Skip to content

Pull requests: NVIDIA/Fuser

New pull request New
198 Open 4,599 Closed
198 Open 4,599 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

optimize non-circular buffered tma load
#5869 opened Jan 23, 2026 by liqiangxl • Draft
3
Optimize TMA inner-reduction and add TMA serial-split
#5867 opened Jan 23, 2026 by tbqh Loading…
12
[IR Refactor] Phase 1: IrContainer/IrStorage Composition Pattern
#5865 opened Jan 22, 2026 by mdavis36 Loading…
28
Add kernel based alltoallv and cuda backend for MoE dispatch and combine
#5863 opened Jan 22, 2026 by samnordmann Loading…
4
tests: update exceptions include
#5859 opened Jan 21, 2026 by wujingyue Loading…
3
MoE Dispatch/Combine first implementation for k=1 and Nccl backend
#5857 opened Jan 21, 2026 by samnordmann Loading…
26
ensure iter domain is divisible by iter unroll factor in tma inner persistent scheduler
#5853 opened Jan 20, 2026 by liqiangxl Loading…
9
Remove unused setDeviceIndex in FusionKernelRuntime::prepareInputs
#5850 opened Jan 20, 2026 by zasdfgbnm • Draft
2
Use symmetric memory in matmul+allreduce reference implementation
#5837 opened Jan 17, 2026 by wujingyue • Draft
5
Remove Ampere-Turing matmul scheduler
#5836 opened Jan 16, 2026 by naoyam Loading…
6
avoid warp diverge in warp specialized kernel
#5830 opened Jan 15, 2026 by liqiangxl Loading…
6
[WIP] Use TensorIndexer by default
#5828 opened Jan 15, 2026 by naoyam • Draft
4
RaggedIterDomain partitioning with multi-dim extents tensor list
#5823 opened Jan 14, 2026 by naoyam Loading…
14
Change from thread idx based predicate to warp idx based
#5821 opened Jan 14, 2026 by liqiangxl • Draft
11
Fix deprecated LLVM PointerType::getUnqual(Type*) for opaque pointer compatibility
#5820 opened Jan 14, 2026 by IvanYashchuk Loading…
3 tasks done
7
Modernize DynamicType with C++20 concepts
#5807 opened Jan 13, 2026 by zasdfgbnm • Draft
3
Replace tuple-based type iteration with C++20 fold expressions
#5798 opened Jan 12, 2026 by csarofeen • Draft
2
[Build Speed] PCH Test
#5795 opened Jan 11, 2026 by csarofeen • Draft
3
TMA inner persistent scheduler heuristics
#5784 opened Jan 9, 2026 by liqiangxl • Draft
1
Migrate from pybind11 to nanobind Direct Bindings Python extension with direct mapping to NvFuser CPP objects.
#5780 opened Jan 8, 2026 by rdspring1 • Draft
3
support dynamic shapes in warp specialized inner outer persistent scheduler
#5765 opened Jan 6, 2026 by liqiangxl Loading…
11
prioritze using smem only when cached inputs are persistent
#5764 opened Jan 6, 2026 by liqiangxl • Draft
9
Add a toy multi-GPU benchmark
#5753 opened Jan 3, 2026 by wujingyue • Draft
13
[Build Speed][WIP] Dynamnic Type, Polymorphic Value, and Precompiled Headers
#5747 opened Jan 2, 2026 by csarofeen • Draft
7
Benchmark for nvfp4 scaled mm
#5737 opened Dec 23, 2025 by protonu Loading…
14
ProTip! Mix and match filters to narrow down what you’re looking for.