-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Pull requests: NVIDIA/cutlass
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[Bugfix][Blackwell] Avoid fmha_bwd workspace int32 overflow
#3614
opened Sep 11, 2026 by
XFDG
Loading…
[Profiler] Skip cuBLAS verification when C and D element types differ
#3613
opened Sep 11, 2026 by
LiRunGuo
Loading…
Restore TMEM allocator class constants for backward compatibility
#3611
opened Sep 11, 2026 by
timmoon10
Member
Loading…
[GitHub CI] Add CUTLASS test infrastructure and CuTeDSL/C++ coverage
#3610
opened Sep 11, 2026 by
zekunf-nv
Collaborator
Loading…
Recover CUDA 12.9 contiguous grouped GEMM performance
#3608
opened Sep 10, 2026 by
myu-guo
Contributor
Loading…
[CuTeDSL] Prevent domain_offset overflow with static strides
#3607
opened Sep 10, 2026 by
kzos
Loading…
[CuTeDSL] Fix swizzled composed layout slice affine abort
#3605
opened Sep 9, 2026 by
Cansult
Loading…
fix: forward-declare copy atoms to make copy.hpp and prefetch.hpp self-contained
#3603
opened Sep 9, 2026 by
Sunjnn
Loading…
Preserve active GQA mailbox contributions during initialization
#3601
opened Sep 8, 2026 by
1sgtpepper
Loading…
[CuTeDSL] reduce example: replace inline PTX with cute.arch.store_async_dsmem
#3593
opened Sep 7, 2026 by
zkyue
Contributor
Loading…
[CuTeDSL] Lower whole-warp warp_reduction to a single redux.sync when op/type/target allow
#3592
opened Sep 7, 2026 by
zkyue
Contributor
Loading…
Leave zero argument bool alone in the CuTeDSL preprocessor
#3590
opened Sep 6, 2026 by
Hashim1999164
Loading…
fix(CuTeDSL): optimize positional-only functions with TVMFFIJitCompiledFunction to eliminate CPU overhead (#3527)
#3589
opened Sep 6, 2026 by
Ammar-Alnagar
Loading…
examples/64: relax exact CC 8.9 gate to inclusive check
#3586
opened Sep 5, 2026 by
SAR2652
Loading…
Size SM100/SM103 array GEMM tensormap workspace from the device SM count
#3584
opened Sep 4, 2026 by
elbourne12345
Loading…
Fix handling of JAX tensor types without TVM FFI and leading dim deduction for all-1-size tensors
#3583
opened Sep 4, 2026 by
dextero
Loading…
Fix Constexpr method receivers in dynamic control-flow regions
#3580
opened Sep 3, 2026 by
nodeeeeee
Loading…
Fix skipped workspace initialization for parallel split-K in GemmUniversalBase
#3578
opened Sep 1, 2026 by
AyaanFaisal21
Loading…
Previous Next
ProTip!
Follow long discussions with comments:>50.