-
Notifications
You must be signed in to change notification settings - Fork 150
Pull requests: pytorch/helion
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[fast-launcher] Opt-in output-tensor pool + codegen rewrite + autotune
CLA Signed
This label is managed by the Meta Open Source bot.
[fast-launcher] Static-kernel launcher (Inductor-style) + bound device
CLA Signed
This label is managed by the Meta Open Source bot.
[tests] Launcher-contract tests + multi-device fix for default_launcher
CLA Signed
This label is managed by the Meta Open Source bot.
#2634
opened May 29, 2026 by
yushangdi
Contributor
Loading…
[autotune] rank TPU matmul candidates by true on-chip time (the big win)
CLA Signed
This label is managed by the Meta Open Source bot.
[pallas] route whole-matrix matmuls through XLA's lax.dot_general
CLA Signed
This label is managed by the Meta Open Source bot.
[pallas] outer-grid matmul strategy for large reduction dimensions
CLA Signed
This label is managed by the Meta Open Source bot.
[pallas] correctness fix: use the right matmul path and full precision on TPU
CLA Signed
This label is managed by the Meta Open Source bot.
[autotune] keep compiler-suggested configs in the running for the final pick
CLA Signed
This label is managed by the Meta Open Source bot.
[autotune] compare each candidate against the current best in the same timing window
CLA Signed
This label is managed by the Meta Open Source bot.
[autotune] double-check the autotuner's winning config before committing to it
CLA Signed
This label is managed by the Meta Open Source bot.
[compile] add ast backend customization hook
CLA Signed
This label is managed by the Meta Open Source bot.
#2624
opened May 28, 2026 by
hinriksnaer
Collaborator
Loading…
[TPU][Pallas] Per-item DMA dispatch for hl.jagged_tile kernels
CLA Signed
This label is managed by the Meta Open Source bot.
#2616
opened May 28, 2026 by
yarongmu-google
Collaborator
•
Draft
[Pallas] Fix atomic add semantics (when targeting an output ref)
CLA Signed
This label is managed by the Meta Open Source bot.
#2615
opened May 27, 2026 by
thcmbs
Collaborator
Loading…
[cute] Move promotion function to cute/device_ir_lowering
CLA Signed
This label is managed by the Meta Open Source bot.
#2613
opened May 27, 2026 by
hinriksnaer
Collaborator
Loading…
[cute] Introduce DeviceIRLowering pipeline and migrate cute
CLA Signed
This label is managed by the Meta Open Source bot.
#2612
opened May 27, 2026 by
hinriksnaer
Collaborator
Loading…
[fast-launcher] Concrete-tensor fast path skips _hashable_dims (Python-only)
CLA Signed
This label is managed by the Meta Open Source bot.
[fast-launcher] Minimal Rust CompiledLauncher with __call__ (Chunk E)
CLA Signed
This label is managed by the Meta Open Source bot.
[fast-launcher] Minimal C CompiledLauncher with tp_call (Chunk E)
CLA Signed
This label is managed by the Meta Open Source bot.
[fast-launcher] Add helion._native optional Rust extension shim
CLA Signed
This label is managed by the Meta Open Source bot.
[fast-launcher] Codegen rewrite to wire the pool into generated wrappers (Python-only)
CLA Signed
This label is managed by the Meta Open Source bot.
[fast-launcher] Opt-in output-tensor pool (Chunk D, Python-only)
CLA Signed
This label is managed by the Meta Open Source bot.
[compile] Migrate Types to a dedicated module
CLA Signed
This label is managed by the Meta Open Source bot.
#2603
opened May 27, 2026 by
hinriksnaer
Collaborator
Loading…
[pallas] fix emit_pipeline's output mapping and indexing
CLA Signed
This label is managed by the Meta Open Source bot.
#2599
opened May 27, 2026 by
cota
Collaborator
Loading…
[pallas] direct-call hot-path squeezes: sig-lock + full_invoke closure baking
CLA Signed
This label is managed by the Meta Open Source bot.
[Pallas] examples: add TPU-optimized jagged_sum
CLA Signed
This label is managed by the Meta Open Source bot.
#2596
opened May 27, 2026 by
yarongmu-google
Collaborator
•
Draft
Previous Next
ProTip!
Follow long discussions with comments:>50.