This is one of the first steps at moving to replace target-dependent

intrinsics with target-indepdent intrinsics.  The first instruction(s) to be 
handled are the vector versions of count leading zeros (ctlz).

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

<rdar://problem/11831778>

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>


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@160200 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 57 insertions, 19 deletions

diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index 404b3ad176..c540d0ddd7 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -4825,7 +4825,7 @@ defm VCLS     : N2VInt_QHS<0b11, 0b11, 0b00, 0b01000, 0,
 //   VCLZ     : Vector Count Leading Zeros
 defm VCLZ     : N2VInt_QHS<0b11, 0b11, 0b00, 0b01001, 0,
                            IIC_VCNTiD, IIC_VCNTiQ, "vclz", "i",
-                           int_arm_neon_vclz>;
+                           ctlz>;
 //   VCNT     : Vector Count One Bits
 def  VCNTd    : N2VDInt<0b11, 0b11, 0b00, 0b00, 0b01010, 0,
                         IIC_VCNTiD, "vcnt", "8",
diff --git a/lib/VMCore/AutoUpgrade.cpp b/lib/VMCore/AutoUpgrade.cpp
index 1defd20728..8febce18c0 100644
--- a/lib/VMCore/AutoUpgrade.cpp
+++ b/lib/VMCore/AutoUpgrade.cpp
@@ -52,6 +52,22 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
 
   switch (Name[0]) {
   default: break;
+  case 'a': {
+    if (Name.startswith("arm.neon.vclz")) {
+      Type* args[2] = {
+        F->arg_begin()->getType(), 
+        Type::getInt1Ty(F->getContext())
+      };
+      // Can't use Intrinsic::getDeclaration here as it adds a ".i1" to
+      // the end of the name. Change name from llvm.arm.neon.vclz.* to
+      //  llvm.ctlz.*
+      FunctionType* fType = FunctionType::get(F->getReturnType(), args, false);
+      NewFn = Function::Create(fType, F->getLinkage(), 
+                               "llvm.ctlz." + Name.substr(14), F->getParent());
+      return true;
+    }
+    break;
+  }
   case 'c': {
     if (Name.startswith("ctlz.") && F->arg_size() == 1) {
       F->setName(Name + ".old");
@@ -295,6 +311,15 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
     CI->eraseFromParent();
     return;
 
+  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(), 
+                                               "llvm.ctlz." + Name.substr(14)));
+    CI->eraseFromParent();
+    return;
+  }
+
   case Intrinsic::x86_xop_vfrcz_ss:
   case Intrinsic::x86_xop_vfrcz_sd:
     CI->replaceAllUsesWith(Builder.CreateCall(NewFn, CI->getArgOperand(1),
diff --git a/test/Bitcode/arm32_neon_vcnt_upgrade.ll b/test/Bitcode/arm32_neon_vcnt_upgrade.ll
new file mode 100644
index 0000000000..b3f2f03d10
--- /dev/null
+++ b/test/Bitcode/arm32_neon_vcnt_upgrade.ll
@@ -0,0 +1,12 @@
+; RUN: llvm-as < %s | llvm-dis | FileCheck %s
+; NB: currently tests only vclz, should also test vcnt and vcls
+
+define <4 x i16> @vclz16(<4 x i16>* %A) nounwind {
+;CHECK: @vclz16
+        %tmp1 = load <4 x i16>* %A
+        %tmp2 = call <4 x i16> @llvm.arm.neon.vclz.v4i16(<4 x i16> %tmp1)
+;CHECK: {{call.*@llvm.ctlz.v4i16\(<4 x i16>.*, i1 false}}
+        ret <4 x i16> %tmp2
+}
+
+declare <4 x i16>  @llvm.arm.neon.vclz.v4i16(<4 x i16>) nounwind readnone
diff --git a/test/CodeGen/ARM/vcnt.ll b/test/CodeGen/ARM/vcnt.ll
index 450f90d03d..7e54524db7 100644
--- a/test/CodeGen/ARM/vcnt.ll
+++ b/test/CodeGen/ARM/vcnt.ll
@@ -1,4 +1,5 @@
 ; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
+; NB: this tests vcnt, vclz, and vcls
 
 define <8 x i8> @vcnt8(<8 x i8>* %A) nounwind {
 ;CHECK: vcnt8:
@@ -21,59 +22,59 @@ declare <16 x i8> @llvm.arm.neon.vcnt.v16i8(<16 x i8>) nounwind readnone
 
 define <8 x i8> @vclz8(<8 x i8>* %A) nounwind {
 ;CHECK: vclz8:
-;CHECK: vclz.i8
+;CHECK: vclz.i8 {{d[0-9]+}}, {{d[0-9]+}}
 	%tmp1 = load <8 x i8>* %A
-	%tmp2 = call <8 x i8> @llvm.arm.neon.vclz.v8i8(<8 x i8> %tmp1)
+	%tmp2 = call <8 x i8> @llvm.ctlz.v8i8(<8 x i8> %tmp1, i1 0)
 	ret <8 x i8> %tmp2
 }
 
 define <4 x i16> @vclz16(<4 x i16>* %A) nounwind {
 ;CHECK: vclz16:
-;CHECK: vclz.i16
+;CHECK: vclz.i16 {{d[0-9]+}}, {{d[0-9]+}}
 	%tmp1 = load <4 x i16>* %A
-	%tmp2 = call <4 x i16> @llvm.arm.neon.vclz.v4i16(<4 x i16> %tmp1)
+	%tmp2 = call <4 x i16> @llvm.ctlz.v4i16(<4 x i16> %tmp1, i1 0)
 	ret <4 x i16> %tmp2
 }
 
 define <2 x i32> @vclz32(<2 x i32>* %A) nounwind {
 ;CHECK: vclz32:
-;CHECK: vclz.i32
+;CHECK: vclz.i32 {{d[0-9]+}}, {{d[0-9]+}}
 	%tmp1 = load <2 x i32>* %A
-	%tmp2 = call <2 x i32> @llvm.arm.neon.vclz.v2i32(<2 x i32> %tmp1)
+	%tmp2 = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %tmp1, i1 0)
 	ret <2 x i32> %tmp2
 }
 
 define <16 x i8> @vclzQ8(<16 x i8>* %A) nounwind {
 ;CHECK: vclzQ8:
-;CHECK: vclz.i8
+;CHECK: vclz.i8 {{q[0-9]+}}, {{q[0-9]+}}
 	%tmp1 = load <16 x i8>* %A
-	%tmp2 = call <16 x i8> @llvm.arm.neon.vclz.v16i8(<16 x i8> %tmp1)
+	%tmp2 = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %tmp1, i1 0)
 	ret <16 x i8> %tmp2
 }
 
 define <8 x i16> @vclzQ16(<8 x i16>* %A) nounwind {
 ;CHECK: vclzQ16:
-;CHECK: vclz.i16
+;CHECK: vclz.i16 {{q[0-9]+}}, {{q[0-9]+}}
 	%tmp1 = load <8 x i16>* %A
-	%tmp2 = call <8 x i16> @llvm.arm.neon.vclz.v8i16(<8 x i16> %tmp1)
+	%tmp2 = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %tmp1, i1 0)
 	ret <8 x i16> %tmp2
 }
 
 define <4 x i32> @vclzQ32(<4 x i32>* %A) nounwind {
 ;CHECK: vclzQ32:
-;CHECK: vclz.i32
+;CHECK: vclz.i32 {{q[0-9]+}}, {{q[0-9]+}}
 	%tmp1 = load <4 x i32>* %A
-	%tmp2 = call <4 x i32> @llvm.arm.neon.vclz.v4i32(<4 x i32> %tmp1)
+	%tmp2 = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %tmp1, i1 0)
 	ret <4 x i32> %tmp2
 }
 
-declare <8 x i8>  @llvm.arm.neon.vclz.v8i8(<8 x i8>) nounwind readnone
-declare <4 x i16> @llvm.arm.neon.vclz.v4i16(<4 x i16>) nounwind readnone
-declare <2 x i32> @llvm.arm.neon.vclz.v2i32(<2 x i32>) nounwind readnone
+declare <8 x i8>  @llvm.ctlz.v8i8(<8 x i8>, i1) nounwind readnone
+declare <4 x i16> @llvm.ctlz.v4i16(<4 x i16>, i1) nounwind readnone
+declare <2 x i32> @llvm.ctlz.v2i32(<2 x i32>, i1) nounwind readnone
 
-declare <16 x i8> @llvm.arm.neon.vclz.v16i8(<16 x i8>) nounwind readnone
-declare <8 x i16> @llvm.arm.neon.vclz.v8i16(<8 x i16>) nounwind readnone
-declare <4 x i32> @llvm.arm.neon.vclz.v4i32(<4 x i32>) nounwind readnone
+declare <16 x i8> @llvm.ctlz.v16i8(<16 x i8>, i1) nounwind readnone
+declare <8 x i16> @llvm.ctlz.v8i16(<8 x i16>, i1) nounwind readnone
+declare <4 x i32> @llvm.ctlz.v4i32(<4 x i32>, i1) nounwind readnone
 
 define <8 x i8> @vclss8(<8 x i8>* %A) nounwind {
 ;CHECK: vclss8: