Issues · NVIDIA/warp · GitHub

Labels Milestones New issue

Add `wp.tile_dot()` for fused element-wise multiply and reduce

feature request

#1364

· daedalus5 opened

on Apr 13, 2026

·

Add `wp.tile_axpy()` for fused scalar-multiply-and-accumulate on tiles

feature request

#1363

· daedalus5 opened

on Apr 13, 2026

·

`@wp.func_native` passes tile parameters by value instead of by reference

#1362

· daedalus5 opened

on Apr 13, 2026

·

Optimize tiled CUDA launches by reading blockIdx.x directly

#1361

· adenzler-nvidia opened

on Apr 13, 2026

Split ASV CI job into three parallel jobs by variability tier

#1359

· c0d1f1ed opened

on Apr 12, 2026

·

Enable LLVM auto-vectorization for CPU kernels

feature request

#1358

· c0d1f1ed opened

on Apr 12, 2026

·

Reduce Python-side overhead for kernel launches using METH_FASTCALL

#1357

· c0d1f1ed opened

on Apr 10, 2026

·

Reduce Python-side overhead for builtin functions using METH_FASTCALL

#1356

· c0d1f1ed opened

on Apr 10, 2026

·

Remove Direct Access to Private API

#1352

· christophercrouzet opened

on Apr 8, 2026

·

create_warp_function fails with parenthesized multi-line lambda bodies

#1351

· shi-eric opened

on Apr 8, 2026

wp.Tape: backward() evaluates at current array state, not forward values — silent gradient corruption

#1350

· jf--- opened

on Apr 7, 2026

·

APIC MVP

feature request

#1349

· c0d1f1ed opened

on Apr 7, 2026

·