How to run hello_cuda ELF when using nvbitfi ?

[WilliamWangPeng]

Hi dear developer,
It's an honor to write a letter here, I have read README, but can't understand how to run another ELF like hello_cuda, I have followed the steps and build the test simple_add.
What can I do and where shall I change to test the hello_cuda ?

thank you
best regards
William

[sergicuen]

Hi William, this is a sort "guide for dummies" I wrote to do simple tests.

1: reuse the simple_add files a) create a new directory for your app in test_apps (e.g. hello_cuda) and copy your cuda code (hello_cuda.cu) b) copy the following files from simple_add : Makefile, run.sh and sdc_check.sh c) modify the files to point to your app

2: modify the script /scripts/params.py

• configure the injection campaign
• include your app in the apps list (check the expected runtime for long execution apps)

3: modify the script /nvbitfi/test.sh Step 0(4) : point to your app datapath

I hope you find it useful

[WilliamWangPeng]

hi dear @sergicuen
thank you for your kind advice and help 👍
I got the following datas in instruction parts in profiler folds

pr: 216452736 nodest: 276306816 others: 1511140512 gppr: 2379714624 gp: 2163261888

Could you please tell me what's the meaning of pr, nodest, gppr and gp ?

thank you
best regards
William

[sergicuen]

Hi, it is related to the instruction type. Take a look to the params.py for more information. regards