Implementing lowering for calls #164
Conversation
parsifal-47
commented
Aug 15, 2024
parsifal-47
commented
- removing the assertion which disallowed tensors of 1x1, I do not see why it was set, it seems that removing does not produce an error, let me know what you think. Thank you!
|
@nhat-nguyen please take a look when you have a chance, thank you! |
|
sure, give me some time to double check that removing the assert is ok |
great, no rush, thank you! |
|
@nhat-nguyen could you please take a look when you have a chance |
include/triton-shared/Conversion/TritonArithToLinalg/ConversionPatterns.hpp
Outdated
Show resolved
Hide resolved
include/triton-shared/Conversion/TritonArithToLinalg/ConversionPatterns.hpp
Outdated
Show resolved
Hide resolved
include/triton-shared/Conversion/TritonArithToLinalg/ConversionPatterns.hpp
Show resolved
Hide resolved
parsifal-47
force-pushed
the
add-call
branch
from
ed90909 to
0121084
Compare
|
@nhat-nguyen your comments for this PR should be resolved, thank you for reviewing! |
| if (argsNeed > args.size()) { | ||
| int missing = argsNeed - args.size(); | ||
| for (int i = 0; i < missing; i++) { | ||
| args.push_back(parentInputs[parentInputs.size() - i - 1]); |
There was a problem hiding this comment.
Hi parsifal, I wonder why the extra arguments are added in reverse order here. Six extra arguments are added by addProgramInfo, and here you add the sixth firstly, and the first finally. Why is the case?
There was a problem hiding this comment.
Hi, thank you for noticing! I do not see any reason for them to be reversed, and I looked at lit test, it clearly shows that they getting backwards, let me fix that quickly
There was a problem hiding this comment.
Thank you @Bolzano983
#200
