1/*
2 * Copyright (c) 2021 - 2023 the ThorVG project. All rights reserved.
3
4 * Permission is hereby granted, free of charge, to any person obtaining a copy
5 * of this software and associated documentation files (the "Software"), to deal
6 * in the Software without restriction, including without limitation the rights
7 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8 * copies of the Software, and to permit persons to whom the Software is
9 * furnished to do so, subject to the following conditions:
10
11 * The above copyright notice and this permission notice shall be included in all
12 * copies or substantial portions of the Software.
13
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
20 * SOFTWARE.
21 */
22
23#ifdef THORVG_NEON_VECTOR_SUPPORT
24
25#include <arm_neon.h>
26
27static inline uint8x8_t ALPHA_BLEND(uint8x8_t c, uint8x8_t a)
28{
29 uint16x8_t t = vmull_u8(c, a);
30 return vshrn_n_u16(t, 8);
31}
32
33
34static void neonRasterPixel32(uint32_t *dst, uint32_t val, uint32_t offset, int32_t len)
35{
36 uint32_t iterations = len / 4;
37 uint32_t neonFilled = iterations * 4;
38
39 dst += offset;
40 uint32x4_t vectorVal = {val, val, val, val};
41
42 for (uint32_t i = 0; i < iterations; ++i) {
43 vst1q_u32(dst, vectorVal);
44 dst += 4;
45 }
46
47 int32_t leftovers = len - neonFilled;
48 while (leftovers--) *dst++ = val;
49}
50
51
52static bool neonRasterTranslucentRle(SwSurface* surface, const SwRleData* rle, uint8_t r, uint8_t g, uint8_t b, uint8_t a)
53{
54 if (surface->channelSize != sizeof(uint32_t)) {
55 TVGERR("SW_ENGINE", "Unsupported Channel Size = %d", surface->channelSize);
56 return false;
57 }
58
59 auto color = surface->blender.join(r, g, b, a);
60 auto span = rle->spans;
61 uint32_t src;
62 uint8x8_t *vDst = nullptr;
63 uint16_t align;
64
65 for (uint32_t i = 0; i < rle->size; ++i) {
66 if (span->coverage < 255) src = ALPHA_BLEND(color, span->coverage);
67 else src = color;
68
69 auto dst = &surface->buf32[span->y * surface->stride + span->x];
70 auto ialpha = IALPHA(src);
71
72 if ((((uint32_t) dst) & 0x7) != 0) {
73 //fill not aligned byte
74 *dst = src + ALPHA_BLEND(*dst, ialpha);
75 vDst = (uint8x8_t*)(dst + 1);
76 align = 1;
77 } else {
78 vDst = (uint8x8_t*) dst;
79 align = 0;
80 }
81
82 uint8x8_t vSrc = (uint8x8_t) vdup_n_u32(src);
83 uint8x8_t vIalpha = vdup_n_u8((uint8_t) ialpha);
84
85 for (uint32_t x = 0; x < (span->len - align) / 2; ++x)
86 vDst[x] = vadd_u8(vSrc, ALPHA_BLEND(vDst[x], vIalpha));
87
88 auto leftovers = (span->len - align) % 2;
89 if (leftovers > 0) dst[span->len - 1] = src + ALPHA_BLEND(dst[span->len - 1], ialpha);
90
91 ++span;
92 }
93 return true;
94}
95
96
97static bool neonRasterTranslucentRect(SwSurface* surface, const SwBBox& region, uint8_t r, uint8_t g, uint8_t b, uint8_t a)
98{
99 if (surface->channelSize != sizeof(uint32_t)) {
100 TVGERR("SW_ENGINE", "Unsupported Channel Size = %d", surface->channelSize);
101 return false;
102 }
103
104 auto color = surface->blender.join(r, g, b, a);
105 auto buffer = surface->buf32 + (region.min.y * surface->stride) + region.min.x;
106 auto h = static_cast<uint32_t>(region.max.y - region.min.y);
107 auto w = static_cast<uint32_t>(region.max.x - region.min.x);
108 auto ialpha = 255 - a;
109
110 auto vColor = vdup_n_u32(color);
111 auto vIalpha = vdup_n_u8((uint8_t) ialpha);
112
113 uint8x8_t* vDst = nullptr;
114 uint32_t align;
115
116 for (uint32_t y = 0; y < h; ++y) {
117 auto dst = &buffer[y * surface->stride];
118
119 if ((((uint32_t) dst) & 0x7) != 0) {
120 //fill not aligned byte
121 *dst = color + ALPHA_BLEND(*dst, ialpha);
122 vDst = (uint8x8_t*) (dst + 1);
123 align = 1;
124 } else {
125 vDst = (uint8x8_t*) dst;
126 align = 0;
127 }
128
129 for (uint32_t x = 0; x < (w - align) / 2; ++x)
130 vDst[x] = vadd_u8((uint8x8_t)vColor, ALPHA_BLEND(vDst[x], vIalpha));
131
132 auto leftovers = (w - align) % 2;
133 if (leftovers > 0) dst[w - 1] = color + ALPHA_BLEND(dst[w - 1], ialpha);
134 }
135 return true;
136}
137
138#endif
139