Skip to content

Pipe quantize kernel through FusionExecutorCache#4760

Merged
zasdfgbnm merged 95 commits intomainfrom
q-fec
Jul 24, 2025
Merged

Pipe quantize kernel through FusionExecutorCache#4760
zasdfgbnm merged 95 commits intomainfrom
q-fec

Conversation

@zasdfgbnm
Copy link
Collaborator

@zasdfgbnm zasdfgbnm commented Jul 9, 2025

Stacked on #4748

@zasdfgbnm
Copy link
Collaborator Author

!test

@zasdfgbnm
Copy link
Collaborator Author

!test

@zasdfgbnm
Copy link
Collaborator Author

!test

@zasdfgbnm
Copy link
Collaborator Author

!test

Base automatically changed from fp4-cast-fec to main July 11, 2025 19:42
@zasdfgbnm
Copy link
Collaborator Author

!test

@zasdfgbnm
Copy link
Collaborator Author

!test

@zasdfgbnm zasdfgbnm requested a review from naoyam July 15, 2025 23:50
namespace nvfuser {

bool MatmulScheduler::canScheduleCompileTime(Fusion* fusion) {
for (auto tv : fusion->allTvs()) {
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm fine with this, but we could also do this uniformly at https://github.com/NVIDIA/Fuser/blob/main/csrc/scheduler/registry.cpp#L25.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would suggest doing it in each scheduler, this way, in the future, we can just remove this per scheduler when we incrementally enable support for fp4.

TensorDomain::getContiguityFilledWith(logical_domain, true)),
x->getDataType().value());
IrBuilder::create<ViewOp>(x, out_tv);
IrBuilder::create<ViewOp>(out_tv, x);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ha, so this has been wrong, right? Thanks for fixing it.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You are right. It's a bug. It was not discovered until we started running the fusion definition.

Copy link
Collaborator

@naoyam naoyam left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@zasdfgbnm zasdfgbnm merged commit a70e0f5 into main Jul 24, 2025
46 of 49 checks passed
@zasdfgbnm zasdfgbnm deleted the q-fec branch July 24, 2025 18:17
wujingyue added a commit that referenced this pull request Jul 26, 2025
This reverts commit a70e0f5.

It broke
```
_bn && pytest tests/python/opinfo/test_legacy_ops.py -k test_correctness_abs_float64 -s
```
nsarka pushed a commit to nsarka/Fuser that referenced this pull request Jul 28, 2025
nsarka pushed a commit to nsarka/Fuser that referenced this pull request Jul 28, 2025
zasdfgbnm added a commit that referenced this pull request Jul 28, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants