Search results

[CUDA] Add experimental green context support for SM carveout

matrix multiplication

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

open source

Status: Open (in progress).

pytorch/pytorchnumber 159104

#159104 In pytorch/pytorch;

· eqy opened on Jul 25, 2025

· 21/45

11

torch.compile doesn't work with torch.mm(out_dtype=...) variant

matrix multiplication

module: inductor

oncall: pt2

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#159023 In pytorch/pytorch;

· yf225 opened on Jul 24, 2025

1



[CUDA][FP8] Expect SM110 to fail FP8 matmul tests

ciflow/trunkTrigger trunk jobs on your pull requestTrigger trunk jobs on your pull request

matrix multiplication

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: floatx (formerly float8)For torch.float8_e5m2 and torch.float8_e4m3 and other sub 8-bit float typesFor torch.float8_e5m2 and torch.float8_e4m3 and other sub 8-bit float types

open source

topic: not user facingtopic categorytopic category

Status: Open (in progress).

pytorch/pytorchnumber 158309

#158309 In pytorch/pytorch;

· eqy opened on Jul 15, 2025

· Approved

· 144/150

7

[Inductor] addmm + activation function fusion

matrix multiplication

module: inductor

open source

topic: not user facingtopic categorytopic category

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open (in progress).

pytorch/pytorchnumber 158137

#158137 In pytorch/pytorch;

· AaronWang04 opened on Jul 11, 2025

· Approved

· 239/256

19

grouped_mm optional zero initialization of the output

matrix multiplication

module: python frontendFor issues relating to PyTorch's Python frontendFor issues relating to PyTorch's Python frontend

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#154557 In pytorch/pytorch;

· IvanKobzarev opened on May 28, 2025

5 comments

[cuBLASLt][cuBLAS] Support 2D bias and beta != 1.0 in cuBLASLt

ci-no-tdDo not run TD on this PRDo not run TD on this PR

ciflow/inductor-rocmTrigger "inductor" config CI on ROCmTrigger "inductor" config CI on ROCm

ciflow/rocmTrigger "default" config CI on ROCmTrigger "default" config CI on ROCm

ciflow/rocm-mi300Trigger "default" config CI on ROCm MI300Trigger "default" config CI on ROCm MI300

ciflow/trunkTrigger trunk jobs on your pull requestTrigger trunk jobs on your pull request

matrix multiplication

Merged

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

open source

Reverted

topic: not user facingtopic categorytopic category

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open (in progress).

pytorch/pytorchnumber 154170

#154170 In pytorch/pytorch;

· eqy opened on May 23, 2025

· Approved

· 157/179

45

matmul uses excessive memory in batch cases with more than 3 dimensions

actionable

has workaround

matrix multiplication

module: memory usagePyTorch is using more memory than it should, or it is leaking memoryPyTorch is using more memory than it should, or it is leaking memory

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#154128 In pytorch/pytorch;

· JackBuck opened on May 22, 2025

3 comments

[cuBLAS] Strange behavior with out_dtype argument in torch.addmm

matrix multiplication

module: bfloat16

module: cublasProblem related to cublas supportProblem related to cublas support

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: halfRelated to float16 half-precision floatsRelated to float16 half-precision floats

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#154086 In pytorch/pytorch;

· eqy opened on May 22, 2025

2 comments



torch.bmm is slow on non-contiguous BF16 CPU tensors

matrix multiplication

module: bfloat16

module: cpuCPU specific problem (e.g., perf, algorithm)CPU specific problem (e.g., perf, algorithm)

module: halfRelated to float16 half-precision floatsRelated to float16 half-precision floats

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#151934 In pytorch/pytorch;

· ar0ck opened on Apr 22, 2025

3 comments

BF16 linear(matmul) operator 100x slower on odd matrix dimension sizes on A100

matrix multiplication

module: cublasProblem related to cublas supportProblem related to cublas support

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: linear algebraIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmulIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmul

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#146817 In pytorch/pytorch;

· piubwd opened on Feb 10, 2025

3 comments

Addmm is not always better than add + mm

matrix multiplication

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#141210 In pytorch/pytorch;

· ConnollyLeon opened on Nov 21, 2024

3D matrix support in _scaled_mm

enhancementNot as big of a feature, but technically not a bug. Should be easy to fixNot as big of a feature, but technically not a bug. Should be easy to fix

matrix multiplication

module: floatx (formerly float8)For torch.float8_e5m2 and torch.float8_e4m3 and other sub 8-bit float typesFor torch.float8_e5m2 and torch.float8_e4m3 and other sub 8-bit float types

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#129803 In pytorch/pytorch;

· rybakov opened on Jun 28, 2024

2 comments

control output type in matmul

matrix multiplication

needs designWe want to add this feature but we need to figure out how firstWe want to add this feature but we need to figure out how first

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#128283 In pytorch/pytorch;

· rybakov opened on Jun 8, 2024

2 comments

triton package that ships with torch is not compatible with torch.ops.matmul

matrix multiplication

module: buildBuild system issuesBuild system issues

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#125539 In pytorch/pytorch;

· jeromeku opened on May 4, 2024

6 comments

RuntimeError: could not create a primitive descriptor for a matmul primitive

matrix multiplication

module: cpuCPU specific problem (e.g., perf, algorithm)CPU specific problem (e.g., perf, algorithm)

module: crashProblem manifests as a hard crash, as opposed to a RuntimeErrorProblem manifests as a hard crash, as opposed to a RuntimeError

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#125247 In pytorch/pytorch;

· buvanesh-git opened on Apr 30, 2024

4 comments

Tensor's storage changes computation outcome for CPU tensors.

matrix multiplication

module: cpuCPU specific problem (e.g., perf, algorithm)CPU specific problem (e.g., perf, algorithm)

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#124934 In pytorch/pytorch;

· stswidwinski opened on Apr 25, 2024

6 comments

Matmul with broadcast makes copy because of reshape to bmm, and it is slow

matrix multiplication

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#124465 In pytorch/pytorch;

· 54limiao opened on Apr 19, 2024

Add support for CUBLAS_COMPUTE_16F for GEMM opeartions in cudaBLAS

enhancementNot as big of a feature, but technically not a bug. Should be easy to fixNot as big of a feature, but technically not a bug. Should be easy to fix

matrix multiplication

module: cublasProblem related to cublas supportProblem related to cublas support

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#123157 In pytorch/pytorch;

· gswxp2 opened on Apr 2, 2024

2 comments

Submatrix multiplication selected by rows or columns

matrix multiplication

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#119624 In pytorch/pytorch;

· neuer93 opened on Feb 10, 2024

1 comment

[discussion] Route pointwise Conv1d/Conv2d to matmul? (also in eager)

matrix multiplication

module: convolutionProblems related to convolutions (THNN, THCUNN, CuDNN)Problems related to convolutions (THNN, THCUNN, CuDNN)

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: mkldnnRelated to Intel IDEEP or oneDNN (a.k.a. mkldnn) integrationRelated to Intel IDEEP or oneDNN (a.k.a. mkldnn) integration

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

module: xnnpack

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#116506 In pytorch/pytorch;

· vadimkantorov opened on Dec 28, 2023

3 comments

The cuda batched GEMM has a poor performance for bigger batch size with smaller matrix size

matrix multiplication

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#113743 In pytorch/pytorch;

· XiaobingSuper opened on Nov 15, 2023

2 comments

resutl of (torch.mm(a,b) does not match result of (a[:part,:], b)

matrix multiplication

module: numerical-stabilityProblems related to numerical stability of operationsProblems related to numerical stability of operations

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#108521 In pytorch/pytorch;

· head-with-nothing opened on Sep 4, 2023

2 comments

[BUG] Poor torch.bmm performance on H100

matrix multiplication

module: cublasProblem related to cublas supportProblem related to cublas support

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#100850 In pytorch/pytorch;

· szhengac opened on May 8, 2023

6 comments

COO @ COO tries to allocate way too much memory on CUDA

matrix multiplication

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: sparseRelated to torch.sparseRelated to torch.sparse

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#95169 In pytorch/pytorch;

· nikitaved opened on Feb 20, 2023

Add rearrange operator to PyTorch natively

actionable

featureA request for a proper, new feature.A request for a proper, new feature.

good first issue

matrix multiplication

module: first class dims

module: functorchPertaining to torch.func or pytorch/functorchPertaining to torch.func or pytorch/functorch

module: python frontendFor issues relating to PyTorch's Python frontendFor issues relating to PyTorch's Python frontend

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#92675 In pytorch/pytorch;

· tchaton opened on Jan 20, 2023

3

21 comments