Where can I find information regarding compiler output with -Xptxas=-v?

Accelerated Computing CUDA CUDA NVCC Compiler
1

Hello, I am trying to understand the kernel resource usage for my application. I use -Xptxas=-v to output kernel resource usage during compilation, but am not sure I understand the numbers entirely. For instance, for an output along the lines of below, is smem reported per block or SM? What about cmem? Are the spills per thread? Is there a guide online that has descriptions for these?

    64 bytes stack frame, 128 bytes spill stores, 164 bytes spill loads
ptxas info    : Used 64 registers, 7136 bytes smem, 432 bytes cmem[0], 64 bytes cmem[2]
2

smem and cmem is per block. spill stores and loads are the sum of the instructions IIRC.

1 Like
3

There’s an explanation in the NVCC docs, here.

1 Like
4

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.