[PATCH] D48789: [X86] Replace (32/64 - n) shift amounts with (neg n) since the shift amount is masked in hardware

[Craig Topper via Phabricator via llvm-commits]

craig.topper updated this revision to Diff 153592.
craig.topper added a reviewer: niravd.
craig.topper added a comment.

Place the new nodes before the original shift amount in the topological order of the DAG. This ensures that all shifts that had the same shift amount will be selected and run this transform before any of the new nodes are selected. The selection DAG should CSE all the new nodes together for multiple shifts.


https://reviews.llvm.org/D48789

Files:
  lib/Target/X86/X86ISelDAGToDAG.cpp
  test/CodeGen/X86/extract-lowbits.ll
  test/CodeGen/X86/schedule-x86-64-shld.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D48789.153592.patch
Type: text/x-patch
Size: 20194 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180629/1a669cb7/attachment.bin>