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
|
/*
* Copyright 2016 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#ifndef SkFixed15_DEFINED
#define SkFixed15_DEFINED
#include "SkTypes.h"
// SkFixed15 is a fixed point value that represents values in [0,1] as [0x0000, 0x8000].
// This mapping allows us to implement most operations in tightly packed 16-bit SIMD,
// most notably multiplying using Q15 multiplication instructions (and a little fixup).
class SkFixed15 {
public:
SkFixed15() = default;
SkFixed15(float val) : fVal(val * 32768) { SkASSERT(0.0f <= val && val <= 1.0f); }
explicit operator float() const { return fVal * (1/32768.0f); }
static SkFixed15 Load(uint16_t val) {
SkASSERT(val <= 32768);
return val;
}
uint16_t store() const { return fVal; }
static SkFixed15 FromU8(uint8_t val) {
return val*128 + (val>>1) // 32768/255 == 128.50196..., which is very close to 128 + 0.5.
+ ((val+1)>>8); // All val but 255 are correct. +1 if val == 255 to get 32768.
}
uint8_t to_u8() const {
// FromU8() and to_u8() roundtrip all bytes.
// There is still much room to tweak this towards the ideal, a rounding scale by 255/32768.
return (fVal - (fVal>>8))>>7;
}
SkFixed15 operator +(SkFixed15 o) const { return fVal + o.fVal; }
SkFixed15 operator -(SkFixed15 o) const { return fVal - o.fVal; }
SkFixed15 operator *(SkFixed15 o) const { return (fVal * o.fVal + (1<<14)) >> 15; }
SkFixed15 operator<<(int bits) const { return fVal << bits; }
SkFixed15 operator>>(int bits) const { return fVal >> bits; }
SkFixed15& operator +=(SkFixed15 o) { return (*this = *this + o); }
SkFixed15& operator -=(SkFixed15 o) { return (*this = *this - o); }
SkFixed15& operator *=(SkFixed15 o) { return (*this = *this * o); }
SkFixed15& operator<<=(int bits) { return (*this = *this << bits); }
SkFixed15& operator>>=(int bits) { return (*this = *this >> bits); }
bool operator==(SkFixed15 o) const { return fVal == o.fVal; }
bool operator!=(SkFixed15 o) const { return fVal != o.fVal; }
bool operator<=(SkFixed15 o) const { return fVal <= o.fVal; }
bool operator>=(SkFixed15 o) const { return fVal >= o.fVal; }
bool operator< (SkFixed15 o) const { return fVal < o.fVal; }
bool operator> (SkFixed15 o) const { return fVal > o.fVal; }
private:
SkFixed15(int val) : fVal(val) {}
uint16_t fVal;
};
// Notes
// - SSSE3+ multiply is _mm_abs_epi16(_mm_mulhrs_epi16(x, y));
// - NEON multipy is vsraq_n_u16(vabsq_s16(vqrdmulhq_s16(x,y)),
// vandq_s16(x,y), 15);
// - Conversion to and from float can be done manually with bit masks and float add/subtract,
// rather than the naive version here involving int<->float conversion and float multiply.
// - On x86, conversion to float is _mm_sub_ps(_mm_unpacklo_epi16(x, _mm_set1_epi16(0x4380)),
// _mm_set1_ps(256.0f)). // 0x43800000
// - On ARM, we can use the vcvtq_n_f32_u32(vmovl_u16(x), 15) to convert to float,
// and vcvtq_n_u32_f32(..., 15) for the other way around.
#endif//SkFixed15_DEFINED
|
