Teach RemoveCopyByCommutingDef to check all aliases, not just subregisters.

This caused a miscompilation in WebKit where %RAX had conflicting defs when
RemoveCopyByCommutingDef was commuting a %EAX use.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@112751 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 42 insertions, 16 deletions

diff --git a/lib/CodeGen/SimpleRegisterCoalescing.cpp b/lib/CodeGen/SimpleRegisterCoalescing.cpp
index 5dd5fdb7b1..b29ea19835 100644
--- a/lib/CodeGen/SimpleRegisterCoalescing.cpp
+++ b/lib/CodeGen/SimpleRegisterCoalescing.cpp
@@ -389,16 +389,12 @@ bool SimpleRegisterCoalescing::RemoveCopyByCommutingDef(const CoalescerPair &CP,
   if (HasOtherReachingDefs(IntA, IntB, AValNo, BValNo))
     return false;
 
-  bool BHasSubRegs = false;
-  if (TargetRegisterInfo::isPhysicalRegister(IntB.reg))
-    BHasSubRegs = *tri_->getSubRegisters(IntB.reg);
-
-  // Abort if the subregisters of IntB.reg have values that are not simply the
+  // Abort if the aliases of IntB.reg have values that are not simply the
   // clobbers from the superreg.
-  if (BHasSubRegs)
-    for (const unsigned *SR = tri_->getSubRegisters(IntB.reg); *SR; ++SR)
-      if (li_->hasInterval(*SR) &&
-          HasOtherReachingDefs(IntA, li_->getInterval(*SR), AValNo, 0))
+  if (TargetRegisterInfo::isPhysicalRegister(IntB.reg))
+    for (const unsigned *AS = tri_->getAliasSet(IntB.reg); *AS; ++AS)
+      if (li_->hasInterval(*AS) &&
+          HasOtherReachingDefs(IntA, li_->getInterval(*AS), AValNo, 0))
         return false;
 
   // If some of the uses of IntA.reg is already coalesced away, return false.
@@ -415,6 +411,8 @@ bool SimpleRegisterCoalescing::RemoveCopyByCommutingDef(const CoalescerPair &CP,
       return false;
   }
 
+  DEBUG(dbgs() << "\tRemoveCopyByCommutingDef: " << *DefMI);
+
   // At this point we have decided that it is legal to do this
   // transformation.  Start by commuting the instruction.
   MachineBasicBlock *MBB = DefMI->getParent();
@@ -478,7 +476,7 @@ bool SimpleRegisterCoalescing::RemoveCopyByCommutingDef(const CoalescerPair &CP,
     if (UseMI->getOperand(0).getReg() != IntB.reg ||
         UseMI->getOperand(0).getSubReg())
       continue;
-        
+
     // This copy will become a noop. If it's defining a new val#,
     // remove that val# as well. However this live range is being
     // extended to the end of the existing live range defined by the copy.
@@ -503,13 +501,13 @@ bool SimpleRegisterCoalescing::RemoveCopyByCommutingDef(const CoalescerPair &CP,
   // Remove val#'s defined by copies that will be coalesced away.
   for (unsigned i = 0, e = BDeadValNos.size(); i != e; ++i) {
     VNInfo *DeadVNI = BDeadValNos[i];
-    if (BHasSubRegs) {
-      for (const unsigned *SR = tri_->getSubRegisters(IntB.reg); *SR; ++SR) {
-        if (!li_->hasInterval(*SR))
+    if (TargetRegisterInfo::isPhysicalRegister(IntB.reg)) {
+      for (const unsigned *AS = tri_->getAliasSet(IntB.reg); *AS; ++AS) {
+        if (!li_->hasInterval(*AS))
           continue;
-        LiveInterval &SRLI = li_->getInterval(*SR);
-        if (const LiveRange *SRLR = SRLI.getLiveRangeContaining(DeadVNI->def))
-          SRLI.removeValNo(SRLR->valno);
+        LiveInterval &ASLI = li_->getInterval(*AS);
+        if (const LiveRange *ASLR = ASLI.getLiveRangeContaining(DeadVNI->def))
+          ASLI.removeValNo(ASLR->valno);
       }
     }
     IntB.removeValNo(BDeadValNos[i]);
diff --git a/test/CodeGen/X86/2010-09-01-RemoveCopyByCommutingDef.ll b/test/CodeGen/X86/2010-09-01-RemoveCopyByCommutingDef.ll
new file mode 100644
index 0000000000..e5542baf2e
--- /dev/null
+++ b/test/CodeGen/X86/2010-09-01-RemoveCopyByCommutingDef.ll
@@ -0,0 +1,28 @@
+; RUN: llc < %s -verify-machineinstrs | FileCheck %s
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+target triple = "x86_64-apple-darwin10.0.0"
+
+; This test exercises the alias checking in SimpleRegisterCoalescing::RemoveCopyByCommutingDef.
+
+define void @f(i32* %w, i32* %h, i8* %_this, i8* %image) nounwind ssp {
+  %x1 = tail call i64 @g(i8* %_this, i8* %image) nounwind ; <i64> [#uses=3]
+  %tmp1 = trunc i64 %x1 to i32                     ; <i32> [#uses=1]
+; CHECK: movl (%r{{.*}}), %
+  %x4 = load i32* %h, align 4                      ; <i32> [#uses=1]
+
+; The imull clobbers a 32-bit register.
+; CHECK: imull %{{...}}, %e[[CLOBBER:..]]
+  %x5 = mul nsw i32 %x4, %tmp1                      ; <i32> [#uses=1]
+
+; So we cannot use the corresponding 64-bit register anymore.
+; CHECK-NOT: shrq $32, %r[[CLOBBER]]
+  %btmp3 = lshr i64 %x1, 32                         ; <i64> [#uses=1]
+  %btmp4 = trunc i64 %btmp3 to i32                  ; <i32> [#uses=1]
+
+; CHECK: idiv
+  %x6 = sdiv i32 %x5, %btmp4                         ; <i32> [#uses=1]
+  store i32 %x6, i32* %w, align 4
+  ret void
+}
+
+declare i64 @g(i8*, i8*)