• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // Copyright 2014 Google Inc. All Rights Reserved.
2 //
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
5 // tree. An additional intellectual property rights grant can be found
6 // in the file PATENTS. All contributing project authors may
7 // be found in the AUTHORS file in the root of the source tree.
8 // -----------------------------------------------------------------------------
9 //
10 // Image transforms and color space conversion methods for lossless decoder.
11 //
12 // Author(s):  Djordje Pesut    (djordje.pesut@imgtec.com)
13 //             Jovan Zelincevic (jovan.zelincevic@imgtec.com)
14 
15 #include "src/dsp/dsp.h"
16 
17 #if defined(WEBP_USE_MIPS_DSP_R2)
18 
19 #include "src/dsp/lossless.h"
20 #include "src/dsp/lossless_common.h"
21 
22 #define MAP_COLOR_FUNCS(FUNC_NAME, TYPE, GET_INDEX, GET_VALUE)                 \
23 static void FUNC_NAME(const TYPE* src,                                         \
24                       const uint32_t* const color_map,                         \
25                       TYPE* dst, int y_start, int y_end,                       \
26                       int width) {                                             \
27   int y;                                                                       \
28   for (y = y_start; y < y_end; ++y) {                                          \
29     int x;                                                                     \
30     for (x = 0; x < (width >> 2); ++x) {                                       \
31       int tmp1, tmp2, tmp3, tmp4;                                              \
32       __asm__ volatile (                                                       \
33       ".ifc        " #TYPE ",  uint8_t                  \n\t"                  \
34         "lbu       %[tmp1],  0(%[src])                  \n\t"                  \
35         "lbu       %[tmp2],  1(%[src])                  \n\t"                  \
36         "lbu       %[tmp3],  2(%[src])                  \n\t"                  \
37         "lbu       %[tmp4],  3(%[src])                  \n\t"                  \
38         "addiu     %[src],   %[src],      4             \n\t"                  \
39       ".endif                                           \n\t"                  \
40       ".ifc        " #TYPE ",  uint32_t                 \n\t"                  \
41         "lw        %[tmp1],  0(%[src])                  \n\t"                  \
42         "lw        %[tmp2],  4(%[src])                  \n\t"                  \
43         "lw        %[tmp3],  8(%[src])                  \n\t"                  \
44         "lw        %[tmp4],  12(%[src])                 \n\t"                  \
45         "ext       %[tmp1],  %[tmp1],     8,        8   \n\t"                  \
46         "ext       %[tmp2],  %[tmp2],     8,        8   \n\t"                  \
47         "ext       %[tmp3],  %[tmp3],     8,        8   \n\t"                  \
48         "ext       %[tmp4],  %[tmp4],     8,        8   \n\t"                  \
49         "addiu     %[src],   %[src],      16            \n\t"                  \
50       ".endif                                           \n\t"                  \
51         "sll       %[tmp1],  %[tmp1],     2             \n\t"                  \
52         "sll       %[tmp2],  %[tmp2],     2             \n\t"                  \
53         "sll       %[tmp3],  %[tmp3],     2             \n\t"                  \
54         "sll       %[tmp4],  %[tmp4],     2             \n\t"                  \
55         "lwx       %[tmp1],  %[tmp1](%[color_map])      \n\t"                  \
56         "lwx       %[tmp2],  %[tmp2](%[color_map])      \n\t"                  \
57         "lwx       %[tmp3],  %[tmp3](%[color_map])      \n\t"                  \
58         "lwx       %[tmp4],  %[tmp4](%[color_map])      \n\t"                  \
59       ".ifc        " #TYPE ",  uint8_t                  \n\t"                  \
60         "ext       %[tmp1],  %[tmp1],     8,        8   \n\t"                  \
61         "ext       %[tmp2],  %[tmp2],     8,        8   \n\t"                  \
62         "ext       %[tmp3],  %[tmp3],     8,        8   \n\t"                  \
63         "ext       %[tmp4],  %[tmp4],     8,        8   \n\t"                  \
64         "sb        %[tmp1],  0(%[dst])                  \n\t"                  \
65         "sb        %[tmp2],  1(%[dst])                  \n\t"                  \
66         "sb        %[tmp3],  2(%[dst])                  \n\t"                  \
67         "sb        %[tmp4],  3(%[dst])                  \n\t"                  \
68         "addiu     %[dst],   %[dst],      4             \n\t"                  \
69       ".endif                                           \n\t"                  \
70       ".ifc        " #TYPE ",  uint32_t                 \n\t"                  \
71         "sw        %[tmp1],  0(%[dst])                  \n\t"                  \
72         "sw        %[tmp2],  4(%[dst])                  \n\t"                  \
73         "sw        %[tmp3],  8(%[dst])                  \n\t"                  \
74         "sw        %[tmp4],  12(%[dst])                 \n\t"                  \
75         "addiu     %[dst],   %[dst],      16            \n\t"                  \
76       ".endif                                           \n\t"                  \
77         : [tmp1]"=&r"(tmp1), [tmp2]"=&r"(tmp2), [tmp3]"=&r"(tmp3),             \
78           [tmp4]"=&r"(tmp4), [src]"+&r"(src), [dst]"+r"(dst)                   \
79         : [color_map]"r"(color_map)                                            \
80         : "memory"                                                             \
81       );                                                                       \
82     }                                                                          \
83     for (x = 0; x < (width & 3); ++x) {                                        \
84       *dst++ = GET_VALUE(color_map[GET_INDEX(*src++)]);                        \
85     }                                                                          \
86   }                                                                            \
87 }
88 
MAP_COLOR_FUNCS(MapARGB_MIPSdspR2,uint32_t,VP8GetARGBIndex,VP8GetARGBValue)89 MAP_COLOR_FUNCS(MapARGB_MIPSdspR2, uint32_t, VP8GetARGBIndex, VP8GetARGBValue)
90 MAP_COLOR_FUNCS(MapAlpha_MIPSdspR2, uint8_t, VP8GetAlphaIndex, VP8GetAlphaValue)
91 
92 #undef MAP_COLOR_FUNCS
93 
94 static WEBP_INLINE uint32_t ClampedAddSubtractFull(uint32_t c0, uint32_t c1,
95                                                    uint32_t c2) {
96   int temp0, temp1, temp2, temp3, temp4, temp5;
97   __asm__ volatile (
98     "preceu.ph.qbr   %[temp1],   %[c0]                 \n\t"
99     "preceu.ph.qbl   %[temp2],   %[c0]                 \n\t"
100     "preceu.ph.qbr   %[temp3],   %[c1]                 \n\t"
101     "preceu.ph.qbl   %[temp4],   %[c1]                 \n\t"
102     "preceu.ph.qbr   %[temp5],   %[c2]                 \n\t"
103     "preceu.ph.qbl   %[temp0],   %[c2]                 \n\t"
104     "subq.ph         %[temp3],   %[temp3],   %[temp5]  \n\t"
105     "subq.ph         %[temp4],   %[temp4],   %[temp0]  \n\t"
106     "addq.ph         %[temp1],   %[temp1],   %[temp3]  \n\t"
107     "addq.ph         %[temp2],   %[temp2],   %[temp4]  \n\t"
108     "shll_s.ph       %[temp1],   %[temp1],   7         \n\t"
109     "shll_s.ph       %[temp2],   %[temp2],   7         \n\t"
110     "precrqu_s.qb.ph %[temp2],   %[temp2],   %[temp1]  \n\t"
111     : [temp0]"=r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
112       [temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5)
113     : [c0]"r"(c0), [c1]"r"(c1), [c2]"r"(c2)
114     : "memory"
115   );
116   return temp2;
117 }
118 
ClampedAddSubtractHalf(uint32_t c0,uint32_t c1,uint32_t c2)119 static WEBP_INLINE uint32_t ClampedAddSubtractHalf(uint32_t c0, uint32_t c1,
120                                                    uint32_t c2) {
121   int temp0, temp1, temp2, temp3, temp4, temp5;
122   __asm__ volatile (
123     "adduh.qb         %[temp5],   %[c0],      %[c1]       \n\t"
124     "preceu.ph.qbr    %[temp3],   %[c2]                   \n\t"
125     "preceu.ph.qbr    %[temp1],   %[temp5]                \n\t"
126     "preceu.ph.qbl    %[temp2],   %[temp5]                \n\t"
127     "preceu.ph.qbl    %[temp4],   %[c2]                   \n\t"
128     "subq.ph          %[temp3],   %[temp1],   %[temp3]    \n\t"
129     "subq.ph          %[temp4],   %[temp2],   %[temp4]    \n\t"
130     "shrl.ph          %[temp5],   %[temp3],   15          \n\t"
131     "shrl.ph          %[temp0],   %[temp4],   15          \n\t"
132     "addq.ph          %[temp3],   %[temp3],   %[temp5]    \n\t"
133     "addq.ph          %[temp4],   %[temp0],   %[temp4]    \n\t"
134     "shra.ph          %[temp3],   %[temp3],   1           \n\t"
135     "shra.ph          %[temp4],   %[temp4],   1           \n\t"
136     "addq.ph          %[temp1],   %[temp1],   %[temp3]    \n\t"
137     "addq.ph          %[temp2],   %[temp2],   %[temp4]    \n\t"
138     "shll_s.ph        %[temp1],   %[temp1],   7           \n\t"
139     "shll_s.ph        %[temp2],   %[temp2],   7           \n\t"
140     "precrqu_s.qb.ph  %[temp1],   %[temp2],   %[temp1]    \n\t"
141     : [temp0]"=r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
142       [temp3]"=&r"(temp3), [temp4]"=r"(temp4), [temp5]"=&r"(temp5)
143     : [c0]"r"(c0), [c1]"r"(c1), [c2]"r"(c2)
144     : "memory"
145   );
146   return temp1;
147 }
148 
Select(uint32_t a,uint32_t b,uint32_t c)149 static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
150   int temp0, temp1, temp2, temp3, temp4, temp5;
151   __asm__ volatile (
152     "cmpgdu.lt.qb %[temp1], %[c],     %[b]             \n\t"
153     "pick.qb      %[temp1], %[b],     %[c]             \n\t"
154     "pick.qb      %[temp2], %[c],     %[b]             \n\t"
155     "cmpgdu.lt.qb %[temp4], %[c],     %[a]             \n\t"
156     "pick.qb      %[temp4], %[a],     %[c]             \n\t"
157     "pick.qb      %[temp5], %[c],     %[a]             \n\t"
158     "subu.qb      %[temp3], %[temp1], %[temp2]         \n\t"
159     "subu.qb      %[temp0], %[temp4], %[temp5]         \n\t"
160     "raddu.w.qb   %[temp3], %[temp3]                   \n\t"
161     "raddu.w.qb   %[temp0], %[temp0]                   \n\t"
162     "subu         %[temp3], %[temp3], %[temp0]         \n\t"
163     "slti         %[temp0], %[temp3], 0x1              \n\t"
164     "movz         %[a],     %[b],     %[temp0]         \n\t"
165     : [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), [temp3]"=&r"(temp3),
166       [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), [temp0]"=&r"(temp0),
167       [a]"+&r"(a)
168     : [b]"r"(b), [c]"r"(c)
169   );
170   return a;
171 }
172 
Average2(uint32_t a0,uint32_t a1)173 static WEBP_INLINE uint32_t Average2(uint32_t a0, uint32_t a1) {
174   __asm__ volatile (
175     "adduh.qb    %[a0], %[a0], %[a1]       \n\t"
176     : [a0]"+r"(a0)
177     : [a1]"r"(a1)
178   );
179   return a0;
180 }
181 
Average3(uint32_t a0,uint32_t a1,uint32_t a2)182 static WEBP_INLINE uint32_t Average3(uint32_t a0, uint32_t a1, uint32_t a2) {
183   return Average2(Average2(a0, a2), a1);
184 }
185 
Average4(uint32_t a0,uint32_t a1,uint32_t a2,uint32_t a3)186 static WEBP_INLINE uint32_t Average4(uint32_t a0, uint32_t a1,
187                                      uint32_t a2, uint32_t a3) {
188   return Average2(Average2(a0, a1), Average2(a2, a3));
189 }
190 
Predictor5_MIPSdspR2(uint32_t left,const uint32_t * const top)191 static uint32_t Predictor5_MIPSdspR2(uint32_t left, const uint32_t* const top) {
192   return Average3(left, top[0], top[1]);
193 }
194 
Predictor6_MIPSdspR2(uint32_t left,const uint32_t * const top)195 static uint32_t Predictor6_MIPSdspR2(uint32_t left, const uint32_t* const top) {
196   return Average2(left, top[-1]);
197 }
198 
Predictor7_MIPSdspR2(uint32_t left,const uint32_t * const top)199 static uint32_t Predictor7_MIPSdspR2(uint32_t left, const uint32_t* const top) {
200   return Average2(left, top[0]);
201 }
202 
Predictor8_MIPSdspR2(uint32_t left,const uint32_t * const top)203 static uint32_t Predictor8_MIPSdspR2(uint32_t left, const uint32_t* const top) {
204   (void)left;
205   return Average2(top[-1], top[0]);
206 }
207 
Predictor9_MIPSdspR2(uint32_t left,const uint32_t * const top)208 static uint32_t Predictor9_MIPSdspR2(uint32_t left, const uint32_t* const top) {
209   (void)left;
210   return Average2(top[0], top[1]);
211 }
212 
Predictor10_MIPSdspR2(uint32_t left,const uint32_t * const top)213 static uint32_t Predictor10_MIPSdspR2(uint32_t left,
214                                       const uint32_t* const top) {
215   return Average4(left, top[-1], top[0], top[1]);
216 }
217 
Predictor11_MIPSdspR2(uint32_t left,const uint32_t * const top)218 static uint32_t Predictor11_MIPSdspR2(uint32_t left,
219                                       const uint32_t* const top) {
220   return Select(top[0], left, top[-1]);
221 }
222 
Predictor12_MIPSdspR2(uint32_t left,const uint32_t * const top)223 static uint32_t Predictor12_MIPSdspR2(uint32_t left,
224                                       const uint32_t* const top) {
225   return ClampedAddSubtractFull(left, top[0], top[-1]);
226 }
227 
Predictor13_MIPSdspR2(uint32_t left,const uint32_t * const top)228 static uint32_t Predictor13_MIPSdspR2(uint32_t left,
229                                       const uint32_t* const top) {
230   return ClampedAddSubtractHalf(left, top[0], top[-1]);
231 }
232 
233 // Add green to blue and red channels (i.e. perform the inverse transform of
234 // 'subtract green').
AddGreenToBlueAndRed_MIPSdspR2(const uint32_t * src,int num_pixels,uint32_t * dst)235 static void AddGreenToBlueAndRed_MIPSdspR2(const uint32_t* src, int num_pixels,
236                                            uint32_t* dst) {
237   uint32_t temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
238   const uint32_t* const p_loop1_end = src + (num_pixels & ~3);
239   const uint32_t* const p_loop2_end = src + num_pixels;
240   __asm__ volatile (
241     ".set       push                                          \n\t"
242     ".set       noreorder                                     \n\t"
243     "beq        %[src],          %[p_loop1_end],     3f       \n\t"
244     " nop                                                     \n\t"
245   "0:                                                         \n\t"
246     "lw         %[temp0],        0(%[src])                    \n\t"
247     "lw         %[temp1],        4(%[src])                    \n\t"
248     "lw         %[temp2],        8(%[src])                    \n\t"
249     "lw         %[temp3],        12(%[src])                   \n\t"
250     "ext        %[temp4],        %[temp0],           8,    8  \n\t"
251     "ext        %[temp5],        %[temp1],           8,    8  \n\t"
252     "ext        %[temp6],        %[temp2],           8,    8  \n\t"
253     "ext        %[temp7],        %[temp3],           8,    8  \n\t"
254     "addiu      %[src],          %[src],             16       \n\t"
255     "addiu      %[dst],          %[dst],             16       \n\t"
256     "replv.ph   %[temp4],        %[temp4]                     \n\t"
257     "replv.ph   %[temp5],        %[temp5]                     \n\t"
258     "replv.ph   %[temp6],        %[temp6]                     \n\t"
259     "replv.ph   %[temp7],        %[temp7]                     \n\t"
260     "addu.qb    %[temp0],        %[temp0],           %[temp4] \n\t"
261     "addu.qb    %[temp1],        %[temp1],           %[temp5] \n\t"
262     "addu.qb    %[temp2],        %[temp2],           %[temp6] \n\t"
263     "addu.qb    %[temp3],        %[temp3],           %[temp7] \n\t"
264     "sw         %[temp0],        -16(%[dst])                  \n\t"
265     "sw         %[temp1],        -12(%[dst])                  \n\t"
266     "sw         %[temp2],        -8(%[dst])                   \n\t"
267     "bne        %[src],          %[p_loop1_end],     0b       \n\t"
268     " sw        %[temp3],        -4(%[dst])                   \n\t"
269   "3:                                                         \n\t"
270     "beq        %[src],          %[p_loop2_end],     2f       \n\t"
271     " nop                                                     \n\t"
272   "1:                                                         \n\t"
273     "lw         %[temp0],        0(%[src])                    \n\t"
274     "addiu      %[src],          %[src],             4        \n\t"
275     "addiu      %[dst],          %[dst],             4        \n\t"
276     "ext        %[temp4],        %[temp0],           8,    8  \n\t"
277     "replv.ph   %[temp4],        %[temp4]                     \n\t"
278     "addu.qb    %[temp0],        %[temp0],           %[temp4] \n\t"
279     "bne        %[src],          %[p_loop2_end],     1b       \n\t"
280     " sw        %[temp0],        -4(%[dst])                   \n\t"
281   "2:                                                         \n\t"
282     ".set       pop                                           \n\t"
283     : [dst]"+&r"(dst), [src]"+&r"(src), [temp0]"=&r"(temp0),
284       [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), [temp3]"=&r"(temp3),
285       [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), [temp6]"=&r"(temp6),
286       [temp7]"=&r"(temp7)
287     : [p_loop1_end]"r"(p_loop1_end), [p_loop2_end]"r"(p_loop2_end)
288     : "memory"
289   );
290 }
291 
TransformColorInverse_MIPSdspR2(const VP8LMultipliers * const m,const uint32_t * src,int num_pixels,uint32_t * dst)292 static void TransformColorInverse_MIPSdspR2(const VP8LMultipliers* const m,
293                                             const uint32_t* src, int num_pixels,
294                                             uint32_t* dst) {
295   int temp0, temp1, temp2, temp3, temp4, temp5;
296   uint32_t argb, argb1, new_red;
297   const uint32_t G_to_R = m->green_to_red_;
298   const uint32_t G_to_B = m->green_to_blue_;
299   const uint32_t R_to_B = m->red_to_blue_;
300   const uint32_t* const p_loop_end = src + (num_pixels & ~1);
301   __asm__ volatile (
302     ".set            push                                    \n\t"
303     ".set            noreorder                               \n\t"
304     "beq             %[src],       %[p_loop_end],  1f        \n\t"
305     " nop                                                    \n\t"
306     "replv.ph        %[temp0],     %[G_to_R]                 \n\t"
307     "replv.ph        %[temp1],     %[G_to_B]                 \n\t"
308     "replv.ph        %[temp2],     %[R_to_B]                 \n\t"
309     "shll.ph         %[temp0],     %[temp0],       8         \n\t"
310     "shll.ph         %[temp1],     %[temp1],       8         \n\t"
311     "shll.ph         %[temp2],     %[temp2],       8         \n\t"
312     "shra.ph         %[temp0],     %[temp0],       8         \n\t"
313     "shra.ph         %[temp1],     %[temp1],       8         \n\t"
314     "shra.ph         %[temp2],     %[temp2],       8         \n\t"
315   "0:                                                        \n\t"
316     "lw              %[argb],      0(%[src])                 \n\t"
317     "lw              %[argb1],     4(%[src])                 \n\t"
318     "sw              %[argb],      0(%[dst])                 \n\t"
319     "sw              %[argb1],     4(%[dst])                 \n\t"
320     "addiu           %[src],       %[src],         8         \n\t"
321     "addiu           %[dst],       %[dst],         8         \n\t"
322     "precrq.qb.ph    %[temp3],     %[argb],        %[argb1]  \n\t"
323     "preceu.ph.qbra  %[temp3],     %[temp3]                  \n\t"
324     "shll.ph         %[temp3],     %[temp3],       8         \n\t"
325     "shra.ph         %[temp3],     %[temp3],       8         \n\t"
326     "mul.ph          %[temp5],     %[temp3],       %[temp0]  \n\t"
327     "mul.ph          %[temp3],     %[temp3],       %[temp1]  \n\t"
328     "precrq.ph.w     %[new_red],   %[argb],        %[argb1]  \n\t"
329     "ins             %[argb1],     %[argb],        16,   16  \n\t"
330     "shra.ph         %[temp5],     %[temp5],       5         \n\t"
331     "shra.ph         %[temp3],     %[temp3],       5         \n\t"
332     "addu.ph         %[new_red],   %[new_red],     %[temp5]  \n\t"
333     "addu.ph         %[argb1],     %[argb1],       %[temp3]  \n\t"
334     "preceu.ph.qbra  %[temp5],     %[new_red]                \n\t"
335     "shll.ph         %[temp4],     %[temp5],       8         \n\t"
336     "shra.ph         %[temp4],     %[temp4],       8         \n\t"
337     "mul.ph          %[temp4],     %[temp4],       %[temp2]  \n\t"
338     "sb              %[temp5],     -2(%[dst])                \n\t"
339     "sra             %[temp5],     %[temp5],       16        \n\t"
340     "shra.ph         %[temp4],     %[temp4],       5         \n\t"
341     "addu.ph         %[argb1],     %[argb1],       %[temp4]  \n\t"
342     "preceu.ph.qbra  %[temp3],     %[argb1]                  \n\t"
343     "sb              %[temp5],     -6(%[dst])                \n\t"
344     "sb              %[temp3],     -4(%[dst])                \n\t"
345     "sra             %[temp3],     %[temp3],       16        \n\t"
346     "bne             %[src],       %[p_loop_end],  0b        \n\t"
347     " sb             %[temp3],     -8(%[dst])                \n\t"
348   "1:                                                        \n\t"
349     ".set            pop                                     \n\t"
350     : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
351       [temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
352       [new_red]"=&r"(new_red), [argb]"=&r"(argb),
353       [argb1]"=&r"(argb1), [dst]"+&r"(dst), [src]"+&r"(src)
354     : [G_to_R]"r"(G_to_R), [R_to_B]"r"(R_to_B),
355       [G_to_B]"r"(G_to_B), [p_loop_end]"r"(p_loop_end)
356     : "memory", "hi", "lo"
357   );
358 
359   // Fall-back to C-version for left-overs.
360   if (num_pixels & 1) VP8LTransformColorInverse_C(m, src, 1, dst);
361 }
362 
ConvertBGRAToRGB_MIPSdspR2(const uint32_t * src,int num_pixels,uint8_t * dst)363 static void ConvertBGRAToRGB_MIPSdspR2(const uint32_t* src,
364                                        int num_pixels, uint8_t* dst) {
365   int temp0, temp1, temp2, temp3;
366   const uint32_t* const p_loop1_end = src + (num_pixels & ~3);
367   const uint32_t* const p_loop2_end = src + num_pixels;
368   __asm__ volatile (
369     ".set       push                                       \n\t"
370     ".set       noreorder                                  \n\t"
371     "beq        %[src],      %[p_loop1_end],    3f         \n\t"
372     " nop                                                  \n\t"
373   "0:                                                      \n\t"
374     "lw         %[temp3],    12(%[src])                    \n\t"
375     "lw         %[temp2],    8(%[src])                     \n\t"
376     "lw         %[temp1],    4(%[src])                     \n\t"
377     "lw         %[temp0],    0(%[src])                     \n\t"
378     "ins        %[temp3],    %[temp2],          24,   8    \n\t"
379     "sll        %[temp2],    %[temp2],          8          \n\t"
380     "rotr       %[temp3],    %[temp3],          16         \n\t"
381     "ins        %[temp2],    %[temp1],          0,    16   \n\t"
382     "sll        %[temp1],    %[temp1],          8          \n\t"
383     "wsbh       %[temp3],    %[temp3]                      \n\t"
384     "balign     %[temp0],    %[temp1],          1          \n\t"
385     "wsbh       %[temp2],    %[temp2]                      \n\t"
386     "wsbh       %[temp0],    %[temp0]                      \n\t"
387     "usw        %[temp3],    8(%[dst])                     \n\t"
388     "rotr       %[temp0],    %[temp0],          16         \n\t"
389     "usw        %[temp2],    4(%[dst])                     \n\t"
390     "addiu      %[src],      %[src],            16         \n\t"
391     "usw        %[temp0],    0(%[dst])                     \n\t"
392     "bne        %[src],      %[p_loop1_end],    0b         \n\t"
393     " addiu     %[dst],      %[dst],            12         \n\t"
394   "3:                                                      \n\t"
395     "beq        %[src],      %[p_loop2_end],    2f         \n\t"
396     " nop                                                  \n\t"
397   "1:                                                      \n\t"
398     "lw         %[temp0],    0(%[src])                     \n\t"
399     "addiu      %[src],      %[src],            4          \n\t"
400     "wsbh       %[temp1],    %[temp0]                      \n\t"
401     "addiu      %[dst],      %[dst],            3          \n\t"
402     "ush        %[temp1],    -2(%[dst])                    \n\t"
403     "sra        %[temp0],    %[temp0],          16         \n\t"
404     "bne        %[src],      %[p_loop2_end],    1b         \n\t"
405     " sb        %[temp0],    -3(%[dst])                    \n\t"
406   "2:                                                      \n\t"
407     ".set       pop                                        \n\t"
408     : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
409       [temp3]"=&r"(temp3), [dst]"+&r"(dst), [src]"+&r"(src)
410     : [p_loop1_end]"r"(p_loop1_end), [p_loop2_end]"r"(p_loop2_end)
411     : "memory"
412   );
413 }
414 
ConvertBGRAToRGBA_MIPSdspR2(const uint32_t * src,int num_pixels,uint8_t * dst)415 static void ConvertBGRAToRGBA_MIPSdspR2(const uint32_t* src,
416                                         int num_pixels, uint8_t* dst) {
417   int temp0, temp1, temp2, temp3;
418   const uint32_t* const p_loop1_end = src + (num_pixels & ~3);
419   const uint32_t* const p_loop2_end = src + num_pixels;
420   __asm__ volatile (
421     ".set       push                                       \n\t"
422     ".set       noreorder                                  \n\t"
423     "beq        %[src],      %[p_loop1_end],    3f         \n\t"
424     " nop                                                  \n\t"
425   "0:                                                      \n\t"
426     "lw         %[temp0],    0(%[src])                     \n\t"
427     "lw         %[temp1],    4(%[src])                     \n\t"
428     "lw         %[temp2],    8(%[src])                     \n\t"
429     "lw         %[temp3],    12(%[src])                    \n\t"
430     "wsbh       %[temp0],    %[temp0]                      \n\t"
431     "wsbh       %[temp1],    %[temp1]                      \n\t"
432     "wsbh       %[temp2],    %[temp2]                      \n\t"
433     "wsbh       %[temp3],    %[temp3]                      \n\t"
434     "addiu      %[src],      %[src],            16         \n\t"
435     "balign     %[temp0],    %[temp0],          1          \n\t"
436     "balign     %[temp1],    %[temp1],          1          \n\t"
437     "balign     %[temp2],    %[temp2],          1          \n\t"
438     "balign     %[temp3],    %[temp3],          1          \n\t"
439     "usw        %[temp0],    0(%[dst])                     \n\t"
440     "usw        %[temp1],    4(%[dst])                     \n\t"
441     "usw        %[temp2],    8(%[dst])                     \n\t"
442     "usw        %[temp3],    12(%[dst])                    \n\t"
443     "bne        %[src],      %[p_loop1_end],    0b         \n\t"
444     " addiu     %[dst],      %[dst],            16         \n\t"
445   "3:                                                      \n\t"
446     "beq        %[src],      %[p_loop2_end],    2f         \n\t"
447     " nop                                                  \n\t"
448   "1:                                                      \n\t"
449     "lw         %[temp0],    0(%[src])                     \n\t"
450     "wsbh       %[temp0],    %[temp0]                      \n\t"
451     "addiu      %[src],      %[src],            4          \n\t"
452     "balign     %[temp0],    %[temp0],          1          \n\t"
453     "usw        %[temp0],    0(%[dst])                     \n\t"
454     "bne        %[src],      %[p_loop2_end],    1b         \n\t"
455     " addiu     %[dst],      %[dst],            4          \n\t"
456   "2:                                                      \n\t"
457     ".set       pop                                        \n\t"
458     : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
459       [temp3]"=&r"(temp3), [dst]"+&r"(dst), [src]"+&r"(src)
460     : [p_loop1_end]"r"(p_loop1_end), [p_loop2_end]"r"(p_loop2_end)
461     : "memory"
462   );
463 }
464 
ConvertBGRAToRGBA4444_MIPSdspR2(const uint32_t * src,int num_pixels,uint8_t * dst)465 static void ConvertBGRAToRGBA4444_MIPSdspR2(const uint32_t* src,
466                                             int num_pixels, uint8_t* dst) {
467   int temp0, temp1, temp2, temp3, temp4, temp5;
468   const uint32_t* const p_loop1_end = src + (num_pixels & ~3);
469   const uint32_t* const p_loop2_end = src + num_pixels;
470   __asm__ volatile (
471     ".set           push                                       \n\t"
472     ".set           noreorder                                  \n\t"
473     "beq            %[src],      %[p_loop1_end],    3f         \n\t"
474     " nop                                                      \n\t"
475   "0:                                                          \n\t"
476     "lw             %[temp0],    0(%[src])                     \n\t"
477     "lw             %[temp1],    4(%[src])                     \n\t"
478     "lw             %[temp2],    8(%[src])                     \n\t"
479     "lw             %[temp3],    12(%[src])                    \n\t"
480     "ext            %[temp4],    %[temp0],          28,   4    \n\t"
481     "ext            %[temp5],    %[temp0],          12,   4    \n\t"
482     "ins            %[temp0],    %[temp4],          0,    4    \n\t"
483     "ext            %[temp4],    %[temp1],          28,   4    \n\t"
484     "ins            %[temp0],    %[temp5],          16,   4    \n\t"
485     "ext            %[temp5],    %[temp1],          12,   4    \n\t"
486     "ins            %[temp1],    %[temp4],          0,    4    \n\t"
487     "ext            %[temp4],    %[temp2],          28,   4    \n\t"
488     "ins            %[temp1],    %[temp5],          16,   4    \n\t"
489     "ext            %[temp5],    %[temp2],          12,   4    \n\t"
490     "ins            %[temp2],    %[temp4],          0,    4    \n\t"
491     "ext            %[temp4],    %[temp3],          28,   4    \n\t"
492     "ins            %[temp2],    %[temp5],          16,   4    \n\t"
493     "ext            %[temp5],    %[temp3],          12,   4    \n\t"
494     "ins            %[temp3],    %[temp4],          0,    4    \n\t"
495     "precr.qb.ph    %[temp1],    %[temp1],          %[temp0]   \n\t"
496     "ins            %[temp3],    %[temp5],          16,   4    \n\t"
497     "addiu          %[src],      %[src],            16         \n\t"
498     "precr.qb.ph    %[temp3],    %[temp3],          %[temp2]   \n\t"
499 #if (WEBP_SWAP_16BIT_CSP == 1)
500     "usw            %[temp1],    0(%[dst])                     \n\t"
501     "usw            %[temp3],    4(%[dst])                     \n\t"
502 #else
503     "wsbh           %[temp1],    %[temp1]                      \n\t"
504     "wsbh           %[temp3],    %[temp3]                      \n\t"
505     "usw            %[temp1],    0(%[dst])                     \n\t"
506     "usw            %[temp3],    4(%[dst])                     \n\t"
507 #endif
508     "bne            %[src],      %[p_loop1_end],    0b         \n\t"
509     " addiu         %[dst],      %[dst],            8          \n\t"
510   "3:                                                          \n\t"
511     "beq            %[src],      %[p_loop2_end],    2f         \n\t"
512     " nop                                                      \n\t"
513   "1:                                                          \n\t"
514     "lw             %[temp0],    0(%[src])                     \n\t"
515     "ext            %[temp4],    %[temp0],          28,   4    \n\t"
516     "ext            %[temp5],    %[temp0],          12,   4    \n\t"
517     "ins            %[temp0],    %[temp4],          0,    4    \n\t"
518     "ins            %[temp0],    %[temp5],          16,   4    \n\t"
519     "addiu          %[src],      %[src],            4          \n\t"
520     "precr.qb.ph    %[temp0],    %[temp0],          %[temp0]   \n\t"
521 #if (WEBP_SWAP_16BIT_CSP == 1)
522     "ush            %[temp0],    0(%[dst])                     \n\t"
523 #else
524     "wsbh           %[temp0],    %[temp0]                      \n\t"
525     "ush            %[temp0],    0(%[dst])                     \n\t"
526 #endif
527     "bne            %[src],      %[p_loop2_end],    1b         \n\t"
528     " addiu         %[dst],      %[dst],            2          \n\t"
529   "2:                                                          \n\t"
530     ".set           pop                                        \n\t"
531     : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
532       [temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
533       [dst]"+&r"(dst), [src]"+&r"(src)
534     : [p_loop1_end]"r"(p_loop1_end), [p_loop2_end]"r"(p_loop2_end)
535     : "memory"
536   );
537 }
538 
ConvertBGRAToRGB565_MIPSdspR2(const uint32_t * src,int num_pixels,uint8_t * dst)539 static void ConvertBGRAToRGB565_MIPSdspR2(const uint32_t* src,
540                                           int num_pixels, uint8_t* dst) {
541   int temp0, temp1, temp2, temp3, temp4, temp5;
542   const uint32_t* const p_loop1_end = src + (num_pixels & ~3);
543   const uint32_t* const p_loop2_end = src + num_pixels;
544   __asm__ volatile (
545     ".set           push                                       \n\t"
546     ".set           noreorder                                  \n\t"
547     "beq            %[src],      %[p_loop1_end],    3f         \n\t"
548     " nop                                                      \n\t"
549   "0:                                                          \n\t"
550     "lw             %[temp0],    0(%[src])                     \n\t"
551     "lw             %[temp1],    4(%[src])                     \n\t"
552     "lw             %[temp2],    8(%[src])                     \n\t"
553     "lw             %[temp3],    12(%[src])                    \n\t"
554     "ext            %[temp4],    %[temp0],          8,    16   \n\t"
555     "ext            %[temp5],    %[temp0],          5,    11   \n\t"
556     "ext            %[temp0],    %[temp0],          3,    5    \n\t"
557     "ins            %[temp4],    %[temp5],          0,    11   \n\t"
558     "ext            %[temp5],    %[temp1],          5,    11   \n\t"
559     "ins            %[temp4],    %[temp0],          0,    5    \n\t"
560     "ext            %[temp0],    %[temp1],          8,    16   \n\t"
561     "ext            %[temp1],    %[temp1],          3,    5    \n\t"
562     "ins            %[temp0],    %[temp5],          0,    11   \n\t"
563     "ext            %[temp5],    %[temp2],          5,    11   \n\t"
564     "ins            %[temp0],    %[temp1],          0,    5    \n\t"
565     "ext            %[temp1],    %[temp2],          8,    16   \n\t"
566     "ext            %[temp2],    %[temp2],          3,    5    \n\t"
567     "ins            %[temp1],    %[temp5],          0,    11   \n\t"
568     "ext            %[temp5],    %[temp3],          5,    11   \n\t"
569     "ins            %[temp1],    %[temp2],          0,    5    \n\t"
570     "ext            %[temp2],    %[temp3],          8,    16   \n\t"
571     "ext            %[temp3],    %[temp3],          3,    5    \n\t"
572     "ins            %[temp2],    %[temp5],          0,    11   \n\t"
573     "append         %[temp0],    %[temp4],          16         \n\t"
574     "ins            %[temp2],    %[temp3],          0,    5    \n\t"
575     "addiu          %[src],      %[src],            16         \n\t"
576     "append         %[temp2],    %[temp1],          16         \n\t"
577 #if (WEBP_SWAP_16BIT_CSP == 1)
578     "usw            %[temp0],    0(%[dst])                     \n\t"
579     "usw            %[temp2],    4(%[dst])                     \n\t"
580 #else
581     "wsbh           %[temp0],    %[temp0]                      \n\t"
582     "wsbh           %[temp2],    %[temp2]                      \n\t"
583     "usw            %[temp0],    0(%[dst])                     \n\t"
584     "usw            %[temp2],    4(%[dst])                     \n\t"
585 #endif
586     "bne            %[src],      %[p_loop1_end],    0b         \n\t"
587     " addiu         %[dst],      %[dst],            8          \n\t"
588   "3:                                                          \n\t"
589     "beq            %[src],      %[p_loop2_end],    2f         \n\t"
590     " nop                                                      \n\t"
591   "1:                                                          \n\t"
592     "lw             %[temp0],    0(%[src])                     \n\t"
593     "ext            %[temp4],    %[temp0],          8,    16   \n\t"
594     "ext            %[temp5],    %[temp0],          5,    11   \n\t"
595     "ext            %[temp0],    %[temp0],          3,    5    \n\t"
596     "ins            %[temp4],    %[temp5],          0,    11   \n\t"
597     "addiu          %[src],      %[src],            4          \n\t"
598     "ins            %[temp4],    %[temp0],          0,    5    \n\t"
599 #if (WEBP_SWAP_16BIT_CSP == 1)
600     "ush            %[temp4],    0(%[dst])                     \n\t"
601 #else
602     "wsbh           %[temp4],    %[temp4]                      \n\t"
603     "ush            %[temp4],    0(%[dst])                     \n\t"
604 #endif
605     "bne            %[src],      %[p_loop2_end],    1b         \n\t"
606     " addiu         %[dst],      %[dst],            2          \n\t"
607   "2:                                                          \n\t"
608     ".set           pop                                        \n\t"
609     : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
610       [temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
611       [dst]"+&r"(dst), [src]"+&r"(src)
612     : [p_loop1_end]"r"(p_loop1_end), [p_loop2_end]"r"(p_loop2_end)
613     : "memory"
614   );
615 }
616 
ConvertBGRAToBGR_MIPSdspR2(const uint32_t * src,int num_pixels,uint8_t * dst)617 static void ConvertBGRAToBGR_MIPSdspR2(const uint32_t* src,
618                                        int num_pixels, uint8_t* dst) {
619   int temp0, temp1, temp2, temp3;
620   const uint32_t* const p_loop1_end = src + (num_pixels & ~3);
621   const uint32_t* const p_loop2_end = src + num_pixels;
622   __asm__ volatile (
623     ".set       push                                         \n\t"
624     ".set       noreorder                                    \n\t"
625     "beq        %[src],      %[p_loop1_end],    3f           \n\t"
626     " nop                                                    \n\t"
627   "0:                                                        \n\t"
628     "lw         %[temp0],    0(%[src])                       \n\t"
629     "lw         %[temp1],    4(%[src])                       \n\t"
630     "lw         %[temp2],    8(%[src])                       \n\t"
631     "lw         %[temp3],    12(%[src])                      \n\t"
632     "ins        %[temp0],    %[temp1],          24,    8     \n\t"
633     "sra        %[temp1],    %[temp1],          8            \n\t"
634     "ins        %[temp1],    %[temp2],          16,    16    \n\t"
635     "sll        %[temp2],    %[temp2],          8            \n\t"
636     "balign     %[temp3],    %[temp2],          1            \n\t"
637     "addiu      %[src],      %[src],            16           \n\t"
638     "usw        %[temp0],    0(%[dst])                       \n\t"
639     "usw        %[temp1],    4(%[dst])                       \n\t"
640     "usw        %[temp3],    8(%[dst])                       \n\t"
641     "bne        %[src],      %[p_loop1_end],    0b           \n\t"
642     " addiu     %[dst],      %[dst],            12           \n\t"
643   "3:                                                        \n\t"
644     "beq        %[src],      %[p_loop2_end],    2f           \n\t"
645     " nop                                                    \n\t"
646   "1:                                                        \n\t"
647     "lw         %[temp0],    0(%[src])                       \n\t"
648     "addiu      %[src],      %[src],            4            \n\t"
649     "addiu      %[dst],      %[dst],            3            \n\t"
650     "ush        %[temp0],    -3(%[dst])                      \n\t"
651     "sra        %[temp0],    %[temp0],          16           \n\t"
652     "bne        %[src],      %[p_loop2_end],    1b           \n\t"
653     " sb        %[temp0],    -1(%[dst])                      \n\t"
654   "2:                                                        \n\t"
655     ".set       pop                                          \n\t"
656     : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
657       [temp3]"=&r"(temp3), [dst]"+&r"(dst), [src]"+&r"(src)
658     : [p_loop1_end]"r"(p_loop1_end), [p_loop2_end]"r"(p_loop2_end)
659     : "memory"
660   );
661 }
662 
663 //------------------------------------------------------------------------------
664 // Entry point
665 
666 extern void VP8LDspInitMIPSdspR2(void);
667 
VP8LDspInitMIPSdspR2(void)668 WEBP_TSAN_IGNORE_FUNCTION void VP8LDspInitMIPSdspR2(void) {
669   VP8LMapColor32b = MapARGB_MIPSdspR2;
670   VP8LMapColor8b = MapAlpha_MIPSdspR2;
671 
672   VP8LPredictors[5] = Predictor5_MIPSdspR2;
673   VP8LPredictors[6] = Predictor6_MIPSdspR2;
674   VP8LPredictors[7] = Predictor7_MIPSdspR2;
675   VP8LPredictors[8] = Predictor8_MIPSdspR2;
676   VP8LPredictors[9] = Predictor9_MIPSdspR2;
677   VP8LPredictors[10] = Predictor10_MIPSdspR2;
678   VP8LPredictors[11] = Predictor11_MIPSdspR2;
679   VP8LPredictors[12] = Predictor12_MIPSdspR2;
680   VP8LPredictors[13] = Predictor13_MIPSdspR2;
681 
682   VP8LAddGreenToBlueAndRed = AddGreenToBlueAndRed_MIPSdspR2;
683   VP8LTransformColorInverse = TransformColorInverse_MIPSdspR2;
684 
685   VP8LConvertBGRAToRGB = ConvertBGRAToRGB_MIPSdspR2;
686   VP8LConvertBGRAToRGBA = ConvertBGRAToRGBA_MIPSdspR2;
687   VP8LConvertBGRAToRGBA4444 = ConvertBGRAToRGBA4444_MIPSdspR2;
688   VP8LConvertBGRAToRGB565 = ConvertBGRAToRGB565_MIPSdspR2;
689   VP8LConvertBGRAToBGR = ConvertBGRAToBGR_MIPSdspR2;
690 }
691 
692 #else  // !WEBP_USE_MIPS_DSP_R2
693 
694 WEBP_DSP_INIT_STUB(VP8LDspInitMIPSdspR2)
695 
696 #endif  // WEBP_USE_MIPS_DSP_R2
697