1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
4 define <3 x i32> @f_29(<12 x i16> %a, <12 x i16> %b) {
6 ; CHECK: # %bb.0: # %entry
7 ; CHECK-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
8 ; CHECK-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
9 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
10 ; CHECK-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
11 ; CHECK-NEXT: movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
12 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
13 ; CHECK-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
14 ; CHECK-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
15 ; CHECK-NEXT: movd {{.*#+}} xmm3 = mem[0],zero,zero,zero
16 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
17 ; CHECK-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
18 ; CHECK-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
19 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
20 ; CHECK-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
21 ; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
22 ; CHECK-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
23 ; CHECK-NEXT: movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
24 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
25 ; CHECK-NEXT: movd %r9d, %xmm0
26 ; CHECK-NEXT: movd %r8d, %xmm3
27 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
28 ; CHECK-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
29 ; CHECK-NEXT: movd %ecx, %xmm0
30 ; CHECK-NEXT: movd %edx, %xmm2
31 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
32 ; CHECK-NEXT: movd %esi, %xmm4
33 ; CHECK-NEXT: movd %edi, %xmm0
34 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
35 ; CHECK-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
36 ; CHECK-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
37 ; CHECK-NEXT: movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
38 ; CHECK-NEXT: pinsrw $1, {{[0-9]+}}(%rsp), %xmm2
39 ; CHECK-NEXT: pinsrw $2, {{[0-9]+}}(%rsp), %xmm2
40 ; CHECK-NEXT: pinsrw $3, {{[0-9]+}}(%rsp), %xmm2
41 ; CHECK-NEXT: movd {{.*#+}} xmm3 = mem[0],zero,zero,zero
42 ; CHECK-NEXT: pinsrw $1, {{[0-9]+}}(%rsp), %xmm3
43 ; CHECK-NEXT: pinsrw $2, {{[0-9]+}}(%rsp), %xmm3
44 ; CHECK-NEXT: pinsrw $3, {{[0-9]+}}(%rsp), %xmm3
45 ; CHECK-NEXT: movdqa %xmm0, %xmm4
46 ; CHECK-NEXT: pmulhuw %xmm1, %xmm4
47 ; CHECK-NEXT: pmullw %xmm1, %xmm0
48 ; CHECK-NEXT: movdqa %xmm0, %xmm1
49 ; CHECK-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
50 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
51 ; CHECK-NEXT: movdqa %xmm3, %xmm4
52 ; CHECK-NEXT: pmulhuw %xmm2, %xmm4
53 ; CHECK-NEXT: pmullw %xmm2, %xmm3
54 ; CHECK-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
55 ; CHECK-NEXT: movdqa %xmm0, %xmm2
56 ; CHECK-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,3],xmm1[1,3]
57 ; CHECK-NEXT: pshufd {{.*#+}} xmm4 = xmm3[1,3,1,3]
58 ; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
59 ; CHECK-NEXT: paddd %xmm2, %xmm0
60 ; CHECK-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,2,1,3]
61 ; CHECK-NEXT: paddd %xmm4, %xmm1
62 ; CHECK-NEXT: movdqa %xmm0, %xmm2
63 ; CHECK-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,3],xmm1[1,3]
64 ; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,3]
65 ; CHECK-NEXT: paddd %xmm2, %xmm0
68 %a32 = zext <12 x i16> %a to <12 x i32>
69 %b32 = zext <12 x i16> %b to <12 x i32>
70 %prod12 = mul nuw <12 x i32> %a32, %b32
71 %odd12 = shufflevector <12 x i32> %prod12, <12 x i32> undef, <6 x i32> <i32
72 1, i32 3, i32 5, i32 7, i32 9, i32 11>
73 %even12 = shufflevector <12 x i32> %prod12, <12 x i32> undef, <6 x i32> <i32
74 0, i32 2, i32 4, i32 6, i32 8, i32 10>
75 %prod6 = add <6 x i32> %odd12, %even12
76 %odd6 = shufflevector <6 x i32> %prod6, <6 x i32> undef, <3 x i32> <i32 1,
78 %even6 = shufflevector <6 x i32> %prod6, <6 x i32> undef, <3 x i32> <i32 0,
80 %result = add <3 x i32> %odd6, %even6