[llvm-commits] [llvm] r44263 - in /llvm/trunk: lib/CodeGen/SelectionDAG/LegalizeDAGTypes.cpp test/CodeGen/Generic/2007-11-21-UndeadIllegalNode.ll
Duncan Sands
baldrick at free.fr
Wed Nov 21 08:43:20 PST 2007
Author: baldrick
Date: Wed Nov 21 10:43:19 2007
New Revision: 44263
URL: http://llvm.org/viewvc/llvm-project?rev=44263&view=rev
Log:
Fix a bug in which node A is replaced by node B, but later
node A gets back into the DAG again because it was hiding in
one of the node maps: make sure that node replacement happens
in those maps too.
Added:
llvm/trunk/test/CodeGen/Generic/2007-11-21-UndeadIllegalNode.ll
Modified:
llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAGTypes.cpp
Modified: llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAGTypes.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAGTypes.cpp?rev=44263&r1=44262&r2=44263&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAGTypes.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAGTypes.cpp Wed Nov 21 10:43:19 2007
@@ -92,7 +92,11 @@
/// ExpandedNodes - For nodes that need to be expanded this map indicates
/// which operands are the expanded version of the input.
DenseMap<SDOperand, std::pair<SDOperand, SDOperand> > ExpandedNodes;
-
+
+ /// ReplacedNodes - For nodes that have been replaced with another,
+ /// indicates the replacement node to use.
+ DenseMap<SDOperand, SDOperand> ReplacedNodes;
+
/// Worklist - This defines a worklist of nodes to process. In order to be
/// pushed onto this worklist, all operands of a node must have already been
/// processed.
@@ -112,12 +116,15 @@
void MarkNewNodes(SDNode *N);
void ReplaceLegalValueWith(SDOperand From, SDOperand To);
-
+
+ void RemapNode(SDOperand &N);
+
SDOperand GetPromotedOp(SDOperand Op) {
- Op = PromotedNodes[Op];
- assert(Op.Val && "Operand wasn't promoted?");
- return Op;
- }
+ SDOperand &PromotedOp = PromotedNodes[Op];
+ RemapNode(PromotedOp);
+ assert(PromotedOp.Val && "Operand wasn't promoted?");
+ return PromotedOp;
+ }
void SetPromotedOp(SDOperand Op, SDOperand Result);
/// GetPromotedZExtOp - Get a promoted operand and zero extend it to the final
@@ -397,7 +404,11 @@
// Anything that used the old node should now use the new one. Note that this
// can potentially cause recursive merging.
DAG.ReplaceAllUsesOfValueWith(From, To);
-
+
+ // The old node may still be present in ExpandedNodes or PromotedNodes.
+ // Inform them about the replacement.
+ ReplacedNodes[From] = To;
+
// Since we just made an unstructured update to the DAG, which could wreak
// general havoc on anything that once used N and now uses Res, walk all users
// of the result, updating their flags.
@@ -414,6 +425,16 @@
}
}
+/// RemapNode - If the specified value was already legalized to another value,
+/// replace it by that value.
+void DAGTypeLegalizer::RemapNode(SDOperand &N) {
+ DenseMap<SDOperand, SDOperand>::iterator I = ReplacedNodes.find(N);
+ if (I != ReplacedNodes.end()) {
+ RemapNode(I->second);
+ N = I->second;
+ }
+}
+
void DAGTypeLegalizer::SetPromotedOp(SDOperand Op, SDOperand Result) {
if (Result.Val->getNodeId() == NewNode)
MarkNewNodes(Result.Val);
@@ -426,6 +447,8 @@
void DAGTypeLegalizer::GetExpandedOp(SDOperand Op, SDOperand &Lo,
SDOperand &Hi) {
std::pair<SDOperand, SDOperand> &Entry = ExpandedNodes[Op];
+ RemapNode(Entry.first);
+ RemapNode(Entry.second);
assert(Entry.first.Val && "Operand isn't expanded");
Lo = Entry.first;
Hi = Entry.second;
Added: llvm/trunk/test/CodeGen/Generic/2007-11-21-UndeadIllegalNode.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Generic/2007-11-21-UndeadIllegalNode.ll?rev=44263&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/Generic/2007-11-21-UndeadIllegalNode.ll (added)
+++ llvm/trunk/test/CodeGen/Generic/2007-11-21-UndeadIllegalNode.ll Wed Nov 21 10:43:19 2007
@@ -0,0 +1,161 @@
+; RUN: llvm-as < %s | llc -o -
+; XFAIL: *
+; Un-XFAIL this once LegalizeDAGTypes is turned on.
+
+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-f80:32:32"
+target triple = "i686-pc-linux-gnu"
+ %struct.RETURN = type { i32, i32 }
+ %struct.ada__finalization__controlled = type { %struct.system__finalization_root__root_controlled }
+ %struct.ada__streams__root_stream_type = type { %struct.ada__tags__dispatch_table* }
+ %struct.ada__strings__unbounded__string_access = type { i8*, %struct.RETURN* }
+ %struct.ada__strings__unbounded__unbounded_string = type { %struct.ada__finalization__controlled, %struct.ada__strings__unbounded__string_access, i32 }
+ %struct.ada__tags__dispatch_table = type { [1 x i32] }
+ %struct.exception = type { i8, i8, i32, i8*, i8*, i32, i8* }
+ %struct.system__finalization_root__root_controlled = type { %struct.ada__streams__root_stream_type, %struct.system__finalization_root__root_controlled*, %struct.system__finalization_root__root_controlled* }
+ %struct.system__standard_library__exception_data = type { i8, i8, i32, i32, %struct.system__standard_library__exception_data*, i32, void ()* }
+ at C.495.7639 = internal constant %struct.RETURN { i32 1, i32 16 } ; <%struct.RETURN*> [#uses=1]
+ at ada__strings__index_error = external global %struct.exception ; <%struct.exception*> [#uses=1]
+ at .str5 = internal constant [16 x i8] c"a-strunb.adb:690" ; <[16 x i8]*> [#uses=1]
+
+declare void @llvm.memcpy.i32(i8*, i8*, i32, i32)
+
+declare void @ada__strings__unbounded__realloc_for_chunk(%struct.ada__strings__unbounded__unbounded_string*, i32)
+
+declare void @__gnat_raise_exception(%struct.system__standard_library__exception_data*, i64)
+
+define void @ada__strings__unbounded__insert__2(%struct.ada__strings__unbounded__unbounded_string* %source, i32 %before, i64 %new_item.0.0) {
+entry:
+ %tmp24636 = lshr i64 %new_item.0.0, 32 ; <i64> [#uses=1]
+ %tmp24637 = trunc i64 %tmp24636 to i32 ; <i32> [#uses=1]
+ %tmp24638 = inttoptr i32 %tmp24637 to %struct.RETURN* ; <%struct.RETURN*> [#uses=2]
+ %tmp25 = getelementptr %struct.RETURN* %tmp24638, i32 0, i32 0 ; <i32*> [#uses=1]
+ %tmp26 = load i32* %tmp25, align 4 ; <i32> [#uses=1]
+ %tmp29 = getelementptr %struct.RETURN* %tmp24638, i32 0, i32 1 ; <i32*> [#uses=1]
+ %tmp30 = load i32* %tmp29, align 4 ; <i32> [#uses=1]
+ %tmp63 = getelementptr %struct.ada__strings__unbounded__unbounded_string* %source, i32 0, i32 1, i32 1 ; <%struct.RETURN**> [#uses=5]
+ %tmp64 = load %struct.RETURN** %tmp63, align 4 ; <%struct.RETURN*> [#uses=1]
+ %tmp65 = getelementptr %struct.RETURN* %tmp64, i32 0, i32 0 ; <i32*> [#uses=1]
+ %tmp66 = load i32* %tmp65, align 4 ; <i32> [#uses=1]
+ %tmp67 = icmp sgt i32 %tmp66, %before ; <i1> [#uses=1]
+ br i1 %tmp67, label %bb77, label %bb
+
+bb: ; preds = %entry
+ %tmp71 = getelementptr %struct.ada__strings__unbounded__unbounded_string* %source, i32 0, i32 2 ; <i32*> [#uses=4]
+ %tmp72 = load i32* %tmp71, align 4 ; <i32> [#uses=1]
+ %tmp73 = add i32 %tmp72, 1 ; <i32> [#uses=1]
+ %tmp74 = icmp slt i32 %tmp73, %before ; <i1> [#uses=1]
+ br i1 %tmp74, label %bb77, label %bb84
+
+bb77: ; preds = %bb, %entry
+ tail call void @__gnat_raise_exception( %struct.system__standard_library__exception_data* bitcast (%struct.exception* @ada__strings__index_error to %struct.system__standard_library__exception_data*), i64 or (i64 zext (i32 ptrtoint ([16 x i8]* @.str5 to i32) to i64), i64 shl (i64 zext (i32 ptrtoint (%struct.RETURN* @C.495.7639 to i32) to i64), i64 32)) )
+ unreachable
+
+bb84: ; preds = %bb
+ %tmp93 = sub i32 %tmp30, %tmp26 ; <i32> [#uses=2]
+ %tmp9394 = sext i32 %tmp93 to i36 ; <i36> [#uses=1]
+ %tmp95 = shl i36 %tmp9394, 3 ; <i36> [#uses=1]
+ %tmp96 = add i36 %tmp95, 8 ; <i36> [#uses=2]
+ %tmp97 = icmp sgt i36 %tmp96, -1 ; <i1> [#uses=1]
+ %tmp100 = select i1 %tmp97, i36 %tmp96, i36 0 ; <i36> [#uses=2]
+ %tmp101 = icmp slt i36 %tmp100, 17179869177 ; <i1> [#uses=1]
+ %tmp100.cast = trunc i36 %tmp100 to i32 ; <i32> [#uses=1]
+ %min102 = select i1 %tmp101, i32 %tmp100.cast, i32 -8 ; <i32> [#uses=1]
+ tail call void @ada__strings__unbounded__realloc_for_chunk( %struct.ada__strings__unbounded__unbounded_string* %source, i32 %min102 )
+ %tmp148 = load i32* %tmp71, align 4 ; <i32> [#uses=4]
+ %tmp152 = add i32 %tmp93, 1 ; <i32> [#uses=2]
+ %tmp153 = icmp sgt i32 %tmp152, -1 ; <i1> [#uses=1]
+ %max154 = select i1 %tmp153, i32 %tmp152, i32 0 ; <i32> [#uses=5]
+ %tmp155 = add i32 %tmp148, %max154 ; <i32> [#uses=5]
+ %tmp315 = getelementptr %struct.ada__strings__unbounded__unbounded_string* %source, i32 0, i32 1, i32 0 ; <i8**> [#uses=4]
+ %tmp328 = load %struct.RETURN** %tmp63, align 4 ; <%struct.RETURN*> [#uses=1]
+ %tmp329 = getelementptr %struct.RETURN* %tmp328, i32 0, i32 0 ; <i32*> [#uses=1]
+ %tmp330 = load i32* %tmp329, align 4 ; <i32> [#uses=4]
+ %tmp324 = add i32 %max154, %before ; <i32> [#uses=3]
+ %tmp331 = sub i32 %tmp324, %tmp330 ; <i32> [#uses=1]
+ %tmp349 = sub i32 %before, %tmp330 ; <i32> [#uses=1]
+ %tmp356 = icmp sgt i32 %tmp331, %tmp349 ; <i1> [#uses=1]
+ %tmp431 = icmp sgt i32 %tmp324, %tmp155 ; <i1> [#uses=2]
+ br i1 %tmp356, label %bb420, label %bb359
+
+bb359: ; preds = %bb84
+ br i1 %tmp431, label %bb481, label %bb382
+
+bb382: ; preds = %bb382, %bb359
+ %indvar = phi i32 [ 0, %bb359 ], [ %indvar.next, %bb382 ] ; <i32> [#uses=2]
+ %max379.pn = phi i32 [ %max154, %bb359 ], [ %L492b.0, %bb382 ] ; <i32> [#uses=1]
+ %before.pn = phi i32 [ %before, %bb359 ], [ 1, %bb382 ] ; <i32> [#uses=1]
+ %L492b.0 = add i32 %before.pn, %max379.pn ; <i32> [#uses=3]
+ %tmp386 = load %struct.RETURN** %tmp63, align 4 ; <%struct.RETURN*> [#uses=1]
+ %tmp387 = getelementptr %struct.RETURN* %tmp386, i32 0, i32 0 ; <i32*> [#uses=1]
+ %tmp388 = load i32* %tmp387, align 4 ; <i32> [#uses=2]
+ %tmp392 = load i8** %tmp315, align 4 ; <i8*> [#uses=2]
+ %R493b.0 = add i32 %indvar, %before ; <i32> [#uses=1]
+ %tmp405 = sub i32 %R493b.0, %tmp388 ; <i32> [#uses=1]
+ %tmp406 = getelementptr i8* %tmp392, i32 %tmp405 ; <i8*> [#uses=1]
+ %tmp407 = load i8* %tmp406, align 1 ; <i8> [#uses=1]
+ %tmp408 = sub i32 %L492b.0, %tmp388 ; <i32> [#uses=1]
+ %tmp409 = getelementptr i8* %tmp392, i32 %tmp408 ; <i8*> [#uses=1]
+ store i8 %tmp407, i8* %tmp409, align 1
+ %tmp414 = icmp eq i32 %L492b.0, %tmp155 ; <i1> [#uses=1]
+ %indvar.next = add i32 %indvar, 1 ; <i32> [#uses=1]
+ br i1 %tmp414, label %bb481, label %bb382
+
+bb420: ; preds = %bb84
+ br i1 %tmp431, label %bb481, label %bb436.preheader
+
+bb436.preheader: ; preds = %bb420
+ %tmp4468 = load i8** %tmp315, align 4 ; <i8*> [#uses=2]
+ %tmp4599 = sub i32 %tmp148, %tmp330 ; <i32> [#uses=1]
+ %tmp46010 = getelementptr i8* %tmp4468, i32 %tmp4599 ; <i8*> [#uses=1]
+ %tmp46111 = load i8* %tmp46010, align 1 ; <i8> [#uses=1]
+ %tmp46212 = sub i32 %tmp155, %tmp330 ; <i32> [#uses=1]
+ %tmp46313 = getelementptr i8* %tmp4468, i32 %tmp46212 ; <i8*> [#uses=1]
+ store i8 %tmp46111, i8* %tmp46313, align 1
+ %exitcond14 = icmp eq i32 %tmp155, %tmp324 ; <i1> [#uses=1]
+ br i1 %exitcond14, label %bb481, label %bb.nph
+
+bb.nph: ; preds = %bb436.preheader
+ %tmp5 = sub i32 %tmp148, %before ; <i32> [#uses=1]
+ br label %bb478
+
+bb478: ; preds = %bb478, %bb.nph
+ %indvar6422 = phi i32 [ 0, %bb.nph ], [ %indvar.next643, %bb478 ] ; <i32> [#uses=1]
+ %indvar.next643 = add i32 %indvar6422, 1 ; <i32> [#uses=4]
+ %L490b.0 = sub i32 %tmp155, %indvar.next643 ; <i32> [#uses=1]
+ %R491b.0 = sub i32 %tmp148, %indvar.next643 ; <i32> [#uses=1]
+ %tmp440 = load %struct.RETURN** %tmp63, align 4 ; <%struct.RETURN*> [#uses=1]
+ %tmp441 = getelementptr %struct.RETURN* %tmp440, i32 0, i32 0 ; <i32*> [#uses=1]
+ %tmp442 = load i32* %tmp441, align 4 ; <i32> [#uses=2]
+ %tmp446 = load i8** %tmp315, align 4 ; <i8*> [#uses=2]
+ %tmp459 = sub i32 %R491b.0, %tmp442 ; <i32> [#uses=1]
+ %tmp460 = getelementptr i8* %tmp446, i32 %tmp459 ; <i8*> [#uses=1]
+ %tmp461 = load i8* %tmp460, align 1 ; <i8> [#uses=1]
+ %tmp462 = sub i32 %L490b.0, %tmp442 ; <i32> [#uses=1]
+ %tmp463 = getelementptr i8* %tmp446, i32 %tmp462 ; <i8*> [#uses=1]
+ store i8 %tmp461, i8* %tmp463, align 1
+ %exitcond = icmp eq i32 %indvar.next643, %tmp5 ; <i1> [#uses=1]
+ br i1 %exitcond, label %bb481, label %bb478
+
+bb481: ; preds = %bb478, %bb436.preheader, %bb420, %bb382, %bb359
+ %tmp577 = add i32 %before, -1 ; <i32> [#uses=3]
+ %tmp578 = add i32 %max154, %tmp577 ; <i32> [#uses=2]
+ %tmp581 = icmp sge i32 %tmp578, %tmp577 ; <i1> [#uses=1]
+ %max582 = select i1 %tmp581, i32 %tmp578, i32 %tmp577 ; <i32> [#uses=1]
+ %tmp584 = sub i32 %max582, %before ; <i32> [#uses=1]
+ %tmp585 = add i32 %tmp584, 1 ; <i32> [#uses=2]
+ %tmp586 = icmp sgt i32 %tmp585, -1 ; <i1> [#uses=1]
+ %max587 = select i1 %tmp586, i32 %tmp585, i32 0 ; <i32> [#uses=1]
+ %tmp591 = load %struct.RETURN** %tmp63, align 4 ; <%struct.RETURN*> [#uses=1]
+ %tmp592 = getelementptr %struct.RETURN* %tmp591, i32 0, i32 0 ; <i32*> [#uses=1]
+ %tmp593 = load i32* %tmp592, align 4 ; <i32> [#uses=1]
+ %tmp597 = load i8** %tmp315, align 4 ; <i8*> [#uses=1]
+ %tmp600621 = trunc i64 %new_item.0.0 to i32 ; <i32> [#uses=1]
+ %tmp600622 = inttoptr i32 %tmp600621 to i8* ; <i8*> [#uses=1]
+ %tmp601 = sub i32 %before, %tmp593 ; <i32> [#uses=1]
+ %tmp602 = getelementptr i8* %tmp597, i32 %tmp601 ; <i8*> [#uses=1]
+ tail call void @llvm.memcpy.i32( i8* %tmp602, i8* %tmp600622, i32 %max587, i32 1 )
+ %tmp606 = load i32* %tmp71, align 4 ; <i32> [#uses=1]
+ %tmp613 = add i32 %tmp606, %max154 ; <i32> [#uses=1]
+ store i32 %tmp613, i32* %tmp71, align 4
+ ret void
+}
More information about the llvm-commits
mailing list