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
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
|
; RUN: llc -mcpu=pwr6 -mattr=+altivec < %s | FileCheck %s
; Check vector extend load expansion with altivec enabled.
target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v128:128:128-n32:64"
target triple = "powerpc64-unknown-linux-gnu"
; Altivec does not provides an sext intruction, so it expands
; a set of vector stores (stvx), bytes load/sign expand/store
; (lbz/stb), and a final vector load (lvx) to load the result
; extended vector.
define <16 x i8> @v16si8_sext_in_reg(<16 x i8> %a) {
%b = trunc <16 x i8> %a to <16 x i4>
%c = sext <16 x i4> %b to <16 x i8>
ret <16 x i8> %c
}
; CHECK: v16si8_sext_in_reg:
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lbz
; CHECK: stb
; CHECK: lvx 2, {{[0-9]+}}, {{[0-9]+}}
; The zero extend uses a more clever logic: a vector splat
; and a logic and to set higher bits to 0.
define <16 x i8> @v16si8_zext_in_reg(<16 x i8> %a) {
%b = trunc <16 x i8> %a to <16 x i4>
%c = zext <16 x i4> %b to <16 x i8>
ret <16 x i8> %c
}
; CHECK: v16si8_zext_in_reg:
; CHECK: vspltisb [[VMASK:[0-9]+]], 15
; CHECK-NEXT: vand 2, 2, [[VMASK]]
; Same as v16si8_sext_in_reg, expands to load/store halfwords (lhz/sth).
define <8 x i16> @v8si16_sext_in_reg(<8 x i16> %a) {
%b = trunc <8 x i16> %a to <8 x i8>
%c = sext <8 x i8> %b to <8 x i16>
ret <8 x i16> %c
}
; CHECK: v8si16_sext_in_reg:
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lhz
; CHECK: sth
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lhz
; CHECK: sth
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lhz
; CHECK: sth
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lhz
; CHECK: sth
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lhz
; CHECK: sth
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lhz
; CHECK: sth
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lhz
; CHECK: sth
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lhz
; CHECK: sth
; CHECK: lvx 2, {{[0-9]+}}, {{[0-9]+}}
; Same as v8si16_sext_in_reg, but instead of creating the mask
; with a splat, loads it from memory.
define <8 x i16> @v8si16_zext_in_reg(<8 x i16> %a) {
%b = trunc <8 x i16> %a to <8 x i8>
%c = zext <8 x i8> %b to <8 x i16>
ret <8 x i16> %c
}
; CHECK: v8si16_zext_in_reg:
; CHECK: ld [[RMASKTOC:[0-9]+]], .LC{{[0-9]+}}@toc(2)
; CHECK-NEXT: lvx [[VMASK:[0-9]+]], {{[0-9]+}}, [[RMASKTOC]]
; CHECK-NEXT: vand 2, 2, [[VMASK]]
; Same as v16si8_sext_in_reg, expands to load halfword (lha) and
; store words (stw).
define <4 x i32> @v4si32_sext_in_reg(<4 x i32> %a) {
%b = trunc <4 x i32> %a to <4 x i16>
%c = sext <4 x i16> %b to <4 x i32>
ret <4 x i32> %c
}
; CHECK: v4si32_sext_in_reg:
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lha
; CHECK: stw
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lha
; CHECK: stw
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lha
; CHECK: stw
; CHECK: stvx 2, {{[0-9]+}}, {{[0-9]+}}
; CHECK: lha
; CHECK: stw
; CHECK: lvx 2, {{[0-9]+}}, {{[0-9]+}}
; Same as v8si16_sext_in_reg.
define <4 x i32> @v4si32_zext_in_reg(<4 x i32> %a) {
%b = trunc <4 x i32> %a to <4 x i16>
%c = zext <4 x i16> %b to <4 x i32>
ret <4 x i32> %c
}
; CHECK: v4si32_zext_in_reg:
; CHECK: vspltisw [[VMASK:[0-9]+]], -16
; CHECK-NEXT: vsrw [[VMASK]], [[VMASK]], [[VMASK]]
; CHECK-NEXT: vand 2, 2, [[VMASK]]
|
