1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
declare <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8>, <16 x i8>) nounwind readnone
declare <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16>, <8 x i16>) nounwind readnone
define <16 x i8> @combine_pavgb_self(<16 x i8> %a0) {
; SSE-LABEL: combine_pavgb_self:
; SSE: # %bb.0:
; SSE-NEXT: retq
;
; AVX-LABEL: combine_pavgb_self:
; AVX: # %bb.0:
; AVX-NEXT: retq
%1 = call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> %a0, <16 x i8> %a0)
ret <16 x i8> %1
}
define <16 x i8> @combine_pavgb_zero(<16 x i8> %a0) {
; SSE-LABEL: combine_pavgb_zero:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: pavgb %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: combine_pavgb_zero:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpavgb %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
%1 = call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> zeroinitializer, <16 x i8> %a0)
ret <16 x i8> %1
}
define <16 x i8> @combine_pavgw_knownbits(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> %a2, <8 x i16> %a3) {
; SSE-LABEL: combine_pavgw_knownbits:
; SSE: # %bb.0:
; SSE-NEXT: pmovsxbw {{.*#+}} xmm4 = [31,31,31,31,31,31,31,31]
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: pand %xmm4, %xmm1
; SSE-NEXT: pavgw %xmm1, %xmm0
; SSE-NEXT: pand %xmm4, %xmm2
; SSE-NEXT: pand %xmm4, %xmm3
; SSE-NEXT: pavgw %xmm2, %xmm3
; SSE-NEXT: packuswb %xmm3, %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_pavgw_knownbits:
; AVX1: # %bb.0:
; AVX1-NEXT: vbroadcastss {{.*#+}} xmm4 = [31,31,31,31,31,31,31,31]
; AVX1-NEXT: vpand %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vpavgw %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm1
; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm2
; AVX1-NEXT: vpavgw %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpackuswb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_pavgw_knownbits:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastw {{.*#+}} xmm4 = [31,31,31,31,31,31,31,31]
; AVX2-NEXT: vpand %xmm4, %xmm0, %xmm0
; AVX2-NEXT: vpand %xmm4, %xmm1, %xmm1
; AVX2-NEXT: vpavgw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpand %xmm4, %xmm2, %xmm1
; AVX2-NEXT: vpand %xmm4, %xmm3, %xmm2
; AVX2-NEXT: vpavgw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpackuswb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%m0 = and <8 x i16> %a0, <i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31>
%m1 = and <8 x i16> %a1, <i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31>
%m2 = and <8 x i16> %a2, <i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31>
%m3 = and <8 x i16> %a3, <i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31, i16 31>
%avg01 = tail call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %m0, <8 x i16> %m1)
%avg23 = tail call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %m2, <8 x i16> %m3)
%shuffle = shufflevector <8 x i16> %avg01, <8 x i16> %avg23, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
%trunc = trunc <16 x i16> %shuffle to <16 x i8>
ret <16 x i8> %trunc
}
define <8 x i16> @combine_pavgw_demandedelts(<8 x i16> %a0, <8 x i16> %a1) {
; SSE-LABEL: combine_pavgw_demandedelts:
; SSE: # %bb.0:
; SSE-NEXT: pavgw %xmm1, %xmm0
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_pavgw_demandedelts:
; AVX1: # %bb.0:
; AVX1-NEXT: vpavgw %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_pavgw_demandedelts:
; AVX2: # %bb.0:
; AVX2-NEXT: vpavgw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %xmm0
; AVX2-NEXT: retq
%s0 = shufflevector <8 x i16> %a0, <8 x i16> poison, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
%avg = tail call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %s0, <8 x i16> %a1)
%shuffle = shufflevector <8 x i16> %avg, <8 x i16> poison, <8 x i32> zeroinitializer
ret <8 x i16> %shuffle
}
|
