|
1 /* |
|
2 * Copyright (c) 2005 |
|
3 * Eric Anholt. All rights reserved. |
|
4 * |
|
5 * Redistribution and use in source and binary forms, with or without |
|
6 * modification, are permitted provided that the following conditions |
|
7 * are met: |
|
8 * 1. Redistributions of source code must retain the above copyright |
|
9 * notice, this list of conditions and the following disclaimer. |
|
10 * 2. Redistributions in binary form must reproduce the above copyright |
|
11 * notice, this list of conditions and the following disclaimer in the |
|
12 * documentation and/or other materials provided with the distribution. |
|
13 * |
|
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND |
|
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
|
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
|
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE |
|
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
|
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
|
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
|
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
|
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
|
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
|
24 * SUCH DAMAGE. |
|
25 */ |
|
26 //Portions Copyright (c) 2008-2009 Nokia Corporation and/or its subsidiary(-ies). All rights reserved. |
|
27 |
|
28 #ifdef HAVE_CONFIG_H |
|
29 #include "config.h" |
|
30 #endif |
|
31 #include <liboilclasses.h> |
|
32 #include <liboilfunction.h> |
|
33 #include <emmintrin.h> |
|
34 #include <liboilcolorspace.h> |
|
35 |
|
36 #define SSE_FUNCTION __attribute__((force_align_arg_pointer)) |
|
37 |
|
38 #define COMPOSITE_ADD(d,s) oil_clamp_255((d) + (s)) |
|
39 |
|
40 SSE_FUNCTION static void |
|
41 composite_add_argb_sse (uint32_t *dest, const uint32_t *src, int n) |
|
42 { |
|
43 /* Initial operations to align the destination pointer */ |
|
44 for (; ((long)dest & 15) && (n > 0); n--) { |
|
45 uint32_t d = *dest, s = *src++; |
|
46 |
|
47 *dest++ = oil_argb( |
|
48 COMPOSITE_ADD(oil_argb_A(d), oil_argb_A(s)), |
|
49 COMPOSITE_ADD(oil_argb_R(d), oil_argb_R(s)), |
|
50 COMPOSITE_ADD(oil_argb_G(d), oil_argb_G(s)), |
|
51 COMPOSITE_ADD(oil_argb_B(d), oil_argb_B(s))); |
|
52 } |
|
53 for (; n >= 4; n -= 4) { |
|
54 __m128i s; |
|
55 s = _mm_loadu_si128((__m128i *)src); |
|
56 *(__m128i *)dest = _mm_adds_epu8(s, *(__m128i *)dest); |
|
57 src += 4; |
|
58 dest += 4; |
|
59 } |
|
60 for (; n > 0; n--) { |
|
61 uint32_t d = *dest, s = *src++; |
|
62 |
|
63 *dest++ = oil_argb( |
|
64 COMPOSITE_ADD(oil_argb_A(d), oil_argb_A(s)), |
|
65 COMPOSITE_ADD(oil_argb_R(d), oil_argb_R(s)), |
|
66 COMPOSITE_ADD(oil_argb_G(d), oil_argb_G(s)), |
|
67 COMPOSITE_ADD(oil_argb_B(d), oil_argb_B(s))); |
|
68 } |
|
69 } |
|
70 OIL_DEFINE_IMPL_FULL (composite_add_argb_sse, composite_add_argb, |
|
71 OIL_IMPL_FLAG_SSE2); |
|
72 |
|
73 SSE_FUNCTION static void |
|
74 composite_add_argb_const_src_sse (uint32_t *dest, const uint32_t *src_1, int n) |
|
75 { |
|
76 __m128i s; |
|
77 uint32_t val = *src_1; |
|
78 |
|
79 /* Initial operations to align the destination pointer */ |
|
80 for (; ((long)dest & 15) && (n > 0); n--) { |
|
81 uint32_t d = *dest; |
|
82 |
|
83 *dest++ = oil_argb( |
|
84 COMPOSITE_ADD(oil_argb_A(d), oil_argb_A(val)), |
|
85 COMPOSITE_ADD(oil_argb_R(d), oil_argb_R(val)), |
|
86 COMPOSITE_ADD(oil_argb_G(d), oil_argb_G(val)), |
|
87 COMPOSITE_ADD(oil_argb_B(d), oil_argb_B(val))); |
|
88 } |
|
89 s = _mm_set1_epi32(val); |
|
90 for (; n >= 4; n -= 4) { |
|
91 __m128i xmm0; |
|
92 xmm0 = _mm_adds_epu8(s, *(__m128i *)dest); |
|
93 _mm_store_si128((__m128i *)dest, xmm0); |
|
94 dest += 4; |
|
95 } |
|
96 for (; n > 0; n--) { |
|
97 uint32_t d = *dest; |
|
98 |
|
99 *dest++ = oil_argb( |
|
100 COMPOSITE_ADD(oil_argb_A(d), oil_argb_A(val)), |
|
101 COMPOSITE_ADD(oil_argb_R(d), oil_argb_R(val)), |
|
102 COMPOSITE_ADD(oil_argb_G(d), oil_argb_G(val)), |
|
103 COMPOSITE_ADD(oil_argb_B(d), oil_argb_B(val))); |
|
104 } |
|
105 } |
|
106 OIL_DEFINE_IMPL_FULL (composite_add_argb_const_src_sse, |
|
107 composite_add_argb_const_src, OIL_IMPL_FLAG_SSE2); |
|
108 |
|
109 SSE_FUNCTION static void |
|
110 composite_add_u8_sse (uint8_t *dest, const uint8_t *src, int n) |
|
111 { |
|
112 /* Initial operations to align the destination pointer */ |
|
113 for (; ((long)dest & 15) && (n > 0); n--) { |
|
114 int x = (int)*dest + *src++; |
|
115 if (x > 255) |
|
116 x = 255; |
|
117 *dest++ = x; |
|
118 } |
|
119 for (; n >= 16; n -= 16) { |
|
120 __m128i d, s; |
|
121 s = _mm_loadu_si128((__m128i *)src); |
|
122 d = _mm_adds_epu8(s, *(__m128i *)dest); |
|
123 _mm_store_si128((__m128i *)dest, d); |
|
124 src += 16; |
|
125 dest += 16; |
|
126 } |
|
127 for (; n > 0; n--) { |
|
128 int x = (int)*dest + *src++; |
|
129 if (x > 255) |
|
130 x = 255; |
|
131 *dest++ = x; |
|
132 } |
|
133 } |
|
134 OIL_DEFINE_IMPL_FULL (composite_add_u8_sse, composite_add_u8, |
|
135 OIL_IMPL_FLAG_SSE2); |
|
136 |
|
137 SSE_FUNCTION static void |
|
138 composite_add_u8_const_src_sse (uint8_t *dest, const uint8_t *src_1, int n) |
|
139 { |
|
140 __m128i s; |
|
141 int val = *src_1; |
|
142 |
|
143 /* Initial operations to align the destination pointer */ |
|
144 for (; ((long)dest & 15) && (n > 0); n--) { |
|
145 int x = *dest + val; |
|
146 if (x > 255) |
|
147 x = 255; |
|
148 *dest++ = x; |
|
149 } |
|
150 s = _mm_set1_epi8(val); |
|
151 for (; n >= 16; n -= 16) { |
|
152 __m128i d; |
|
153 d = _mm_adds_epu8(*(__m128i *)dest, s); |
|
154 _mm_store_si128((__m128i *)dest, d); |
|
155 dest += 16; |
|
156 } |
|
157 for (; n > 0; n--) { |
|
158 int x = *dest + val; |
|
159 if (x > 255) |
|
160 x = 255; |
|
161 *dest++ = x; |
|
162 } |
|
163 } |
|
164 OIL_DEFINE_IMPL_FULL (composite_add_u8_const_src_sse, |
|
165 composite_add_u8_const_src, OIL_IMPL_FLAG_SSE2); |
|
166 |
|
167 |
|
168 #ifdef __SYMBIAN32__ |
|
169 |
|
170 OilFunctionImpl* __oil_function_impl_composite_add_argb_sse, composite_add_argb() { |
|
171 return &_oil_function_impl_composite_add_argb_sse, composite_add_argb; |
|
172 } |
|
173 #endif |
|
174 |
|
175 #ifdef __SYMBIAN32__ |
|
176 |
|
177 OilFunctionImpl* __oil_function_impl_composite_add_argb_const_src_sse() { |
|
178 return &_oil_function_impl_composite_add_argb_const_src_sse; |
|
179 } |
|
180 #endif |
|
181 |
|
182 #ifdef __SYMBIAN32__ |
|
183 |
|
184 OilFunctionImpl* __oil_function_impl_composite_add_u8_sse, composite_add_u8() { |
|
185 return &_oil_function_impl_composite_add_u8_sse, composite_add_u8; |
|
186 } |
|
187 #endif |
|
188 |
|
189 #ifdef __SYMBIAN32__ |
|
190 |
|
191 OilFunctionImpl* __oil_function_impl_composite_add_u8_const_src_sse() { |
|
192 return &_oil_function_impl_composite_add_u8_const_src_sse; |
|
193 } |
|
194 #endif |
|
195 |