v0.4.1

[coderfeli]

What's Changed

• [Skills] Add review-pr: advisory PR review adapted from aiter and ATOM by @jhinpan in [Skills] Add review-pr: advisory PR review adapted from aiter and ATOM #1279
• [Kernel][gfx1250] Support K tails in the BF16 GEMM by @jhinpan in [Kernel][gfx1250] Support K tails in the BF16 GEMM #1230
• [fix] Make make_tile treat every argument list as modes by @xudoyuan in [fix] Make make_tile treat every argument list as modes #1284
• [fix] Support waves_per_eu min,max range for full VGPR utilization by @jli-melchior in [fix] Support waves_per_eu min,max range for full VGPR utilization #1184
• [Tool][Debug] Inject hand-edited assembly into a JIT cache entry by @Phil-amd in [Tool][Debug] Inject hand-edited assembly into a JIT cache entry #1188
• [fix][CI] Run aiter's install_triton.sh to match aiter's Triton version by @jli-melchior in [fix][CI] Run aiter's install_triton.sh to match aiter's Triton version #1300
• [ROCDL] Gate the TDM tile_dim0 span check on padding by @xudoyuan in [ROCDL] Gate the TDM tile_dim0 span check on padding #1302
• [Skills] Ensure flydsl-kernel-authoring follows Anthropic best practices by @danielholanda in [Skills] Ensure flydsl-kernel-authoring follows Anthropic best practices #1305
• [Skills] Add evals and skill cards for kernel authoring and trace analysis by @danielholanda in [Skills] Add evals and skill cards for kernel authoring and trace analysis #1304
• [llvm] bump llvm to 1df311a6 by @jli-melchior in [llvm] bump llvm to 1df311a6 #1243
• Align bf16 implicit Conv3d with aiter by @huizzhan in Align bf16 implicit Conv3d with aiter #1217
• [Fix] Bitcast small floats across converted ops in convert-fly-to-rocdl by @Phil-amd in [Fix] Bitcast small floats across converted ops in convert-fly-to-rocdl #1290
• [gfx1250] Add A8W8/A8W4/A4W4 compute-bound GEMM by @aoli26 in [gfx1250] Add A8W8/A8W4/A4W4 compute-bound GEMM #971
• [Core] Add four concrete layout categories by @sjfeng1999 in [Core] Add four concrete layout categories #1292
• [CI] Fix the SGLang nightly setup after sgl-kernel moved into sglang.kernels.aot by @jhinpan in [CI] Fix the SGLang nightly setup after sgl-kernel moved into sglang.kernels.aot #1287
• [Kernel] Reuse typed APIs in GLM and Kimi-K3 monokernels by @coderfeli in [Kernel] Reuse typed APIs in GLM and Kimi-K3 monokernels #1310
• [compiler] [llvm] xnack perf by @jli-melchior in [compiler] [llvm] xnack perf #1308
• [Release] Bump version to 0.4.1 by @jli-melchior in [Release] Bump version to 0.4.1 #1312
• 1,Balance the prologue's extra wave_m == 1 barrier so both groups' ep… by @solinzby1 in 1,Balance the prologue's extra wave_m == 1 barrier so both groups' ep… #1306
• [Dialect][ROCDL] Add gfx120 integer WMMA and an INT4 GEMM by @qiangpan2 in [Dialect][ROCDL] Add gfx120 integer WMMA and an INT4 GEMM #1266
• fix(coop): preserve declared operator element types by @sjfeng1999 in fix(coop): preserve declared operator element types #1185
• [Kernel] Remove redundant scalar, vector, and IR wrappers by @coderfeli in [Kernel] Remove redundant scalar, vector, and IR wrappers #1262
• [Kernel] Replace raw memory glue and manual reductions with fx APIs by @coderfeli in [Kernel] Replace raw memory glue and manual reductions with fx APIs #1315
• [Skills] review-pr: review a PR when a maintainer adds the ai-review label by @jhinpan in [Skills] review-pr: review a PR when a maintainer adds the ai-review label #1311
• [Skills] review-pr: strip the API key and show why model calls failed by @jhinpan in [Skills] review-pr: strip the API key and show why model calls failed #1317
• kernel-trace-analysis: fix SKILL.md nits so the evals pass by @danielholanda in kernel-trace-analysis: fix SKILL.md nits so the evals pass #1322
• [Ext][Coop] Add more block-level primitives by @sjfeng1999 in [Ext][Coop] Add more block-level primitives #1314
• [LLVM] Roll back to 941a04e and cherry-pick unclaused VMEM patches by @jli-melchior in [LLVM] Roll back to 941a04e and cherry-pick unclaused VMEM patches #1327

New Contributors

• @danielholanda made their first contribution in [Skills] Ensure flydsl-kernel-authoring follows Anthropic best practices #1305
• @huizzhan made their first contribution in Align bf16 implicit Conv3d with aiter #1217

Full Changelog: v0.4.0...v0.4.1

---

This discussion was created from the release v0.4.1.