[llvm-commits] [llvm] r110268 - in /llvm/trunk: lib/Target/X86/X86ISelDAGToDAG.cpp test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll

[Eli Friedman]

Author: efriedma
Date: Wed Aug  4 17:40:58 2010
New Revision: 110268

URL: http://llvm.org/viewvc/llvm-project?rev=110268&view=rev
Log:
PR7814: Truncates cannot be ignored for signed comparisons.


Added:
    llvm/trunk/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll
Modified:
    llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp

Modified: llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp?rev=110268&r1=110267&r2=110268&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp Wed Aug  4 17:40:58 2010
@@ -1841,7 +1841,8 @@
 
     // Look for (X86cmp (and $op, $imm), 0) and see if we can convert it to
     // use a smaller encoding.
-    if (N0.getOpcode() == ISD::TRUNCATE && N0.hasOneUse())
+    if (N0.getOpcode() == ISD::TRUNCATE && N0.hasOneUse() &&
+        HasNoSignedComparisonUses(Node))
       // Look past the truncate if CMP is the only use of it.
       N0 = N0.getOperand(0);
     if (N0.getNode()->getOpcode() == ISD::AND && N0.getNode()->hasOneUse() &&

Added: llvm/trunk/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll?rev=110268&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll (added)
+++ llvm/trunk/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll Wed Aug  4 17:40:58 2010
@@ -0,0 +1,36 @@
+; RUN: llc < %s -march=x86-64 | FileCheck %s
+; PR7814
+
+ at g_16 = global i64 -3738643449681751625, align 8  ; <i64*> [#uses=1]
+ at g_38 = global i32 0, align 4                     ; <i32*> [#uses=2]
+ at .str = private constant [4 x i8] c"%d\0A\00"     ; <[4 x i8]*> [#uses=1]
+
+define i32 @main() nounwind {
+entry:
+  %tmp = load i64* @g_16                          ; <i64> [#uses=1]
+  %not.lnot = icmp ne i64 %tmp, 0                 ; <i1> [#uses=1]
+  %conv = sext i1 %not.lnot to i64                ; <i64> [#uses=1]
+  %and = and i64 %conv, 150                       ; <i64> [#uses=1]
+  %conv.i = trunc i64 %and to i8                  ; <i8> [#uses=1]
+  %cmp = icmp sgt i8 %conv.i, 0                   ; <i1> [#uses=1]
+  br i1 %cmp, label %if.then, label %entry.if.end_crit_edge
+
+; CHECK: andl	$150
+; CHECK-NEXT: testb
+; CHECK-NEXT: jg
+
+entry.if.end_crit_edge:                           ; preds = %entry
+  %tmp4.pre = load i32* @g_38                     ; <i32> [#uses=1]
+  br label %if.end
+
+if.then:                                          ; preds = %entry
+  store i32 1, i32* @g_38
+  br label %if.end
+
+if.end:                                           ; preds = %entry.if.end_crit_edge, %if.then
+  %tmp4 = phi i32 [ %tmp4.pre, %entry.if.end_crit_edge ], [ 1, %if.then ] ; <i32> [#uses=1]
+  %call5 = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([4 x i8]* @.str, i64 0, i64 0), i32 %tmp4) nounwind ; <i32> [#uses=0]
+  ret i32 0
+}
+
+declare i32 @printf(i8* nocapture, ...) nounwind