Ecosyste.ms: Issues

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

GitHub / rocm/composable_kernel issues and pull requests

#1854 - Fix pk_int4 cast and add pk_int4 dtype in ck tile

Pull Request - State: open - Opened by bartekxk 7 days ago

#1853 - CK Tile GEMM Compute V2 (2 LDS Ping Pong mechanism)

Pull Request - State: open - Opened by ThomasNing 7 days ago

#1852 - Add pre_softmax fnctor

Pull Request - State: open - Opened by amd-hhashemi 7 days ago

#1851 - Fix ck_tile gemm benchmarking scripts

Pull Request - State: closed - Opened by illsilin 7 days ago

#1850 - Enable ck_tile gemms build in CI by default.

Pull Request - State: closed - Opened by illsilin 7 days ago

#1849 - turn on the ck_tile gemm tests by default

Pull Request - State: closed - Opened by illsilin 8 days ago

#1848 - Bump rocm-docs-core from 1.14.1 to 1.15.0 in /docs/sphinx

Pull Request - State: closed - Opened by dependabot[bot] 8 days ago
Labels: documentation, dependencies, ci:docs-only

#1847 - add new pk_i4 cvt for fp16 and bf16

Pull Request - State: open - Opened by zjing14 8 days ago

#1846 - Add flex attention example

Pull Request - State: open - Opened by tenpercent 9 days ago - 1 comment

#1845 - Support for dtypes (fp8, bf8, bf16 and fp16) for the ck_tile/03_gemm example.

Pull Request - State: open - Opened by kylasa 9 days ago - 1 comment

#1843 - [CK Tile] Spatially local GEMM tile partitioner.

Pull Request - State: closed - Opened by aosewski 10 days ago

#1842 - [CK TILE] Implement cschuflle algorithm

Pull Request - State: closed - Opened by bartekxk 10 days ago

#1841 - [CK Tile][Feature] Enable a block ping-poing scheduling pipeline in CK Tile GEMM

Issue - State: open - Opened by zjing14 10 days ago
Labels: Under Investigation, feature request

#1840 - [CK_TILE] moe sorting ex kernel to support expert > 128

Pull Request - State: open - Opened by carlushuang 12 days ago

#1839 - Added Int4 mixed batch gemm support

Pull Request - State: open - Opened by mtgu0705 12 days ago

#1838 - Cka8w8 uc newpipe

Pull Request - State: closed - Opened by aska-0096 15 days ago

#1837 - data type (fp8, bf8) support for gemm example in CK_Tile

Pull Request - State: closed - Opened by kylasa 15 days ago - 1 comment

#1832 - Bump rocm-docs-core from 1.13.0 to 1.14.1 in /docs/sphinx

Pull Request - State: closed - Opened by dependabot[bot] 16 days ago
Labels: documentation, dependencies, ci:docs-only

#1831 - Add Conv NGCHW client example

Pull Request - State: closed - Opened by bartekxk 16 days ago

#1830 - [CK_TILE] moe-smoothquant support fp8 output

Pull Request - State: closed - Opened by carlushuang 17 days ago

#1829 - Add OCP FP8 support in CK_TILE

Pull Request - State: closed - Opened by andriy-ca 17 days ago - 3 comments

#1828 - [CK-Tile] Refactor the is_same_v conditions

Pull Request - State: closed - Opened by mozga-amd 17 days ago

#1827 - Refactor file structure dev

Pull Request - State: closed - Opened by aledudek 19 days ago - 1 comment

#1826 - [CK_TILE] Refactor ck_tile file structure

Pull Request - State: closed - Opened by aledudek 19 days ago - 3 comments

#1825 - Added bf16 instances grouped gemm fixed nk

Pull Request - State: closed - Opened by deepsek 19 days ago

#1824 - Add make_kernel_pt for specific architecture compilation guards

Pull Request - State: open - Opened by alugorey 20 days ago - 4 comments

#1823 - Add bf16 instances for grouped gemm fixed nk

Pull Request - State: closed - Opened by deepsek 20 days ago - 1 comment

#1822 - Add bf16 instances grouped gemm fixed nk

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

#1821 - [CK_TILE] Add error threshold calculation for gemm examples

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

#1820 - fix a bug for int4 scale weight only kernel

Pull Request - State: closed - Opened by mtgu0705 21 days ago
Labels: bug, CI - Testing

#1819 - Implementing Test Filters for Smoke and Regression Tests

Pull Request - State: closed - Opened by AviralGoelAMD 22 days ago - 1 comment
Labels: good first issue

#1818 - Fix and optimize dynamic unary elementwise

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

#1817 - Change flag to CK_GFX90A_DENORM_WORKAROUND

Pull Request - State: closed - Opened by darren-amd 22 days ago

#1816 - Disable inductor codegen tests on legacy OS

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

#1815 - Add rounding for float to bf16 conversion as default (rel-6.2)

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

#1813 - Prec param new

Pull Request - State: open - Opened by aledudek 24 days ago

#1812 - Add rounding for float to bf16 conversion as default

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

#1811 - [CK_TILE] Use the GEMM example prec argument

Pull Request - State: open - Opened by aledudek 26 days ago - 1 comment

#1810 - Update for fmha_fwd qs_ks_vs pipeline

Pull Request - State: closed - Opened by qianfengz 26 days ago

#1809 - CK Tile Gemm API and heuristics changes

Pull Request - State: open - Opened by jakpiase 28 days ago

#1807 - enable int4 scale (weight only) kernel

Pull Request - State: closed - Opened by mtgu0705 29 days ago

#1806 - Created a branch int4_pr_based_on_JingPR

Pull Request - State: closed - Opened by mtgu0705 29 days ago

#1805 - Revert "[Draft] Revive QsKsVs FMHA pipeline"

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

#1804 - Disable building DPP kernels by default

Pull Request - State: closed - Opened by darren-amd 30 days ago

#1803 - updated int4_pr_debug

Pull Request - State: closed - Opened by mtgu0705 30 days ago
Labels: bug, code quality, CI - Testing

#1802 - [CK_TILE] Add Various Fusion Functions to RMSNorm

Pull Request - State: closed - Opened by ruanjm 30 days ago - 1 comment
Labels: feature request

#1801 - Update LICENSE to 2025 (#1797)

Pull Request - State: open - Opened by spolifroni-amd about 1 month ago
Labels: documentation, ci:docs-only

#1799 - Update in GridSize() and using GridSize() for splitkv kernel

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

#1799 - Update in GridSize() and using GridSize() for splitkv kernel

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

#1798 - Bump rocm-docs-core from 1.12.1 to 1.13.0 in /docs/sphinx

Pull Request - State: closed - Opened by dependabot[bot] about 1 month ago
Labels: documentation, dependencies, ci:docs-only

#1798 - Bump rocm-docs-core from 1.12.1 to 1.13.0 in /docs/sphinx

Pull Request - State: closed - Opened by dependabot[bot] about 1 month ago
Labels: documentation, dependencies, ci:docs-only

#1797 - Update LICENSE to 2025

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

#1796 - Fix parsing instances for pt inductor

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

#1795 - Cross GPU Reduce Operator Initial Development

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

#1794 - Add CK_TIME_KERNEL as toggleable CMake Variable

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

#1793 - [CK_TILE] Support moe with up gemm

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

#1792 - terminology clean-up

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

#1791 - [CK_TILE] Add GetName for GEMM kernels

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

#1790 - Fix universal gemm profiler for pk_i4_t

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

#1789 - [CK_TILE] fmha fwd splitkv optimization for decode (seqlen_q=1)

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

#1788 - Bump rocm-docs-core from 1.12.0 to 1.12.1 in /docs/sphinx

Pull Request - State: closed - Opened by dependabot[bot] about 1 month ago
Labels: documentation, dependencies, ci:docs-only

#1787 - Add afagaj to CODEOWNERS

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

#1786 - Implement the fp16xint4 scale weight only kernel for Ali

Pull Request - State: closed - Opened by mtgu0705 about 1 month ago
Labels: enhancement, CI - Testing, priority

#1785 - [CK_TILE] Sync fmha fwd splitkv minor optimizations

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

#1784 - Ck tile/layernorm: implement naive reduce, opt performance

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

#1783 - Add NGCHW bf16 grouped conv fwd instances

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

#1782 - [Issue]: Does not honor cmake BUILD_SHARED

Issue - State: open - Opened by trixirt about 1 month ago - 2 comments
Labels: Under Investigation

#1782 - [Issue]: Does not honor cmake BUILD_SHARED

Issue - State: open - Opened by trixirt about 1 month ago - 2 comments
Labels: Under Investigation

#1781 - [Issue]: RFE cmake BUILD_EXAMPLES

Issue - State: open - Opened by trixirt about 1 month ago - 1 comment
Labels: Under Investigation

#1781 - [Issue]: RFE cmake BUILD_EXAMPLES

Issue - State: open - Opened by trixirt about 1 month ago - 1 comment
Labels: Under Investigation

#1780 - [Issue]: CK_TIME_KERNEL used by default

Issue - State: closed - Opened by trixirt about 1 month ago - 5 comments
Labels: Under Investigation

#1779 - [CK_TILE] Adjust kBlockSize of reduce example for better perf

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

#1778 - Remove using partitioner for all fmha kernels

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

#1777 - [Issue]: Some kernel pass AB0B1 and output as std::vector<const void*>

Issue - State: closed - Opened by Jay19751103 about 1 month ago - 3 comments
Labels: Under Investigation

#1776 - CK Tile GEMM CICD fixed & register block method refactor

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

#1775 - [CK_TILE] Fix fmha fwd splitkv codegen error

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

#1774 - Dev/merge u8w8

Pull Request - State: closed - Opened by coderfeli about 2 months ago - 1 comment

#1773 - [Issue]: `lld: error: undefined hidden symbol: unsigned short ck::atomic_add`

Issue - State: closed - Opened by tjtanaa about 2 months ago - 5 comments
Labels: Under Investigation

#1772 - Grouped convolution backward weight special vector size loads

Pull Request - State: closed - Opened by bartekxk about 2 months ago

#1772 - Grouped convolution backward weight special vector size loads

Pull Request - State: open - Opened by bartekxk about 2 months ago

#1771 - [CK_TILE] optimize moe-sorting kernel

Pull Request - State: closed - Opened by carlushuang about 2 months ago

#1770 - Promote develop into amd-develop

Pull Request - State: closed - Opened by illsilin about 2 months ago

#1769 - fix typo for CK_USE_OCP_FP8

Pull Request - State: closed - Opened by illsilin about 2 months ago

#1768 - hot-fix missing flags

Pull Request - State: closed - Opened by carlushuang about 2 months ago

#1767 - Promote latest CK develop

Pull Request - State: closed - Opened by illsilin about 2 months ago

#1766 - fix profiler_grouped_gemm

Pull Request - State: closed - Opened by illsilin about 2 months ago

#1765 - [Draft] Revive QsKsVs FMHA pipeline

Pull Request - State: closed - Opened by tenpercent about 2 months ago - 1 comment

#1764 - fix: preprocessor directives logic error if/else

Pull Request - State: closed - Opened by deepsek about 2 months ago

#1763 - device_prop.hpp: move static map to helper function and initialize there

Pull Request - State: open - Opened by coconutruben about 2 months ago - 2 comments

#1763 - device_prop.hpp: move static map to helper function and initialize there

Pull Request - State: open - Opened by coconutruben about 2 months ago - 1 comment

#1762 - Jing's contribution: prototype of mixed precision gemm FP16/BF16xint4 GEMM

Pull Request - State: closed - Opened by aosewski about 2 months ago - 7 comments
Labels: enhancement, CI - Testing, external contribution

#1762 - Jing's contribution: prototype of mixed precision gemm FP16/BF16xint4 GEMM

Pull Request - State: closed - Opened by aosewski about 2 months ago - 7 comments
Labels: enhancement, CI - Testing, external contribution