[AArch64]Fix the problem that can't select mul of v1i64/v2i64 types.

E.g. Can't select such IR:
     %tmp = mul <2 x i64> %a, %b

llvm-svn: 198188

lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -385,6 +385,18 @@ AArch64TargetLowering::AArch64TargetLowering(AArch64TargetMachine &TM)
           setTruncStoreAction(VT, VT1, Expand);
       }
     }
+
+    // There is no v1i64/v2i64 multiply, expand v1i64/v2i64 to GPR i64 multiply.
+    // FIXME: For a v2i64 multiply, we copy VPR to GPR and do 2 i64 multiplies,
+    // and then copy back to VPR. This solution may be optimized by Following 3
+    // NEON instructions:
+    //        pmull  v2.1q, v0.1d, v1.1d
+    //        pmull2 v3.1q, v0.2d, v1.2d
+    //        ins    v2.d[1], v3.d[0]
+    // As currently we can't verify the correctness of such assumption, we can
+    // do such optimization in the future.
+    setOperationAction(ISD::MUL, MVT::v1i64, Expand);
+    setOperationAction(ISD::MUL, MVT::v2i64, Expand);
   }
 }
 

test/CodeGen/AArch64/neon-mul-div.ll

@@ -37,6 +37,21 @@ define <4 x i32> @mul4x32(<4 x i32> %A, <4 x i32> %B) {
 	ret <4 x i32> %tmp3
 }
 
+define <1 x i64> @mul1xi64(<1 x i64> %A, <1 x i64> %B) {
+;CHECK-LABEL: mul1xi64:
+;CHECK: mul x{{[0-9]+}}, x{{[0-9]+}}, x{{[0-9]+}}
+  %tmp3 = mul <1 x i64> %A, %B;
+  ret <1 x i64> %tmp3
+}
+
+define <2 x i64> @mul2xi64(<2 x i64> %A, <2 x i64> %B) {
+;CHECK-LABEL: mul2xi64:
+;CHECK: mul x{{[0-9]+}}, x{{[0-9]+}}, x{{[0-9]+}}
+;CHECK: mul x{{[0-9]+}}, x{{[0-9]+}}, x{{[0-9]+}}
+  %tmp3 = mul <2 x i64> %A, %B;
+  ret <2 x i64> %tmp3
+}
+
  define <2 x float> @mul2xfloat(<2 x float> %A, <2 x float> %B) {
 ;CHECK: fmul {{v[0-9]+}}.2s, {{v[0-9]+}}.2s, {{v[0-9]+}}.2s
 	%tmp3 = fmul <2 x float> %A, %B;