Welcome to OStack Knowledge Sharing Community for programmer and developer-Open, Learning and Share
Welcome To Ask or Share your Answers For Others

Categories

0 votes
538 views
in Technique[技术] by (71.8m points)

profiling - Is it possible to collect performance metrics in CUDA per region inside the kernel?

I have the following fragment of code for which I want to count the number of executed instructions.

#include <cstdio>
#include <cuda.h>

__global__ void kernel() {
    double value = 10;
    // region 1 start
    for (int i = 0; i < 100000; i++) {
        value *= 1.000001;
    }
    // region 1 end
    // region 2 start
    for (int i = 0; i < 100000; i++) {
        value *= 1.000001;
    }
    // region 2 end
    printf("value: %lg
", value); // prevent compiler optimizations to remove entire kernel content
}

int main() {
    kernel<<<32,32>>>();
    cudaDeviceSynchronize();
}

I am doing it with nsight-compute using the following command:

nvcc sample.cu -arch=sm_75 && /opt/cuda/nsight_compute/ncu --metrics smsp__inst_executed ./a.out

As part of output I see the collected metric for the entire kernel:

  kernel(), 2021-Jan-24 12:10:11, Context 1, Stream 7
    Section: Command line profiler metrics
    ---------------------------------------------------------------------- --------------- ------------------------------
    smsp__inst_executed.avg                                                           inst                       41871.01
    smsp__inst_executed.max                                                           inst                         209356
    smsp__inst_executed.min                                                           inst                              0
    smsp__inst_executed.sum                                                           inst                        6699361
    ---------------------------------------------------------------------- --------------- ------------------------------

Is it possible to collect this metric for region 1 and 2 individually as commented in the code instead of the whole kernel?

question from:https://stackoverflow.com/questions/65869878/is-it-possible-to-collect-performance-metrics-in-cuda-per-region-inside-the-kern

与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…
Welcome To Ask or Share your Answers For Others

1 Answer

0 votes
by (71.8m points)
Waitting for answers

与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…
Welcome to OStack Knowledge Sharing Community for programmer and developer-Open, Learning and Share
Click Here to Ask a Question

...