More replacing of target-dependent intrinsics with target-indepdent

intrinsics.  The second instruction(s) to be handled are the vector versions 
of count set bits (ctpop).

The changes here are to clang so that it generates a target independent 
vector ctpop when it sees an ARM dependent vector bits set count.  The changes 
in llvm are to match the target independent vector ctpop and in 
VMCore/AutoUpgrade.cpp to update any existing bc files containing ARM 
dependent vector pop counts with target-independent ctpops.  There are also 
changes to an existing test case in llvm for ARM vector count instructions and 
to a test for the bitcode upgrade.

<rdar://problem/11892519>

There is deliberately no test for the change to clang, as so far as I know, no
consensus has been reached regarding how to test neon instructions in clang;
q.v. <rdar://problem/8762292>

llvm-svn: 160410

llvm/lib/Target/ARM/ARMInstrNEON.td

@@ -4829,10 +4829,10 @@ defm VCLZ     : N2VInt_QHS<0b11, 0b11, 0b00, 0b01001, 0,
 //   VCNT     : Vector Count One Bits
 def  VCNTd    : N2VDInt<0b11, 0b11, 0b00, 0b00, 0b01010, 0,
                         IIC_VCNTiD, "vcnt", "8",
-                        v8i8, v8i8, int_arm_neon_vcnt>;
+                        v8i8, v8i8, ctpop>;
 def  VCNTq    : N2VQInt<0b11, 0b11, 0b00, 0b00, 0b01010, 0,
                         IIC_VCNTiQ, "vcnt", "8",
-                        v16i8, v16i8, int_arm_neon_vcnt>;
+                        v16i8, v16i8, ctpop>;
 
 // Vector Swap
 def  VSWPd    : N2VX<0b11, 0b11, 0b00, 0b10, 0b00000, 0, 0,

llvm/lib/VMCore/AutoUpgrade.cpp

@@ -66,6 +66,11 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
                                "llvm.ctlz." + Name.substr(14), F->getParent());
       return true;
     }
+    if (Name.startswith("arm.neon.vcnt")) {
+      NewFn = Intrinsic::getDeclaration(F->getParent(), Intrinsic::ctpop,
+                                        F->arg_begin()->getType());
+      return true;
+    }
     break;
   }
   case 'c': {
@@ -314,11 +319,16 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
   case Intrinsic::arm_neon_vclz: {
     // Change name from llvm.arm.neon.vclz.* to llvm.ctlz.*
     CI->replaceAllUsesWith(Builder.CreateCall2(NewFn, CI->getArgOperand(0),
-                                               Builder.getFalse(), 
+                                               Builder.getFalse(),
                                                "llvm.ctlz." + Name.substr(14)));
     CI->eraseFromParent();
     return;
   }
+  case Intrinsic::ctpop: {
+    CI->replaceAllUsesWith(Builder.CreateCall(NewFn, CI->getArgOperand(0)));
+    CI->eraseFromParent();
+    return;
+  }
 
   case Intrinsic::x86_xop_vfrcz_ss:
   case Intrinsic::x86_xop_vfrcz_sd:

llvm/test/Bitcode/arm32_neon_vcnt_upgrade.ll

@@ -1,5 +1,5 @@
 ; RUN: llvm-as < %s | llvm-dis | FileCheck %s
-; NB: currently tests only vclz, should also test vcnt and vcls
+; Tests vclz and vcnt
 
 define <4 x i16> @vclz16(<4 x i16>* %A) nounwind {
 ;CHECK: @vclz16
@@ -9,4 +9,13 @@ define <4 x i16> @vclz16(<4 x i16>* %A) nounwind {
         ret <4 x i16> %tmp2
 }
 
+define <8 x i8> @vcnt8(<8 x i8>* %A) nounwind {
+;CHECK: @vcnt8
+        %tmp1 = load <8 x i8>* %A
+        %tmp2 = call <8 x i8> @llvm.arm.neon.vcnt.v8i8(<8 x i8> %tmp1)
+;CHECK: call <8 x i8> @llvm.ctpop.v8i8(<8 x i8>
+        ret <8 x i8> %tmp2
+}
+
 declare <4 x i16>  @llvm.arm.neon.vclz.v4i16(<4 x i16>) nounwind readnone
+declare <8 x i8>  @llvm.arm.neon.vcnt.v8i8(<8 x i8>) nounwind readnone

llvm/test/CodeGen/ARM/vcnt.ll

@@ -3,22 +3,22 @@
 
 define <8 x i8> @vcnt8(<8 x i8>* %A) nounwind {
 ;CHECK: vcnt8:
-;CHECK: vcnt.8
+;CHECK: vcnt.8 {{d[0-9]+}}, {{d[0-9]+}}
 	%tmp1 = load <8 x i8>* %A
-	%tmp2 = call <8 x i8> @llvm.arm.neon.vcnt.v8i8(<8 x i8> %tmp1)
+	%tmp2 = call <8 x i8> @llvm.ctpop.v8i8(<8 x i8> %tmp1)
 	ret <8 x i8> %tmp2
 }
 
 define <16 x i8> @vcntQ8(<16 x i8>* %A) nounwind {
 ;CHECK: vcntQ8:
-;CHECK: vcnt.8
+;CHECK: vcnt.8 {{q[0-9]+}}, {{q[0-9]+}}
 	%tmp1 = load <16 x i8>* %A
-	%tmp2 = call <16 x i8> @llvm.arm.neon.vcnt.v16i8(<16 x i8> %tmp1)
+	%tmp2 = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %tmp1)
 	ret <16 x i8> %tmp2
 }
 
-declare <8 x i8>  @llvm.arm.neon.vcnt.v8i8(<8 x i8>) nounwind readnone
+declare <8 x i8>  @llvm.ctpop.v8i8(<8 x i8>) nounwind readnone
-declare <16 x i8> @llvm.arm.neon.vcnt.v16i8(<16 x i8>) nounwind readnone
+declare <16 x i8> @llvm.ctpop.v16i8(<16 x i8>) nounwind readnone
 
 define <8 x i8> @vclz8(<8 x i8>* %A) nounwind {
 ;CHECK: vclz8: