MFMAF64FlopsHigh,MFMAI8Ops,MFMAFI8OpsLow,MFMAI8OpsHigh
|
||||
0,1388.9282,1388.3175,1389.5389,5017.3882,5014.4512,5020.3252,9227.0234,9226.3682,9227.6787,17716.459,17713.33,17719.588,20945.281,20884.613,21005.949,20273.305,20272.646,20273.963,170494.08,170490.92,170497.23,164927.81,164923.56,164932.06,41438.559,41437.895,41439.223,41493.992,41493.492,41494.492,166407.34,165915.83,166898.86
|
||||
1,1388.9865,1388.4292,1389.5437,5029.7632,5027.7769,5031.7495,9237.8516,9237.1484,9238.5547,18239.699,18237.943,18241.455,20986.152,20919.436,21052.869,20301.568,20300.916,20302.221,170908.56,170904.03,170913.09,165157.95,165153.84,165162.06,41499.078,41498.238,41499.918,41553.703,41552.781,41554.625,166922.34,166918.17,166926.52
|
||||
2,1388.8414,1388.2102,1389.4727,5035.4419,5033.2456,5037.6382,9262.9658,9262.335,9263.5967,18220.949,18220.014,18221.885,21057.277,21056.908,21057.646,20353.641,20352.898,20354.383,171148.84,171144.11,171153.58,165526.91,165522.73,165531.08,41582.828,41582.098,41583.559,41640.613,41639.422,41641.805,166980.31,166387.7,167572.92
|
||||
3,1389.2573,1388.7434,1389.7712,5032.5112,5030.8784,5034.144,9235.8096,9235.0342,9236.585,18324.967,18324.154,18325.779,21029.967,21029.445,21030.488,20304.945,20304.5,20305.391,171032.27,171027.84,171036.69,165140.88,165137.02,165144.73,41421.703,41268.262,41575.145,41566.469,41565.211,41567.727,166934.39,166929.88,166938.91
|
||||
|
@@ -0,0 +1,2 @@
|
||||
workload_name,command,ip_blocks,timestamp,version,hostname,cpu_model,sbios,linux_distro,linux_kernel_version,amd_gpu_kernel_version,cpu_memory,gpu_memory,rocm_version,vbios,compute_partition,memory_partition,gpu_model,gpu_arch,gpu_l1,gpu_l2,cu_per_gpu,simd_per_cu,se_per_gpu,wave_size,workgroup_max_size,max_waves_per_cu,max_sclk,max_mclk,cur_sclk,cur_mclk,total_l2_chan,lds_banks_per_cu,sqc_per_gpu,pipes_per_gpu,hbm_bw,num_xcd,num_hbm_channels
|
||||
path,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF|roofline,Thu 21 Mar 2024 04:16:46 PM (CDT),2,t007-002.hpcfund,AMD EPYC 7V13 64-Core Processor,American Megatrends Inc.0602,Rocky Linux 9.1 (Blue Onyx),5.14.0-162.18.1.el9_1.x86_64,,527650760,,6.0.2-115,113-D67301-059,NA,NA,MI200,gfx90a,16,8192,104,4,8,64,1024,32,1700,1600,1700,1600,32,32,56,4,1638.4,1,32
|
||||
|
@@ -0,0 +1,4 @@
|
||||
Dispatch_ID,Kernel_Name,GPU_ID,queue-id,queue-index,pid,tid,Grid_Size,Workgroup_Size,LDS_Per_Workgroup,Scratch_Per_Workitem,Arch_VGPR,Accum_VGPR,SGPR,wave_size,sig,obj,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
|
||||
0,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,0,4135228,4135228,1048576,256,0,0,8,0,16,64,0x0,0x7f765deb0ec0,1412081937480123,1412081937504773,1412081937524933,1412081937540988
|
||||
1,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,2,4135228,4135228,1048576,256,0,0,8,0,16,64,0x0,0x7f765deb0ec0,1412081937536329,1412081937544293,1412081937559973,1412081937615799
|
||||
2,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,4,4135228,4135228,1048576,256,0,0,8,0,16,64,0x0,0x7f765deb0ec0,1412081937570474,1412081937618533,1412081937635173,1412081937636388
|
||||
|
Reference in New Issue
Block a user