[all-commits] [llvm/llvm-project] f8e133: [SLP] Support internal users of splat loads

[vporpo via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: f8e1337115623cb879f734940fd9dfeb29a611e7
      https://github.com/llvm/llvm-project/commit/f8e1337115623cb879f734940fd9dfeb29a611e7
  Author: Vasileios Porpodas <vporpodas at google.com>
  Date:   2022-04-18 (Mon, 18 Apr 2022)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/test/Transforms/SLPVectorizer/X86/lookahead.ll

  Log Message:
  -----------
  [SLP] Support internal users of splat loads

Until now we would only accept a broadcast load pattern if it is only used
by a single vector of instructions.

This patch relaxes this, and allows for the broadcast to have more than one
user vector, as long as all of its uses are internal to the SLP graph and
vectorized.

Differential Revision: https://reviews.llvm.org/D121940