I have installed VSCE in my VS Code. I have set up my launch as follows:
{
// Use IntelliSense to learn about possible attributes.
// Hover to view descriptions of existing attributes.
// For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
"version": "0.2.0",
"configurations": [
{
"name": "(cuda-gdb) Launch Gelu",
"type": "cuda-gdb",
"request": "launch",
"program": "${workspaceFolder}/tensorflowopgeluexample/gelu_kernel",
"args": "",
"breakOnLaunch": true,
"cwd": "${workspaceFolder}/tensorflowopgeluexample"
},
{
"name": "(cuda-gdb) Launch Leaf Algo",
"type": "cuda-gdb",
"request": "launch",
"program": "${workspaceFolder}/Virgo_Algo/Blazer/MercuryImageComputer/KT/leaf/M31/BrightField/bin/LeafStandalone/Debug/12/LeafStandAlone.x86-64",
"args": "-noForcedPatches ${workspaceFolder}/virgo_algo_preq/data/JobInfo_108",
"stopAtEntry": true,
"cwd": "${workspaceFolder}/Virgo_Algo/Blazer/MercuryImageComputer/KT/leaf/M31/BrightField/bin/LeafStandalone/Debug/12"
}
]
}
I can use the debug functionality to debug my application within VS Code. However, I found that the cuda-gdb started by VS Code is not as flexible as started from command line. For instance, as started from command line, cuda-gdb can switch between blocks and threads within a kernel and print local variables, as indicated by the following link:
https://developer.download.nvidia.com/GTC/PDF/1062_Satoor.pdf
The same cuda-gdb started from VS Code cannot achieve that. Is there a way to set up VS Code so that cuda-gdb is more flexible from the GUI?