1; RUN: llc < %s -mtriple=x86_64-linux -join-physregs | FileCheck %s
2; CHECK-NOT:     mov
3; CHECK:     paddw
4; CHECK-NOT:     mov
5; CHECK:     paddw
6; CHECK-NOT:     paddw
7; CHECK-NOT:     mov
8
9; The 2-addr pass should ensure that identical code is produced for these functions
10; no extra copy should be generated.
11
12define <2 x i64> @test1(<2 x i64> %x, <2 x i64> %y) nounwind  {
13entry:
14	%tmp6 = bitcast <2 x i64> %y to <8 x i16>		; <<8 x i16>> [#uses=1]
15	%tmp8 = bitcast <2 x i64> %x to <8 x i16>		; <<8 x i16>> [#uses=1]
16	%tmp9 = add <8 x i16> %tmp8, %tmp6		; <<8 x i16>> [#uses=1]
17	%tmp10 = bitcast <8 x i16> %tmp9 to <2 x i64>		; <<2 x i64>> [#uses=1]
18	ret <2 x i64> %tmp10
19}
20
21define <2 x i64> @test2(<2 x i64> %x, <2 x i64> %y) nounwind  {
22entry:
23	%tmp6 = bitcast <2 x i64> %x to <8 x i16>		; <<8 x i16>> [#uses=1]
24	%tmp8 = bitcast <2 x i64> %y to <8 x i16>		; <<8 x i16>> [#uses=1]
25	%tmp9 = add <8 x i16> %tmp8, %tmp6		; <<8 x i16>> [#uses=1]
26	%tmp10 = bitcast <8 x i16> %tmp9 to <2 x i64>		; <<2 x i64>> [#uses=1]
27	ret <2 x i64> %tmp10
28}
29
30
31; The coalescer should commute the add to avoid a copy.
32define <4 x float> @test3(<4 x float> %V) {
33entry:
34        %tmp8 = shufflevector <4 x float> %V, <4 x float> undef,
35                                        <4 x i32> < i32 3, i32 2, i32 1, i32 0 >
36        %add = fadd <4 x float> %tmp8, %V
37        ret <4 x float> %add
38}
39
40