[all-commits] [llvm/llvm-project] a67f43: [NVPTX] Add range attributes for cluster rank intr...

Daniel Donenfeld via All-commits all-commits at lists.llvm.org
Wed Sep 30 14:39:42 PDT 2026


  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: a67f4353a08c319a7fef6da1cc0dc9f24e22fa10
      https://github.com/llvm/llvm-project/commit/a67f4353a08c319a7fef6da1cc0dc9f24e22fa10
  Author: Daniel Donenfeld <ddonenfeld at nvidia.com>
  Date:   2026-09-30 (Wed, 30 Sep 2026)

  Changed paths:
    M llvm/include/llvm/IR/IntrinsicsNVVM.td
    M llvm/test/CodeGen/NVPTX/intr-range.ll
    A llvm/test/Transforms/InstCombine/NVPTX/intrinsic-range.ll

  Log Message:
  -----------
  [NVPTX] Add range attributes for cluster rank intrinsics to preserve signed non-negativity (#224727)

Adds `range` properties in `IntrinsicsNVVM.td` for:

* `llvm.nvvm.read.ptx.sreg.cluster.ctarank`
* `llvm.nvvm.read.ptx.sreg.cluster.nctarank`

The CTA rank is encoded in a single byte within a shared pointer, so the
max value of ctarank is 255 and the number of CTAs is one greater. This
enables InstCombine to apply signed-division optimizations (e.g. sdiv X,
16 -\> shift-based form) that require X \>= 0.

---------

Co-authored-by: Alex MacLean <amaclean at nvidia.com>



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications


More information about the All-commits mailing list