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
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
|
// +build ignore
package main
import (
. "github.com/mmcloughlin/avo/build"
. "github.com/mmcloughlin/avo/operand"
. "github.com/mmcloughlin/avo/reg"
. "github.com/segmentio/asm/build/internal/x86"
"github.com/segmentio/asm/cpu"
)
func init() {
ConstraintExpr("!purego")
}
func main() {
TEXT("ValidPrintString", NOSPLIT, "func(s string) bool")
Doc("ValidPrintString returns true if s contains only printable ASCII characters.")
p := Mem{Base: Load(Param("s").Base(), GP64())}
n := Load(Param("s").Len(), GP64())
ret, _ := ReturnIndex(0).Resolve()
m1 := GP64()
m2 := GP64()
m3 := GP64()
val := GP32()
tmp := GP32()
CMPQ(n, U8(16)) // if n < 16:
JB(LabelRef("init_x86")) // goto init_x86
JumpIfFeature("init_avx", cpu.AVX2) // goto init_avx if supported
Label("init_x86")
CMPQ(n, U8(8)) // if n < 8:
JB(LabelRef("cmp4")) // goto cmp4
MOVQ(U64(0xDFDFDFDFDFDFDFE0), m1)
MOVQ(U64(0x0101010101010101), m2)
MOVQ(U64(0x8080808080808080), m3)
Label("cmp8")
valid8(p, n, m1, m2, m3) // ZF = [compare 8 bytes]
JNE(LabelRef("done")) // return ZF if ZF == 0
CMPQ(n, U8(8)) // if n < 8:
JB(LabelRef("cmp4")) // goto cmp4
JMP(LabelRef("cmp8")) // loop cmp8
Label("cmp4")
CMPQ(n, U8(4)) // if n < 4:
JB(LabelRef("cmp3")) // goto cmp3
valid4(p, n) // ZF = [compare 4 bytes]
JNE(LabelRef("done")) // return ZF if ZF == 0
Label("cmp3")
CMPQ(n, U8(3)) // if n < 3:
JB(LabelRef("cmp2")) // goto cmp2
MOVWLZX(p, tmp) // tmp = p[0:2]
MOVBLZX(p.Offset(2), val) // val = p[2:3]
SHLL(U8(16), val) // val <<= 16
ORL(tmp, val) // val = tmp | val
ORL(U32(0x20000000), val) // val = 0x20000000 | val
JMP(LabelRef("final"))
Label("cmp2")
CMPQ(n, U8(2)) // if n < 2:
JB(LabelRef("cmp1")) // goto cmp1
MOVWLZX(p, val) // val = p[0:2]
ORL(U32(0x20200000), val) // val = 0x20200000 | val
JMP(LabelRef("final"))
Label("cmp1")
CMPQ(n, U8(0)) // if n == 0:
JE(LabelRef("done")) // return true
MOVBLZX(p, val) // val = p[0:1]
ORL(U32(0x20202000), val) // val = 0x20202000 | val
Label("final")
setup4(val) // [update val register]
TESTL(U32(0x80808080), val) // ZF = (0x80808080 & val) == 0
Label("done")
SETEQ(ret.Addr) // return ZF
RET() // ...
Label("init_avx")
min := VecBroadcast(U8(0x1F), YMM())
max := VecBroadcast(U8(0x7E), YMM())
vec := NewVectorizer(14, func(l VectorLane) Register {
v0 := l.Read(p)
v1 := l.Alloc()
VPCMPGTB(min, v0, v1) // v1 = bytes that are greater than the min-1 (i.e. valid at lower end)
VPCMPGTB(max, v0, v0) // v0 = bytes that are greater than the max (i.e. invalid at upper end)
VPANDN(v1, v0, v0) // y2 & ~y3 mask should be full unless there's an invalid byte
return v0
}).Reduce(ReduceAnd) // merge all comparisons together
cmpAVX := func(spec Spec, lanes int, incr bool) {
sz := int(spec.Size())
out := vec.Compile(spec, lanes)[0] // [compare sz*lanes bytes]
if incr {
ADDQ(U8(sz*lanes), p.Base) // p += sz*lanes
SUBQ(U8(sz*lanes), n) // n -= sz*lanes
}
VPMOVMSKB(out, tmp) // tmp[0,1,2,...] = y0[0,8,16,...]
XORL(U32(^uint32(0)>>(32-sz)), tmp) // ZF = (tmp == 0xFFFFFFFF)
}
Label("cmp128")
CMPQ(n, U8(128)) // if n < 128:
JB(LabelRef("cmp64")) // goto cmp64
cmpAVX(S256, 4, true) // ZF = [compare 128 bytes]
JNE(LabelRef("done")) // return if ZF == 0
JMP(LabelRef("cmp128")) // loop cmp128
Label("cmp64")
CMPQ(n, U8(64)) // if n < 64:
JB(LabelRef("cmp32")) // goto cmp32
cmpAVX(S256, 2, true) // ZF = [compare 64 bytes]
JNE(LabelRef("done")) // return if ZF == 0
Label("cmp32")
CMPQ(n, U8(32)) // if n < 32:
JB(LabelRef("cmp16")) // goto cmp16
cmpAVX(S256, 1, true) // ZF = [compare 32 bytes]
JNE(LabelRef("done")) // return if ZF == 0
Label("cmp16")
// Convert YMM masks to XMM
min = min.(Vec).AsX()
max = max.(Vec).AsX()
CMPQ(n, U8(16)) // if n <= 16:
JLE(LabelRef("cmp_tail")) // goto cmp_tail
cmpAVX(S128, 1, true) // ZF = [compare 16 bytes]
JNE(LabelRef("done")) // return if ZF == 0
Label("cmp_tail")
// At this point, we have <= 16 bytes to compare, but we know the total input
// is >= 16 bytes. Move the pointer to the *last* 16 bytes of the input so we
// can skip the fallback.
SUBQ(Imm(16), n) // n -= 16
ADDQ(n, p.Base) // p += n
cmpAVX(S128, 1, false) // ZF = [compare 16 bytes]
JMP(LabelRef("done")) // return ZF
Generate()
}
func valid4(p Mem, n Register) {
val := GP32()
MOVL(p, val) // val = p[0:4]
setup4(val) // [update val register]
ADDQ(U8(4), p.Base) // p += 4
SUBQ(U8(4), n) // n -= 4
TESTL(U32(0x80808080), val) // ZF = (0x80808080 & val) == 0
}
func setup4(val Register) {
nval := GP32()
tmp1 := GP32()
tmp2 := GP32()
MOVL(val, nval) // nval = val
LEAL(Mem{Disp: 0xDFDFDFE0, Base: val}, tmp1) // tmp1 = val + 0xDFDFDFE0
NOTL(nval) // nval = ^nval
ANDL(nval, tmp1) // tmp1 = nval & tmp1
LEAL(Mem{Disp: 0x01010101, Base: val}, tmp2) // tmp2 = val + 0x01010101
ORL(tmp2, val) // val = val | tmp2
ORL(tmp1, val) // val = val | tmp1
}
func valid8(p Mem, n, m1, m2, m3 Register) {
val := GP64()
nval := GP64()
tmp1 := GP64()
tmp2 := GP64()
MOVQ(p, val) // val = p[0:8]
MOVQ(val, nval) // nval = val
LEAQ(Mem{Base: val, Index: m1, Scale: 1}, tmp1) // tmp1 = val + m1
NOTQ(nval) // nval = ^nval
ANDQ(nval, tmp1) // tmp1 = nval & tmp1
LEAQ(Mem{Base: val, Index: m2, Scale: 1}, tmp2) // tmp2 = val + m2
ORQ(tmp2, val) // val = val | tmp2
ORQ(tmp1, val) // val = val | tmp1
ADDQ(U8(8), p.Base) // p += 8
SUBQ(U8(8), n) // n -= 8
TESTQ(m3, val) // ZF = (m3 & val) == 0
}
|
