Merging r195476:

------------------------------------------------------------------------
r195476 | hliao | 2013-11-22 09:56:57 -0800 (Fri, 22 Nov 2013) | 6 lines

Fix PR18014

- When simplifying the mask generation for BLEND, check whether that mask is
  also consumed by other non-BLEND insns. If true, skip that simplification.

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_34@195602 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 9df0232..627aa86 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -17007,6 +17007,15 @@ static SDValue PerformSELECTCombine(SDNode *N, SelectionDAG &DAG,
     if (BitWidth == 1)
       return SDValue();
 
+    // Check all uses of that condition operand to check whether it will be
+    // consumed by non-BLEND instructions, which may depend on all bits are set
+    // properly.
+    for (SDNode::use_iterator I = Cond->use_begin(),
+                              E = Cond->use_end(); I != E; ++I)
+      if (I->getOpcode() != ISD::VSELECT)
+        // TODO: Add other opcodes eventually lowered into BLEND.
+        return SDValue();
+
     assert(BitWidth >= 8 && BitWidth <= 64 && "Invalid mask size");
     APInt DemandedMask = APInt::getHighBitsSet(BitWidth, 1);
 

diff --git a/test/CodeGen/X86/pr18014.ll b/test/CodeGen/X86/pr18014.ll
new file mode 100644 (file)
index 0000000..e3860b8
--- /dev/null
+++ b/test/CodeGen/X86/pr18014.ll
@@ -0,0 +1,16 @@
+; RUN: llc < %s -mtriple=x86_64-linux-pc -mcpu=penryn | FileCheck %s
+
+; Ensure PSRAD is generated as the condition is consumed by both PADD and
+; BLENDVPS. PAND requires all bits setting properly.
+
+define <4 x i32> @foo(<4 x i32>* %p, <4 x i1> %cond, <4 x i32> %v1, <4 x i32> %v2, <4 x i32> %v3) {
+  %sext_cond = sext <4 x i1> %cond to <4 x i32>
+  %t1 = add <4 x i32> %v1, %sext_cond
+  %t2 = select <4 x i1> %cond, <4 x i32> %v1, <4 x i32> %v2
+  store <4 x i32> %t2, <4 x i32>* %p
+  ret <4 x i32> %t1
+; CHECK: foo
+; CHECK: pslld
+; CHECK: psrad
+; CHECK: ret
+}