[all-commits] [llvm/llvm-project] c0cbe3: [NFC][DAGCombine][X86][AArch64] Tests for 'A - (A ...

[Roman Lebedev via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: c0cbe3fbb7d6b3f197b05a47a2cc25ccb7977629
      https://github.com/llvm/llvm-project/commit/c0cbe3fbb7d6b3f197b05a47a2cc25ccb7977629
  Author: Roman Lebedev <lebedev.ri at gmail.com>
  Date:   2020-01-03 (Fri, 03 Jan 2020)

  Changed paths:
    A llvm/test/CodeGen/AArch64/align-down.ll
    A llvm/test/CodeGen/X86/align-down.ll

  Log Message:
  -----------
  [NFC][DAGCombine][X86][AArch64] Tests for 'A - (A & (B - 1))' pattern (PR44448)

https://rise4fun.com/Alive/ZVdp

Name: ptr - (ptr & (alignment-1))  ->  ptr & (0 - alignment)
  %mask = add i64 %alignment, -1
  %bias = and i64 %ptr, %mask
  %r = sub i64 %ptr, %bias
=>
  %highbitmask = sub i64 0, %alignment
  %r = and i64 %ptr, %highbitmask

The main motivational pattern involes pointer-typed values,
so this transform can't really be done in middle-end.

See
  https://bugs.llvm.org/show_bug.cgi?id=44448
  https://reviews.llvm.org/D71499


  Commit: 8dab0a4a7d691f2704f1079538e0ef29548db159
      https://github.com/llvm/llvm-project/commit/8dab0a4a7d691f2704f1079538e0ef29548db159
  Author: Roman Lebedev <lebedev.ri at gmail.com>
  Date:   2020-01-03 (Fri, 03 Jan 2020)

  Changed paths:
    M llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
    M llvm/test/CodeGen/AArch64/align-down.ll
    M llvm/test/CodeGen/X86/align-down.ll

  Log Message:
  -----------
  [DAGCombine][X86][AArch64] 'A - (A & (B - 1))' -> 'A & (0 - B)' fold (PR44448)

While we do manage to fold integer-typed IR in middle-end,
we can't do that for the main motivational case of pointers.

There is @llvm.ptrmask() intrinsic which may or may not be helpful,
but i'm not sure it is fully considered canonical yet,
not everything is fully aware of it likely.

https://rise4fun.com/Alive/ZVdp

Name: ptr - (ptr & (alignment-1))  ->  ptr & (0 - alignment)
  %mask = add i64 %alignment, -1
  %bias = and i64 %ptr, %mask
  %r = sub i64 %ptr, %bias
=>
  %highbitmask = sub i64 0, %alignment
  %r = and i64 %ptr, %highbitmask

See
  https://bugs.llvm.org/show_bug.cgi?id=44448
  https://reviews.llvm.org/D71499


Compare: https://github.com/llvm/llvm-project/compare/2e033244417c...8dab0a4a7d69