Hi, i use nsight compute to analysis memory workload, and i am confused about metric lts__t_sectors_aperture_sysmem_op_read.sum and lts__d_sectors_fill_sysmem.sum. What is the difference?
I found some tips about these metric:
- lts__d_sectors_fill_sysmem.sum: sectors filled from system memory
- lts__t_sectors_aperture_sysmem_op_read.sum: of LTS sectors accessing system memory (sysmem) for read
If i want to know dram(sysmem) read transactions, which metric is correct? i found from https://docs.nvidia.com/nsight-compute/NsightComputeCli/index.html nsight compute metric lts__t_sectors_aperture_sysmem_op_read.sum is euqal to sysmem_read_transactions , but my test result is not like this.
nvprof sysmem_read_transactions value is 73727
nsight compute metric lts__d_sectors_fill_sysmem.sum is 79264
nsight compute metric lts__t_sectors_aperture_sysmem_op_read.sum is 409745
maybe lts__d_sectors_fill_sysmem.sum is equal to sysmem_read_transactions
Thanks