[all-commits] [llvm/llvm-project] ce586b: [msan] Handle AVX512 fpclass (floating-point class...

[Thurston Dang via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: ce586b57f4a8e32d47df58dc1b11730f07ffc834
      https://github.com/llvm/llvm-project/commit/ce586b57f4a8e32d47df58dc1b11730f07ffc834
  Author: Thurston Dang <thurston at google.com>
  Date:   2026-07-15 (Wed, 15 Jul 2026)

  Changed paths:
    M llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
    M llvm/test/Instrumentation/MemorySanitizer/X86/avx512dq-intrinsics.ll

  Log Message:
  -----------
  [msan] Handle AVX512 fpclass (floating-point classification) for packed double/float (#209652)

Each bit of the classifier constant specifies whether a particular
classifier is enabled. If Classifiers == 0 (no classifiers are enabled),
the output is trivially known to be zero, thus the output is fully
initialized.

Otherwise, each bit of the output is the bitwise OR of one or more
classifications; we approximate each bit of the output shadow based on
whether the corresponding input element is fully initialized (without
which the classification is potentially unknown). This is only
approximate, because some classifications do not rely on all the bits of
the input element.

This patch only handles packed double/single-precision floating-point.
Handling the scalar equivalents is more complicated because those
intrinsics encode a mask.



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