alpha_processing_mips_dsp_r2.c (10249B)
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 // Utilities for processing transparent channel. 11 // 12 // Author(s): Branimir Vasic (branimir.vasic@imgtec.com) 13 // Djordje Pesut (djordje.pesut@imgtec.com) 14 15 #include "src/dsp/dsp.h" 16 17 #if defined(WEBP_USE_MIPS_DSP_R2) 18 19 static int DispatchAlpha_MIPSdspR2(const uint8_t* alpha, int alpha_stride, 20 int width, int height, 21 uint8_t* dst, int dst_stride) { 22 uint32_t alpha_mask = 0xffffffff; 23 int i, j, temp0; 24 25 for (j = 0; j < height; ++j) { 26 uint8_t* pdst = dst; 27 const uint8_t* palpha = alpha; 28 for (i = 0; i < (width >> 2); ++i) { 29 int temp1, temp2, temp3; 30 31 __asm__ volatile ( 32 "ulw %[temp0], 0(%[palpha]) \n\t" 33 "addiu %[palpha], %[palpha], 4 \n\t" 34 "addiu %[pdst], %[pdst], 16 \n\t" 35 "srl %[temp1], %[temp0], 8 \n\t" 36 "srl %[temp2], %[temp0], 16 \n\t" 37 "srl %[temp3], %[temp0], 24 \n\t" 38 "and %[alpha_mask], %[alpha_mask], %[temp0] \n\t" 39 "sb %[temp0], -16(%[pdst]) \n\t" 40 "sb %[temp1], -12(%[pdst]) \n\t" 41 "sb %[temp2], -8(%[pdst]) \n\t" 42 "sb %[temp3], -4(%[pdst]) \n\t" 43 : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), 44 [temp3]"=&r"(temp3), [palpha]"+r"(palpha), [pdst]"+r"(pdst), 45 [alpha_mask]"+r"(alpha_mask) 46 : 47 : "memory" 48 ); 49 } 50 51 for (i = 0; i < (width & 3); ++i) { 52 __asm__ volatile ( 53 "lbu %[temp0], 0(%[palpha]) \n\t" 54 "addiu %[palpha], %[palpha], 1 \n\t" 55 "sb %[temp0], 0(%[pdst]) \n\t" 56 "and %[alpha_mask], %[alpha_mask], %[temp0] \n\t" 57 "addiu %[pdst], %[pdst], 4 \n\t" 58 : [temp0]"=&r"(temp0), [palpha]"+r"(palpha), [pdst]"+r"(pdst), 59 [alpha_mask]"+r"(alpha_mask) 60 : 61 : "memory" 62 ); 63 } 64 alpha += alpha_stride; 65 dst += dst_stride; 66 } 67 68 __asm__ volatile ( 69 "ext %[temp0], %[alpha_mask], 0, 16 \n\t" 70 "srl %[alpha_mask], %[alpha_mask], 16 \n\t" 71 "and %[alpha_mask], %[alpha_mask], %[temp0] \n\t" 72 "ext %[temp0], %[alpha_mask], 0, 8 \n\t" 73 "srl %[alpha_mask], %[alpha_mask], 8 \n\t" 74 "and %[alpha_mask], %[alpha_mask], %[temp0] \n\t" 75 : [temp0]"=&r"(temp0), [alpha_mask]"+r"(alpha_mask) 76 : 77 ); 78 79 return (alpha_mask != 0xff); 80 } 81 82 static void MultARGBRow_MIPSdspR2(uint32_t* const ptr, int width, 83 int inverse) { 84 int x; 85 const uint32_t c_00ffffff = 0x00ffffffu; 86 const uint32_t c_ff000000 = 0xff000000u; 87 const uint32_t c_8000000 = 0x00800000u; 88 const uint32_t c_8000080 = 0x00800080u; 89 for (x = 0; x < width; ++x) { 90 const uint32_t argb = ptr[x]; 91 if (argb < 0xff000000u) { // alpha < 255 92 if (argb <= 0x00ffffffu) { // alpha == 0 93 ptr[x] = 0; 94 } else { 95 int temp0, temp1, temp2, temp3, alpha; 96 __asm__ volatile ( 97 "srl %[alpha], %[argb], 24 \n\t" 98 "replv.qb %[temp0], %[alpha] \n\t" 99 "and %[temp0], %[temp0], %[c_00ffffff] \n\t" 100 "beqz %[inverse], 0f \n\t" 101 "divu $zero, %[c_ff000000], %[alpha] \n\t" 102 "mflo %[temp0] \n\t" 103 "0: \n\t" 104 "andi %[temp1], %[argb], 0xff \n\t" 105 "ext %[temp2], %[argb], 8, 8 \n\t" 106 "ext %[temp3], %[argb], 16, 8 \n\t" 107 "mul %[temp1], %[temp1], %[temp0] \n\t" 108 "mul %[temp2], %[temp2], %[temp0] \n\t" 109 "mul %[temp3], %[temp3], %[temp0] \n\t" 110 "precrq.ph.w %[temp1], %[temp2], %[temp1] \n\t" 111 "addu %[temp3], %[temp3], %[c_8000000] \n\t" 112 "addu %[temp1], %[temp1], %[c_8000080] \n\t" 113 "precrq.ph.w %[temp3], %[argb], %[temp3] \n\t" 114 "precrq.qb.ph %[temp1], %[temp3], %[temp1] \n\t" 115 : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), 116 [temp3]"=&r"(temp3), [alpha]"=&r"(alpha) 117 : [inverse]"r"(inverse), [c_00ffffff]"r"(c_00ffffff), 118 [c_8000000]"r"(c_8000000), [c_8000080]"r"(c_8000080), 119 [c_ff000000]"r"(c_ff000000), [argb]"r"(argb) 120 : "memory", "hi", "lo" 121 ); 122 ptr[x] = temp1; 123 } 124 } 125 } 126 } 127 128 #ifdef WORDS_BIGENDIAN 129 static void PackARGB_MIPSdspR2(const uint8_t* a, const uint8_t* r, 130 const uint8_t* g, const uint8_t* b, int len, 131 uint32_t* out) { 132 int temp0, temp1, temp2, temp3, offset; 133 const int rest = len & 1; 134 const uint32_t* const loop_end = out + len - rest; 135 const int step = 4; 136 __asm__ volatile ( 137 "xor %[offset], %[offset], %[offset] \n\t" 138 "beq %[loop_end], %[out], 0f \n\t" 139 "2: \n\t" 140 "lbux %[temp0], %[offset](%[a]) \n\t" 141 "lbux %[temp1], %[offset](%[r]) \n\t" 142 "lbux %[temp2], %[offset](%[g]) \n\t" 143 "lbux %[temp3], %[offset](%[b]) \n\t" 144 "ins %[temp1], %[temp0], 16, 16 \n\t" 145 "ins %[temp3], %[temp2], 16, 16 \n\t" 146 "addiu %[out], %[out], 4 \n\t" 147 "precr.qb.ph %[temp0], %[temp1], %[temp3] \n\t" 148 "sw %[temp0], -4(%[out]) \n\t" 149 "addu %[offset], %[offset], %[step] \n\t" 150 "bne %[loop_end], %[out], 2b \n\t" 151 "0: \n\t" 152 "beq %[rest], $zero, 1f \n\t" 153 "lbux %[temp0], %[offset](%[a]) \n\t" 154 "lbux %[temp1], %[offset](%[r]) \n\t" 155 "lbux %[temp2], %[offset](%[g]) \n\t" 156 "lbux %[temp3], %[offset](%[b]) \n\t" 157 "ins %[temp1], %[temp0], 16, 16 \n\t" 158 "ins %[temp3], %[temp2], 16, 16 \n\t" 159 "precr.qb.ph %[temp0], %[temp1], %[temp3] \n\t" 160 "sw %[temp0], 0(%[out]) \n\t" 161 "1: \n\t" 162 : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), 163 [temp3]"=&r"(temp3), [offset]"=&r"(offset), [out]"+&r"(out) 164 : [a]"r"(a), [r]"r"(r), [g]"r"(g), [b]"r"(b), [step]"r"(step), 165 [loop_end]"r"(loop_end), [rest]"r"(rest) 166 : "memory" 167 ); 168 } 169 #endif // WORDS_BIGENDIAN 170 171 static void PackRGB_MIPSdspR2(const uint8_t* r, const uint8_t* g, 172 const uint8_t* b, int len, int step, 173 uint32_t* out) { 174 int temp0, temp1, temp2, offset; 175 const int rest = len & 1; 176 const int a = 0xff; 177 const uint32_t* const loop_end = out + len - rest; 178 __asm__ volatile ( 179 "xor %[offset], %[offset], %[offset] \n\t" 180 "beq %[loop_end], %[out], 0f \n\t" 181 "2: \n\t" 182 "lbux %[temp0], %[offset](%[r]) \n\t" 183 "lbux %[temp1], %[offset](%[g]) \n\t" 184 "lbux %[temp2], %[offset](%[b]) \n\t" 185 "ins %[temp0], %[a], 16, 16 \n\t" 186 "ins %[temp2], %[temp1], 16, 16 \n\t" 187 "addiu %[out], %[out], 4 \n\t" 188 "precr.qb.ph %[temp0], %[temp0], %[temp2] \n\t" 189 "sw %[temp0], -4(%[out]) \n\t" 190 "addu %[offset], %[offset], %[step] \n\t" 191 "bne %[loop_end], %[out], 2b \n\t" 192 "0: \n\t" 193 "beq %[rest], $zero, 1f \n\t" 194 "lbux %[temp0], %[offset](%[r]) \n\t" 195 "lbux %[temp1], %[offset](%[g]) \n\t" 196 "lbux %[temp2], %[offset](%[b]) \n\t" 197 "ins %[temp0], %[a], 16, 16 \n\t" 198 "ins %[temp2], %[temp1], 16, 16 \n\t" 199 "precr.qb.ph %[temp0], %[temp0], %[temp2] \n\t" 200 "sw %[temp0], 0(%[out]) \n\t" 201 "1: \n\t" 202 : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), 203 [offset]"=&r"(offset), [out]"+&r"(out) 204 : [a]"r"(a), [r]"r"(r), [g]"r"(g), [b]"r"(b), [step]"r"(step), 205 [loop_end]"r"(loop_end), [rest]"r"(rest) 206 : "memory" 207 ); 208 } 209 210 //------------------------------------------------------------------------------ 211 // Entry point 212 213 extern void WebPInitAlphaProcessingMIPSdspR2(void); 214 215 WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingMIPSdspR2(void) { 216 WebPDispatchAlpha = DispatchAlpha_MIPSdspR2; 217 WebPMultARGBRow = MultARGBRow_MIPSdspR2; 218 #ifdef WORDS_BIGENDIAN 219 WebPPackARGB = PackARGB_MIPSdspR2; 220 #endif 221 WebPPackRGB = PackRGB_MIPSdspR2; 222 } 223 224 #else // !WEBP_USE_MIPS_DSP_R2 225 226 WEBP_DSP_INIT_STUB(WebPInitAlphaProcessingMIPSdspR2) 227 228 #endif // WEBP_USE_MIPS_DSP_R2