Mark x86's V_SET0 and V_SETALLONES with isSimpleLoad, and teach X86's

foldMemoryOperand how to "fold" them, by converting them into constant-pool
loads. When they aren't folded, they use xorps/cmpeqd, but for example when
register pressure is high, they may now be folded as memory operands, which
reduces register pressure.

Also, mark V_SET0 isAsCheapAsAMove so that two-address-elimination will
remat it instead of copying zeros around (V_SETALLONES was already marked).

llvm-svn: 60461

1 files changed, 0 insertions, 9 deletions

diff --git a/llvm/utils/TableGen/CodeGenDAGPatterns.cpp b/llvm/utils/TableGen/CodeGenDAGPatterns.cpp
index b3ed27b8d89..859a8ccac8d 100644
--- a/llvm/utils/TableGen/CodeGenDAGPatterns.cpp
+++ b/llvm/utils/TableGen/CodeGenDAGPatterns.cpp
@@ -1759,15 +1759,6 @@ static void InferFromPattern(const CodeGenInstruction &Inst,
     MayLoad = true;
   }
 
-  // Sanity-check the isSimpleLoad flag.
-  if (Inst.isSimpleLoad) {
-    if (!MayLoad)
-      fprintf(stderr,
-              "Warning: mayLoad flag not set or inferred for instruction '%s'"
-              " which has isSimpleLoad set.\n",
-              Inst.TheDef->getName().c_str());
-  }
-
   if (Inst.neverHasSideEffects) {
     if (HadPattern)
       fprintf(stderr, "Warning: neverHasSideEffects set on instruction '%s' "