Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / triton-lang/triton issues and pull requests

#5009 - [BACKEND][NVIDIA] Add kWidth support in MMAv3

Pull Request - State: closed - Opened by ggengnv 18 days ago - 5 comments

#5008 - [AMD] Skip scaled_dot tests for gfx11 and gfx12

Pull Request - State: closed - Opened by AlexAUT 18 days ago

#5007 - Fold fp_to_fp op with zero constant input

Pull Request - State: open - Opened by oplavsic 18 days ago

#5006 - [FRONTEND] Fix transpose with tuple dims

Pull Request - State: closed - Opened by peterbell10 18 days ago

#5005 - [AMD][BACKEND] Switch to code object v5

Pull Request - State: open - Opened by AlexAUT 18 days ago

#5002 - [AMD] remove redundant LDS bypass checks

Pull Request - State: closed - Opened by binarman 19 days ago - 2 comments

#5000 - different between do_not_specialize and tl.constexpr

Issue - State: closed - Opened by zhanglei1172 20 days ago - 1 comment

#4999 - Hopper support for mma -> mma layout conversion?

Issue - State: closed - Opened by hypnopump 20 days ago - 1 comment

#4998 - [AMD] Restructure ReorderInstructions pass

Pull Request - State: closed - Opened by antiagainst 21 days ago - 1 comment

#4997 - [CI] bump yapf version from be72557 to 7e21823

Pull Request - State: closed - Opened by XuehaiPan 21 days ago

#4996 - [Frontend][Backend] Implement support for scale_dot(-, bf16)

Pull Request - State: closed - Opened by lezcano 22 days ago

#4995 - libtriton.so undefined symbol: Pyxxx

Issue - State: closed - Opened by helloworldstone 22 days ago

#4994 - [AMD] Add initial support for scaled_dot(mxfp8, fp8)

Pull Request - State: closed - Opened by antiagainst 22 days ago - 1 comment

#4993 - [AMD] getBackwardSlice variant with handling for op regions

Pull Request - State: open - Opened by karthik-man 23 days ago - 1 comment

#4992 - Interpreter fails with float16 atomic_add

Issue - State: closed - Opened by davidberard98 23 days ago - 5 comments
Labels: enhancement

#4991 - [BACKEND] Improve detection of register to register conversion

Pull Request - State: closed - Opened by Jokeren 23 days ago

#4990 - Pytest on Ampere platform has errors, see the log below

Issue - State: closed - Opened by bingyizh233 23 days ago - 1 comment

#4989 - about DCE for ModuleAxisInfoAnalysis

Issue - State: open - Opened by Shaquille-Wu 23 days ago - 2 comments

#4987 - [Proton] Adding Sorting of Kernels

Pull Request - State: closed - Opened by CRobeck 24 days ago - 4 comments

#4983 - [AMD] Enable shared->MFMA dot operand conversion through LinearLayout

Pull Request - State: closed - Opened by binarman 24 days ago - 1 comment

#4981 - [TEST] float16 test for test_tensor_atomic_rmw

Pull Request - State: closed - Opened by davidberard98 24 days ago - 3 comments

#4980 - [Backend] Fix when trying to convert an mma<!tt.ptr<f32>> into blocked

Pull Request - State: closed - Opened by lezcano 24 days ago - 6 comments

#4978 - Pypi build for aarch64

Issue - State: open - Opened by surak 24 days ago - 2 comments

#4970 - Where can I find a triton 2.0.0 dev verison?

Issue - State: open - Opened by hulihan-start 25 days ago - 1 comment

#4968 - [Backend] Add LL::quotient and remove uses of divideRight and sublayoutIsIdentity

Pull Request - State: closed - Opened by lezcano 25 days ago - 2 comments

#4966 - [AMD] Add pass to convert tt.load/tt.store to buffer operations

Pull Request - State: closed - Opened by giuseros 25 days ago - 3 comments

#4962 - Document that tl.reduce assumes associativity/commutativity

Issue - State: open - Opened by bertmaher 26 days ago - 1 comment

#4955 - [Frontend] [BC breaking] Always follow C semantics on %

Pull Request - State: closed - Opened by brod4910 28 days ago

#4951 - [BACKEND] Replace `isMmaToDotShortcut` with linear layout based logic

Pull Request - State: closed - Opened by Jokeren 29 days ago - 2 comments

#4950 - [Backend] Pipeline scale_dot

Pull Request - State: closed - Opened by lezcano 29 days ago

#4940 - [AMD] Reland instruction scheduling hint changes

Pull Request - State: closed - Opened by ravil-mobile about 1 month ago - 2 comments

#4937 - [AMD]Add fast_expf to libdevice

Pull Request - State: open - Opened by knwng about 1 month ago

#4936 - [BUILD] Add instrumentation libs to `.gitignore`

Pull Request - State: closed - Opened by Jokeren about 1 month ago

#4935 - [AMD] Reland sinking the 2nd tt.load after local_load's

Pull Request - State: closed - Opened by zhanglx13 about 1 month ago - 1 comment

#4934 - Fix 3xTF32 precision issues

Pull Request - State: closed - Opened by alexsamardzic about 1 month ago

#4933 - Fix 3xTF32 precision issues

Pull Request - State: closed - Opened by ThomasRaoux about 1 month ago - 1 comment

#4932 - Added an option to dump IR to files

Pull Request - State: open - Opened by arakhmati-openai about 1 month ago

#4930 - [backend] Update to llvm/llvm-project@b5cc222d7429

Pull Request - State: closed - Opened by ravil-mobile about 1 month ago

#4929 - Accelerating custom sparsity patterns on GPU with triton

Issue - State: open - Opened by abhishektyaagi about 1 month ago - 2 comments

#4928 - Update GCC aarch64 toolchain to 14.2.0-1

Pull Request - State: closed - Opened by antiagainst about 1 month ago

#4927 - [BACKEND] Update to llvm/llvm-project@b5cc222d7429

Pull Request - State: closed - Opened by ravil-mobile about 1 month ago

#4926 - Fix 3xTF32 precision issues

Pull Request - State: closed - Opened by alexsamardzic about 1 month ago - 12 comments

#4925 - [AMD] Emit vectorized 16-bit float LLVM atomic ops

Pull Request - State: closed - Opened by joviliast about 1 month ago - 9 comments

#4924 - [frontend] Remove Complex Regex for MLIR Parsing

Pull Request - State: open - Opened by SamGinzburg about 1 month ago

#4923 - [BACKEND] Update LLVM to llvm/llvm-project@1d40fefb08e9b11

Pull Request - State: closed - Opened by ravil-mobile about 1 month ago - 1 comment

#4922 - LLVM ERROR: mma16816 data type not supported

Issue - State: closed - Opened by mobicham about 1 month ago - 8 comments

#4921 - Return `num_warmups`, `num_reps` and `use_cuda_graph` fields of `Autotuner`

Pull Request - State: open - Opened by anmyachev about 1 month ago - 1 comment

#4920 - [AMD] unrevert #4901; revert #4823

Pull Request - State: closed - Opened by ptillet about 1 month ago

#4919 - [AMD] revert optimizations

Pull Request - State: closed - Opened by ptillet about 1 month ago

#4918 - log1p

Issue - State: closed - Opened by nimz about 1 month ago - 1 comment

#4917 - [CI] Run CI on all PRs

Pull Request - State: closed - Opened by peterbell10 about 1 month ago - 3 comments

#4916 - Add tensor descriptor API backed by device-side TMA creation

Pull Request - State: closed - Opened by peterbell10 about 1 month ago

#4915 - [Frontend] Factor out block shape validation function

Pull Request - State: closed - Opened by peterbell10 about 1 month ago

#4914 - Fix order

Pull Request - State: open - Opened by rawnhenry about 1 month ago - 8 comments

#4913 - Update pre-commit config

Pull Request - State: closed - Opened by anmyachev about 1 month ago

#4912 - [AMD] Fix gfx12 warp size and fix wmma in maybeDeduplicate

Pull Request - State: closed - Opened by AlexAUT about 1 month ago

#4911 - Windows fix

Pull Request - State: closed - Opened by wkpark about 1 month ago

#4910 - Add a tt.pointer_range_32 specialization for AMD backend

Pull Request - State: open - Opened by giuseros about 1 month ago - 2 comments

#4909 - Will tl.load faster without boundary_check?

Issue - State: open - Opened by MARD1NO about 1 month ago

#4908 - `triton-nightly` package not updated since July 17

Issue - State: open - Opened by 152334H about 1 month ago

#4907 - [WIP] [AMD] Apply basic optimizations to vecAdd on MI300X

Pull Request - State: closed - Opened by zhanglx13 about 1 month ago

#4906 - Poor performance on Ampere vs. Ada with bitpacked weights

Issue - State: open - Opened by mobicham about 1 month ago - 18 comments

#4905 - [Frontend] Re-enable NumPy 2.0 semantics for add, sub, mul.

Pull Request - State: closed - Opened by lezcano about 1 month ago

#4904 - [Backend] Implement `scaled_dot(mxfp4, fp8)`

Pull Request - State: closed - Opened by lezcano about 1 month ago

#4903 - Introduce amdgpu.buffer_load and amdgpu.buffer_store

Pull Request - State: open - Opened by giuseros about 1 month ago - 1 comment

#4901 - [AMD] Fix "keep Q tensor in VGPRS" optimization

Pull Request - State: closed - Opened by oplavsic about 1 month ago - 2 comments

#4900 - chore: update core.py

Pull Request - State: closed - Opened by eltociear about 1 month ago - 1 comment

#4898 - only cpu

Issue - State: closed - Opened by JocelynPanPan about 1 month ago - 1 comment

#4896 - [IR] Add poison value to triton IR and use in frontend in place of undef

Pull Request - State: closed - Opened by peterbell10 about 1 month ago - 1 comment

#4895 - [BACKEND] Small fixes for dot operand properties

Pull Request - State: closed - Opened by Jokeren about 1 month ago - 2 comments

#4894 - Fix type hint in setup.py

Pull Request - State: closed - Opened by kbumsik about 1 month ago

#4893 - [Backend] Update scf.if result uses in RewriteTensorPointer pass

Pull Request - State: closed - Opened by yiqian1 about 1 month ago

#4891 - [Linear Layouts] Implement LL conversion for DotOperand(version=2)

Pull Request - State: closed - Opened by lezcano about 1 month ago - 2 comments

#4890 - [Pipeliner] Fix epilogue peeling for num_stages=3+

Pull Request - State: closed - Opened by sjw36 about 1 month ago - 1 comment

#4889 - Auto import backends in `triton.language.extra`

Pull Request - State: closed - Opened by kbumsik about 1 month ago

#4888 - Add string representation for AttrsDescriptor

Pull Request - State: closed - Opened by alexbaden about 1 month ago - 4 comments

#4886 - [AMD] Skip incompatible layouts in RemoveLayoutConversions pass

Pull Request - State: open - Opened by binarman about 1 month ago

#4884 - Unpack the int8 incorrectly under irregular dimensions

Issue - State: closed - Opened by kuviki about 1 month ago - 1 comment

#4883 - LLVM ERROR: operation destroyed but still has uses

Issue - State: closed - Opened by jansel about 1 month ago
Labels: bug

#4882 - [frontend] Pretty-print `ptxas` command on failure

Pull Request - State: closed - Opened by bertmaher about 1 month ago

#4881 - [AMD][Pipeliner] Improve clustering and add prefetch

Pull Request - State: closed - Opened by sjw36 about 1 month ago - 1 comment

#4880 - [SWP][Tests] Add one test that triggers SWP error and improve test logic

Pull Request - State: closed - Opened by sfzhu93 about 1 month ago - 1 comment

#4879 - Weird tl.trans dims error

Issue - State: closed - Opened by Edenzzzz about 1 month ago

#4878 - How can I add some dialects for optimization?

Issue - State: open - Opened by zhananran about 1 month ago - 4 comments

#4876 - different gemm performance behaviors on A100 and RTX 4060

Issue - State: open - Opened by brisker about 1 month ago