rocm-systems/tests/workloads/vcopy/MI100/pmc_perf_7.csv

1	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_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	DispatchNs	Start_Timestamp	End_Timestamp	CompleteNs
2	0	vecCopy(double*, double*, double*, int, int)	2	0	0	2503406	2503406	1048576	256	0	0	8	8	16	64	0x0	0x7fe43a3dce80	0	0	0	16384	0	0	0	0	0.0	0.0	0.0	0.0	16384.0	0.0	0	81920	0.0	0.0	28555.0	123147.0	5255473651421229	5255477523602709	5255477523626869	5255473659148252
3	1	vecCopy(double*, double*, double*, int, int)	2	0	2	2503406	2503406	1048576	256	0	0	8	8	16	64	0x0	0x7fe43a3dce80	0	0	0	16384	0	0	0	0	0.0	0.0	0.0	0.0	16384.0	0.0	0	81920	0.0	0.0	24406.0	117740.0	5255473659556374	5255477523727189	5255477523745909	5255473660045560
4	2	vecCopy(double*, double*, double*, int, int)	2	0	4	2503406	2503406	1048576	256	0	0	8	8	16	64	0x0	0x7fe43a3dce80	0	0	0	16384	0	0	0	0	0.0	0.0	0.0	0.0	16384.0	0.0	0	81920	0.0	0.0	30143.0	117619.0	5255473660464542	5255477523767029	5255477523785909	5255473660902821