From: Chris Lattner <sabre@nondot.org>
Date: Wed, 8 Jul 2009 00:31:33 +0000 (+0000)
Subject: dag combine sext(setcc) -> vsetcc before legalize.  To make this safe,
X-Git-Url: http://demsky.eecs.uci.edu/git/?a=commitdiff_plain;h=2b7a271c713ff1db83990f691126bc33d6c59b52;p=oota-llvm.git

dag combine sext(setcc) -> vsetcc before legalize.  To make this safe,
VSETCC must define all bits, which is different than it was documented
to before.  Since all targets that implement VSETCC already have this
behavior, and we don't optimize based on this, just change the
documentation.  We now get nice code for vec_compare.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@74978 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/include/llvm/CodeGen/SelectionDAGNodes.h b/include/llvm/CodeGen/SelectionDAGNodes.h
index 975253751c8..ac345af2b23 100644
--- a/include/llvm/CodeGen/SelectionDAGNodes.h
+++ b/include/llvm/CodeGen/SelectionDAGNodes.h
@@ -363,12 +363,11 @@ namespace ISD {
     // them with (op #2) as a CondCodeSDNode.
     SETCC,
 
-    // Vector SetCC operator - This evaluates to a vector of integer elements
-    // with the high bit in each element set to true if the comparison is true
-    // and false if the comparison is false.  All other bits in each element
-    // are undefined.  The operands to this are the left and right operands
-    // to compare (ops #0, and #1) and the condition code to compare them with
-    // (op #2) as a CondCodeSDNode.
+    // RESULT = VSETCC(LHS, RHS, COND) operator - This evaluates to a vector of
+    // integer elements with all bits of the result elements set to true if the
+    // comparison is true or all cleared if the comparison is false.  The
+    // operands to this are the left and right operands to compare (LHS/RHS) and
+    // the condition code to compare them with (COND) as a CondCodeSDNode.
     VSETCC,
 
     // SHL_PARTS/SRA_PARTS/SRL_PARTS - These operators are used for expanded
diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 609ec82c5ad..b10751b0478 100644
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -3071,14 +3071,32 @@ SDValue DAGCombiner::visitSIGN_EXTEND(SDNode *N) {
     }
   }
 
-  // sext(setcc x, y, cc) -> (select_cc x, y, -1, 0, cc)
   if (N0.getOpcode() == ISD::SETCC) {
+    // sext(setcc) -> sext_in_reg(vsetcc) for vectors.
+    if (VT.isVector() &&
+        // We know that the # elements of the results is the same as the
+        // # elements of the compare (and the # elements of the compare result
+        // for that matter).  Check to see that they are the same size.  If so,
+        // we know that the element size of the sext'd result matches the
+        // element size of the compare operands.
+        VT.getSizeInBits() == N0.getOperand(0).getValueType().getSizeInBits() &&
+      
+        // Only do this before legalize for now.
+        !LegalOperations) {
+      return DAG.getVSetCC(N->getDebugLoc(), VT, N0.getOperand(0),
+                           N0.getOperand(1),
+                           cast<CondCodeSDNode>(N0.getOperand(2))->get());
+    }
+    
+    // sext(setcc x, y, cc) -> (select_cc x, y, -1, 0, cc)
     SDValue SCC =
       SimplifySelectCC(N->getDebugLoc(), N0.getOperand(0), N0.getOperand(1),
                        DAG.getConstant(~0ULL, VT), DAG.getConstant(0, VT),
                        cast<CondCodeSDNode>(N0.getOperand(2))->get(), true);
     if (SCC.getNode()) return SCC;
   }
+  
+  
 
   // fold (sext x) -> (zext x) if the sign bit is known zero.
   if ((!LegalOperations || TLI.isOperationLegal(ISD::ZERO_EXTEND, VT)) &&
diff --git a/test/CodeGen/X86/vec_compare.ll b/test/CodeGen/X86/vec_compare.ll
index 7a8dfd68cfd..74e34d0202d 100644
--- a/test/CodeGen/X86/vec_compare.ll
+++ b/test/CodeGen/X86/vec_compare.ll
@@ -1,4 +1,4 @@
-; RUN: llvm-as < %s | llc -march=x86 -mcpu=yonah | grep pcmpgtd
+; RUN: llvm-as < %s | llc -march=x86 -mcpu=yonah | grep pcmpgtd | count 2
 
 define <4 x i32> @test(<4 x i32> %A, <4 x i32> %B) nounwind {
 	%C = vicmp sgt <4 x i32> %A, %B