You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: 06-H_Overlap_Communication_and_Computation_MPI/.master/Instructions.md
+2-1Lines changed: 2 additions & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -31,7 +31,8 @@ Use the Nsight System profiler to profile the starting point version non-Overlap
31
31
1. Start by compiling and running the application with `make run`
32
32
1. Record an Nsight Systems profile, using the appropriate Makefile target (`make profile`)
33
33
1. Open the recorded profile in the GUI
34
-
- Either: Install Nsight Systems locally, and transfer the .qdrep/.nsys-rep file
34
+
- Either: Install Nsight Systems locally, and transfer the .nsys-rep file.
35
+
-*Note*: Right-click in file-browser, choose "Download" from context menu
35
36
- Or: By running Xpra in your browser: In Jupyter, select "File > New Launcher" and "Xpra Desktop", which will open in a new tab. Don't forget to source the environment in your `xterm`.
36
37
1. Familiarize yourself with the different rows and the traces they represent.
37
38
- See if you can correlate a CUDA API kernel launch call and the resulting kernel execution on the device
0 commit comments