[PATCH][X86] Lower unsigned vsetcc to psubus in certain cases

[Adam Nemet]

This is another improvement for SingleSource/Benchmarks/Misc/lowercase.c.

The current approach to lower a vsetult is to flip the sign bit of the operands, swap the operands and then use a (signed) pcmpgt.  psubus (unsigned saturating subtract) can be used to emulate a vsetult more efficiently:

+    case ISD::SETULT: {
+      // If the comparison is against a constant we can turn this into a
+      // setule.  With psubus, setule does not require a swap.  This is
+      // beneficial because the constant in the register is no longer
+      // destructed as the destination so it can be hoisted out of a loop.

I also enable lowering via psubus in a few other cases where it’s clearly beneficial: setule and setuge if minu/maxu cannot be used.  You can see the improvements in the diff to the existing test vec_setcc.ll.

I am leaving the other case alone for now because it’s harder to reason about profitability in those cases.

Tested on x86_64-apple-darwin.

Please apply if it looks good.

Adam

<rdar://problem/14338765>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: vsetcc-psubus-v1.patch
Type: application/octet-stream
Size: 9688 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140224/ea19df7d/attachment.obj>