Hi It seems that smsp__inst_executed.sum is not the total executed instructions as I expect because as you can see, it is less that control and ldst instructions. smsp__inst_executed.sum inst 766803.000000 788421.000000 777612.000000 smsp__sass_thread_inst_executed_op_control_pred_on.sum inst 5500945.000000 5504005.000000 5502475.000000 smsp__sass_thread_inst_executed_op_memory_pred_on.sum inst 3948027.000000 3953653.000000 3950840.000000 I only can think that smsp__inst_executed reflects warp instructions. So total of thread instructions is 766803*32. Any thought on that?

Thanks it seems that I didn’t get notified (or missed that). Please delete this topic.

Getting number of total executed instructions

felix_dt May 28, 2021, 12:04pm 2

Topic		Replies	Views
Inst_executed and thread_inst_executed Nsight Compute	3	2328	September 17, 2020
Small number of thread instructions Nsight Compute	2	813	January 21, 2022
Difference sm__cycles_elapsed/smsp__cycles_elapsed and sm__inst_executed/smsp__inst_executed? Nsight Compute performance-metrics	5	2334	February 2, 2022
Doubt regarding definition of "inst_executed" metric - nvprof Visual Profiler and nvprof	1	1400	July 5, 2017
Instructions Executed per Thread Teaching & Curriculum Support	0	1052	June 4, 2015
Question about number of instructions Nsight Compute	1	639	June 24, 2022
Difference between thread_inst_executed metrics Nsight Compute performance-metrics	5	1216	June 27, 2022
Question about smsp and sm Nsight Compute	4	1786	September 17, 2020
“inst_executed” metric on - nvprof. What does it mean? Visual Profiler and nvprof	0	630	March 22, 2021
About instruction per warp metric CUDA Programming and Performance	5	1028	March 24, 2020