1/*
2 * Copyright (C) 2012 University of Szeged
3 * Copyright (C) 2012 Gabor Rapcsanyi
4 * Copyright (C) 2014 Adobe Systems Incorporated. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 *    notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY UNIVERSITY OF SZEGED ``AS IS'' AND ANY
16 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
18 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL UNIVERSITY OF SZEGED OR
19 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
20 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
21 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
22 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
23 * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
25 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 */
27
28#ifndef FEBlendNEON_h
29#define FEBlendNEON_h
30
31#if ENABLE(FILTERS) && HAVE(ARM_NEON_INTRINSICS)
32
33#include "FEBlend.h"
34#include <arm_neon.h>
35
36namespace WebCore {
37
38class FEBlendUtilitiesNEON {
39public:
40    static inline uint16x8_t div255(uint16x8_t num, uint16x8_t sixteenConst255, uint16x8_t sixteenConstOne)
41    {
42        uint16x8_t quotient = vshrq_n_u16(num, 8);
43        uint16x8_t remainder = vaddq_u16(vsubq_u16(num, vmulq_u16(sixteenConst255, quotient)), sixteenConstOne);
44        return vaddq_u16(quotient, vshrq_n_u16(remainder, 8));
45    }
46
47    static inline uint16x8_t normal(uint16x8_t pixelA, uint16x8_t pixelB, uint16x8_t alphaA, uint16x8_t,
48                                    uint16x8_t sixteenConst255, uint16x8_t sixteenConstOne)
49    {
50        uint16x8_t tmp1 = vsubq_u16(sixteenConst255, alphaA);
51        uint16x8_t tmp2 = vmulq_u16(tmp1, pixelB);
52        uint16x8_t tmp3 = div255(tmp2, sixteenConst255, sixteenConstOne);
53        return vaddq_u16(tmp3, pixelA);
54    }
55
56    static inline uint16x8_t multiply(uint16x8_t pixelA, uint16x8_t pixelB, uint16x8_t alphaA, uint16x8_t alphaB,
57                                      uint16x8_t sixteenConst255, uint16x8_t sixteenConstOne)
58    {
59        uint16x8_t tmp1 = vsubq_u16(sixteenConst255, alphaA);
60        uint16x8_t tmp2 = vmulq_u16(tmp1, pixelB);
61        uint16x8_t tmp3 = vaddq_u16(vsubq_u16(sixteenConst255, alphaB), pixelB);
62        uint16x8_t tmp4 = vmulq_u16(tmp3, pixelA);
63        uint16x8_t tmp5 = vaddq_u16(tmp2, tmp4);
64        return div255(tmp5, sixteenConst255, sixteenConstOne);
65    }
66
67    static inline uint16x8_t screen(uint16x8_t pixelA, uint16x8_t pixelB, uint16x8_t, uint16x8_t,
68                                    uint16x8_t sixteenConst255, uint16x8_t sixteenConstOne)
69    {
70        uint16x8_t tmp1 = vaddq_u16(pixelA, pixelB);
71        uint16x8_t tmp2 = vmulq_u16(pixelA, pixelB);
72        uint16x8_t tmp3 = div255(tmp2, sixteenConst255, sixteenConstOne);
73        return vsubq_u16(tmp1, tmp3);
74    }
75
76    static inline uint16x8_t darken(uint16x8_t pixelA, uint16x8_t pixelB, uint16x8_t alphaA, uint16x8_t alphaB,
77                                    uint16x8_t sixteenConst255, uint16x8_t sixteenConstOne)
78    {
79        uint16x8_t tmp1 = vsubq_u16(sixteenConst255, alphaA);
80        uint16x8_t tmp2 = vmulq_u16(tmp1, pixelB);
81        uint16x8_t tmp3 = div255(tmp2, sixteenConst255, sixteenConstOne);
82        uint16x8_t tmp4 = vaddq_u16(tmp3, pixelA);
83
84        uint16x8_t tmp5 = vsubq_u16(sixteenConst255, alphaB);
85        uint16x8_t tmp6 = vmulq_u16(tmp5, pixelA);
86        uint16x8_t tmp7 = div255(tmp6, sixteenConst255, sixteenConstOne);
87        uint16x8_t tmp8 = vaddq_u16(tmp7, pixelB);
88
89        return vminq_u16(tmp4, tmp8);
90    }
91
92    static inline uint16x8_t lighten(uint16x8_t pixelA, uint16x8_t pixelB, uint16x8_t alphaA, uint16x8_t alphaB,
93                                     uint16x8_t sixteenConst255, uint16x8_t sixteenConstOne)
94    {
95        uint16x8_t tmp1 = vsubq_u16(sixteenConst255, alphaA);
96        uint16x8_t tmp2 = vmulq_u16(tmp1, pixelB);
97        uint16x8_t tmp3 = div255(tmp2, sixteenConst255, sixteenConstOne);
98        uint16x8_t tmp4 = vaddq_u16(tmp3, pixelA);
99
100        uint16x8_t tmp5 = vsubq_u16(sixteenConst255, alphaB);
101        uint16x8_t tmp6 = vmulq_u16(tmp5, pixelA);
102        uint16x8_t tmp7 = div255(tmp6, sixteenConst255, sixteenConstOne);
103        uint16x8_t tmp8 = vaddq_u16(tmp7, pixelB);
104
105        return vmaxq_u16(tmp4, tmp8);
106    }
107};
108
109void FEBlend::platformApplySoftware()
110{
111    FilterEffect* in = inputEffect(0);
112    FilterEffect* in2 = inputEffect(1);
113
114    Uint8ClampedArray* dstPixelArray = createPremultipliedImageResult();
115    if (!dstPixelArray)
116        return;
117
118    IntRect effectADrawingRect = requestedRegionOfInputImageData(in->absolutePaintRect());
119    RefPtr<Uint8ClampedArray> srcPixelArrayA = in->asPremultipliedImage(effectADrawingRect);
120
121    IntRect effectBDrawingRect = requestedRegionOfInputImageData(in2->absolutePaintRect());
122    RefPtr<Uint8ClampedArray> srcPixelArrayB = in2->asPremultipliedImage(effectBDrawingRect);
123
124    unsigned pixelArrayLength = srcPixelArrayA->length();
125    ASSERT(pixelArrayLength == srcPixelArrayB->length());
126
127    if (pixelArrayLength >= 8) {
128        platformApplyNEON(srcPixelArrayA->data(), srcPixelArrayB->data(), dstPixelArray->data(), pixelArrayLength);
129        return
130    }
131    // If there is just one pixel we expand it to two.
132    ASSERT(pixelArrayLength > 0);
133    uint32_t sourceA[2] = {0, 0};
134    uint32_t sourceBAndDest[2] = {0, 0};
135
136    sourceA[0] = reinterpret_cast<uint32_t*>(srcPixelArrayA->data())[0];
137    sourceBAndDest[0] = reinterpret_cast<uint32_t*>(srcPixelArrayB->data())[0];
138    platformApplyNEON(reinterpret_cast<uint8_t*>(sourceA), reinterpret_cast<uint8_t*>(sourceBAndDest), reinterpret_cast<uint8_t*>(sourceBAndDest), 8);
139    reinterpret_cast<uint32_t*>(dstPixelArray->data())[0] = sourceBAndDest[0];
140}
141
142void FEBlend::platformApplyNEON(unsigned char* srcPixelArrayA, unsigned char* srcPixelArrayB, unsigned char* dstPixelArray,
143                                unsigned colorArrayLength)
144{
145    uint8_t* sourcePixelA = reinterpret_cast<uint8_t*>(srcPixelArrayA);
146    uint8_t* sourcePixelB = reinterpret_cast<uint8_t*>(srcPixelArrayB);
147    uint8_t* destinationPixel = reinterpret_cast<uint8_t*>(dstPixelArray);
148
149    uint16x8_t sixteenConst255 = vdupq_n_u16(255);
150    uint16x8_t sixteenConstOne = vdupq_n_u16(1);
151
152    unsigned colorOffset = 0;
153    while (colorOffset < colorArrayLength) {
154        unsigned char alphaA1 = srcPixelArrayA[colorOffset + 3];
155        unsigned char alphaB1 = srcPixelArrayB[colorOffset + 3];
156        unsigned char alphaA2 = srcPixelArrayA[colorOffset + 7];
157        unsigned char alphaB2 = srcPixelArrayB[colorOffset + 7];
158
159        uint16x8_t doubblePixelA = vmovl_u8(vld1_u8(sourcePixelA + colorOffset));
160        uint16x8_t doubblePixelB = vmovl_u8(vld1_u8(sourcePixelB + colorOffset));
161        uint16x8_t alphaA = vcombine_u16(vdup_n_u16(alphaA1), vdup_n_u16(alphaA2));
162        uint16x8_t alphaB = vcombine_u16(vdup_n_u16(alphaB1), vdup_n_u16(alphaB2));
163
164        uint16x8_t result;
165        switch (m_mode) {
166        case BlendModeNormal:
167            result = FEBlendUtilitiesNEON::normal(doubblePixelA, doubblePixelB, alphaA, alphaB, sixteenConst255, sixteenConstOne);
168            break;
169        case BlendModeMultiply:
170            result = FEBlendUtilitiesNEON::multiply(doubblePixelA, doubblePixelB, alphaA, alphaB, sixteenConst255, sixteenConstOne);
171            break;
172        case BlendModeScreen:
173            result = FEBlendUtilitiesNEON::screen(doubblePixelA, doubblePixelB, alphaA, alphaB, sixteenConst255, sixteenConstOne);
174            break;
175        case BlendModeDarken:
176            result = FEBlendUtilitiesNEON::darken(doubblePixelA, doubblePixelB, alphaA, alphaB, sixteenConst255, sixteenConstOne);
177            break;
178        case BlendModeLighten:
179            result = FEBlendUtilitiesNEON::lighten(doubblePixelA, doubblePixelB, alphaA, alphaB, sixteenConst255, sixteenConstOne);
180            break;
181        default:
182            result = vdupq_n_u16(0);
183            break;
184        }
185
186        vst1_u8(destinationPixel + colorOffset, vmovn_u16(result));
187
188        unsigned char alphaR1 = 255 - ((255 - alphaA1) * (255 - alphaB1)) / 255;
189        unsigned char alphaR2 = 255 - ((255 - alphaA2) * (255 - alphaB2)) / 255;
190
191        dstPixelArray[colorOffset + 3] = alphaR1;
192        dstPixelArray[colorOffset + 7] = alphaR2;
193
194        colorOffset += 8;
195        if (colorOffset > colorArrayLength) {
196            ASSERT(colorOffset - 4 == colorArrayLength);
197            colorOffset = colorArrayLength - 8;
198        }
199    }
200}
201
202} // namespace WebCore
203
204#endif // ENABLE(FILTERS) && HAVE(ARM_NEON_INTRINSICS)
205
206#endif // FEBlendNEON_h
207