| 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 "./dsp.h" |
| 16 | |
| 17 | #if defined(WEBP_USE_MIPS_DSP_R2) |
| 18 | |
| 19 | static int DispatchAlpha(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(uint32_t* const ptr, int width, int inverse) { |
| 83 | int x; |
| 84 | const uint32_t c_00ffffff = 0x00ffffffu; |
| 85 | const uint32_t c_ff000000 = 0xff000000u; |
| 86 | const uint32_t c_8000000 = 0x00800000u; |
| 87 | const uint32_t c_8000080 = 0x00800080u; |
| 88 | for (x = 0; x < width; ++x) { |
| 89 | const uint32_t argb = ptr[x]; |
| 90 | if (argb < 0xff000000u) { // alpha < 255 |
| 91 | if (argb <= 0x00ffffffu) { // alpha == 0 |
| 92 | ptr[x] = 0; |
| 93 | } else { |
| 94 | int temp0, temp1, temp2, temp3, alpha; |
| 95 | __asm__ volatile ( |
| 96 | "srl %[alpha], %[argb], 24 \n\t" |
| 97 | "replv.qb %[temp0], %[alpha] \n\t" |
| 98 | "and %[temp0], %[temp0], %[c_00ffffff] \n\t" |
| 99 | "beqz %[inverse], 0f \n\t" |
| 100 | "divu $zero, %[c_ff000000], %[alpha] \n\t" |
| 101 | "mflo %[temp0] \n\t" |
| 102 | "0: \n\t" |
| 103 | "andi %[temp1], %[argb], 0xff \n\t" |
| 104 | "ext %[temp2], %[argb], 8, 8 \n\t" |
| 105 | "ext %[temp3], %[argb], 16, 8 \n\t" |
| 106 | "mul %[temp1], %[temp1], %[temp0] \n\t" |
| 107 | "mul %[temp2], %[temp2], %[temp0] \n\t" |
| 108 | "mul %[temp3], %[temp3], %[temp0] \n\t" |
| 109 | "precrq.ph.w %[temp1], %[temp2], %[temp1] \n\t" |
| 110 | "addu %[temp3], %[temp3], %[c_8000000] \n\t" |
| 111 | "addu %[temp1], %[temp1], %[c_8000080] \n\t" |
| 112 | "precrq.ph.w %[temp3], %[argb], %[temp3] \n\t" |
| 113 | "precrq.qb.ph %[temp1], %[temp3], %[temp1] \n\t" |
| 114 | : [temp0]"=&r" (temp0), [temp1]"=&r" (temp1), [temp2]"=&r" (temp2), |
| 115 | [temp3]"=&r" (temp3), [alpha]"=&r" (alpha) |
| 116 | : [inverse]"r" (inverse), [c_00ffffff]"r" (c_00ffffff), |
| 117 | [c_8000000]"r" (c_8000000), [c_8000080]"r" (c_8000080), |
| 118 | [c_ff000000]"r" (c_ff000000), [argb]"r" (argb) |
| 119 | : "memory" , "hi" , "lo" |
| 120 | ); |
| 121 | ptr[x] = temp1; |
| 122 | } |
| 123 | } |
| 124 | } |
| 125 | } |
| 126 | |
| 127 | //------------------------------------------------------------------------------ |
| 128 | // Entry point |
| 129 | |
| 130 | extern void WebPInitAlphaProcessingMIPSdspR2(void); |
| 131 | |
| 132 | WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingMIPSdspR2(void) { |
| 133 | WebPDispatchAlpha = DispatchAlpha; |
| 134 | WebPMultARGBRow = MultARGBRow; |
| 135 | } |
| 136 | |
| 137 | #else // !WEBP_USE_MIPS_DSP_R2 |
| 138 | |
| 139 | WEBP_DSP_INIT_STUB(WebPInitAlphaProcessingMIPSdspR2) |
| 140 | |
| 141 | #endif // WEBP_USE_MIPS_DSP_R2 |
| 142 | |