I’m using CUDA 11.5 and Visual Studio 2019 Professional (16.9.12). I’m running on an RTX Virtual Workstation (see here: AWS Marketplace: NVIDIA RTX Virtual Workstation - WinServer 2019) on AWS with a g4dn.xlarge instance (Has a Tesla T4). My project includes the glm library and I’m using glm vectors and matrices extensively.
When I set a breakpoint within my kernel and run the Next-Gen debugger, I can’t inspect many of the values of local objects. This includes glm objects, but also structs that I’ve written. The values seem to be set correctly since I can manipulate them and store them in more basic data types. However, when I try to inspect structs and glm objects, values show up as { =unexpected .} or {???}
I didn’t experience this when running the Legacy debugger on my laptop. What could be the issue?
Same problem here with VS2022 (17.8.10) and CUDA 12.4.
I can’t see the values for GLM vector (glm::vec3) and other GLM data structures.
Trying to debug with a DELL laptop on a NVIDIA RTX A2000 laptop GPU.
I managed to see the glm::vec3 coordinates by the following debugger watch line “(__ global __ float*)p1)”. To see local variables you need to add “__ local __”.
By setting the vector display elements in the NSight options to 3 I can see something similar to:
0xb04a00200 { 50.0000000, 0.00000000, 0.00000000 }|float*|
||[0]|50.0000000|float|
||[1]|0.00000000|float|
||[2]|0.00000000|float|
However, it is very unpractical.
Did you find the issue with GLM variable values (glm::fvec3 vector for exemple) being always replaced with “unexpected” in Nsight Debugging Next Gen ?
I’m using latest VS2022 with latest CUDA (12.6)
I can’t debug anything with all that “unexpected” values in my breakpoints :(
I have a Dell laptop with a Nvidia GeForce MX250 card compatible with Cuda
I have included glm like that in my c++ project : define CUDA_VERSION 12000
// Ensure GLM configuration is compatible with CUDA define GLM_FORCE_CUDA include <glm/glm.hpp>