Closed chinmaydd closed 3 weeks ago
ping. Could someone help us here? Thanks !
Adding @TorreZuk to see if there are any unforeseeable issues with this change.
Okay git blame confusion, I think the original .bat was buggy so the hipcc without extension was the wrapper for perl I used, thus the use of 'perl' + which( ). So when you later switched to .bat taking precedence the perl should have been removed. There is also a .exe now but without windows CI I will not suggest anything further.
Okay git blame confusion, I think the original .bat was buggy so the hipcc without extension was the wrapper for perl I used, thus the use of 'perl' + which( ). So when you later switched to .bat taking precedence the perl should have been removed. There is also a .exe now but without windows CI I will not suggest anything further.
We've been waiting on windows CI for a while... I don't recall what the reasoning was for the ordering but this changed when migrating to amdclang++.
Well fine with this merging to unblock, but think it probably worth a new tech. debt ticket to move off hipcc entirely. Was the reason this was left as amdclang++ couldn't provide the hip version info?
@TorreZuk thanks for the review. Could someone with access merge this? Thanks again !
@ellosel thanks! If you dont mind me asking, what is the tensile_tag.txt
update cadence in rocBLAS ?
Summary: Invoke "hipcc.bat" script instead of "perl hipcc". Invoking "perl hipcc" is not ideal, because there is no guarantee that hipcc is actually a perl script. Submitting in lieu of https://github.com/ROCm/Tensile/pull/2037 .
Outcomes:
What is the result of this change? What components of the project does it affect?
At this point this is NFC - hipcc.bat script calls perl script.
Notable changes: N/A
Testing and Environment:
What environment are you targeting (OS, ROCm version, Python versions, etc.)? Windows