[Thumb] Fix crash in Thumb1RegisterInfo::rewriteFrameIndex

This function can, for some offsets from the SP, split one instruction
into two. Since it re-uses the original instruction as the first
instruction of the result, we need ensure its result register is not
marked as dead before we use it in the second instruction.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220194 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/ARM/Thumb1RegisterInfo.cpp

@@ -418,6 +418,7 @@ rewriteFrameIndex(MachineBasicBlock::iterator II, unsigned FrameRegIdx,
       }
       Offset = (Offset - Mask * Scale);
       MachineBasicBlock::iterator NII = std::next(II);
+      MI.getOperand(0).setIsDead(false);
       emitThumbRegPlusImmediate(MBB, NII, dl, DestReg, DestReg, Offset, TII,
                                 *this);
     } else {

test/CodeGen/Thumb/large-stack.ll

@@ -33,3 +33,22 @@ define i32 @test3() {
     %tmp1 = load i32* %tmp
     ret i32 %tmp1
 }
+
+; Here, the adds get optimized out because they are dead, but the calculation
+; of the address of stack_a is dead but not optimized out. When the address
+; calculation gets expanded to two instructions, we need to avoid reading a
+; dead register.
+; No CHECK lines (just test for crashes), as we hope this will be optimised
+; better in future.
+define i32 @test4() {
+entry:
+  %stack_a = alloca i8, align 1
+  %stack_b = alloca [256 x i32*], align 4
+  %int = ptrtoint i8* %stack_a to i32
+  %add = add i32 %int, 1
+  br label %block2
+
+block2:
+  %add2 = add i32 %add, 1
+  ret i32 0
+}