Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(78)

Issues Search
    My Issues | Starred     Open | Closed | All

Side by Side Diff: third_party/libwebp/dsp/upsampling_neon.c

Issue 116213006: Update libwebp to 0.4.0 (Closed) Base URL: svn://svn.chromium.org/chrome/trunk/src
Patch Set: After Blink Roll Created 6 years, 11 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch | Annotate | Revision Log
« no previous file with comments | « third_party/libwebp/dsp/upsampling.c ('k') | third_party/libwebp/dsp/upsampling_sse2.c » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
1 // Copyright 2011 Google Inc. All Rights Reserved. 1 // Copyright 2011 Google Inc. All Rights Reserved.
2 // 2 //
3 // Use of this source code is governed by a BSD-style license 3 // Use of this source code is governed by a BSD-style license
4 // that can be found in the COPYING file in the root of the source 4 // that can be found in the COPYING file in the root of the source
5 // tree. An additional intellectual property rights grant can be found 5 // tree. An additional intellectual property rights grant can be found
6 // in the file PATENTS. All contributing project authors may 6 // in the file PATENTS. All contributing project authors may
7 // be found in the AUTHORS file in the root of the source tree. 7 // be found in the AUTHORS file in the root of the source tree.
8 // ----------------------------------------------------------------------------- 8 // -----------------------------------------------------------------------------
9 // 9 //
10 // NEON version of YUV to RGB upsampling functions. 10 // NEON version of YUV to RGB upsampling functions.
11 // 11 //
12 // Author: mans@mansr.com (Mans Rullgard) 12 // Author: mans@mansr.com (Mans Rullgard)
13 // Based on SSE code by: somnath@google.com (Somnath Banerjee) 13 // Based on SSE code by: somnath@google.com (Somnath Banerjee)
14 14
15 #include "./dsp.h" 15 #include "./dsp.h"
16 16
17 #if defined(__cplusplus) || defined(c_plusplus)
18 extern "C" {
19 #endif
20
21 #if defined(WEBP_USE_NEON) 17 #if defined(WEBP_USE_NEON)
22 18
23 #include <assert.h> 19 #include <assert.h>
24 #include <arm_neon.h> 20 #include <arm_neon.h>
25 #include <string.h> 21 #include <string.h>
26 #include "./yuv.h" 22 #include "./yuv.h"
27 23
28 #ifdef FANCY_UPSAMPLING 24 #ifdef FANCY_UPSAMPLING
29 25
26 //-----------------------------------------------------------------------------
27 // U/V upsampling
28
30 // Loads 9 pixels each from rows r1 and r2 and generates 16 pixels. 29 // Loads 9 pixels each from rows r1 and r2 and generates 16 pixels.
31 #define UPSAMPLE_16PIXELS(r1, r2, out) { \ 30 #define UPSAMPLE_16PIXELS(r1, r2, out) { \
32 uint8x8_t a = vld1_u8(r1); \ 31 uint8x8_t a = vld1_u8(r1); \
33 uint8x8_t b = vld1_u8(r1 + 1); \ 32 uint8x8_t b = vld1_u8(r1 + 1); \
34 uint8x8_t c = vld1_u8(r2); \ 33 uint8x8_t c = vld1_u8(r2); \
35 uint8x8_t d = vld1_u8(r2 + 1); \ 34 uint8x8_t d = vld1_u8(r2 + 1); \
36 \ 35 \
37 uint16x8_t al = vshll_n_u8(a, 1); \ 36 uint16x8_t al = vshll_n_u8(a, 1); \
38 uint16x8_t bl = vshll_n_u8(b, 1); \ 37 uint16x8_t bl = vshll_n_u8(b, 1); \
39 uint16x8_t cl = vshll_n_u8(c, 1); \ 38 uint16x8_t cl = vshll_n_u8(c, 1); \
(...skipping 38 matching lines...) Expand 10 before | Expand all | Expand 10 after
78 #define UPSAMPLE_LAST_BLOCK(tb, bb, num_pixels, out) { \ 77 #define UPSAMPLE_LAST_BLOCK(tb, bb, num_pixels, out) { \
79 uint8_t r1[9], r2[9]; \ 78 uint8_t r1[9], r2[9]; \
80 memcpy(r1, (tb), (num_pixels)); \ 79 memcpy(r1, (tb), (num_pixels)); \
81 memcpy(r2, (bb), (num_pixels)); \ 80 memcpy(r2, (bb), (num_pixels)); \
82 /* replicate last byte */ \ 81 /* replicate last byte */ \
83 memset(r1 + (num_pixels), r1[(num_pixels) - 1], 9 - (num_pixels)); \ 82 memset(r1 + (num_pixels), r1[(num_pixels) - 1], 9 - (num_pixels)); \
84 memset(r2 + (num_pixels), r2[(num_pixels) - 1], 9 - (num_pixels)); \ 83 memset(r2 + (num_pixels), r2[(num_pixels) - 1], 9 - (num_pixels)); \
85 Upsample16Pixels(r1, r2, out); \ 84 Upsample16Pixels(r1, r2, out); \
86 } 85 }
87 86
88 #define CY 76283 87 //-----------------------------------------------------------------------------
89 #define CVR 89858 88 // YUV->RGB conversion
90 #define CUG 22014
91 #define CVG 45773
92 #define CUB 113618
93 89
94 static const int16_t coef[4] = { CVR / 4, CUG, CVG / 2, CUB / 4 }; 90 static const int16_t kCoeffs[4] = { kYScale, kVToR, kUToG, kVToG };
91
92 #define v255 vmov_n_u8(255)
93
94 #define STORE_Rgb(out, r, g, b) do { \
95 const uint8x8x3_t r_g_b = {{ r, g, b }}; \
96 vst3_u8(out, r_g_b); \
97 } while (0)
98
99 #define STORE_Bgr(out, r, g, b) do { \
100 const uint8x8x3_t b_g_r = {{ b, g, r }}; \
101 vst3_u8(out, b_g_r); \
102 } while (0)
103
104 #define STORE_Rgba(out, r, g, b) do { \
105 const uint8x8x4_t r_g_b_v255 = {{ r, g, b, v255 }}; \
106 vst4_u8(out, r_g_b_v255); \
107 } while (0)
108
109 #define STORE_Bgra(out, r, g, b) do { \
110 const uint8x8x4_t b_g_r_v255 = {{ b, g, r, v255 }}; \
111 vst4_u8(out, b_g_r_v255); \
112 } while (0)
95 113
96 #define CONVERT8(FMT, XSTEP, N, src_y, src_uv, out, cur_x) { \ 114 #define CONVERT8(FMT, XSTEP, N, src_y, src_uv, out, cur_x) { \
97 int i; \ 115 int i; \
98 for (i = 0; i < N; i += 8) { \ 116 for (i = 0; i < N; i += 8) { \
99 int off = ((cur_x) + i) * XSTEP; \ 117 const int off = ((cur_x) + i) * XSTEP; \
100 uint8x8_t y = vld1_u8(src_y + (cur_x) + i); \ 118 uint8x8_t y = vld1_u8((src_y) + (cur_x) + i); \
101 uint8x8_t u = vld1_u8((src_uv) + i); \ 119 uint8x8_t u = vld1_u8((src_uv) + i); \
102 uint8x8_t v = vld1_u8((src_uv) + i + 16); \ 120 uint8x8_t v = vld1_u8((src_uv) + i + 16); \
103 int16x8_t yy = vreinterpretq_s16_u16(vsubl_u8(y, u16)); \ 121 const int16x8_t yy = vreinterpretq_s16_u16(vsubl_u8(y, u16)); \
104 int16x8_t uu = vreinterpretq_s16_u16(vsubl_u8(u, u128)); \ 122 const int16x8_t uu = vreinterpretq_s16_u16(vsubl_u8(u, u128)); \
105 int16x8_t vv = vreinterpretq_s16_u16(vsubl_u8(v, u128)); \ 123 const int16x8_t vv = vreinterpretq_s16_u16(vsubl_u8(v, u128)); \
106 \ 124 int32x4_t yl = vmull_lane_s16(vget_low_s16(yy), cf16, 0); \
107 int16x8_t ud = vshlq_n_s16(uu, 1); \ 125 int32x4_t yh = vmull_lane_s16(vget_high_s16(yy), cf16, 0); \
108 int16x8_t vd = vshlq_n_s16(vv, 1); \ 126 const int32x4_t rl = vmlal_lane_s16(yl, vget_low_s16(vv), cf16, 1);\
109 \ 127 const int32x4_t rh = vmlal_lane_s16(yh, vget_high_s16(vv), cf16, 1);\
110 int32x4_t vrl = vqdmlal_lane_s16(vshll_n_s16(vget_low_s16(vv), 1), \ 128 int32x4_t gl = vmlsl_lane_s16(yl, vget_low_s16(uu), cf16, 2); \
111 vget_low_s16(vd), cf16, 0); \ 129 int32x4_t gh = vmlsl_lane_s16(yh, vget_high_s16(uu), cf16, 2); \
112 int32x4_t vrh = vqdmlal_lane_s16(vshll_n_s16(vget_high_s16(vv), 1), \ 130 const int32x4_t bl = vmovl_s16(vget_low_s16(uu)); \
113 vget_high_s16(vd), cf16, 0); \ 131 const int32x4_t bh = vmovl_s16(vget_high_s16(uu)); \
114 int16x8_t vr = vcombine_s16(vrshrn_n_s32(vrl, 16), \ 132 gl = vmlsl_lane_s16(gl, vget_low_s16(vv), cf16, 3); \
115 vrshrn_n_s32(vrh, 16)); \ 133 gh = vmlsl_lane_s16(gh, vget_high_s16(vv), cf16, 3); \
116 \ 134 yl = vmlaq_lane_s32(yl, bl, cf32, 0); \
117 int32x4_t vl = vmovl_s16(vget_low_s16(vv)); \ 135 yh = vmlaq_lane_s32(yh, bh, cf32, 0); \
118 int32x4_t vh = vmovl_s16(vget_high_s16(vv)); \ 136 /* vrshrn_n_s32() already incorporates the rounding constant */ \
119 int32x4_t ugl = vmlal_lane_s16(vl, vget_low_s16(uu), cf16, 1); \ 137 y = vqmovun_s16(vcombine_s16(vrshrn_n_s32(rl, YUV_FIX2), \
120 int32x4_t ugh = vmlal_lane_s16(vh, vget_high_s16(uu), cf16, 1); \ 138 vrshrn_n_s32(rh, YUV_FIX2))); \
121 int32x4_t gcl = vqdmlal_lane_s16(ugl, vget_low_s16(vv), cf16, 2); \ 139 u = vqmovun_s16(vcombine_s16(vrshrn_n_s32(gl, YUV_FIX2), \
122 int32x4_t gch = vqdmlal_lane_s16(ugh, vget_high_s16(vv), cf16, 2); \ 140 vrshrn_n_s32(gh, YUV_FIX2))); \
123 int16x8_t gc = vcombine_s16(vrshrn_n_s32(gcl, 16), \ 141 v = vqmovun_s16(vcombine_s16(vrshrn_n_s32(yl, YUV_FIX2), \
124 vrshrn_n_s32(gch, 16)); \ 142 vrshrn_n_s32(yh, YUV_FIX2))); \
125 \ 143 STORE_ ## FMT(out + off, y, u, v); \
126 int32x4_t ubl = vqdmlal_lane_s16(vshll_n_s16(vget_low_s16(uu), 1), \
127 vget_low_s16(ud), cf16, 3); \
128 int32x4_t ubh = vqdmlal_lane_s16(vshll_n_s16(vget_high_s16(uu), 1), \
129 vget_high_s16(ud), cf16, 3); \
130 int16x8_t ub = vcombine_s16(vrshrn_n_s32(ubl, 16), \
131 vrshrn_n_s32(ubh, 16)); \
132 \
133 int32x4_t rl = vaddl_s16(vget_low_s16(yy), vget_low_s16(vr)); \
134 int32x4_t rh = vaddl_s16(vget_high_s16(yy), vget_high_s16(vr)); \
135 int32x4_t gl = vsubl_s16(vget_low_s16(yy), vget_low_s16(gc)); \
136 int32x4_t gh = vsubl_s16(vget_high_s16(yy), vget_high_s16(gc)); \
137 int32x4_t bl = vaddl_s16(vget_low_s16(yy), vget_low_s16(ub)); \
138 int32x4_t bh = vaddl_s16(vget_high_s16(yy), vget_high_s16(ub)); \
139 \
140 rl = vmulq_lane_s32(rl, cf32, 0); \
141 rh = vmulq_lane_s32(rh, cf32, 0); \
142 gl = vmulq_lane_s32(gl, cf32, 0); \
143 gh = vmulq_lane_s32(gh, cf32, 0); \
144 bl = vmulq_lane_s32(bl, cf32, 0); \
145 bh = vmulq_lane_s32(bh, cf32, 0); \
146 \
147 y = vqmovun_s16(vcombine_s16(vrshrn_n_s32(rl, 16), \
148 vrshrn_n_s32(rh, 16))); \
149 u = vqmovun_s16(vcombine_s16(vrshrn_n_s32(gl, 16), \
150 vrshrn_n_s32(gh, 16))); \
151 v = vqmovun_s16(vcombine_s16(vrshrn_n_s32(bl, 16), \
152 vrshrn_n_s32(bh, 16))); \
153 STR_ ## FMT(out + off, y, u, v); \
154 } \ 144 } \
155 } 145 }
156 146
157 #define v255 vmov_n_u8(255) 147 #define CONVERT1(FUNC, XSTEP, N, src_y, src_uv, rgb, cur_x) { \
158
159 #define STR_Rgb(out, r, g, b) do { \
160 const uint8x8x3_t r_g_b = {{ r, g, b }}; \
161 vst3_u8(out, r_g_b); \
162 } while (0)
163
164 #define STR_Bgr(out, r, g, b) do { \
165 const uint8x8x3_t b_g_r = {{ b, g, r }}; \
166 vst3_u8(out, b_g_r); \
167 } while (0)
168
169 #define STR_Rgba(out, r, g, b) do { \
170 const uint8x8x4_t r_g_b_v255 = {{ r, g, b, v255 }}; \
171 vst4_u8(out, r_g_b_v255); \
172 } while (0)
173
174 #define STR_Bgra(out, r, g, b) do { \
175 const uint8x8x4_t b_g_r_v255 = {{ b, g, r, v255 }}; \
176 vst4_u8(out, b_g_r_v255); \
177 } while (0)
178
179 #define CONVERT1(FMT, XSTEP, N, src_y, src_uv, rgb, cur_x) { \
180 int i; \ 148 int i; \
181 for (i = 0; i < N; i++) { \ 149 for (i = 0; i < N; i++) { \
182 int off = ((cur_x) + i) * XSTEP; \ 150 const int off = ((cur_x) + i) * XSTEP; \
183 int y = src_y[(cur_x) + i]; \ 151 const int y = src_y[(cur_x) + i]; \
184 int u = (src_uv)[i]; \ 152 const int u = (src_uv)[i]; \
185 int v = (src_uv)[i + 16]; \ 153 const int v = (src_uv)[i + 16]; \
186 VP8YuvTo ## FMT(y, u, v, rgb + off); \ 154 FUNC(y, u, v, rgb + off); \
187 } \ 155 } \
188 } 156 }
189 157
190 #define CONVERT2RGB_8(FMT, XSTEP, top_y, bottom_y, uv, \ 158 #define CONVERT2RGB_8(FMT, XSTEP, top_y, bottom_y, uv, \
191 top_dst, bottom_dst, cur_x, len) { \ 159 top_dst, bottom_dst, cur_x, len) { \
192 if (top_y) { \ 160 CONVERT8(FMT, XSTEP, len, top_y, uv, top_dst, cur_x) \
193 CONVERT8(FMT, XSTEP, len, top_y, uv, top_dst, cur_x) \ 161 if (bottom_y != NULL) { \
194 } \
195 if (bottom_y) { \
196 CONVERT8(FMT, XSTEP, len, bottom_y, (uv) + 32, bottom_dst, cur_x) \ 162 CONVERT8(FMT, XSTEP, len, bottom_y, (uv) + 32, bottom_dst, cur_x) \
197 } \ 163 } \
198 } 164 }
199 165
200 #define CONVERT2RGB_1(FMT, XSTEP, top_y, bottom_y, uv, \ 166 #define CONVERT2RGB_1(FUNC, XSTEP, top_y, bottom_y, uv, \
201 top_dst, bottom_dst, cur_x, len) { \ 167 top_dst, bottom_dst, cur_x, len) { \
202 if (top_y) { \ 168 CONVERT1(FUNC, XSTEP, len, top_y, uv, top_dst, cur_x); \
203 CONVERT1(FMT, XSTEP, len, top_y, uv, top_dst, cur_x); \ 169 if (bottom_y != NULL) { \
204 } \ 170 CONVERT1(FUNC, XSTEP, len, bottom_y, (uv) + 32, bottom_dst, cur_x); \
205 if (bottom_y) { \
206 CONVERT1(FMT, XSTEP, len, bottom_y, (uv) + 32, bottom_dst, cur_x); \
207 } \ 171 } \
208 } 172 }
209 173
210 #define NEON_UPSAMPLE_FUNC(FUNC_NAME, FMT, XSTEP) \ 174 #define NEON_UPSAMPLE_FUNC(FUNC_NAME, FMT, XSTEP) \
211 static void FUNC_NAME(const uint8_t *top_y, const uint8_t *bottom_y, \ 175 static void FUNC_NAME(const uint8_t *top_y, const uint8_t *bottom_y, \
212 const uint8_t *top_u, const uint8_t *top_v, \ 176 const uint8_t *top_u, const uint8_t *top_v, \
213 const uint8_t *cur_u, const uint8_t *cur_v, \ 177 const uint8_t *cur_u, const uint8_t *cur_v, \
214 uint8_t *top_dst, uint8_t *bottom_dst, int len) { \ 178 uint8_t *top_dst, uint8_t *bottom_dst, int len) { \
215 int block; \ 179 int block; \
216 /* 16 byte aligned array to cache reconstructed u and v */ \ 180 /* 16 byte aligned array to cache reconstructed u and v */ \
217 uint8_t uv_buf[2 * 32 + 15]; \ 181 uint8_t uv_buf[2 * 32 + 15]; \
218 uint8_t *const r_uv = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15); \ 182 uint8_t *const r_uv = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15); \
219 const int uv_len = (len + 1) >> 1; \ 183 const int uv_len = (len + 1) >> 1; \
220 /* 9 pixels must be read-able for each block */ \ 184 /* 9 pixels must be read-able for each block */ \
221 const int num_blocks = (uv_len - 1) >> 3; \ 185 const int num_blocks = (uv_len - 1) >> 3; \
222 const int leftover = uv_len - num_blocks * 8; \ 186 const int leftover = uv_len - num_blocks * 8; \
223 const int last_pos = 1 + 16 * num_blocks; \ 187 const int last_pos = 1 + 16 * num_blocks; \
224 \ 188 \
225 const int u_diag = ((top_u[0] + cur_u[0]) >> 1) + 1; \ 189 const int u_diag = ((top_u[0] + cur_u[0]) >> 1) + 1; \
226 const int v_diag = ((top_v[0] + cur_v[0]) >> 1) + 1; \ 190 const int v_diag = ((top_v[0] + cur_v[0]) >> 1) + 1; \
227 \ 191 \
228 const int16x4_t cf16 = vld1_s16(coef); \ 192 const int16x4_t cf16 = vld1_s16(kCoeffs); \
229 const int32x2_t cf32 = vmov_n_s32(CY); \ 193 const int32x2_t cf32 = vmov_n_s32(kUToB); \
230 const uint8x8_t u16 = vmov_n_u8(16); \ 194 const uint8x8_t u16 = vmov_n_u8(16); \
231 const uint8x8_t u128 = vmov_n_u8(128); \ 195 const uint8x8_t u128 = vmov_n_u8(128); \
232 \ 196 \
233 /* Treat the first pixel in regular way */ \ 197 /* Treat the first pixel in regular way */ \
234 if (top_y) { \ 198 assert(top_y != NULL); \
199 { \
235 const int u0 = (top_u[0] + u_diag) >> 1; \ 200 const int u0 = (top_u[0] + u_diag) >> 1; \
236 const int v0 = (top_v[0] + v_diag) >> 1; \ 201 const int v0 = (top_v[0] + v_diag) >> 1; \
237 VP8YuvTo ## FMT(top_y[0], u0, v0, top_dst); \ 202 VP8YuvTo ## FMT(top_y[0], u0, v0, top_dst); \
238 } \ 203 } \
239 if (bottom_y) { \ 204 if (bottom_y != NULL) { \
240 const int u0 = (cur_u[0] + u_diag) >> 1; \ 205 const int u0 = (cur_u[0] + u_diag) >> 1; \
241 const int v0 = (cur_v[0] + v_diag) >> 1; \ 206 const int v0 = (cur_v[0] + v_diag) >> 1; \
242 VP8YuvTo ## FMT(bottom_y[0], u0, v0, bottom_dst); \ 207 VP8YuvTo ## FMT(bottom_y[0], u0, v0, bottom_dst); \
243 } \ 208 } \
244 \ 209 \
245 for (block = 0; block < num_blocks; ++block) { \ 210 for (block = 0; block < num_blocks; ++block) { \
246 UPSAMPLE_16PIXELS(top_u, cur_u, r_uv); \ 211 UPSAMPLE_16PIXELS(top_u, cur_u, r_uv); \
247 UPSAMPLE_16PIXELS(top_v, cur_v, r_uv + 16); \ 212 UPSAMPLE_16PIXELS(top_v, cur_v, r_uv + 16); \
248 CONVERT2RGB_8(FMT, XSTEP, top_y, bottom_y, r_uv, \ 213 CONVERT2RGB_8(FMT, XSTEP, top_y, bottom_y, r_uv, \
249 top_dst, bottom_dst, 16 * block + 1, 16); \ 214 top_dst, bottom_dst, 16 * block + 1, 16); \
250 top_u += 8; \ 215 top_u += 8; \
251 cur_u += 8; \ 216 cur_u += 8; \
252 top_v += 8; \ 217 top_v += 8; \
253 cur_v += 8; \ 218 cur_v += 8; \
254 } \ 219 } \
255 \ 220 \
256 UPSAMPLE_LAST_BLOCK(top_u, cur_u, leftover, r_uv); \ 221 UPSAMPLE_LAST_BLOCK(top_u, cur_u, leftover, r_uv); \
257 UPSAMPLE_LAST_BLOCK(top_v, cur_v, leftover, r_uv + 16); \ 222 UPSAMPLE_LAST_BLOCK(top_v, cur_v, leftover, r_uv + 16); \
258 CONVERT2RGB_1(FMT, XSTEP, top_y, bottom_y, r_uv, \ 223 CONVERT2RGB_1(VP8YuvTo ## FMT, XSTEP, top_y, bottom_y, r_uv, \
259 top_dst, bottom_dst, last_pos, len - last_pos); \ 224 top_dst, bottom_dst, last_pos, len - last_pos); \
260 } 225 }
261 226
262 // NEON variants of the fancy upsampler. 227 // NEON variants of the fancy upsampler.
263 NEON_UPSAMPLE_FUNC(UpsampleRgbLinePairNEON, Rgb, 3) 228 NEON_UPSAMPLE_FUNC(UpsampleRgbLinePairNEON, Rgb, 3)
264 NEON_UPSAMPLE_FUNC(UpsampleBgrLinePairNEON, Bgr, 3) 229 NEON_UPSAMPLE_FUNC(UpsampleBgrLinePairNEON, Bgr, 3)
265 NEON_UPSAMPLE_FUNC(UpsampleRgbaLinePairNEON, Rgba, 4) 230 NEON_UPSAMPLE_FUNC(UpsampleRgbaLinePairNEON, Rgba, 4)
266 NEON_UPSAMPLE_FUNC(UpsampleBgraLinePairNEON, Bgra, 4) 231 NEON_UPSAMPLE_FUNC(UpsampleBgraLinePairNEON, Bgra, 4)
267 232
268 #endif // FANCY_UPSAMPLING 233 #endif // FANCY_UPSAMPLING
269 234
270 #endif // WEBP_USE_NEON 235 #endif // WEBP_USE_NEON
271 236
272 //------------------------------------------------------------------------------ 237 //------------------------------------------------------------------------------
273 238
239 #ifdef FANCY_UPSAMPLING
240
274 extern WebPUpsampleLinePairFunc WebPUpsamplers[/* MODE_LAST */]; 241 extern WebPUpsampleLinePairFunc WebPUpsamplers[/* MODE_LAST */];
275 242
276 void WebPInitUpsamplersNEON(void) { 243 void WebPInitUpsamplersNEON(void) {
277 #if defined(WEBP_USE_NEON) 244 #if defined(WEBP_USE_NEON)
278 WebPUpsamplers[MODE_RGB] = UpsampleRgbLinePairNEON; 245 WebPUpsamplers[MODE_RGB] = UpsampleRgbLinePairNEON;
279 WebPUpsamplers[MODE_RGBA] = UpsampleRgbaLinePairNEON; 246 WebPUpsamplers[MODE_RGBA] = UpsampleRgbaLinePairNEON;
280 WebPUpsamplers[MODE_BGR] = UpsampleBgrLinePairNEON; 247 WebPUpsamplers[MODE_BGR] = UpsampleBgrLinePairNEON;
281 WebPUpsamplers[MODE_BGRA] = UpsampleBgraLinePairNEON; 248 WebPUpsamplers[MODE_BGRA] = UpsampleBgraLinePairNEON;
282 #endif // WEBP_USE_NEON 249 #endif // WEBP_USE_NEON
283 } 250 }
284 251
285 void WebPInitPremultiplyNEON(void) { 252 void WebPInitPremultiplyNEON(void) {
286 #if defined(WEBP_USE_NEON) 253 #if defined(WEBP_USE_NEON)
287 WebPUpsamplers[MODE_rgbA] = UpsampleRgbaLinePairNEON; 254 WebPUpsamplers[MODE_rgbA] = UpsampleRgbaLinePairNEON;
288 WebPUpsamplers[MODE_bgrA] = UpsampleBgraLinePairNEON; 255 WebPUpsamplers[MODE_bgrA] = UpsampleBgraLinePairNEON;
289 #endif // WEBP_USE_NEON 256 #endif // WEBP_USE_NEON
290 } 257 }
291 258
292 #if defined(__cplusplus) || defined(c_plusplus) 259 #else
293 } // extern "C" 260
294 #endif 261 // this empty function is to avoid an empty .o
262 void WebPInitPremultiplyNEON(void) {}
263
264 #endif // FANCY_UPSAMPLING
265
OLDNEW
« no previous file with comments | « third_party/libwebp/dsp/upsampling.c ('k') | third_party/libwebp/dsp/upsampling_sse2.c » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698