Passing flag to nvcc compiler in nsight

Accelerated Computing CUDA CUDA Programming and Performance
1

I am trying to show the registers per thread and shared memory per block of my GPU for a specific kernel. Using the command line nvcc i can do this by using the –ptxas-options=-v option.

But nsight IDE doesn’t allow me to manually put a flag to pass to nvcc. There is xptxas option in properties > build stages. I tried to put -v but it doesn’t show the above parameters which the –ptxas-options=-v shows.

How can i manually pass flag to nvcc in nsight IDE ?

2

nsight eclipse edition or nsight visual studio edition?

3

Nsight version 9.2 (eclipse)
Operating sys Ubuntu

4
  • For the project in question, select the project in the left hand side project explorer pane.
  • Right-click on the selected project, and select "Properties".
  • In the left hand side navigation pane of the dialog box that opens up, select "Build" to expand the sub-tree.
  • Under the "Build" sub-tree, click on "Settings"
  • On the right hand side, click on "Tool Settings" tab.
  • Within this tab, you will see another navigation pane on the left, select "Build Stages"
  • Now on the right hand side of this tab, you will see various options panes. The 4th options pane will be "ptaxas options -Xptxas" (yes it is misspelled). You cannot directly edit in this box, you need to look to the right where you see a small green plus sign, hover your mouse over this, the tooltip help should say "Add...". Click on this icon.
  • A new dialog box opens up. Enter -v into this editable box, click "OK"
  • Click "Apply"
  • Click "OK" to close the Properties dialog.
  • Select Build->Build Project
  • In the Console tab at the bottom, you should see the ptxas verbose output you asked for.
1 Like
5

Thank you.