GitHub · Where software is built

Labels Milestones New issue

cap grid size for distributed kernel

#1646

· shunting314 opened

on Mar 11, 2026

hl.triton_kernel codegen at the wrong places

#1642

· shunting314 opened

on Mar 10, 2026

Suboptimal config from autotuning

#1640

· gmagogsfm opened

on Mar 10, 2026

Control recompilations/specialization tighter from helion, including triton recompilations

language support

#1539

· Chillee opened

on Feb 23, 2026

[RFC] Helion Autotuning service for instantaneous auto-tuning times

#1518

· msaroufim opened

on Feb 19, 2026

[AMD 350X] Helion Layernorm kernel performance is worse compared to inductor

#1425

· mengluy0125 opened

on Feb 11, 2026

Constrained autotuning

#1415

· ethche opened

on Feb 10, 2026

Overlapping compilation time with benchmarking for the autotuner

#1400

· ethche opened

on Feb 9, 2026

[RFC] Helion kernel + Inductor ops prologue / epilogue fusion

PyTorch interop

#1346

· yf225 opened

on Jan 23, 2026

Autotuning with tensors >= 2^32 errors when block pointers are used (and why isn't `HELION_AUTOTUNE_IGNORE_ERRORS=1` on by default?)

#1335

· Chillee opened

on Jan 13, 2026

While doing accuracy checks, run with determinism mode enabled (i.e. map `torch.empty` to be all nans or something)

#1334

· Chillee opened

on Jan 13, 2026

Enable User-Driven Early Termination and Caching in Autotuning

#1331

· rafaykhurram opened

on Jan 9, 2026