NVIDIA Fuser issues - Githubissues

NVIDIA / Fuser

A Fusion Code Generator for NVIDIA GPUs (commonly known as "nvFuser")

Other

271 stars 53 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Additional IDs are lost when mutating a TensorDomain

#3409 jacobhinkle closed 6 days ago
0
Clean-up: reduction producer IterDomains aren't mapped anyway.

#3408 wujingyue closed 6 days ago
1
Support 2D pointwise scheduler with autotuning

#3407 rdspring1 opened 1 week ago
1
[WIP] Schedule Hopper MMA without input broadcasts

#3406 jacobhinkle opened 1 week ago
0
Make sorting deterministic.

#3405 wujingyue closed 1 week ago
3
Parametrize Hopper matmul tests

#3404 jacobhinkle opened 1 week ago
1
Reject fusions with CPU scalar outputs in `KernelExecutor`

#3403 Priya2698 closed 2 days ago
12
Skip SetUp when the communicator isn't available.

#3402 wujingyue closed 1 week ago
1
Prevent marking inplace update target as candidate for alias analysis

#3401 jjsjann123 closed 6 days ago
5
warp-specialization

#3400 zasdfgbnm opened 1 week ago
0
AllocationDomainPass sets correct allocation orders for SdpaFwdOp/SdpaBwdOp.

#3399 wujingyue closed 1 week ago
3
Change tests to call runPass directly.

#3398 wujingyue closed 1 week ago
1
Replace PERMISSIVE graph with BROADCAST for matmul scheduler

#3397 jacobhinkle closed 1 week ago
1
Non-deterministic output tensor names

#3396 naoyam closed 1 week ago
2
Add support to store the outputs of the Mma operator using stmatrix

#3395 protonu opened 1 week ago
1
Add thunder benchmarks

#3394 Priya2698 opened 1 week ago
4
add knob controls unroll on top of vectorization in inner reduction

#3393 liqiangxl closed 1 week ago
3
Ring-based decomposition for Allgather+GEMM overlap ATen implementation

#3392 nsarka opened 1 week ago
0
Accept axis mapping when defining MmaOp

#3391 jacobhinkle closed 1 week ago
7
Test multiple memory formats

#3390 wujingyue closed 1 week ago
3
Failed to create/execute the same FusionDefinition multiple times.

#3389 wujingyue opened 1 week ago
0
Update input sizes for `test_many_segment_benchmark` to ensure kernel reuse

#3388 Priya2698 closed 1 week ago
1
Enable IdModel tensor indexer when expanded IDs are reshaped

#3387 naoyam closed 1 week ago
7
SdpaFwdOp::evaluate produces a tensor whose stride order doesn't match the output allocation domain.

#3386 wujingyue closed 1 week ago
3
Clean up IdModel options

#3385 naoyam closed 1 week ago
3
Remove debug prints

#3384 wujingyue closed 1 week ago
1
Use TMA with reduction for Hopper Split-K

#3383 jacobhinkle closed 1 week ago
2
Use structured binding for better readability

#3382 wujingyue closed 1 week ago
1
Remove MatMulTileOptions::instruction_tile

#3381 jacobhinkle opened 1 week ago
1
Rename addSetsForCacheReads to cacheOperandsToRegisters

#3380 jacobhinkle closed 1 week ago
1
in extent substitution preseg pass, use DisjointSets of extents which is constructed from DisjointSets of IterDomains

#3379 liqiangxl closed 1 week ago
5
return in()->toInlineString() in LoadStoreOp::toInlineString

#3378 liqiangxl closed 1 week ago
2
Change a functor to a regular function.

#3377 wujingyue closed 1 week ago
1
Inspect all IDs instead of just loop in ParallelDimensionMap

#3376 jacobhinkle closed 1 week ago
2
Validate allocation sizes and strides more widely.

#3375 wujingyue closed 1 week ago
4
Indexing error with HF's Qwen 2 model

#3374 naoyam closed 1 week ago
2
Fix the legacy loop indexing traversal

#3373 naoyam closed 1 week ago
2
Enable MmaOp to receive unbroadcasted inputs

#3372 jacobhinkle closed 1 week ago
0
Explicit check build flag

#3371 jjsjann123 closed 2 weeks ago
2
Redo #3326.

#3370 wujingyue closed 2 weeks ago
2
Two-hop mutations are not supported. Found registrations from 5 to 5 to 5

#3369 wujingyue closed 1 week ago
6
Add NVFUSER_DUMP=python_definition_segments

#3368 jacobhinkle closed 2 weeks ago
3
[Do not merge] Overlap AG+GEMM benchmark

#3367 samnordmann opened 2 weeks ago
0
[Proposal] Support root->logical transforms in Fusion inputs

#3366 jacobhinkle closed 2 weeks ago
5
Updating readme

#3365 jjsjann123 closed 1 hour ago
4
Fix launch configuration error with ScatterGatherTest.TorchGatherAllRankAllSelectedDim

#3364 naoyam closed 2 weeks ago
1
Accidentally lost in #3028

#3363 naoyam closed 1 week ago
1
Renaming from #3263 part2

#3362 naoyam closed 2 weeks ago
2
Use SimplyfingIrBuilder for circular buffering

#3361 naoyam closed 2 weeks ago
1
LoadStoreOp::toInlineString should not assume it's a tensor op

#3360 naoyam closed 1 week ago
7

Previous Next