OLD | NEW |
1 /* | 1 /* |
2 * Copyright 2015 Google Inc. | 2 * Copyright 2015 Google Inc. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license that can be | 4 * Use of this source code is governed by a BSD-style license that can be |
5 * found in the LICENSE file. | 5 * found in the LICENSE file. |
6 */ | 6 */ |
7 | 7 |
8 inline SkPMFloat& SkPMFloat::operator=(const SkPMFloat& that) { | 8 inline SkPMFloat& SkPMFloat::operator=(const SkPMFloat& that) { |
9 fColors = that.fColors; | 9 fColors = that.fColors; |
10 return *this; | 10 return *this; |
11 } | 11 } |
12 | 12 |
13 // For SkPMFloat(SkPMColor), we widen our 8 bit components (fix8) to 8-bit compo
nents in 32 bits | 13 // For SkPMFloat(SkPMColor), we widen our 8 bit components (fix8) to 8-bit compo
nents in 32 bits |
14 // (fix8_32), then convert those to floats. | 14 // (fix8_32), then convert those to floats. |
15 | 15 |
16 // get() does the opposite, working from floats to 8-bit-in-32-bits, then back t
o packed 8 bit. | 16 // get() does the opposite, working from floats to 8-bit-in-32-bits, then back t
o packed 8 bit. |
17 | 17 |
18 // clamped() is the same as _SSE2: floats to 8-in-32, to 8-in-16, to packed 8 bi
t, with | 18 // clamped() is the same as _SSE2: floats to 8-in-32, to 8-in-16, to packed 8 bi
t, with |
19 // _mm_packus_epi16() both clamping and narrowing. | 19 // _mm_packus_epi16() both clamping and narrowing. |
20 | 20 |
21 inline SkPMFloat::SkPMFloat(SkPMColor c) { | 21 inline SkPMFloat::SkPMFloat(SkPMColor c) { |
22 SkPMColorAssert(c); | 22 SkPMColorAssert(c); |
23 const int _ = 255; // _ means to zero that byte. | 23 const int _ = 255; // _ means to zero that byte. |
24 __m128i fix8 = _mm_set_epi32(0,0,0,c), | 24 __m128i fix8 = _mm_set_epi32(0,0,0,c), |
25 fix8_32 = _mm_shuffle_epi8(fix8, _mm_set_epi8(_,_,_,3, _,_,_,2, _,_,
_,1, _,_,_,0)); | 25 fix8_32 = _mm_shuffle_epi8(fix8, _mm_set_epi8(_,_,_,3, _,_,_,2, _,_,
_,1, _,_,_,0)); |
26 fColors = _mm_cvtepi32_ps(fix8_32); | 26 fColors = _mm_cvtepi32_ps(fix8_32); |
27 SkASSERT(this->isValid()); | 27 SkASSERT(this->isValid()); |
28 } | 28 } |
29 | 29 |
30 inline SkPMColor SkPMFloat::get() const { | 30 inline SkPMColor SkPMFloat::trunc() const { |
31 SkASSERT(this->isValid()); | |
32 const int _ = 255; // _ means to zero that byte. | 31 const int _ = 255; // _ means to zero that byte. |
33 // We don't use _mm_cvtps_epi32, because we want precise control over how 0.
5 rounds (up). | 32 __m128i fix8_32 = _mm_cvttps_epi32(fColors), |
34 __m128i fix8_32 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), fColors)), | |
35 fix8 = _mm_shuffle_epi8(fix8_32, _mm_set_epi8(_,_,_,_, _,_,_,_, _
,_,_,_, 12,8,4,0)); | 33 fix8 = _mm_shuffle_epi8(fix8_32, _mm_set_epi8(_,_,_,_, _,_,_,_, _
,_,_,_, 12,8,4,0)); |
36 SkPMColor c = _mm_cvtsi128_si32(fix8); | 34 SkPMColor c = _mm_cvtsi128_si32(fix8); |
37 SkPMColorAssert(c); | 35 SkPMColorAssert(c); |
38 return c; | 36 return c; |
39 } | 37 } |
40 | 38 |
| 39 inline SkPMColor SkPMFloat::get() const { |
| 40 SkASSERT(this->isValid()); |
| 41 return SkPMFloat(Sk4f(0.5f) + *this).trunc(); |
| 42 } |
| 43 |
41 inline SkPMColor SkPMFloat::clamped() const { | 44 inline SkPMColor SkPMFloat::clamped() const { |
42 // We don't use _mm_cvtps_epi32, because we want precise control over how 0.
5 rounds (up). | 45 // We don't use _mm_cvtps_epi32, because we want precise control over how 0.
5 rounds (up). |
43 __m128i fix8_32 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), fColors)), | 46 __m128i fix8_32 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), fColors)), |
44 fix8_16 = _mm_packus_epi16(fix8_32, fix8_32), | 47 fix8_16 = _mm_packus_epi16(fix8_32, fix8_32), |
45 fix8 = _mm_packus_epi16(fix8_16, fix8_16); | 48 fix8 = _mm_packus_epi16(fix8_16, fix8_16); |
46 SkPMColor c = _mm_cvtsi128_si32(fix8); | 49 SkPMColor c = _mm_cvtsi128_si32(fix8); |
47 SkPMColorAssert(c); | 50 SkPMColorAssert(c); |
48 return c; | 51 return c; |
49 } | 52 } |
50 | 53 |
(...skipping 26 matching lines...) Expand all Loading... |
77 c2 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), c.fColors)), | 80 c2 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), c.fColors)), |
78 c3 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), d.fColors)); | 81 c3 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), d.fColors)); |
79 __m128i c3210 = _mm_packus_epi16(_mm_packus_epi16(c0, c1), | 82 __m128i c3210 = _mm_packus_epi16(_mm_packus_epi16(c0, c1), |
80 _mm_packus_epi16(c2, c3)); | 83 _mm_packus_epi16(c2, c3)); |
81 _mm_storeu_si128((__m128i*)colors, c3210); | 84 _mm_storeu_si128((__m128i*)colors, c3210); |
82 SkPMColorAssert(colors[0]); | 85 SkPMColorAssert(colors[0]); |
83 SkPMColorAssert(colors[1]); | 86 SkPMColorAssert(colors[1]); |
84 SkPMColorAssert(colors[2]); | 87 SkPMColorAssert(colors[2]); |
85 SkPMColorAssert(colors[3]); | 88 SkPMColorAssert(colors[3]); |
86 } | 89 } |
OLD | NEW |