Nvrtc compiler summary

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

one simple question regarding the output of the nvrtc compiler. With cuda version 9 or 10 (not sure) the nvrtc compiler prints some useful information to the compiler output window (Windows/VS2022). It shows information like global/shared/const memory usage, registers per kernel and much more. Is there any way to get this information back with the latest nvrtc?

Thanks a lot.

2

I think for those things you would want to specify the ptxas option -v . (edited for future readers, I had mistakenly used capital V)

3

Thanks for your reply. I’m not able to get this working. Here are the options I use - which works without the --ptxas-options option set. The last lines are the output from the compiler:

nvrtc ver: 12.3

-arch=sm_86
-D_CUDA
-D_HAVE128
-restrict
-std=c++20
-extra-device-vectorization
-device-int128
-D=_DEBUG
-lineinfo
--ptxas-options=-V

ptxas fatal : Internal error: exit status 0

Using ‘V’ as option gives a different error:

-arch=sm_61
-D_CUDA
-D_HAVE128
-restrict
-std=c++20
-extra-device-vectorization
-device-int128
-D=_DEBUG
-lineinfo
--ptxas-options=V

ptxas fatal : Input file ‘V’ could not be opened
ptxas fatal : Ptx assembly aborted due to errors

Thanks.

4

I think it needs to be a lower case “v”.

5

Hi,

yes, that seems to be correct - below from NVIDIA CUDA Compiler Driver:

4.2.9.1.22. --verbose (-v)

Enable verbose mode which prints code generation statistics.

4.2.9.1.23. --version (-V)

Semantics same as nvcc option --version.

For all how run into the same problem - this is the option I use and I do get what I was hoping for.

--ptxas-options=-v

Thanks to all.

6

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