ROCm/FlyDSL

View on GitHub

FlyDSL is the Python front‑end of the project: a Flexible Layout Python DSL for expressing tiling, partitioning, data movement, and kernel structure at a high level.

Stars: 282
Forks: 120
Open beginner issues: 0
Indexed issues: 46
Avg merge: 1d 21h
Merged PRs (30d): 67

Dominant language: Python
License: No license data
Last GitHub push: Sep 20, 2026
Latest indexed: Sep 20, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: No beginner labels indexed

Beginner friendly Everything

Open All

46 open issues indexed

[Feature]: Megamoe for TP Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/FlyDSL#1173 · Opened Sep 20, 2026
[Documentation]: more docs about API details. Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/FlyDSL#1172 · Opened Sep 20, 2026
[Issue]: Tensor.fill fails for singleton layouts with multiple unit strides Open

Difficulty 3/5 1-2 days Newbie friendliness 76/100

ROCm/FlyDSL#1165 · 1 reaction · Opened Sep 18, 2026
[Issue]: gfx950: a dropped `s_mov_b64` in an SGPR spill-staging chain puts an undefined register in a buffer descriptor base Open

Difficulty 5/5 Over a week Newbie friendliness 42/100

ROCm/FlyDSL#1087 · 4 comments · Opened Sep 2, 2026
[Issue]: Every kernel is compiled twice Open

ROCm/FlyDSL#1054 · 1 assignee · Opened Aug 21, 2026
[Proposals Welcome] Grow `flydsl.extension` into a library of common GPU building blocks Open

enhancement FlyDSL/Customization

ROCm/FlyDSL#1016 · 1 assignee · Opened Aug 17, 2026
[Feature]: conv3d_implicit — accept and emit NDHWC so chained convs skip the layout round-trip Open

Difficulty 4/5 3-5 days Newbie friendliness 68/100

ROCm/FlyDSL#993 · 3 comments · Opened Aug 10, 2026
Optimize gfx1250 MoE operators Open

ROCm/FlyDSL#962 · 1 comment · 1 assignee · Opened Aug 4, 2026
[Issue]: lld lookup failure in different docker images for gfx1250 Open

ROCm/FlyDSL#946 · 1 assignee · Opened Jul 31, 2026
Expression-layer arithmetic cleanup Open

FlyDSL/Customization

ROCm/FlyDSL#934 · 1 assignee · Opened Jul 30, 2026
[Feature]: Intra-kernel profiling Open

ROCm/FlyDSL#831 · 2 comments · 2 reactions · 1 assignee · Opened Jul 12, 2026
fx.gemm uses more VGPR than raw MFMA, and fails to compile with BufferCopy* atoms for fp8 M=64/N=16/K=128 Open

ROCm/FlyDSL#821 · 2 comments · 1 assignee · Opened Jul 9, 2026
SDK-Free Wheel Packaging and Backend Autodetection Open

ROCm/FlyDSL#813 · 2 reactions · 1 assignee · Opened Jul 7, 2026
[FlyDSL]: gfx1250 FMHA kernel tuning Open

ROCm/FlyDSL#810 · 1 comment · 1 assignee · Opened Jul 7, 2026
[Feature]: [FlyDSL] mega moe support Open

ROCm/FlyDSL#775 · 1 assignee · Opened Jun 30, 2026
[Feature] Make autotune a safe, adopted path Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/FlyDSL#770 · 2 comments · Opened Jun 30, 2026
[Issue]: MXFP6 support for A operand removed due to recent changes in #754 Open

ROCm/FlyDSL#767 · 1 assignee · Opened Jun 29, 2026
[RFC] FlyDSL kernels in downstream libraries Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

ROCm/FlyDSL#749 · 4 comments · Opened Jun 25, 2026
[Issue]: support Megamoe Open

FlyDSL/Customization

ROCm/FlyDSL#740 · 2 comments · 1 assignee · Opened Jun 25, 2026
[Issue]: zipped divide crashes on some cases Open

ROCm/FlyDSL#739 · 3 comments · 1 assignee · Opened Jun 25, 2026
[Issue]: retile() crashed with low-level error messages Open

ROCm/FlyDSL#732 · 1 comment · 2 assignees · Opened Jun 24, 2026
[Feature]: Support Mega-Moe V1 Open

FlyDSL/Customization

ROCm/FlyDSL#727 · 3 comments · 1 assignee · Opened Jun 23, 2026
[Feature]: moe gemm1 support persistent mode Open

FlyDSL/Customization

ROCm/FlyDSL#726 · 2 comments · 1 assignee · Opened Jun 23, 2026
dsv4 a4w4 moe support Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

ROCm/FlyDSL#719 · Opened Jun 23, 2026
[Feature]: minimax M3 moe support Open

FlyDSL/Customization

Difficulty 5/5 Over a week Newbie friendliness 30/100

ROCm/FlyDSL#717 · Opened Jun 23, 2026
Preserve logical signedness for Pointer and Tensor integer dtypes Open

Difficulty 3/5 1-2 days Newbie friendliness 64/100

ROCm/FlyDSL#701 · 1 comment · Opened Jun 17, 2026
[Feature]: Troubleshooting for potential issues related to internal control flow value modification Open

FlyDSL/Customization

ROCm/FlyDSL#688 · 1 assignee · Opened Jun 16, 2026
[Feature]: [Feature]: "early return" check in control flow Open

FlyDSL/Customization

ROCm/FlyDSL#687 · 1 assignee · Opened Jun 16, 2026
Compiler AST Rewriter optimization Open

FlyDSL/Customization

ROCm/FlyDSL#686 · 1 assignee · Opened Jun 16, 2026
[ROCm Test-First CI]: Day 0 FlyDSL MI455X UALoE72 Helios Rack open source github CI by AMD Advancing AI July 22 2026 Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/FlyDSL#663 · 2 comments · Opened Jun 6, 2026
Device printf invisible in Jupyter / piped stdout (libc block-buffering of runtime stdout) Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

ROCm/FlyDSL#653 · Opened Jun 4, 2026
Full docstring + minimal-example coverage for the Python package Open

FlyDSL/Customization

ROCm/FlyDSL#652 · 1 comment · 1 assignee · Opened Jun 4, 2026
FlyDSL Development Roadmap (v0.4) Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/FlyDSL#633 · 1 reaction · Opened Jun 2, 2026
FlyDSL Development Roadmap (v0.3) Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

ROCm/FlyDSL#632 · 1 comment · Opened Jun 2, 2026
FlyDSL Development Roadmap (v0.2) Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/FlyDSL#631 · Opened Jun 2, 2026
[Feature]: More robust AOT Open

ROCm/FlyDSL#621 · 1 assignee · Opened Jun 2, 2026
[Feature]:flydsl integration testing with sglang/vllm Open

ROCm/FlyDSL#603 · 1 comment · 1 assignee · Opened Jun 1, 2026
[Compiler]: Adding verification to the layout-algebra ops — what's the right approach? Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

ROCm/FlyDSL#583 · Opened May 28, 2026
Add a first-class global load op/helper Open

Difficulty 5/5 Over a week Newbie friendliness 48/100

ROCm/FlyDSL#548 · 1 reaction · Opened May 20, 2026
[Feature]: More low-level AMD/CDNA controls on SoTA Kernels Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

ROCm/FlyDSL#515 · 1 comment · 1 reaction · Opened May 12, 2026
ExtUIOp/zext in PA SWA chunked buffer path causes gfx950 correctness failure Open

Difficulty 4/5 3-5 days Newbie friendliness 38/100

ROCm/FlyDSL#510 · 2 comments · Opened May 12, 2026
In-process JIT caches can reuse stale compiled artifacts after helper changes Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

ROCm/FlyDSL#453 · Opened Apr 29, 2026
[Feature]: ATOM integration in FlyDSL CI Open

frameworks-devops

ROCm/FlyDSL#425 · 1 assignee · Opened Apr 22, 2026
[Feature]: SSA Phi Explosion: VGPR Waste and Spills When Large Accumulators Cross `scf.IfOp` Boundaries Open

ROCm/FlyDSL#400 · 3 reactions · 2 assignees · Opened Apr 14, 2026
[Issue]: unary ops(NOT, ~ and negation,-) on vector result in core dump during compilation Open

ROCm/FlyDSL#353 · 1 assignee · Opened Apr 7, 2026
[Feature]: Move benchmark/perf test from test folder to FlyDSL python source Open

ROCm/FlyDSL#304 · 1 comment · 2 assignees · Opened Mar 27, 2026

ROCm/FlyDSL

Get new issues in your inbox