[all-commits] [llvm/llvm-project] 210d23: ValueTracking: Teach computeKnownFPClass to look a...

[Matt Arsenault via All-commits]

  Branch: refs/heads/users/arsenm/valuetracking/computeKnownFPClass-bitcast-integer-min-max
  Home:   https://github.com/llvm/llvm-project
  Commit: 210d2383071449f3dc12ed6e98cd6e4f284abc70
      https://github.com/llvm/llvm-project/commit/210d2383071449f3dc12ed6e98cd6e4f284abc70
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2026-03-04 (Wed, 04 Mar 2026)

  Changed paths:
    M llvm/lib/Analysis/ValueTracking.cpp
    A llvm/test/Transforms/Attributor/nofpclass-bitcast.ll

  Log Message:
  -----------
  ValueTracking: Teach computeKnownFPClass to look at bitcast + integer max

The returned class will still be one of the bitpatterns.

This pattern is used in rocm device libraries in assorted functions, e.g.,
https://github.com/ROCm/llvm-project/blob/amd-staging/amd/device-libs/ocml/src/rlen3F.cl#L20

I believe it is blocking the eliminationg of finite checks in some of the more
complex functions.



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