[PATCH] D47346: [X86] Add a new pass to fold extension into load instructions in previous BB

[Guozhi Wei via Phabricator via llvm-commits]

Carrot created this revision.
Carrot added a reviewer: craig.topper.
Herald added a subscriber: mgorny.

Compile the attached new test case, original llvm generates

  movzwl  a, %esi          // %esi already contains zero extended value
  calll   v1  
  jmp     .LBB0_3

.LBB0_2:

  movzwl  a+2, %esi    // %esi already contains zero extended value
  calll   v2  

.LBB0_3:

  movzwl  %si, %eax    // another zext, we should avoid it.
  cmpl    $4, %eax

In source code all related values are 16 bit. In lowering phase, function X86TargetLowering::EmitCmp intentionally creates zero extension before cmp in order to avoid 16 bit immediate, which is slow on modern x86. Later in X86FixupBWInsts.cpp 16 bit loads are changed to load and extension, it makes the extension before cmp redundant.

The extension created by X86TargetLowering::EmitCmp is beneficial, we need to fold it into previous load instructions to get best performance. The function optimizeLoadInstr() can only move the load into its single user in the same BB, so it can fold the simple ext(load) pair, but it can't do the folding cross BB, and it can't fold ext into load, so it can't be used for this purpose.

So I write this new pass X86FoldXBBExtLoad.cpp to fold 16bit extension into previous load instructions.


Repository:
  rL LLVM

https://reviews.llvm.org/D47346

Files:
  lib/Target/X86/CMakeLists.txt
  lib/Target/X86/X86.h
  lib/Target/X86/X86FoldXBBExtLoad.cpp
  lib/Target/X86/X86TargetMachine.cpp
  test/CodeGen/X86/O3-pipeline.ll
  test/CodeGen/X86/bmi-intrinsics-fast-isel.ll
  test/CodeGen/X86/fold-xbb-ext-load.ll
  test/CodeGen/X86/jump_sign.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D47346.148453.patch
Type: text/x-patch
Size: 11442 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180524/1b348963/attachment.bin>