fix PR3537: if resetting bbi back to the start of a block, we need to

forget about already inserted expressions.

llvm-svn: 64362

llvm/lib/Transforms/Scalar/CodeGenPrepare.cpp

@@ -1241,11 +1241,13 @@ bool CodeGenPrepare::OptimizeMemoryInst(Instruction *MemoryInst, Value *Addr,
   // computation.
   Value *&SunkAddr = SunkAddrs[Addr];
   if (SunkAddr) {
-    DEBUG(cerr << "CGP: Reusing nonlocal addrmode: " << AddrMode << "\n");
+    DEBUG(cerr << "CGP: Reusing nonlocal addrmode: " << AddrMode << " for "
+               << *MemoryInst);
     if (SunkAddr->getType() != Addr->getType())
       SunkAddr = new BitCastInst(SunkAddr, Addr->getType(), "tmp", InsertPt);
   } else {
-    DEBUG(cerr << "CGP: SINKING nonlocal addrmode: " << AddrMode << "\n");
+    DEBUG(cerr << "CGP: SINKING nonlocal addrmode: " << AddrMode << " for "
+               << *MemoryInst);
     const Type *IntPtrTy = TLI->getTargetData()->getIntPtrType();
 
     Value *Result = 0;
@@ -1505,9 +1507,12 @@ bool CodeGenPrepare::OptimizeBlock(BasicBlock &BB) {
       if (TLI && isa<InlineAsm>(CI->getCalledValue()))
         if (const TargetAsmInfo *TAI =
             TLI->getTargetMachine().getTargetAsmInfo()) {
-          if (TAI->ExpandInlineAsm(CI))
+          if (TAI->ExpandInlineAsm(CI)) {
             BBI = BB.begin();
-          else
+            // Avoid processing instructions out of order, which could cause
+            // reuse before a value is defined.
+            SunkAddrs.clear();
+          } else
             // Sink address computing for memory operands into the block.
             MadeChange |= OptimizeInlineAsmInst(I, &(*CI), SunkAddrs);
         }

llvm/test/CodeGen/X86/2009-02-11-codegenprepare-reuse.ll

+; RUN: llvm-as < %s | llc
+; PR3537
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
+target triple = "i386-apple-darwin9.6"
+	%struct.GetBitContext = type <{ i8*, i8*, i32, i32 }>
+
+define i32 @alac_decode_frame() nounwind {
+entry:
+	%tmp2 = load i8** null		; <i8*> [#uses=2]
+	%tmp34 = getelementptr i8* %tmp2, i32 4		; <i8*> [#uses=2]
+	%tmp5.i424 = bitcast i8* %tmp34 to i8**		; <i8**> [#uses=2]
+	%tmp15.i = getelementptr i8* %tmp2, i32 12		; <i8*> [#uses=1]
+	%0 = bitcast i8* %tmp15.i to i32*		; <i32*> [#uses=1]
+	br i1 false, label %if.then43, label %if.end47
+
+if.then43:		; preds = %entry
+	ret i32 0
+
+if.end47:		; preds = %entry
+	%tmp5.i590 = load i8** %tmp5.i424		; <i8*> [#uses=0]
+	store i32 19, i32* %0
+	%tmp6.i569 = load i8** %tmp5.i424		; <i8*> [#uses=0]
+	%1 = call i32 asm "bswap   $0", "=r,0,~{dirflag},~{fpsr},~{flags}"(i32 0) nounwind		; <i32> [#uses=0]
+	br i1 false, label %bb.nph, label %if.then63
+
+if.then63:		; preds = %if.end47
+	unreachable
+
+bb.nph:		; preds = %if.end47
+	%2 = bitcast i8* %tmp34 to %struct.GetBitContext*		; <%struct.GetBitContext*> [#uses=1]
+	%call9.i = call fastcc i32 @decode_scalar(%struct.GetBitContext* %2, i32 0, i32 0, i32 0) nounwind		; <i32> [#uses=0]
+	unreachable
+}
+
+declare fastcc i32 @decode_scalar(%struct.GetBitContext* nocapture, i32, i32, i32) nounwind