Reimplement (part of) the or -> add optimization. Matching 'or' into 'add'

is general goodness because it allows ORs to be converted to LEA to avoid
inserting copies.  However, this is bad because it makes the generated .s
file less obvious and gives valgrind heartburn (tons of false positives in
bitfield code).

While the general fix should be in valgrind, we can at least try to avoid
emitting ADD instructions that *don't* get promoted to LEA.  This is more
work because it requires introducing pseudo instructions to represents
"add that knows the bits are disjoint", but hey, people really love valgrind.

This fixes this testcase:
https://bugs.kde.org/show_bug.cgi?id=242137#c20

the add r/i cases are coming next.

llvm-svn: 116007

1 files changed, 8 insertions, 0 deletions

diff --git a/llvm/lib/Target/X86/X86MCInstLower.cpp b/llvm/lib/Target/X86/X86MCInstLower.cpp
index 29ee2bfe8bf..662717bae71 100644
--- a/llvm/lib/Target/X86/X86MCInstLower.cpp
+++ b/llvm/lib/Target/X86/X86MCInstLower.cpp
@@ -347,6 +347,7 @@ void X86MCInstLower::Lower(const MachineInstr *MI, MCInst &OutMI) const {
   }
   
   // Handle a few special cases to eliminate operand modifiers.
+ReSimplify:
   switch (OutMI.getOpcode()) {
   case X86::LEA64_32r: // Handle 'subreg rewriting' for the lea64_32mem operand.
     lower_lea64_32mem(&OutMI, 1);
@@ -433,6 +434,13 @@ void X86MCInstLower::Lower(const MachineInstr *MI, MCInst &OutMI) const {
     break;
   }
 
+  // These are pseudo-ops for OR to help with the OR->ADD transformation.  We do
+  // this with an ugly goto in case the resultant OR uses EAX and needs the
+  // short form.
+  case X86::ADD16rr_DB: OutMI.setOpcode(X86::OR16rr); goto ReSimplify;
+  case X86::ADD32rr_DB: OutMI.setOpcode(X86::OR32rr); goto ReSimplify;
+  case X86::ADD64rr_DB: OutMI.setOpcode(X86::OR64rr); goto ReSimplify;
+      
   // The assembler backend wants to see branches in their small form and relax
   // them to their large form.  The JIT can only handle the large form because
   // it does not do relaxation.  For now, translate the large form to the