[PATCH] D48765: [X86] The TEST instruction is eliminated when BSF/TZCNT is used

[Ivan Kulagin via Phabricator via llvm-commits]

ikulagin created this revision.
ikulagin added reviewers: davide, craig.topper.
Herald added a subscriber: llvm-commits.

These changes cover the PR#31399. 
Now the ffs(x) function is lowered to (x != 0) ? llvm.cttz(x) + 1 : 0
and it corresponds to the following llvm code:

  %cnt = tail call i32 @llvm.cttz.i32(i32 %v, i1 true)
  %tobool = icmp eq i32 %v, 0
  %.op = add nuw nsw i32 %cnt, 1
  %add = select i1 %tobool, i32 0, i32 %.op

and x86 asm code:

  bsfl          %edi, %ecx
  addl         $1, %ecx
  testl         %edi, %edi
  movl        $0, %eax
  cmovnel  %ecx, %eax

In this case the 'test' instruction can't be eliminated because
the 'add' instruction modifies the EFLAGS, namely, ZF flag
that is set by the 'bsf' instruction when 'x' is zero.

I have moved the 'add' instruction below the 'cmov' instruction
at the peephole optimization stage during the compare instruction
optimization (optimizeCompareInstr), i.e., to implement
the following transformation:

  bsfl    %edi, %ecx 
  addl    $1, %ecx       --------* c1 = 1 
  testl   %edi, %edi                |
  movl    $0, %eax                 *  c2 = 0 -> c2 = c2 - c1 = -1 
                                               *  transform to movl  $-1, %eax
  cmovnel %ecx, %eax          |
                                 <--------|

It produces the following code:

  bsfl          %edi, %ecx
  movl        $-1, %eax
  cmovnel  %ecx, %eax
  addl         $1, %eax


Repository:
  rL LLVM

https://reviews.llvm.org/D48765

Files:
  lib/Target/X86/X86InstrInfo.cpp
  test/CodeGen/X86/peep-test-4.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D48765.153465.patch
Type: text/x-patch
Size: 7698 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180629/17843c68/attachment.bin>