Update workloads (#290)

* updating workloads Signed-off-by: Jose Santos <josantos@amd.com> * updating workloads Signed-off-by: Jose Santos <josantos@amd.com> * Adding generate workloads script Signed-off-by: Jose Santos <josantos@amd.com> * Changing output workloads dir in shell script Signed-off-by: Jose Santos <josantos@amd.com> --------- Signed-off-by: Jose Santos <josantos@amd.com> [ROCm/rocprofiler-compute commit: 2d1f73ad11]
2024-03-01 17:12:44 -06:00
parent 5868974f74
commit 78db87b369
360 changed files with 958 additions and 828 deletions
@@ -0,0 +1,62 @@
+#!/bin/bash
+declare -A commands=(
+    [path]=' '
+    [no_roof]='--no-roof'
+    [kernel_names]='--roof-only --kernel-names'
+    [device_filter]='--device 0'
+    [kernel]='--kernel "vecCopy(double*, double*, double*, int, int) [clone .kd]"'
+    [kernel_summaries]='--kernel-summaries vcopy'
+    [ipblocks_SQ]='-b SQ'
+    [ipblocks_SQC]='-b SQC'
+    [ipblocks_TA]='-b TA'
+    [ipblocks_TD]='-b TD'
+    [ipblocks_TCP]='-b TCP'
+    [ipblocks_TCC]='-b TCC'
+    [ipblocks_SPI]='-b SPI'
+    [ipblocks_CPC]='-b CPC'
+    [ipblocks_CPF]='-b CPF'
+    [ipblocks_SQ_CPC]='-b SQ CPC'
+    [ipblocks_SQ_TA]='-b SQ TA'
+    [ipblocks_SQ_SPI]='-b SQ SPI'
+    [ipblocks_SQ_SQC_TCP_CPC]='-b SQ SQC TCP CPC'
+    [ipblocks_SQ_SPI_TA_TCC_CPF]='-b SQ SPI TA TCC CPF'
+    [dispatch_0]='--dispatch 0'
+    [dispatch_0_1]='--dispatch 0:2'
+    [dispatch_2]='--dispatch 1'
+    [kernel_verbose_0]='--kernel-verbose 0'
+    [kernel_verbose_1]='--kernel-verbose 1'
+    [kernel_verbose_2]='--kernel-verbose 2'
+    [kernel_verbose_3]='--kernel-verbose 3'
+    [kernel_verbose_4]='--kernel-verbose 4'
+    [kernel_verbose_5]='--kernel-verbose 5'
+    [join_type_grid]='--join-type grid'
+    [join_type_kernel]='--join-type kernel'
+    [sort_dispatches]='--roof-only --sort dispatches'
+    [sort_kernels]='--roof-only --sort kernels'
+    [mem_levels_HBM]='--roof-only --mem-level HBM'
+    [mem_levels_L2]='--roof-only --mem-level L2'
+    [mem_levels_vL1D]='--roof-only --mem-level vL1D'
+    [mem_levels_LDS]='--roof-only --mem-level LDS'
+    [mem_levels_HBM_LDS]='--roof-only --mem-level HBM LDS'
+    [mem_levels_vL1d_LDS]='--roof-only --mem-level vL1D LDS'
+    [mem_levels_L2_vL1d_LDS]='--roof-only --mem-level L2 vL1D LDS'
+    #########################################################
+    #           Attempt to break omniperf                   #
+    #########################################################
+    [dispatch_7]='--dispatch 7'
+    [dispatch_6_8]='--dispatch 6:8'
+    [dispatch_inv]='--dispatch invalid'
+    [kernel_substr]='--kernel vecCopy'
+    [kernel_inv_str]='--kernel vecPaste'
+    [kernel_inv_int]='--kernel 42'
+    [device_inv]='--device invalid' # does not generate a workload
+    [device_inv_int]='--device -1'
+    )
+soc=MI100
+echo "starting"
+for key in "${!commands[@]}"; do
+    echo profiling $key;
+    command="${commands[$key]}"
+    echo "$key = ./src/omniperf profile -n $key ${dirs[@]}"
+    ./src/omniperf profile -n $key $command -p tests/workloads/$key/$soc  -- ./tests/vcopy -n 1048576 -b 256 -i 3 ; 
+echo "done" ; done
@@ -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,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2869770,2869770,1048576,256,0,0,8,8,16,64,0x0,0x7eff6a3d2e80,48732,48732,16384,65536,12984,1684952,2187869534755189,2187863086843344,2187863086866864,2187869542260328
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2869770,2869770,1048576,256,0,0,8,8,16,64,0x0,0x7eff6a3d2e80,43698,43698,16384,65536,8101,1048580,2187869542283442,2187863086979504,2187863086998544,2187869542602320
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2869770,2869770,1048576,256,0,0,8,8,16,64,0x0,0x7eff6a3d2e80,42890,42890,16384,65536,8070,1048588,2187869542634450,2187863087020944,2187863087039824,2187869542818836
@@ -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,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2869952,2869952,1048576,256,0,0,8,8,16,64,0x0,0x7f72072bae80,0,0,0,2187869977668644,2187863086843344,2187863086866864,2187869985176257
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2869952,2869952,1048576,256,0,0,8,8,16,64,0x0,0x7f72072bae80,0,0,0,2187869985195062,2187863086979504,2187863086998544,2187869985509663
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2869952,2869952,1048576,256,0,0,8,8,16,64,0x0,0x7f72072bae80,0,0,0,2187869985538427,2187863087020944,2187863087039824,2187869985711863
@@ -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,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2867946,2867946,1048576,256,0,0,8,8,16,64,0x0,0x7fde64f1ae80,65536,193254,24880920,2187864659556927,2187863086843344,2187863086866864,2187864667075622
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2867946,2867946,1048576,256,0,0,8,8,16,64,0x0,0x7fde64f1ae80,65536,177776,22778168,2187864667094427,2187863086979504,2187863086998544,2187864667403628
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2867946,2867946,1048576,256,0,0,8,8,16,64,0x0,0x7fde64f1ae80,65536,177040,22689848,2187864667432131,2187863087020944,2187863087039824,2187864667600928
@@ -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,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2870680,2870680,1048576,256,0,0,8,8,16,64,0x0,0x7f9a920b2e80,32768,646946,82808524,2187872008481374,2187863086843344,2187863086866864,2187872016007002
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2870680,2870680,1048576,256,0,0,8,8,16,64,0x0,0x7f9a920b2e80,32768,665028,85108136,2187872016024845,2187863086979504,2187863086998544,2187872016349314
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2870680,2870680,1048576,256,0,0,8,8,16,64,0x0,0x7f9a920b2e80,32768,662924,84848200,2187872016378269,2187863087020944,2187863087039824,2187872016554119
@@ -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,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,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2870862,2870862,1048576,256,0,0,8,8,16,64,0x0,0x7febf83ace80,49355,49355,18227,394848,16384,25114743,232859,0,100948108,2187872453712459,2187863086843344,2187863086866864,2187872461243888
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2870862,2870862,1048576,256,0,0,8,8,16,64,0x0,0x7febf83ace80,44034,44034,15357,352280,16384,24796619,227952,0,99683264,2187872461273072,2187863086979504,2187863086998544,2187872461597281
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2870862,2870862,1048576,256,0,0,8,8,16,64,0x0,0x7febf83ace80,44730,44730,15579,357848,16384,24297029,232090,0,97687376,2187872461636034,2187863087020944,2187863087039824,2187872461826611
@@ -0,0 +1,5 @@
+pmc: GRBM_COUNT GRBM_GUI_ACTIVE SQ_WAVES SQ_IFETCH SQ_IFETCH_LEVEL  SQ_ACCUM_PREV_HIRES
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc: SQ_INSTS_LDS SQ_INST_LEVEL_LDS  SQ_ACCUM_PREV_HIRES
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc: SQ_INSTS_SMEM  SQ_INST_LEVEL_SMEM SQ_ACCUM_PREV_HIRES
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc: SQ_INSTS_VMEM  SQ_INST_LEVEL_VMEM SQ_ACCUM_PREV_HIRES
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc: 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
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQ_CYCLES SQ_BUSY_CYCLES SQ_BUSY_CU_CYCLES SQ_WAVES SQ_WAVE_CYCLES SQC_TC_INST_REQ SQC_TC_DATA_READ_REQ SQC_TC_DATA_WRITE_REQ GRBM_COUNT GRBM_GUI_ACTIVE TCP_GATE_EN1_sum TCP_GATE_EN2_sum TCP_TD_TCP_STALL_CYCLES_sum TCP_TCR_TCP_STALL_CYCLES_sum TA_TA_BUSY_sum TA_BUFFER_WAVEFRONTS_sum TD_TD_BUSY_sum TD_TC_STALL_sum SPI_CSN_WINDOW_VALID SPI_CSN_BUSY CPC_CPC_STAT_BUSY CPC_CPC_STAT_IDLE CPF_CPF_STAT_BUSY CPF_CPF_STAT_STALL TCC_CYCLE_sum TCC_BUSY_sum TCC_PROBE_sum TCC_PROBE_ALL_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQC_TC_DATA_ATOMIC_REQ SQC_TC_STALL SQC_TC_REQ SQC_DCACHE_REQ_READ_16 SQC_ICACHE_REQ SQC_ICACHE_HITS SQC_ICACHE_MISSES SQC_ICACHE_MISSES_DUPLICATE GRBM_SPI_BUSY TCP_READ_TAGCONFLICT_STALL_CYCLES_sum TCP_WRITE_TAGCONFLICT_STALL_CYCLES_sum TCP_ATOMIC_TAGCONFLICT_STALL_CYCLES_sum TCP_TA_TCP_STATE_READ_sum TA_BUFFER_READ_WAVEFRONTS_sum TA_BUFFER_WRITE_WAVEFRONTS_sum TD_COALESCABLE_WAVEFRONT_sum TD_LOAD_WAVEFRONT_sum SPI_CSN_NUM_THREADGROUPS SPI_CSN_WAVE CPC_CPC_TCIU_BUSY CPC_CPC_TCIU_IDLE CPF_CPF_TCIU_BUSY CPF_CPF_TCIU_STALL TCC_NC_REQ_sum TCC_UC_REQ_sum TCC_CC_REQ_sum TCC_RW_REQ_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  TCC_EA_ATOMIC_LEVEL_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  TCC_ATOMIC[0] TCC_CYCLE[0] TCC_EA_ATOMIC[0] TCC_EA_ATOMIC_LEVEL[0] TCC_ATOMIC[1] TCC_CYCLE[1] TCC_EA_ATOMIC[1] TCC_EA_ATOMIC_LEVEL[1] TCC_ATOMIC[2] TCC_CYCLE[2] TCC_EA_ATOMIC[2] TCC_EA_ATOMIC_LEVEL[2] TCC_ATOMIC[3] TCC_CYCLE[3] TCC_EA_ATOMIC[3] TCC_EA_ATOMIC_LEVEL[3] TCC_ATOMIC[4] TCC_CYCLE[4] TCC_EA_ATOMIC[4] TCC_EA_ATOMIC_LEVEL[4] TCC_ATOMIC[5] TCC_CYCLE[5] TCC_EA_ATOMIC[5] TCC_EA_ATOMIC_LEVEL[5] TCC_ATOMIC[6] TCC_CYCLE[6] TCC_EA_ATOMIC[6] TCC_EA_ATOMIC_LEVEL[6] TCC_ATOMIC[7] TCC_CYCLE[7] TCC_EA_ATOMIC[7] TCC_EA_ATOMIC_LEVEL[7] TCC_ATOMIC[8] TCC_CYCLE[8] TCC_EA_ATOMIC[8] TCC_EA_ATOMIC_LEVEL[8] TCC_ATOMIC[9] TCC_CYCLE[9] TCC_EA_ATOMIC[9] TCC_EA_ATOMIC_LEVEL[9] TCC_ATOMIC[10] TCC_CYCLE[10] TCC_EA_ATOMIC[10] TCC_EA_ATOMIC_LEVEL[10] TCC_ATOMIC[11] TCC_CYCLE[11] TCC_EA_ATOMIC[11] TCC_EA_ATOMIC_LEVEL[11] TCC_ATOMIC[12] TCC_CYCLE[12] TCC_EA_ATOMIC[12] TCC_EA_ATOMIC_LEVEL[12] TCC_ATOMIC[13] TCC_CYCLE[13] TCC_EA_ATOMIC[13] TCC_EA_ATOMIC_LEVEL[13] TCC_ATOMIC[14] TCC_CYCLE[14] TCC_EA_ATOMIC[14] TCC_EA_ATOMIC_LEVEL[14] TCC_ATOMIC[15] TCC_CYCLE[15] TCC_EA_ATOMIC[15] TCC_EA_ATOMIC_LEVEL[15] TCC_ATOMIC[16] TCC_CYCLE[16] TCC_EA_ATOMIC[16] TCC_EA_ATOMIC_LEVEL[16] TCC_ATOMIC[17] TCC_CYCLE[17] TCC_EA_ATOMIC[17] TCC_EA_ATOMIC_LEVEL[17] TCC_ATOMIC[18] TCC_CYCLE[18] TCC_EA_ATOMIC[18] TCC_EA_ATOMIC_LEVEL[18] TCC_ATOMIC[19] TCC_CYCLE[19] TCC_EA_ATOMIC[19] TCC_EA_ATOMIC_LEVEL[19] TCC_ATOMIC[20] TCC_CYCLE[20] TCC_EA_ATOMIC[20] TCC_EA_ATOMIC_LEVEL[20] TCC_ATOMIC[21] TCC_CYCLE[21] TCC_EA_ATOMIC[21] TCC_EA_ATOMIC_LEVEL[21] TCC_ATOMIC[22] TCC_CYCLE[22] TCC_EA_ATOMIC[22] TCC_EA_ATOMIC_LEVEL[22] TCC_ATOMIC[23] TCC_CYCLE[23] TCC_EA_ATOMIC[23] TCC_EA_ATOMIC_LEVEL[23] TCC_ATOMIC[24] TCC_CYCLE[24] TCC_EA_ATOMIC[24] TCC_EA_ATOMIC_LEVEL[24] TCC_ATOMIC[25] TCC_CYCLE[25] TCC_EA_ATOMIC[25] TCC_EA_ATOMIC_LEVEL[25] TCC_ATOMIC[26] TCC_CYCLE[26] TCC_EA_ATOMIC[26] TCC_EA_ATOMIC_LEVEL[26] TCC_ATOMIC[27] TCC_CYCLE[27] TCC_EA_ATOMIC[27] TCC_EA_ATOMIC_LEVEL[27] TCC_ATOMIC[28] TCC_CYCLE[28] TCC_EA_ATOMIC[28] TCC_EA_ATOMIC_LEVEL[28] TCC_ATOMIC[29] TCC_CYCLE[29] TCC_EA_ATOMIC[29] TCC_EA_ATOMIC_LEVEL[29] TCC_ATOMIC[30] TCC_CYCLE[30] TCC_EA_ATOMIC[30] TCC_EA_ATOMIC_LEVEL[30] TCC_ATOMIC[31] TCC_CYCLE[31] TCC_EA_ATOMIC[31] TCC_EA_ATOMIC_LEVEL[31]
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  TCC_EA_RDREQ[0] TCC_EA_RDREQ_32B[0] TCC_EA_RDREQ_DRAM_CREDIT_STALL[0] TCC_EA_RDREQ_GMI_CREDIT_STALL[0] TCC_EA_RDREQ[1] TCC_EA_RDREQ_32B[1] TCC_EA_RDREQ_DRAM_CREDIT_STALL[1] TCC_EA_RDREQ_GMI_CREDIT_STALL[1] TCC_EA_RDREQ[2] TCC_EA_RDREQ_32B[2] TCC_EA_RDREQ_DRAM_CREDIT_STALL[2] TCC_EA_RDREQ_GMI_CREDIT_STALL[2] TCC_EA_RDREQ[3] TCC_EA_RDREQ_32B[3] TCC_EA_RDREQ_DRAM_CREDIT_STALL[3] TCC_EA_RDREQ_GMI_CREDIT_STALL[3] TCC_EA_RDREQ[4] TCC_EA_RDREQ_32B[4] TCC_EA_RDREQ_DRAM_CREDIT_STALL[4] TCC_EA_RDREQ_GMI_CREDIT_STALL[4] TCC_EA_RDREQ[5] TCC_EA_RDREQ_32B[5] TCC_EA_RDREQ_DRAM_CREDIT_STALL[5] TCC_EA_RDREQ_GMI_CREDIT_STALL[5] TCC_EA_RDREQ[6] TCC_EA_RDREQ_32B[6] TCC_EA_RDREQ_DRAM_CREDIT_STALL[6] TCC_EA_RDREQ_GMI_CREDIT_STALL[6] TCC_EA_RDREQ[7] TCC_EA_RDREQ_32B[7] TCC_EA_RDREQ_DRAM_CREDIT_STALL[7] TCC_EA_RDREQ_GMI_CREDIT_STALL[7] TCC_EA_RDREQ[8] TCC_EA_RDREQ_32B[8] TCC_EA_RDREQ_DRAM_CREDIT_STALL[8] TCC_EA_RDREQ_GMI_CREDIT_STALL[8] TCC_EA_RDREQ[9] TCC_EA_RDREQ_32B[9] TCC_EA_RDREQ_DRAM_CREDIT_STALL[9] TCC_EA_RDREQ_GMI_CREDIT_STALL[9] TCC_EA_RDREQ[10] TCC_EA_RDREQ_32B[10] TCC_EA_RDREQ_DRAM_CREDIT_STALL[10] TCC_EA_RDREQ_GMI_CREDIT_STALL[10] TCC_EA_RDREQ[11] TCC_EA_RDREQ_32B[11] TCC_EA_RDREQ_DRAM_CREDIT_STALL[11] TCC_EA_RDREQ_GMI_CREDIT_STALL[11] TCC_EA_RDREQ[12] TCC_EA_RDREQ_32B[12] TCC_EA_RDREQ_DRAM_CREDIT_STALL[12] TCC_EA_RDREQ_GMI_CREDIT_STALL[12] TCC_EA_RDREQ[13] TCC_EA_RDREQ_32B[13] TCC_EA_RDREQ_DRAM_CREDIT_STALL[13] TCC_EA_RDREQ_GMI_CREDIT_STALL[13] TCC_EA_RDREQ[14] TCC_EA_RDREQ_32B[14] TCC_EA_RDREQ_DRAM_CREDIT_STALL[14] TCC_EA_RDREQ_GMI_CREDIT_STALL[14] TCC_EA_RDREQ[15] TCC_EA_RDREQ_32B[15] TCC_EA_RDREQ_DRAM_CREDIT_STALL[15] TCC_EA_RDREQ_GMI_CREDIT_STALL[15] TCC_EA_RDREQ[16] TCC_EA_RDREQ_32B[16] TCC_EA_RDREQ_DRAM_CREDIT_STALL[16] TCC_EA_RDREQ_GMI_CREDIT_STALL[16] TCC_EA_RDREQ[17] TCC_EA_RDREQ_32B[17] TCC_EA_RDREQ_DRAM_CREDIT_STALL[17] TCC_EA_RDREQ_GMI_CREDIT_STALL[17] TCC_EA_RDREQ[18] TCC_EA_RDREQ_32B[18] TCC_EA_RDREQ_DRAM_CREDIT_STALL[18] TCC_EA_RDREQ_GMI_CREDIT_STALL[18] TCC_EA_RDREQ[19] TCC_EA_RDREQ_32B[19] TCC_EA_RDREQ_DRAM_CREDIT_STALL[19] TCC_EA_RDREQ_GMI_CREDIT_STALL[19] TCC_EA_RDREQ[20] TCC_EA_RDREQ_32B[20] TCC_EA_RDREQ_DRAM_CREDIT_STALL[20] TCC_EA_RDREQ_GMI_CREDIT_STALL[20] TCC_EA_RDREQ[21] TCC_EA_RDREQ_32B[21] TCC_EA_RDREQ_DRAM_CREDIT_STALL[21] TCC_EA_RDREQ_GMI_CREDIT_STALL[21] TCC_EA_RDREQ[22] TCC_EA_RDREQ_32B[22] TCC_EA_RDREQ_DRAM_CREDIT_STALL[22] TCC_EA_RDREQ_GMI_CREDIT_STALL[22] TCC_EA_RDREQ[23] TCC_EA_RDREQ_32B[23] TCC_EA_RDREQ_DRAM_CREDIT_STALL[23] TCC_EA_RDREQ_GMI_CREDIT_STALL[23] TCC_EA_RDREQ[24] TCC_EA_RDREQ_32B[24] TCC_EA_RDREQ_DRAM_CREDIT_STALL[24] TCC_EA_RDREQ_GMI_CREDIT_STALL[24] TCC_EA_RDREQ[25] TCC_EA_RDREQ_32B[25] TCC_EA_RDREQ_DRAM_CREDIT_STALL[25] TCC_EA_RDREQ_GMI_CREDIT_STALL[25] TCC_EA_RDREQ[26] TCC_EA_RDREQ_32B[26] TCC_EA_RDREQ_DRAM_CREDIT_STALL[26] TCC_EA_RDREQ_GMI_CREDIT_STALL[26] TCC_EA_RDREQ[27] TCC_EA_RDREQ_32B[27] TCC_EA_RDREQ_DRAM_CREDIT_STALL[27] TCC_EA_RDREQ_GMI_CREDIT_STALL[27] TCC_EA_RDREQ[28] TCC_EA_RDREQ_32B[28] TCC_EA_RDREQ_DRAM_CREDIT_STALL[28] TCC_EA_RDREQ_GMI_CREDIT_STALL[28] TCC_EA_RDREQ[29] TCC_EA_RDREQ_32B[29] TCC_EA_RDREQ_DRAM_CREDIT_STALL[29] TCC_EA_RDREQ_GMI_CREDIT_STALL[29] TCC_EA_RDREQ[30] TCC_EA_RDREQ_32B[30] TCC_EA_RDREQ_DRAM_CREDIT_STALL[30] TCC_EA_RDREQ_GMI_CREDIT_STALL[30] TCC_EA_RDREQ[31] TCC_EA_RDREQ_32B[31] TCC_EA_RDREQ_DRAM_CREDIT_STALL[31] TCC_EA_RDREQ_GMI_CREDIT_STALL[31]
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  TCC_EA_RDREQ_IO_CREDIT_STALL[0] TCC_EA_RDREQ_LEVEL[0] TCC_EA_WRREQ[0] TCC_EA_WRREQ_64B[0] TCC_EA_RDREQ_IO_CREDIT_STALL[1] TCC_EA_RDREQ_LEVEL[1] TCC_EA_WRREQ[1] TCC_EA_WRREQ_64B[1] TCC_EA_RDREQ_IO_CREDIT_STALL[2] TCC_EA_RDREQ_LEVEL[2] TCC_EA_WRREQ[2] TCC_EA_WRREQ_64B[2] TCC_EA_RDREQ_IO_CREDIT_STALL[3] TCC_EA_RDREQ_LEVEL[3] TCC_EA_WRREQ[3] TCC_EA_WRREQ_64B[3] TCC_EA_RDREQ_IO_CREDIT_STALL[4] TCC_EA_RDREQ_LEVEL[4] TCC_EA_WRREQ[4] TCC_EA_WRREQ_64B[4] TCC_EA_RDREQ_IO_CREDIT_STALL[5] TCC_EA_RDREQ_LEVEL[5] TCC_EA_WRREQ[5] TCC_EA_WRREQ_64B[5] TCC_EA_RDREQ_IO_CREDIT_STALL[6] TCC_EA_RDREQ_LEVEL[6] TCC_EA_WRREQ[6] TCC_EA_WRREQ_64B[6] TCC_EA_RDREQ_IO_CREDIT_STALL[7] TCC_EA_RDREQ_LEVEL[7] TCC_EA_WRREQ[7] TCC_EA_WRREQ_64B[7] TCC_EA_RDREQ_IO_CREDIT_STALL[8] TCC_EA_RDREQ_LEVEL[8] TCC_EA_WRREQ[8] TCC_EA_WRREQ_64B[8] TCC_EA_RDREQ_IO_CREDIT_STALL[9] TCC_EA_RDREQ_LEVEL[9] TCC_EA_WRREQ[9] TCC_EA_WRREQ_64B[9] TCC_EA_RDREQ_IO_CREDIT_STALL[10] TCC_EA_RDREQ_LEVEL[10] TCC_EA_WRREQ[10] TCC_EA_WRREQ_64B[10] TCC_EA_RDREQ_IO_CREDIT_STALL[11] TCC_EA_RDREQ_LEVEL[11] TCC_EA_WRREQ[11] TCC_EA_WRREQ_64B[11] TCC_EA_RDREQ_IO_CREDIT_STALL[12] TCC_EA_RDREQ_LEVEL[12] TCC_EA_WRREQ[12] TCC_EA_WRREQ_64B[12] TCC_EA_RDREQ_IO_CREDIT_STALL[13] TCC_EA_RDREQ_LEVEL[13] TCC_EA_WRREQ[13] TCC_EA_WRREQ_64B[13] TCC_EA_RDREQ_IO_CREDIT_STALL[14] TCC_EA_RDREQ_LEVEL[14] TCC_EA_WRREQ[14] TCC_EA_WRREQ_64B[14] TCC_EA_RDREQ_IO_CREDIT_STALL[15] TCC_EA_RDREQ_LEVEL[15] TCC_EA_WRREQ[15] TCC_EA_WRREQ_64B[15] TCC_EA_RDREQ_IO_CREDIT_STALL[16] TCC_EA_RDREQ_LEVEL[16] TCC_EA_WRREQ[16] TCC_EA_WRREQ_64B[16] TCC_EA_RDREQ_IO_CREDIT_STALL[17] TCC_EA_RDREQ_LEVEL[17] TCC_EA_WRREQ[17] TCC_EA_WRREQ_64B[17] TCC_EA_RDREQ_IO_CREDIT_STALL[18] TCC_EA_RDREQ_LEVEL[18] TCC_EA_WRREQ[18] TCC_EA_WRREQ_64B[18] TCC_EA_RDREQ_IO_CREDIT_STALL[19] TCC_EA_RDREQ_LEVEL[19] TCC_EA_WRREQ[19] TCC_EA_WRREQ_64B[19] TCC_EA_RDREQ_IO_CREDIT_STALL[20] TCC_EA_RDREQ_LEVEL[20] TCC_EA_WRREQ[20] TCC_EA_WRREQ_64B[20] TCC_EA_RDREQ_IO_CREDIT_STALL[21] TCC_EA_RDREQ_LEVEL[21] TCC_EA_WRREQ[21] TCC_EA_WRREQ_64B[21] TCC_EA_RDREQ_IO_CREDIT_STALL[22] TCC_EA_RDREQ_LEVEL[22] TCC_EA_WRREQ[22] TCC_EA_WRREQ_64B[22] TCC_EA_RDREQ_IO_CREDIT_STALL[23] TCC_EA_RDREQ_LEVEL[23] TCC_EA_WRREQ[23] TCC_EA_WRREQ_64B[23] TCC_EA_RDREQ_IO_CREDIT_STALL[24] TCC_EA_RDREQ_LEVEL[24] TCC_EA_WRREQ[24] TCC_EA_WRREQ_64B[24] TCC_EA_RDREQ_IO_CREDIT_STALL[25] TCC_EA_RDREQ_LEVEL[25] TCC_EA_WRREQ[25] TCC_EA_WRREQ_64B[25] TCC_EA_RDREQ_IO_CREDIT_STALL[26] TCC_EA_RDREQ_LEVEL[26] TCC_EA_WRREQ[26] TCC_EA_WRREQ_64B[26] TCC_EA_RDREQ_IO_CREDIT_STALL[27] TCC_EA_RDREQ_LEVEL[27] TCC_EA_WRREQ[27] TCC_EA_WRREQ_64B[27] TCC_EA_RDREQ_IO_CREDIT_STALL[28] TCC_EA_RDREQ_LEVEL[28] TCC_EA_WRREQ[28] TCC_EA_WRREQ_64B[28] TCC_EA_RDREQ_IO_CREDIT_STALL[29] TCC_EA_RDREQ_LEVEL[29] TCC_EA_WRREQ[29] TCC_EA_WRREQ_64B[29] TCC_EA_RDREQ_IO_CREDIT_STALL[30] TCC_EA_RDREQ_LEVEL[30] TCC_EA_WRREQ[30] TCC_EA_WRREQ_64B[30] TCC_EA_RDREQ_IO_CREDIT_STALL[31] TCC_EA_RDREQ_LEVEL[31] TCC_EA_WRREQ[31] TCC_EA_WRREQ_64B[31]
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  TCC_EA_WRREQ_DRAM_CREDIT_STALL[0] TCC_EA_WRREQ_GMI_CREDIT_STALL[0] TCC_EA_WRREQ_IO_CREDIT_STALL[0] TCC_EA_WRREQ_LEVEL[0] TCC_EA_WRREQ_DRAM_CREDIT_STALL[1] TCC_EA_WRREQ_GMI_CREDIT_STALL[1] TCC_EA_WRREQ_IO_CREDIT_STALL[1] TCC_EA_WRREQ_LEVEL[1] TCC_EA_WRREQ_DRAM_CREDIT_STALL[2] TCC_EA_WRREQ_GMI_CREDIT_STALL[2] TCC_EA_WRREQ_IO_CREDIT_STALL[2] TCC_EA_WRREQ_LEVEL[2] TCC_EA_WRREQ_DRAM_CREDIT_STALL[3] TCC_EA_WRREQ_GMI_CREDIT_STALL[3] TCC_EA_WRREQ_IO_CREDIT_STALL[3] TCC_EA_WRREQ_LEVEL[3] TCC_EA_WRREQ_DRAM_CREDIT_STALL[4] TCC_EA_WRREQ_GMI_CREDIT_STALL[4] TCC_EA_WRREQ_IO_CREDIT_STALL[4] TCC_EA_WRREQ_LEVEL[4] TCC_EA_WRREQ_DRAM_CREDIT_STALL[5] TCC_EA_WRREQ_GMI_CREDIT_STALL[5] TCC_EA_WRREQ_IO_CREDIT_STALL[5] TCC_EA_WRREQ_LEVEL[5] TCC_EA_WRREQ_DRAM_CREDIT_STALL[6] TCC_EA_WRREQ_GMI_CREDIT_STALL[6] TCC_EA_WRREQ_IO_CREDIT_STALL[6] TCC_EA_WRREQ_LEVEL[6] TCC_EA_WRREQ_DRAM_CREDIT_STALL[7] TCC_EA_WRREQ_GMI_CREDIT_STALL[7] TCC_EA_WRREQ_IO_CREDIT_STALL[7] TCC_EA_WRREQ_LEVEL[7] TCC_EA_WRREQ_DRAM_CREDIT_STALL[8] TCC_EA_WRREQ_GMI_CREDIT_STALL[8] TCC_EA_WRREQ_IO_CREDIT_STALL[8] TCC_EA_WRREQ_LEVEL[8] TCC_EA_WRREQ_DRAM_CREDIT_STALL[9] TCC_EA_WRREQ_GMI_CREDIT_STALL[9] TCC_EA_WRREQ_IO_CREDIT_STALL[9] TCC_EA_WRREQ_LEVEL[9] TCC_EA_WRREQ_DRAM_CREDIT_STALL[10] TCC_EA_WRREQ_GMI_CREDIT_STALL[10] TCC_EA_WRREQ_IO_CREDIT_STALL[10] TCC_EA_WRREQ_LEVEL[10] TCC_EA_WRREQ_DRAM_CREDIT_STALL[11] TCC_EA_WRREQ_GMI_CREDIT_STALL[11] TCC_EA_WRREQ_IO_CREDIT_STALL[11] TCC_EA_WRREQ_LEVEL[11] TCC_EA_WRREQ_DRAM_CREDIT_STALL[12] TCC_EA_WRREQ_GMI_CREDIT_STALL[12] TCC_EA_WRREQ_IO_CREDIT_STALL[12] TCC_EA_WRREQ_LEVEL[12] TCC_EA_WRREQ_DRAM_CREDIT_STALL[13] TCC_EA_WRREQ_GMI_CREDIT_STALL[13] TCC_EA_WRREQ_IO_CREDIT_STALL[13] TCC_EA_WRREQ_LEVEL[13] TCC_EA_WRREQ_DRAM_CREDIT_STALL[14] TCC_EA_WRREQ_GMI_CREDIT_STALL[14] TCC_EA_WRREQ_IO_CREDIT_STALL[14] TCC_EA_WRREQ_LEVEL[14] TCC_EA_WRREQ_DRAM_CREDIT_STALL[15] TCC_EA_WRREQ_GMI_CREDIT_STALL[15] TCC_EA_WRREQ_IO_CREDIT_STALL[15] TCC_EA_WRREQ_LEVEL[15] TCC_EA_WRREQ_DRAM_CREDIT_STALL[16] TCC_EA_WRREQ_GMI_CREDIT_STALL[16] TCC_EA_WRREQ_IO_CREDIT_STALL[16] TCC_EA_WRREQ_LEVEL[16] TCC_EA_WRREQ_DRAM_CREDIT_STALL[17] TCC_EA_WRREQ_GMI_CREDIT_STALL[17] TCC_EA_WRREQ_IO_CREDIT_STALL[17] TCC_EA_WRREQ_LEVEL[17] TCC_EA_WRREQ_DRAM_CREDIT_STALL[18] TCC_EA_WRREQ_GMI_CREDIT_STALL[18] TCC_EA_WRREQ_IO_CREDIT_STALL[18] TCC_EA_WRREQ_LEVEL[18] TCC_EA_WRREQ_DRAM_CREDIT_STALL[19] TCC_EA_WRREQ_GMI_CREDIT_STALL[19] TCC_EA_WRREQ_IO_CREDIT_STALL[19] TCC_EA_WRREQ_LEVEL[19] TCC_EA_WRREQ_DRAM_CREDIT_STALL[20] TCC_EA_WRREQ_GMI_CREDIT_STALL[20] TCC_EA_WRREQ_IO_CREDIT_STALL[20] TCC_EA_WRREQ_LEVEL[20] TCC_EA_WRREQ_DRAM_CREDIT_STALL[21] TCC_EA_WRREQ_GMI_CREDIT_STALL[21] TCC_EA_WRREQ_IO_CREDIT_STALL[21] TCC_EA_WRREQ_LEVEL[21] TCC_EA_WRREQ_DRAM_CREDIT_STALL[22] TCC_EA_WRREQ_GMI_CREDIT_STALL[22] TCC_EA_WRREQ_IO_CREDIT_STALL[22] TCC_EA_WRREQ_LEVEL[22] TCC_EA_WRREQ_DRAM_CREDIT_STALL[23] TCC_EA_WRREQ_GMI_CREDIT_STALL[23] TCC_EA_WRREQ_IO_CREDIT_STALL[23] TCC_EA_WRREQ_LEVEL[23] TCC_EA_WRREQ_DRAM_CREDIT_STALL[24] TCC_EA_WRREQ_GMI_CREDIT_STALL[24] TCC_EA_WRREQ_IO_CREDIT_STALL[24] TCC_EA_WRREQ_LEVEL[24] TCC_EA_WRREQ_DRAM_CREDIT_STALL[25] TCC_EA_WRREQ_GMI_CREDIT_STALL[25] TCC_EA_WRREQ_IO_CREDIT_STALL[25] TCC_EA_WRREQ_LEVEL[25] TCC_EA_WRREQ_DRAM_CREDIT_STALL[26] TCC_EA_WRREQ_GMI_CREDIT_STALL[26] TCC_EA_WRREQ_IO_CREDIT_STALL[26] TCC_EA_WRREQ_LEVEL[26] TCC_EA_WRREQ_DRAM_CREDIT_STALL[27] TCC_EA_WRREQ_GMI_CREDIT_STALL[27] TCC_EA_WRREQ_IO_CREDIT_STALL[27] TCC_EA_WRREQ_LEVEL[27] TCC_EA_WRREQ_DRAM_CREDIT_STALL[28] TCC_EA_WRREQ_GMI_CREDIT_STALL[28] TCC_EA_WRREQ_IO_CREDIT_STALL[28] TCC_EA_WRREQ_LEVEL[28] TCC_EA_WRREQ_DRAM_CREDIT_STALL[29] TCC_EA_WRREQ_GMI_CREDIT_STALL[29] TCC_EA_WRREQ_IO_CREDIT_STALL[29] TCC_EA_WRREQ_LEVEL[29] TCC_EA_WRREQ_DRAM_CREDIT_STALL[30] TCC_EA_WRREQ_GMI_CREDIT_STALL[30] TCC_EA_WRREQ_IO_CREDIT_STALL[30] TCC_EA_WRREQ_LEVEL[30] TCC_EA_WRREQ_DRAM_CREDIT_STALL[31] TCC_EA_WRREQ_GMI_CREDIT_STALL[31] TCC_EA_WRREQ_IO_CREDIT_STALL[31] TCC_EA_WRREQ_LEVEL[31]
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  TCC_HIT[0] TCC_MISS[0] TCC_READ[0] TCC_REQ[0] TCC_HIT[1] TCC_MISS[1] TCC_READ[1] TCC_REQ[1] TCC_HIT[2] TCC_MISS[2] TCC_READ[2] TCC_REQ[2] TCC_HIT[3] TCC_MISS[3] TCC_READ[3] TCC_REQ[3] TCC_HIT[4] TCC_MISS[4] TCC_READ[4] TCC_REQ[4] TCC_HIT[5] TCC_MISS[5] TCC_READ[5] TCC_REQ[5] TCC_HIT[6] TCC_MISS[6] TCC_READ[6] TCC_REQ[6] TCC_HIT[7] TCC_MISS[7] TCC_READ[7] TCC_REQ[7] TCC_HIT[8] TCC_MISS[8] TCC_READ[8] TCC_REQ[8] TCC_HIT[9] TCC_MISS[9] TCC_READ[9] TCC_REQ[9] TCC_HIT[10] TCC_MISS[10] TCC_READ[10] TCC_REQ[10] TCC_HIT[11] TCC_MISS[11] TCC_READ[11] TCC_REQ[11] TCC_HIT[12] TCC_MISS[12] TCC_READ[12] TCC_REQ[12] TCC_HIT[13] TCC_MISS[13] TCC_READ[13] TCC_REQ[13] TCC_HIT[14] TCC_MISS[14] TCC_READ[14] TCC_REQ[14] TCC_HIT[15] TCC_MISS[15] TCC_READ[15] TCC_REQ[15] TCC_HIT[16] TCC_MISS[16] TCC_READ[16] TCC_REQ[16] TCC_HIT[17] TCC_MISS[17] TCC_READ[17] TCC_REQ[17] TCC_HIT[18] TCC_MISS[18] TCC_READ[18] TCC_REQ[18] TCC_HIT[19] TCC_MISS[19] TCC_READ[19] TCC_REQ[19] TCC_HIT[20] TCC_MISS[20] TCC_READ[20] TCC_REQ[20] TCC_HIT[21] TCC_MISS[21] TCC_READ[21] TCC_REQ[21] TCC_HIT[22] TCC_MISS[22] TCC_READ[22] TCC_REQ[22] TCC_HIT[23] TCC_MISS[23] TCC_READ[23] TCC_REQ[23] TCC_HIT[24] TCC_MISS[24] TCC_READ[24] TCC_REQ[24] TCC_HIT[25] TCC_MISS[25] TCC_READ[25] TCC_REQ[25] TCC_HIT[26] TCC_MISS[26] TCC_READ[26] TCC_REQ[26] TCC_HIT[27] TCC_MISS[27] TCC_READ[27] TCC_REQ[27] TCC_HIT[28] TCC_MISS[28] TCC_READ[28] TCC_REQ[28] TCC_HIT[29] TCC_MISS[29] TCC_READ[29] TCC_REQ[29] TCC_HIT[30] TCC_MISS[30] TCC_READ[30] TCC_REQ[30] TCC_HIT[31] TCC_MISS[31] TCC_READ[31] TCC_REQ[31]
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  TCC_RW_REQ[0] TCC_TOO_MANY_EA_WRREQS_STALL[0] TCC_WRITE[0] TCC_RW_REQ[1] TCC_TOO_MANY_EA_WRREQS_STALL[1] TCC_WRITE[1] TCC_RW_REQ[2] TCC_TOO_MANY_EA_WRREQS_STALL[2] TCC_WRITE[2] TCC_RW_REQ[3] TCC_TOO_MANY_EA_WRREQS_STALL[3] TCC_WRITE[3] TCC_RW_REQ[4] TCC_TOO_MANY_EA_WRREQS_STALL[4] TCC_WRITE[4] TCC_RW_REQ[5] TCC_TOO_MANY_EA_WRREQS_STALL[5] TCC_WRITE[5] TCC_RW_REQ[6] TCC_TOO_MANY_EA_WRREQS_STALL[6] TCC_WRITE[6] TCC_RW_REQ[7] TCC_TOO_MANY_EA_WRREQS_STALL[7] TCC_WRITE[7] TCC_RW_REQ[8] TCC_TOO_MANY_EA_WRREQS_STALL[8] TCC_WRITE[8] TCC_RW_REQ[9] TCC_TOO_MANY_EA_WRREQS_STALL[9] TCC_WRITE[9] TCC_RW_REQ[10] TCC_TOO_MANY_EA_WRREQS_STALL[10] TCC_WRITE[10] TCC_RW_REQ[11] TCC_TOO_MANY_EA_WRREQS_STALL[11] TCC_WRITE[11] TCC_RW_REQ[12] TCC_TOO_MANY_EA_WRREQS_STALL[12] TCC_WRITE[12] TCC_RW_REQ[13] TCC_TOO_MANY_EA_WRREQS_STALL[13] TCC_WRITE[13] TCC_RW_REQ[14] TCC_TOO_MANY_EA_WRREQS_STALL[14] TCC_WRITE[14] TCC_RW_REQ[15] TCC_TOO_MANY_EA_WRREQS_STALL[15] TCC_WRITE[15] TCC_RW_REQ[16] TCC_TOO_MANY_EA_WRREQS_STALL[16] TCC_WRITE[16] TCC_RW_REQ[17] TCC_TOO_MANY_EA_WRREQS_STALL[17] TCC_WRITE[17] TCC_RW_REQ[18] TCC_TOO_MANY_EA_WRREQS_STALL[18] TCC_WRITE[18] TCC_RW_REQ[19] TCC_TOO_MANY_EA_WRREQS_STALL[19] TCC_WRITE[19] TCC_RW_REQ[20] TCC_TOO_MANY_EA_WRREQS_STALL[20] TCC_WRITE[20] TCC_RW_REQ[21] TCC_TOO_MANY_EA_WRREQS_STALL[21] TCC_WRITE[21] TCC_RW_REQ[22] TCC_TOO_MANY_EA_WRREQS_STALL[22] TCC_WRITE[22] TCC_RW_REQ[23] TCC_TOO_MANY_EA_WRREQS_STALL[23] TCC_WRITE[23] TCC_RW_REQ[24] TCC_TOO_MANY_EA_WRREQS_STALL[24] TCC_WRITE[24] TCC_RW_REQ[25] TCC_TOO_MANY_EA_WRREQS_STALL[25] TCC_WRITE[25] TCC_RW_REQ[26] TCC_TOO_MANY_EA_WRREQS_STALL[26] TCC_WRITE[26] TCC_RW_REQ[27] TCC_TOO_MANY_EA_WRREQS_STALL[27] TCC_WRITE[27] TCC_RW_REQ[28] TCC_TOO_MANY_EA_WRREQS_STALL[28] TCC_WRITE[28] TCC_RW_REQ[29] TCC_TOO_MANY_EA_WRREQS_STALL[29] TCC_WRITE[29] TCC_RW_REQ[30] TCC_TOO_MANY_EA_WRREQS_STALL[30] TCC_WRITE[30] TCC_RW_REQ[31] TCC_TOO_MANY_EA_WRREQS_STALL[31] TCC_WRITE[31]
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQC_DCACHE_INPUT_VALID_READYB SQC_DCACHE_ATOMIC SQC_DCACHE_REQ_READ_8 SQC_DCACHE_REQ SQC_DCACHE_HITS SQC_DCACHE_MISSES SQC_DCACHE_MISSES_DUPLICATE SQC_DCACHE_REQ_READ_1 TCP_VOLATILE_sum TCP_TOTAL_ACCESSES_sum TCP_TOTAL_READ_sum TCP_TOTAL_WRITE_sum TA_BUFFER_ATOMIC_WAVEFRONTS_sum TA_BUFFER_TOTAL_CYCLES_sum TD_ATOMIC_WAVEFRONT_sum TD_STORE_WAVEFRONT_sum SPI_RA_REQ_NO_ALLOC SPI_RA_REQ_NO_ALLOC_CSN CPC_CPC_STAT_STALL CPC_UTCL1_STALL_ON_TRANSLATION CPF_CPF_STAT_IDLE CPF_CPF_TCIU_IDLE TCC_REQ_sum TCC_STREAMING_REQ_sum TCC_HIT_sum TCC_MISS_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQC_DCACHE_REQ_READ_2 SQC_DCACHE_REQ_READ_4 SQ_INSTS_VMEM_WR SQ_INSTS_VMEM_RD SQ_INSTS_VMEM SQ_INSTS_SALU SQ_INSTS_VSKIPPED SQ_INSTS_SMEM TCP_TOTAL_ATOMIC_WITH_RET_sum TCP_TOTAL_ATOMIC_WITHOUT_RET_sum TCP_TOTAL_WRITEBACK_INVALIDATES_sum TCP_TOTAL_CACHE_ACCESSES_sum TA_BUFFER_COALESCED_READ_CYCLES_sum TA_BUFFER_COALESCED_WRITE_CYCLES_sum SPI_RA_RES_STALL_CSN SPI_RA_TMP_STALL_CSN CPC_CPC_UTCL2IU_BUSY CPC_CPC_UTCL2IU_IDLE CPF_CMP_UTCL1_STALL_ON_TRANSLATION TCC_READ_sum TCC_WRITE_sum TCC_ATOMIC_sum TCC_WRITEBACK_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQ_INSTS_FLAT SQ_INSTS_LDS SQ_INSTS_GDS SQ_INSTS_EXP_GDS SQ_INSTS_BRANCH SQ_INSTS_SENDMSG SQ_INSTS SQ_WAIT_ANY TCP_UTCL1_TRANSLATION_MISS_sum TCP_UTCL1_TRANSLATION_HIT_sum TCP_UTCL1_PERMISSION_MISS_sum TCP_UTCL1_REQUEST_sum TA_ADDR_STALLED_BY_TC_CYCLES_sum TA_TOTAL_WAVEFRONTS_sum SPI_RA_WAVE_SIMD_FULL_CSN SPI_RA_VGPR_SIMD_FULL_CSN CPC_CPC_UTCL2IU_STALL CPC_ME1_BUSY_FOR_PACKET_DECODE TCC_EA_WRREQ_sum TCC_EA_WRREQ_64B_sum TCC_EA_WR_UNCACHED_32B_sum TCC_EA_WRREQ_DRAM_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQ_WAIT_INST_ANY SQ_ACTIVE_INST_ANY SQ_INSTS_VALU SQ_ACTIVE_INST_VMEM SQ_ACTIVE_INST_LDS SQ_ACTIVE_INST_VALU SQ_ACTIVE_INST_SCA SQ_ACTIVE_INST_EXP_GDS TCP_TCP_LATENCY_sum TCP_TCC_READ_REQ_LATENCY_sum TCP_TCC_WRITE_REQ_LATENCY_sum TCP_TCC_READ_REQ_sum TA_ADDR_STALLED_BY_TD_CYCLES_sum TA_DATA_STALLED_BY_TC_CYCLES_sum SPI_RA_SGPR_SIMD_FULL_CSN SPI_RA_LDS_CU_FULL_CSN CPC_ME1_DC0_SPI_BUSY TCC_EA_WRREQ_STALL_sum TCC_EA_WRREQ_IO_CREDIT_STALL_sum TCC_EA_WRREQ_GMI_CREDIT_STALL_sum TCC_EA_WRREQ_DRAM_CREDIT_STALL_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQ_ACTIVE_INST_MISC SQ_ACTIVE_INST_FLAT SQ_INST_CYCLES_VMEM_WR SQ_INST_CYCLES_VMEM_RD SQ_INST_CYCLES_SMEM SQ_INST_CYCLES_SALU SQ_THREAD_CYCLES_VALU SQ_IFETCH TCP_TCC_WRITE_REQ_sum TCP_TCC_ATOMIC_WITH_RET_REQ_sum TCP_TCC_ATOMIC_WITHOUT_RET_REQ_sum TCP_TCC_NC_READ_REQ_sum TA_FLAT_WAVEFRONTS_sum TA_FLAT_READ_WAVEFRONTS_sum SPI_RA_BAR_CU_FULL_CSN SPI_RA_TGLIM_CU_FULL_CSN TCC_EA_RDREQ_sum TCC_EA_RDREQ_32B_sum TCC_EA_RD_UNCACHED_32B_sum TCC_EA_RDREQ_DRAM_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQ_LDS_BANK_CONFLICT SQ_LDS_ADDR_CONFLICT SQ_LDS_UNALIGNED_STALL SQ_WAVES_EQ_64 SQ_WAVES_LT_64 SQ_WAVES_LT_48 SQ_WAVES_LT_32 SQ_WAVES_LT_16 TCP_TCC_NC_WRITE_REQ_sum TCP_TCC_NC_ATOMIC_REQ_sum TCP_TCC_UC_READ_REQ_sum TCP_TCC_UC_WRITE_REQ_sum TA_FLAT_WRITE_WAVEFRONTS_sum TA_FLAT_ATOMIC_WAVEFRONTS_sum SPI_RA_WVLIM_STALL_CSN SPI_SWC_CSC_WR TCC_EA_RDREQ_IO_CREDIT_STALL_sum TCC_EA_RDREQ_GMI_CREDIT_STALL_sum TCC_EA_RDREQ_DRAM_CREDIT_STALL_sum TCC_TAG_STALL_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  SQ_ITEMS SQ_LDS_MEM_VIOLATIONS SQ_LDS_ATOMIC_RETURN SQ_LDS_IDX_ACTIVE SQ_WAVES_RESTORED SQ_WAVES_SAVED SQ_INSTS_SMEM_NORM TCP_TCC_UC_ATOMIC_REQ_sum TCP_TCC_CC_READ_REQ_sum TCP_TCC_CC_WRITE_REQ_sum TCP_TCC_CC_ATOMIC_REQ_sum SPI_VWC_CSC_WR SPI_RA_BULKY_CU_FULL_CSN TCC_NORMAL_WRITEBACK_sum TCC_ALL_TC_OP_WB_WRITEBACK_sum TCC_NORMAL_EVICT_sum TCC_ALL_TC_OP_INV_EVICT_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:  TCP_TCC_RW_READ_REQ_sum TCP_TCC_RW_WRITE_REQ_sum TCP_TCC_RW_ATOMIC_REQ_sum TCP_PENDING_STALL_CYCLES_sum TCC_TOO_MANY_EA_WRREQS_STALL_sum TCC_EA_ATOMIC_sum TCC_EA_RDREQ_LEVEL_sum TCC_EA_WRREQ_LEVEL_sum
+
+gpu:
+range:
+kernel:
@@ -0,0 +1,5 @@
+pmc:
+
+gpu:
+range:
+kernel:
@@ -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
+device_filter,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF,Fri Mar  1 16:46:13 2024 (CST),1,t006-003.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,,527651240,,5.7.1-98,113-D3431401-100,NA,NA,MI100,gfx908,16,8192,120,4,8,64,1024,40,1502,1200,1502,1200,32,32,64,4,1228.8,1
@@ -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]",2,0,0,2867400,2867400,1048576,256,0,0,8,8,16,64,0x0,0x7f1016c04e80,2187863086818021,2187863086843344,2187863086866864,2187863086880578
+1,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,2,2867400,2867400,1048576,256,0,0,8,8,16,64,0x0,0x7f1016c04e80,2187863086879366,2187863086979504,2187863086998544,2187863086999782
+2,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,4,2867400,2867400,1048576,256,0,0,8,8,16,64,0x0,0x7f1016c04e80,2187863087011163,2187863087020944,2187863087039824,2187863087040899
@@ -1,4 +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,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,181827,181827,1048576,256,0,0,8,8,16,64,0x0,0x7f07663bae80,47978,47978,16384,65536,14494,1861536,6402045414567,6401146042762,6401146066762,6402053157530
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,181827,181827,1048576,256,0,0,8,8,16,64,0x0,0x7f07663bae80,44012,44012,16384,65536,8160,1048580,6402053180624,6401146167882,6401146186922,6402053517320
-2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,181827,181827,1048576,256,0,0,8,8,16,64,0x0,0x7f07663bae80,43085,43085,16384,65536,8145,1048584,6402053549281,6401146207402,6401146226282,6402053732076
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2858631,2858631,1048576,256,0,0,8,8,16,64,0x0,0x7f306d022e80,48693,48693,16384,65536,14587,1861268,2187837651454006,2187843682406959,2187843682431279,2187837659153650
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2858631,2858631,1048576,256,0,0,8,8,16,64,0x0,0x7f306d022e80,43450,43450,16384,65536,8185,1048576,2187837659175581,2187843682524719,2187843682543599,2187837659498637
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2858631,2858631,1048576,256,0,0,8,8,16,64,0x0,0x7f306d022e80,42429,42429,16384,65536,8118,1048580,2187837659530768,2187843682565519,2187843682585039,2187837659713381
@@ -1,4 +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,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,178576,178576,1048576,256,0,0,8,8,16,64,0x0,0x7f74bd5b6e80,0,0,0,6386463651065,6401146042762,6401146066762,6386471177349
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,178576,178576,1048576,256,0,0,8,8,16,64,0x0,0x7f74bd5b6e80,0,0,0,6386471197998,6401146167882,6401146186922,6386471552869
-2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,178576,178576,1048576,256,0,0,8,8,16,64,0x0,0x7f74bd5b6e80,0,0,0,6386471582034,6401146207402,6401146226282,6386471757676
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2862637,2862637,1048576,256,0,0,8,8,16,64,0x0,0x7f5be8850e80,0,0,0,2187848852549207,2187843682406959,2187843682431279,2187848859975488
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2862637,2862637,1048576,256,0,0,8,8,16,64,0x0,0x7f5be8850e80,0,0,0,2187848859994684,2187843682524719,2187843682543599,2187848860303414
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2862637,2862637,1048576,256,0,0,8,8,16,64,0x0,0x7f5be8850e80,0,0,0,2187848860332038,2187843682565519,2187843682585039,2187848860506154
@@ -1,4 +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,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,182838,182838,1048576,256,0,0,8,8,16,64,0x0,0x7faf17528e80,65536,221118,28351800,6406815203658,6401146042762,6401146066762,6406822956018
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,182838,182838,1048576,256,0,0,8,8,16,64,0x0,0x7faf17528e80,65536,203702,26080032,6406822974133,6401146167882,6401146186922,6406823418142
-2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,182838,182838,1048576,256,0,0,8,8,16,64,0x0,0x7faf17528e80,65536,203144,25999920,6406823446055,6401146207402,6401146226282,6406823622218
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2861181,2861181,1048576,256,0,0,8,8,16,64,0x0,0x7f4afbf56e80,65536,210560,26916328,2187844811633072,2187843682406959,2187843682431279,2187844819158530
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2861181,2861181,1048576,256,0,0,8,8,16,64,0x0,0x7f4afbf56e80,65536,180192,23099952,2187844819178597,2187843682524719,2187843682543599,2187844819476596
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2861181,2861181,1048576,256,0,0,8,8,16,64,0x0,0x7f4afbf56e80,65536,192032,24611128,2187844819505521,2187843682565519,2187843682585039,2187844819677133
@@ -1,4 +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,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,178778,178778,1048576,256,0,0,8,8,16,64,0x0,0x7fc7e6412e80,32768,659106,84360532,6387368867286,6401146042762,6401146066762,6387376507826
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,178778,178778,1048576,256,0,0,8,8,16,64,0x0,0x7fc7e6412e80,32768,653457,83634844,6387376528044,6401146167882,6401146186922,6387376826258
-2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,178778,178778,1048576,256,0,0,8,8,16,64,0x0,0x7fc7e6412e80,32768,670775,85843888,6387376854291,6401146207402,6401146226282,6387377024353
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2859725,2859725,1048576,256,0,0,8,8,16,64,0x0,0x7f77365b0e80,32768,648628,83041940,2187840584839972,2187843682406959,2187843682431279,2187840592444006
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2859725,2859725,1048576,256,0,0,8,8,16,64,0x0,0x7f77365b0e80,32768,665455,85189652,2187840592464465,2187843682524719,2187843682543599,2187840592804623
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2859725,2859725,1048576,256,0,0,8,8,16,64,0x0,0x7f77365b0e80,32768,630264,80689312,2187840592833738,2187843682565519,2187843682585039,2187840593022011
@@ -1,4 +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,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,"vecCopy(double*, double*, double*, int, int) ",2,0,0,180006,180006,1048576,256,0,0,8,8,16,64,0x0,0x7ff7b376ee80,49008,49008,17099,392072,16384,25315374,242746,0,101780108,6393202548827,6401146042762,6401146066762,6393210198774
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,180006,180006,1048576,256,0,0,8,8,16,64,0x0,0x7ff7b376ee80,42135,42135,12945,337088,16384,24545355,232922,0,98697060,6393210228730,6401146167882,6401146186922,6393210558173
-2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,180006,180006,1048576,256,0,0,8,8,16,64,0x0,0x7ff7b376ee80,41623,41623,13083,332992,16384,24635089,231002,0,99051052,6393210596947,6401146207402,6401146226282,6393210786946
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2861545,2861545,1048576,256,0,0,8,8,16,64,0x0,0x7f66b3458e80,49547,49547,16868,396384,16384,26740947,242792,0,107460160,2187845706739501,2187843682406959,2187843682431279,2187845714403057
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2861545,2861545,1048576,256,0,0,8,8,16,64,0x0,0x7f66b3458e80,43543,43543,13929,348352,16384,25132807,235648,0,101028984,2187845714431621,2187843682524719,2187843682543599,2187845714748586
+2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,2861545,2861545,1048576,256,0,0,8,8,16,64,0x0,0x7f66b3458e80,43359,43359,14054,346880,16384,24431177,229443,0,98220436,2187845714786918,2187843682565519,2187843682585039,2187845714981864
@@ -1,4 +0,0 @@
-Dispatch_ID,Kernel_Name,GPU_ID
-0,"vecCopy(double*, double*, double*, int, int) ",2
-1,"vecCopy(double*, double*, double*, int, int) ",2
-2,"vecCopy(double*, double*, double*, int, int) ",2
@@ -1,2 +1,2 @@
-workload_name,command,host_name,host_cpu,sbios,host_distro,host_kernel,host_rocmver,date,gpu_soc,vbios,numSE,numCU,numSIMD,waveSize,maxWavesPerCU,maxWorkgroupSize,L1,L2,sclk,mclk,cur_sclk,cur_mclk,L2Banks,totalL2Banks,LDSBanks,name,numSQC,numPipes,hbmBW,compute_partition,memory_partition,ip_blocks
-device_inv_int,./tests/vcopy -n 1048576 -b 256 -i 3,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_64,5.7.1-98,Mon Feb 19 13:25:16 2024 (CST),gfx908,113-D3431401-100,8,120,4,64,40,1024,16,8192,1502,1200,1502,1200,32,32,32,MI100,30,4,1200,,,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
+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
+device_inv_int,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF,Fri Mar  1 16:45:48 2024 (CST),1,t006-003.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,,527651240,,5.7.1-98,113-D3431401-100,NA,NA,MI100,gfx908,16,8192,120,4,8,64,1024,40,1502,1200,1502,1200,32,32,64,4,1228.8,1
@@ -1,4 +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) ",2,0,0,181625,181625,1048576,256,0,0,8,8,16,64,0x0,0x7f124867ce80,6401146016248,6401146042762,6401146066762,6401146079037
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,181625,181625,1048576,256,0,0,8,8,16,64,0x0,0x7f124867ce80,6401146079558,6401146167882,6401146186922,6401146188174
-2,"vecCopy(double*, double*, double*, int, int) ",2,0,4,181625,181625,1048576,256,0,0,8,8,16,64,0x0,0x7f124867ce80,6401146198824,6401146207402,6401146226282,6401146227678
+0,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,0,2860817,2860817,1048576,256,0,0,8,8,16,64,0x0,0x7fd32b2cce80,2187843682380846,2187843682406959,2187843682431279,2187843682443404
+1,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,2,2860817,2860817,1048576,256,0,0,8,8,16,64,0x0,0x7fd32b2cce80,2187843682444325,2187843682524719,2187843682543599,2187843682544844
+2,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,4,2860817,2860817,1048576,256,0,0,8,8,16,64,0x0,0x7fd32b2cce80,2187843682556025,2187843682565519,2187843682585039,2187843682586092
@@ -1,2 +1,2 @@
 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,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,103052,103052,1048576,256,0,0,8,8,16,64,0x0,0x7fe965116e80,47544,47544,16384,65536,13103,1693740,4876162203440,4881310843226,4881310867386,4876170172288
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2847318,2847318,1048576,256,0,0,8,8,16,64,0x0,0x7f6648898e80,47274,47274,16384,65536,13112,1663272,2187803528527851,2187803983015672,2187803983039992,2187803536033431
@@ -1,2 +1,2 @@
 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,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,101830,101830,1048576,256,0,0,8,8,16,64,0x0,0x7fb30b508e80,0,0,0,4870225346608,4881310843226,4881310867386,4870233229204
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2846769,2846769,1048576,256,0,0,8,8,16,64,0x0,0x7fe51c612e80,0,0,0,2187801917565820,2187803983015672,2187803983039992,2187801925224307
@@ -1,2 +1,2 @@
 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,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,100812,100812,1048576,256,0,0,8,8,16,64,0x0,0x7f3f44bd2e80,65536,195450,25045552,4865251370909,4881310843226,4881310867386,4865258954789
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2849331,2849331,1048576,256,0,0,8,8,16,64,0x0,0x7f6f424dce80,65536,205200,26259424,2187809164143107,2187803983015672,2187803983039992,2187809171645862
@@ -1,2 +1,2 @@
 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,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,103458,103458,1048576,256,0,0,8,8,16,64,0x0,0x7fa2e8b18e80,32768,633554,81092720,4878063446499,4881310843226,4881310867386,4878071065706
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2849148,2849148,1048576,256,0,0,8,8,16,64,0x0,0x7f37e6006e80,32768,652733,83542268,2187808723799978,2187803983015672,2187803983039992,2187808731263349
@@ -1,2 +1,2 @@
 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,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,"vecCopy(double*, double*, double*, int, int) ",2,0,0,99996,99996,1048576,256,0,0,8,8,16,64,0x0,0x7fec3b190e80,47808,47808,16527,382472,16384,24462222,233988,0,98360336,4861216749685,4881310843226,4881310867386,4861224344736
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2847684,2847684,1048576,256,0,0,8,8,16,64,0x0,0x7f54e0f84e80,47118,47118,17095,376952,16384,24302477,224867,0,97701064,2187804428513108,2187803983015672,2187803983039992,2187804436177596
@@ -1,2 +0,0 @@
-Dispatch_ID,Kernel_Name,GPU_ID
-0,"vecCopy(double*, double*, double*, int, int) ",2
@@ -1,2 +1,2 @@
-workload_name,command,host_name,host_cpu,sbios,host_distro,host_kernel,host_rocmver,date,gpu_soc,vbios,numSE,numCU,numSIMD,waveSize,maxWavesPerCU,maxWorkgroupSize,L1,L2,sclk,mclk,cur_sclk,cur_mclk,L2Banks,totalL2Banks,LDSBanks,name,numSQC,numPipes,hbmBW,compute_partition,memory_partition,ip_blocks
-dispatch_0,./tests/vcopy -n 1048576 -b 256 -i 3,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_64,5.7.1-98,Mon Feb 19 12:59:52 2024 (CST),gfx908,113-D3431401-100,8,120,4,64,40,1024,16,8192,1502,1200,1502,1200,32,32,32,MI100,30,4,1200,,,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
+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
+dispatch_0,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF,Fri Mar  1 16:45:11 2024 (CST),1,t006-003.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,,527651240,,5.7.1-98,113-D3431401-100,NA,NA,MI100,gfx908,16,8192,120,4,8,64,1024,40,1502,1200,1502,1200,32,32,64,4,1228.8,1
@@ -1,2 +1,2 @@
 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) ",2,0,0,104067,104067,1048576,256,0,0,8,8,16,64,0x0,0x7fb012a3ce80,4881310814096,4881310843226,4881310867386,4881310880141
+0,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,0,2847501,2847501,1048576,256,0,0,8,8,16,64,0x0,0x7face032ee80,2187803982990937,2187803983015672,2187803983039992,2187803983054696
@@ -1,3 +1,3 @@
 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,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,106306,106306,1048576,256,0,0,8,8,16,64,0x0,0x7f07ca94ee80,47170,47170,16384,65536,14343,1843816,4893155122265,4896137259986,4896137284466,4893162800916
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,106306,106306,1048576,256,0,0,8,8,16,64,0x0,0x7f07ca94ee80,41313,41313,16384,65536,8217,1048584,4893162823870,4896137389586,4896137408466,4893163144176
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2854821,2854821,1048576,256,0,0,8,8,16,64,0x0,0x7fd201bc8e80,48587,48587,16384,65536,13335,1699196,2187825353651312,2187817997797403,2187817997821723,2187825361131083
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2854821,2854821,1048576,256,0,0,8,8,16,64,0x0,0x7fd201bc8e80,43788,43788,16384,65536,8358,1048576,2187825361154758,2187817997919963,2187817997938843,2187825361465401
@@ -1,3 +1,3 @@
 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,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,108543,108543,1048576,256,0,0,8,8,16,64,0x0,0x7ffb9d048e80,0,0,0,4903924355712,4896137259986,4896137284466,4903931980111
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,108543,108543,1048576,256,0,0,8,8,16,64,0x0,0x7ffb9d048e80,0,0,0,4903931998325,4896137389586,4896137408466,4903932324242
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2852254,2852254,1048576,256,0,0,8,8,16,64,0x0,0x7fdce63b0e80,0,0,0,2187818435420620,2187817997797403,2187817997821723,2187818443006390
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2852254,2852254,1048576,256,0,0,8,8,16,64,0x0,0x7fdce63b0e80,0,0,0,2187818443027440,2187817997919963,2187817997938843,2187818443325199
@@ -1,3 +1,3 @@
 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,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,105885,105885,1048576,256,0,0,8,8,16,64,0x0,0x7f06d787ae80,65536,221164,28292080,4891310211821,4896137259986,4896137284466,4891317996641
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,105885,105885,1048576,256,0,0,8,8,16,64,0x0,0x7f06d787ae80,65536,172286,22054296,4891318015607,4896137389586,4896137408466,4891318319862
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2853357,2853357,1048576,256,0,0,8,8,16,64,0x0,0x7fece8768e80,65536,177888,22849344,2187821117349071,2187817997797403,2187817997821723,2187821124970508
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2853357,2853357,1048576,256,0,0,8,8,16,64,0x0,0x7fece8768e80,65536,167696,21526976,2187821124989614,2187817997919963,2187817997938843,2187821125313562
@@ -1,3 +1,3 @@
 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,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,107122,107122,1048576,256,0,0,8,8,16,64,0x0,0x7f8c2ec5ce80,32768,652466,83517588,4897035318717,4896137259986,4896137284466,4897042979986
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,107122,107122,1048576,256,0,0,8,8,16,64,0x0,0x7f8c2ec5ce80,32768,666290,85282176,4897042999433,4896137389586,4896137408466,4897043314900
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2852991,2852991,1048576,256,0,0,8,8,16,64,0x0,0x7f100dc28e80,32768,657746,84179812,2187820219055599,2187817997797403,2187817997821723,2187820226818161
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2852991,2852991,1048576,256,0,0,8,8,16,64,0x0,0x7f100dc28e80,32768,678033,86771904,2187820226837949,2187817997919963,2187817997938843,2187820227160915
@@ -1,3 +1,3 @@
 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,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,"vecCopy(double*, double*, double*, int, int) ",2,0,0,105274,105274,1048576,256,0,0,8,8,16,64,0x0,0x7f5a39766e80,48042,48042,17103,384344,16384,24957228,232500,0,100356280,4888569436024,4896137259986,4896137284466,4888577137587
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,105274,105274,1048576,256,0,0,8,8,16,64,0x0,0x7f5a39766e80,41327,41327,12734,330624,16384,23699957,225369,0,95346784,4888577166532,4896137389586,4896137408466,4888577502207
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2851521,2851521,1048576,256,0,0,8,8,16,64,0x0,0x7fa106240e80,47947,47947,16547,383584,16384,24538596,231412,0,98657844,2187816647872285,2187817997797403,2187817997821723,2187816655447195
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2851521,2851521,1048576,256,0,0,8,8,16,64,0x0,0x7fa106240e80,42133,42133,13660,337072,16384,24431973,228800,0,98212048,2187816655475638,2187817997919963,2187817997938843,2187816655805868
@@ -1,3 +0,0 @@
-Dispatch_ID,Kernel_Name,GPU_ID
-0,"vecCopy(double*, double*, double*, int, int) ",2
-1,"vecCopy(double*, double*, double*, int, int) ",2
@@ -1,2 +1,2 @@
-workload_name,command,host_name,host_cpu,sbios,host_distro,host_kernel,host_rocmver,date,gpu_soc,vbios,numSE,numCU,numSIMD,waveSize,maxWavesPerCU,maxWorkgroupSize,L1,L2,sclk,mclk,cur_sclk,cur_mclk,L2Banks,totalL2Banks,LDSBanks,name,numSQC,numPipes,hbmBW,compute_partition,memory_partition,ip_blocks
-dispatch_0_1,./tests/vcopy -n 1048576 -b 256 -i 3,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_64,5.7.1-98,Mon Feb 19 13:00:17 2024 (CST),gfx908,113-D3431401-100,8,120,4,64,40,1024,16,8192,1502,1200,1502,1200,32,32,32,MI100,30,4,1200,,,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
+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
+dispatch_0_1,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF,Fri Mar  1 16:45:25 2024 (CST),1,t006-003.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,,527651240,,5.7.1-98,113-D3431401-100,NA,NA,MI100,gfx908,16,8192,120,4,8,64,1024,40,1502,1200,1502,1200,32,32,64,4,1228.8,1
@@ -1,3 +1,3 @@
 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) ",2,0,0,106919,106919,1048576,256,0,0,8,8,16,64,0x0,0x7f8475eece80,4896137232330,4896137259986,4896137284466,4896137296401
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,106919,106919,1048576,256,0,0,8,8,16,64,0x0,0x7f8475eece80,4896137297253,4896137389586,4896137408466,4896137410026
+0,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,0,2852071,2852071,1048576,256,0,0,8,8,16,64,0x0,0x7f0f1fe14e80,2187817997771027,2187817997797403,2187817997821723,2187817997835769
+1,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,2,2852071,2852071,1048576,256,0,0,8,8,16,64,0x0,0x7f0f1fe14e80,2187817997834266,2187817997919963,2187817997938843,2187817997944092
@@ -1,2 +1,2 @@
 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,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,95615,95615,1048576,256,0,0,8,8,16,64,0x0,0x7f8cc503ee80,38738,38738,16384,65536,8147,1048608,4839256684968,4855077119856,4855077138416,4839264410295
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2845401,2845401,1048576,256,0,0,8,8,16,64,0x0,0x7fcadc3dce80,38565,38565,16384,65536,8226,1048620,2187796746637903,2187794678392809,2187794678412009,2187796754235446
@@ -1,2 +1,2 @@
 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,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,95412,95412,1048576,256,0,0,8,8,16,64,0x0,0x7f9b7976ee80,0,0,0,4838332549840,4855077119856,4855077138416,4838340188534
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2843019,2843019,1048576,256,0,0,8,8,16,64,0x0,0x7febe3572e80,0,0,0,2187790583277714,2187794678392809,2187794678412009,2187790590832306
@@ -1,2 +1,2 @@
 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,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,96834,96834,1048576,256,0,0,8,8,16,64,0x0,0x7feee7c56e80,65536,86724,11090536,4845394358857,4855077119856,4855077138416,4845401906929
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2843570,2843570,1048576,256,0,0,8,8,16,64,0x0,0x7f59f3376e80,65536,89910,11518096,2187791957065822,2187794678392809,2187794678412009,2187791964652224
@@ -1,2 +1,2 @@
 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,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,98257,98257,1048576,256,0,0,8,8,16,64,0x0,0x7feaac5dce80,32768,648935,83060448,4851859046316,4855077119856,4855077138416,4851866625217
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2842271,2842271,1048576,256,0,0,8,8,16,64,0x0,0x7fd8069d0e80,32768,639748,81911640,2187788268757096,2187794678392809,2187794678412009,2187788276577677
@@ -1,2 +1,2 @@
 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,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
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,96022,96022,1048576,256,0,0,8,8,16,64,0x0,0x7f9cc9a4ae80,39238,39238,12449,313912,16384,21856220,212342,0,87959476,4841278994784,4855077119856,4855077138416,4841286607339
+1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,2844302,2844302,1048576,256,0,0,8,8,16,64,0x0,0x7f98b83bee80,39353,39353,12306,314832,16384,22386567,216458,0,90050164,2187793773743550,2187794678392809,2187794678412009,2187793781232750
@@ -1,2 +0,0 @@
-Dispatch_ID,Kernel_Name,GPU_ID
-1,"vecCopy(double*, double*, double*, int, int) ",2
@@ -1,2 +1,2 @@
-workload_name,command,host_name,host_cpu,sbios,host_distro,host_kernel,host_rocmver,date,gpu_soc,vbios,numSE,numCU,numSIMD,waveSize,maxWavesPerCU,maxWorkgroupSize,L1,L2,sclk,mclk,cur_sclk,cur_mclk,L2Banks,totalL2Banks,LDSBanks,name,numSQC,numPipes,hbmBW,compute_partition,memory_partition,ip_blocks
-dispatch_2,./tests/vcopy -n 1048576 -b 256 -i 3,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_64,5.7.1-98,Mon Feb 19 12:59:27 2024 (CST),gfx908,113-D3431401-100,8,120,4,64,40,1024,16,8192,1502,1200,1502,1200,32,32,32,MI100,30,4,1200,,,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
+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
+dispatch_2,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF,Fri Mar  1 16:44:58 2024 (CST),1,t006-003.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,,527651240,,5.7.1-98,113-D3431401-100,NA,NA,MI100,gfx908,16,8192,120,4,8,64,1024,40,1502,1200,1502,1200,32,32,64,4,1228.8,1
@@ -1,2 +1,2 @@
 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
-1,"vecCopy(double*, double*, double*, int, int) ",2,0,2,98866,98866,1048576,256,0,0,8,8,16,64,0x0,0x7fc716e2ee80,4855077106740,4855077119856,4855077138416,4855077149110
+1,"vecCopy(double*, double*, double*, int, int) [clone .kd]",2,0,2,2844668,2844668,1048576,256,0,0,8,8,16,64,0x0,0x7f3fa4688e80,2187794678379498,2187794678392809,2187794678412009,2187794678425044
@@ -1,2 +1,2 @@
-workload_name,command,host_name,host_cpu,sbios,host_distro,host_kernel,host_rocmver,date,gpu_soc,vbios,numSE,numCU,numSIMD,waveSize,maxWavesPerCU,maxWorkgroupSize,L1,L2,sclk,mclk,cur_sclk,cur_mclk,L2Banks,totalL2Banks,LDSBanks,name,numSQC,numPipes,hbmBW,compute_partition,memory_partition,ip_blocks
-dispatch_6_8,"./tests/vcopy -n 1048576 -b 256 -i 3",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_64,5.7.1-98,Mon Feb 19 13:26:04 2024 (CST),gfx908,113-D3431401-100,8,120,4,64,40,1024,16,8192,1502,1200,1502,1200,32,32,32,MI100,30,4,1200,NA,NA,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
+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
+dispatch_6_8,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF,Fri Mar  1 16:48:07 2024 (CST),1,t006-003.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,,527651240,,5.7.1-98,113-D3431401-100,NA,NA,MI100,gfx908,16,8192,120,4,8,64,1024,40,1502,1200,1502,1200,32,32,64,4,1228.8,1
@@ -1,2 +1,2 @@
-workload_name,command,host_name,host_cpu,sbios,host_distro,host_kernel,host_rocmver,date,gpu_soc,vbios,numSE,numCU,numSIMD,waveSize,maxWavesPerCU,maxWorkgroupSize,L1,L2,sclk,mclk,cur_sclk,cur_mclk,L2Banks,totalL2Banks,LDSBanks,name,numSQC,numPipes,hbmBW,compute_partition,memory_partition,ip_blocks
-dispatch_7,"./tests/vcopy -n 1048576 -b 256 -i 3",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_64,5.7.1-98,Mon Feb 19 13:24:53 2024 (CST),gfx908,113-D3431401-100,8,120,4,64,40,1024,16,8192,1502,1200,1502,1200,32,32,32,MI100,30,4,1200,NA,NA,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF
+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
+dispatch_7,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF,Fri Mar  1 16:44:46 2024 (CST),1,t006-003.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,,527651240,,5.7.1-98,113-D3431401-100,NA,NA,MI100,gfx908,16,8192,120,4,8,64,1024,40,1502,1200,1502,1200,32,32,64,4,1228.8,1
@@ -1,2 +1,2 @@
 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,GRBM_COUNT,GRBM_GUI_ACTIVE,SQ_WAVES,SQ_IFETCH,SQ_IFETCH_LEVEL,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,196371,196371,1048576,256,0,0,8,8,16,64,0x0,0x7f741569ee80,48123,48123,16384,65536,13130,1706668,6474031491551,6478590216987,6478590240667,6474038988830
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2802076,2802076,1048576,256,0,0,8,8,16,64,0x0,0x7f75dfc16e80,49408,49408,16384,65536,14474,1839408,2187670903710904,2187670460808016,2187670460832816,2187670911232314
@@ -1,2 +1,2 @@
 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,SQ_INSTS_LDS,SQ_INST_LEVEL_LDS,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,195356,195356,1048576,256,0,0,8,8,16,64,0x0,0x7f4d4f096e80,0,0,0,6469038131727,6478590216987,6478590240667,6469045656948
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2800963,2800963,1048576,256,0,0,8,8,16,64,0x0,0x7f52a8504e80,0,0,0,2187667942143695,2187670460808016,2187670460832816,2187667949765323
@@ -1,2 +1,2 @@
 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,SQ_INSTS_SMEM,SQ_INST_LEVEL_SMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,197183,197183,1048576,256,0,0,8,8,16,64,0x0,0x7f7d30072e80,65536,190396,24388520,6477678328572,6478590216987,6478590240667,6477685882998
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2799865,2799865,1048576,256,0,0,8,8,16,64,0x0,0x7fdba7ec8e80,65536,212188,27179520,2187664983133496,2187670460808016,2187670460832816,2187664990601125
@@ -1,2 +1,2 @@
 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,SQ_INSTS_VMEM,SQ_INST_LEVEL_VMEM,SQ_ACCUM_PREV_HIRES,DispatchNs,Start_Timestamp,End_Timestamp,CompleteNs
-0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,193326,193326,1048576,256,0,0,8,8,16,64,0x0,0x7f9a2967ce80,32768,609182,77979468,6458900830969,6478590216987,6478590240667,6458908615711
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2800048,2800048,1048576,256,0,0,8,8,16,64,0x0,0x7f49a05c0e80,32768,657081,84118816,2187665433303653,2187670460808016,2187670460832816,2187665440779667
@@ -1,2 +1,2 @@
 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,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,"vecCopy(double*, double*, double*, int, int) ",2,0,0,194950,194950,1048576,256,0,0,8,8,16,64,0x0,0x7fe35c9ece80,47398,47398,18065,379192,16384,24324908,230465,0,97830880,6466970179659,6478590216987,6478590240667,6466977946989
+0,"vecCopy(double*, double*, double*, int, int) ",2,0,0,2800414,2800414,1048576,256,0,0,8,8,16,64,0x0,0x7f3a26ac0e80,47715,47715,15910,381728,16384,25251126,233100,0,101488776,2187666341957413,2187670460808016,2187670460832816,2187666349596293
@@ -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
+dispatch_inv,./tests/vcopy -n 1048576 -b 256 -i 3,SQ|LDS|SQC|TA|TD|TCP|TCC|SPI|CPC|CPF,Fri Mar  1 16:42:50 2024 (CST),1,t006-003.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,,527651240,,5.7.1-98,113-D3431401-100,NA,NA,MI100,gfx908,16,8192,120,4,8,64,1024,40,1502,1200,1502,1200,32,32,64,4,1228.8,1
--- a/Show More
+++ b/Show More