[x86] Fix register class of folded load index reg.

cooperp · cooperp · commit d31583ddfbc1 · 2015-05-06T21:37:19.000Z
When folding a load in to another instruction, we need to fix the class of the index register
Otherwise, it could be something like GR64 not GR64_NOSP and would fail the machine verifier.

llvm-svn: 236644
diff --git a/llvm/lib/Target/X86/X86FastISel.cpp b/llvm/lib/Target/X86/X86FastISel.cpp
@@ -3534,6 +3534,23 @@ bool X86FastISel::tryToFoldLoadIntoMI(MachineInstr *MI, unsigned OpNo,
   if (!Result)
     return false;
 
+  // The index register could be in the wrong register class.  Unfortunately,
+  // foldMemoryOperandImpl could have commuted the instruction so its not enough
+  // to just look at OpNo + the offset to the index reg.  We actually need to
+  // scan the instruction to find the index reg and see if its the correct reg
+  // class.
+  for (MIOperands MO(Result); MO.isValid(); ++MO) {
+    if (!MO->isReg() || MO->isDef() || MO->getReg() != AM.IndexReg)
+      continue;
+    // Found the index reg, now try to rewrite it.
+    unsigned OpNo = MO.getOperandNo();
+    unsigned IndexReg = constrainOperandRegClass(Result->getDesc(),
+                                                 MO->getReg(), OpNo);
+    if (IndexReg == MO->getReg())
+      continue;
+    MO->setReg(IndexReg);
+  }
+
   Result->addMemOperand(*FuncInfo.MF, createMachineMemOperandFor(LI));
   FuncInfo.MBB->insert(FuncInfo.InsertPt, Result);
   MI->eraseFromParent();
diff --git a/llvm/test/CodeGen/X86/fast-isel-movsbl-indexreg.ll b/llvm/test/CodeGen/X86/fast-isel-movsbl-indexreg.ll
@@ -0,0 +1,20 @@
+; RUN: llc %s -o - -verify-machineinstrs -fast-isel=true | FileCheck %s
+
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-unknown"
+
+; The index register on the folded memory operand was incorrect.  Ensure we generate
+; movsbl in fast-isel, but also that it passes verification which will check the register
+; class.
+
+; CHECK: movsbl
+
+@table = external hidden global [64 x i8], align 16
+
+define i32 @test(i32 %x, i64 %offset) {
+bb:
+  %tmp37 = getelementptr inbounds [64 x i8], [64 x i8]* @table, i64 0, i64 %offset
+  %tmp38 = load i8, i8* %tmp37, align 1
+  %tmp39 = sext i8 %tmp38 to i32
+  ret i32 %tmp39
+}
