michael@0: /* michael@0: * Copyright 2011 The LibYuv Project Authors. All rights reserved. michael@0: * michael@0: * Use of this source code is governed by a BSD-style license michael@0: * that can be found in the LICENSE file in the root of the source michael@0: * tree. An additional intellectual property rights grant can be found michael@0: * in the file PATENTS. All contributing project authors may michael@0: * be found in the AUTHORS file in the root of the source tree. michael@0: */ michael@0: michael@0: #include "libyuv/row.h" michael@0: michael@0: #include // For memcpy and memset. michael@0: michael@0: #include "libyuv/basic_types.h" michael@0: michael@0: #ifdef __cplusplus michael@0: namespace libyuv { michael@0: extern "C" { michael@0: #endif michael@0: michael@0: // llvm x86 is poor at ternary operator, so use branchless min/max. michael@0: michael@0: #define USE_BRANCHLESS 1 michael@0: #if USE_BRANCHLESS michael@0: static __inline int32 clamp0(int32 v) { michael@0: return ((-(v) >> 31) & (v)); michael@0: } michael@0: michael@0: static __inline int32 clamp255(int32 v) { michael@0: return (((255 - (v)) >> 31) | (v)) & 255; michael@0: } michael@0: michael@0: static __inline uint32 Clamp(int32 val) { michael@0: int v = clamp0(val); michael@0: return (uint32)(clamp255(v)); michael@0: } michael@0: michael@0: static __inline uint32 Abs(int32 v) { michael@0: int m = v >> 31; michael@0: return (v + m) ^ m; michael@0: } michael@0: #else // USE_BRANCHLESS michael@0: static __inline int32 clamp0(int32 v) { michael@0: return (v < 0) ? 0 : v; michael@0: } michael@0: michael@0: static __inline int32 clamp255(int32 v) { michael@0: return (v > 255) ? 255 : v; michael@0: } michael@0: michael@0: static __inline uint32 Clamp(int32 val) { michael@0: int v = clamp0(val); michael@0: return (uint32)(clamp255(v)); michael@0: } michael@0: michael@0: static __inline uint32 Abs(int32 v) { michael@0: return (v < 0) ? -v : v; michael@0: } michael@0: #endif // USE_BRANCHLESS michael@0: michael@0: #ifdef LIBYUV_LITTLE_ENDIAN michael@0: #define WRITEWORD(p, v) *(uint32*)(p) = v michael@0: #else michael@0: static inline void WRITEWORD(uint8* p, uint32 v) { michael@0: p[0] = (uint8)(v & 255); michael@0: p[1] = (uint8)((v >> 8) & 255); michael@0: p[2] = (uint8)((v >> 16) & 255); michael@0: p[3] = (uint8)((v >> 24) & 255); michael@0: } michael@0: #endif michael@0: michael@0: void RGB24ToARGBRow_C(const uint8* src_rgb24, uint8* dst_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_rgb24[0]; michael@0: uint8 g = src_rgb24[1]; michael@0: uint8 r = src_rgb24[2]; michael@0: dst_argb[0] = b; michael@0: dst_argb[1] = g; michael@0: dst_argb[2] = r; michael@0: dst_argb[3] = 255u; michael@0: dst_argb += 4; michael@0: src_rgb24 += 3; michael@0: } michael@0: } michael@0: michael@0: void RAWToARGBRow_C(const uint8* src_raw, uint8* dst_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 r = src_raw[0]; michael@0: uint8 g = src_raw[1]; michael@0: uint8 b = src_raw[2]; michael@0: dst_argb[0] = b; michael@0: dst_argb[1] = g; michael@0: dst_argb[2] = r; michael@0: dst_argb[3] = 255u; michael@0: dst_argb += 4; michael@0: src_raw += 3; michael@0: } michael@0: } michael@0: michael@0: void RGB565ToARGBRow_C(const uint8* src_rgb565, uint8* dst_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_rgb565[0] & 0x1f; michael@0: uint8 g = (src_rgb565[0] >> 5) | ((src_rgb565[1] & 0x07) << 3); michael@0: uint8 r = src_rgb565[1] >> 3; michael@0: dst_argb[0] = (b << 3) | (b >> 2); michael@0: dst_argb[1] = (g << 2) | (g >> 4); michael@0: dst_argb[2] = (r << 3) | (r >> 2); michael@0: dst_argb[3] = 255u; michael@0: dst_argb += 4; michael@0: src_rgb565 += 2; michael@0: } michael@0: } michael@0: michael@0: void ARGB1555ToARGBRow_C(const uint8* src_argb1555, uint8* dst_argb, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_argb1555[0] & 0x1f; michael@0: uint8 g = (src_argb1555[0] >> 5) | ((src_argb1555[1] & 0x03) << 3); michael@0: uint8 r = (src_argb1555[1] & 0x7c) >> 2; michael@0: uint8 a = src_argb1555[1] >> 7; michael@0: dst_argb[0] = (b << 3) | (b >> 2); michael@0: dst_argb[1] = (g << 3) | (g >> 2); michael@0: dst_argb[2] = (r << 3) | (r >> 2); michael@0: dst_argb[3] = -a; michael@0: dst_argb += 4; michael@0: src_argb1555 += 2; michael@0: } michael@0: } michael@0: michael@0: void ARGB4444ToARGBRow_C(const uint8* src_argb4444, uint8* dst_argb, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_argb4444[0] & 0x0f; michael@0: uint8 g = src_argb4444[0] >> 4; michael@0: uint8 r = src_argb4444[1] & 0x0f; michael@0: uint8 a = src_argb4444[1] >> 4; michael@0: dst_argb[0] = (b << 4) | b; michael@0: dst_argb[1] = (g << 4) | g; michael@0: dst_argb[2] = (r << 4) | r; michael@0: dst_argb[3] = (a << 4) | a; michael@0: dst_argb += 4; michael@0: src_argb4444 += 2; michael@0: } michael@0: } michael@0: michael@0: void ARGBToRGB24Row_C(const uint8* src_argb, uint8* dst_rgb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_argb[0]; michael@0: uint8 g = src_argb[1]; michael@0: uint8 r = src_argb[2]; michael@0: dst_rgb[0] = b; michael@0: dst_rgb[1] = g; michael@0: dst_rgb[2] = r; michael@0: dst_rgb += 3; michael@0: src_argb += 4; michael@0: } michael@0: } michael@0: michael@0: void ARGBToRAWRow_C(const uint8* src_argb, uint8* dst_rgb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_argb[0]; michael@0: uint8 g = src_argb[1]; michael@0: uint8 r = src_argb[2]; michael@0: dst_rgb[0] = r; michael@0: dst_rgb[1] = g; michael@0: dst_rgb[2] = b; michael@0: dst_rgb += 3; michael@0: src_argb += 4; michael@0: } michael@0: } michael@0: michael@0: void ARGBToRGB565Row_C(const uint8* src_argb, uint8* dst_rgb, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint8 b0 = src_argb[0] >> 3; michael@0: uint8 g0 = src_argb[1] >> 2; michael@0: uint8 r0 = src_argb[2] >> 3; michael@0: uint8 b1 = src_argb[4] >> 3; michael@0: uint8 g1 = src_argb[5] >> 2; michael@0: uint8 r1 = src_argb[6] >> 3; michael@0: WRITEWORD(dst_rgb, b0 | (g0 << 5) | (r0 << 11) | michael@0: (b1 << 16) | (g1 << 21) | (r1 << 27)); michael@0: dst_rgb += 4; michael@0: src_argb += 8; michael@0: } michael@0: if (width & 1) { michael@0: uint8 b0 = src_argb[0] >> 3; michael@0: uint8 g0 = src_argb[1] >> 2; michael@0: uint8 r0 = src_argb[2] >> 3; michael@0: *(uint16*)(dst_rgb) = b0 | (g0 << 5) | (r0 << 11); michael@0: } michael@0: } michael@0: michael@0: void ARGBToARGB1555Row_C(const uint8* src_argb, uint8* dst_rgb, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint8 b0 = src_argb[0] >> 3; michael@0: uint8 g0 = src_argb[1] >> 3; michael@0: uint8 r0 = src_argb[2] >> 3; michael@0: uint8 a0 = src_argb[3] >> 7; michael@0: uint8 b1 = src_argb[4] >> 3; michael@0: uint8 g1 = src_argb[5] >> 3; michael@0: uint8 r1 = src_argb[6] >> 3; michael@0: uint8 a1 = src_argb[7] >> 7; michael@0: *(uint32*)(dst_rgb) = michael@0: b0 | (g0 << 5) | (r0 << 10) | (a0 << 15) | michael@0: (b1 << 16) | (g1 << 21) | (r1 << 26) | (a1 << 31); michael@0: dst_rgb += 4; michael@0: src_argb += 8; michael@0: } michael@0: if (width & 1) { michael@0: uint8 b0 = src_argb[0] >> 3; michael@0: uint8 g0 = src_argb[1] >> 3; michael@0: uint8 r0 = src_argb[2] >> 3; michael@0: uint8 a0 = src_argb[3] >> 7; michael@0: *(uint16*)(dst_rgb) = michael@0: b0 | (g0 << 5) | (r0 << 10) | (a0 << 15); michael@0: } michael@0: } michael@0: michael@0: void ARGBToARGB4444Row_C(const uint8* src_argb, uint8* dst_rgb, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint8 b0 = src_argb[0] >> 4; michael@0: uint8 g0 = src_argb[1] >> 4; michael@0: uint8 r0 = src_argb[2] >> 4; michael@0: uint8 a0 = src_argb[3] >> 4; michael@0: uint8 b1 = src_argb[4] >> 4; michael@0: uint8 g1 = src_argb[5] >> 4; michael@0: uint8 r1 = src_argb[6] >> 4; michael@0: uint8 a1 = src_argb[7] >> 4; michael@0: *(uint32*)(dst_rgb) = michael@0: b0 | (g0 << 4) | (r0 << 8) | (a0 << 12) | michael@0: (b1 << 16) | (g1 << 20) | (r1 << 24) | (a1 << 28); michael@0: dst_rgb += 4; michael@0: src_argb += 8; michael@0: } michael@0: if (width & 1) { michael@0: uint8 b0 = src_argb[0] >> 4; michael@0: uint8 g0 = src_argb[1] >> 4; michael@0: uint8 r0 = src_argb[2] >> 4; michael@0: uint8 a0 = src_argb[3] >> 4; michael@0: *(uint16*)(dst_rgb) = michael@0: b0 | (g0 << 4) | (r0 << 8) | (a0 << 12); michael@0: } michael@0: } michael@0: michael@0: static __inline int RGBToY(uint8 r, uint8 g, uint8 b) { michael@0: return (66 * r + 129 * g + 25 * b + 0x1080) >> 8; michael@0: } michael@0: michael@0: static __inline int RGBToU(uint8 r, uint8 g, uint8 b) { michael@0: return (112 * b - 74 * g - 38 * r + 0x8080) >> 8; michael@0: } michael@0: static __inline int RGBToV(uint8 r, uint8 g, uint8 b) { michael@0: return (112 * r - 94 * g - 18 * b + 0x8080) >> 8; michael@0: } michael@0: michael@0: #define MAKEROWY(NAME, R, G, B, BPP) \ michael@0: void NAME ## ToYRow_C(const uint8* src_argb0, uint8* dst_y, int width) { \ michael@0: int x; \ michael@0: for (x = 0; x < width; ++x) { \ michael@0: dst_y[0] = RGBToY(src_argb0[R], src_argb0[G], src_argb0[B]); \ michael@0: src_argb0 += BPP; \ michael@0: dst_y += 1; \ michael@0: } \ michael@0: } \ michael@0: void NAME ## ToUVRow_C(const uint8* src_rgb0, int src_stride_rgb, \ michael@0: uint8* dst_u, uint8* dst_v, int width) { \ michael@0: const uint8* src_rgb1 = src_rgb0 + src_stride_rgb; \ michael@0: int x; \ michael@0: for (x = 0; x < width - 1; x += 2) { \ michael@0: uint8 ab = (src_rgb0[B] + src_rgb0[B + BPP] + \ michael@0: src_rgb1[B] + src_rgb1[B + BPP]) >> 2; \ michael@0: uint8 ag = (src_rgb0[G] + src_rgb0[G + BPP] + \ michael@0: src_rgb1[G] + src_rgb1[G + BPP]) >> 2; \ michael@0: uint8 ar = (src_rgb0[R] + src_rgb0[R + BPP] + \ michael@0: src_rgb1[R] + src_rgb1[R + BPP]) >> 2; \ michael@0: dst_u[0] = RGBToU(ar, ag, ab); \ michael@0: dst_v[0] = RGBToV(ar, ag, ab); \ michael@0: src_rgb0 += BPP * 2; \ michael@0: src_rgb1 += BPP * 2; \ michael@0: dst_u += 1; \ michael@0: dst_v += 1; \ michael@0: } \ michael@0: if (width & 1) { \ michael@0: uint8 ab = (src_rgb0[B] + src_rgb1[B]) >> 1; \ michael@0: uint8 ag = (src_rgb0[G] + src_rgb1[G]) >> 1; \ michael@0: uint8 ar = (src_rgb0[R] + src_rgb1[R]) >> 1; \ michael@0: dst_u[0] = RGBToU(ar, ag, ab); \ michael@0: dst_v[0] = RGBToV(ar, ag, ab); \ michael@0: } \ michael@0: } michael@0: michael@0: MAKEROWY(ARGB, 2, 1, 0, 4) michael@0: MAKEROWY(BGRA, 1, 2, 3, 4) michael@0: MAKEROWY(ABGR, 0, 1, 2, 4) michael@0: MAKEROWY(RGBA, 3, 2, 1, 4) michael@0: MAKEROWY(RGB24, 2, 1, 0, 3) michael@0: MAKEROWY(RAW, 0, 1, 2, 3) michael@0: #undef MAKEROWY michael@0: michael@0: // JPeg uses a variation on BT.601-1 full range michael@0: // y = 0.29900 * r + 0.58700 * g + 0.11400 * b michael@0: // u = -0.16874 * r - 0.33126 * g + 0.50000 * b + center michael@0: // v = 0.50000 * r - 0.41869 * g - 0.08131 * b + center michael@0: // BT.601 Mpeg range uses: michael@0: // b 0.1016 * 255 = 25.908 = 25 michael@0: // g 0.5078 * 255 = 129.489 = 129 michael@0: // r 0.2578 * 255 = 65.739 = 66 michael@0: // JPeg 8 bit Y (not used): michael@0: // b 0.11400 * 256 = 29.184 = 29 michael@0: // g 0.58700 * 256 = 150.272 = 150 michael@0: // r 0.29900 * 256 = 76.544 = 77 michael@0: // JPeg 7 bit Y: michael@0: // b 0.11400 * 128 = 14.592 = 15 michael@0: // g 0.58700 * 128 = 75.136 = 75 michael@0: // r 0.29900 * 128 = 38.272 = 38 michael@0: // JPeg 8 bit U: michael@0: // b 0.50000 * 255 = 127.5 = 127 michael@0: // g -0.33126 * 255 = -84.4713 = -84 michael@0: // r -0.16874 * 255 = -43.0287 = -43 michael@0: // JPeg 8 bit V: michael@0: // b -0.08131 * 255 = -20.73405 = -20 michael@0: // g -0.41869 * 255 = -106.76595 = -107 michael@0: // r 0.50000 * 255 = 127.5 = 127 michael@0: michael@0: static __inline int RGBToYJ(uint8 r, uint8 g, uint8 b) { michael@0: return (38 * r + 75 * g + 15 * b + 64) >> 7; michael@0: } michael@0: michael@0: static __inline int RGBToUJ(uint8 r, uint8 g, uint8 b) { michael@0: return (127 * b - 84 * g - 43 * r + 0x8080) >> 8; michael@0: } michael@0: static __inline int RGBToVJ(uint8 r, uint8 g, uint8 b) { michael@0: return (127 * r - 107 * g - 20 * b + 0x8080) >> 8; michael@0: } michael@0: michael@0: #define AVGB(a, b) (((a) + (b) + 1) >> 1) michael@0: michael@0: #define MAKEROWYJ(NAME, R, G, B, BPP) \ michael@0: void NAME ## ToYJRow_C(const uint8* src_argb0, uint8* dst_y, int width) { \ michael@0: int x; \ michael@0: for (x = 0; x < width; ++x) { \ michael@0: dst_y[0] = RGBToYJ(src_argb0[R], src_argb0[G], src_argb0[B]); \ michael@0: src_argb0 += BPP; \ michael@0: dst_y += 1; \ michael@0: } \ michael@0: } \ michael@0: void NAME ## ToUVJRow_C(const uint8* src_rgb0, int src_stride_rgb, \ michael@0: uint8* dst_u, uint8* dst_v, int width) { \ michael@0: const uint8* src_rgb1 = src_rgb0 + src_stride_rgb; \ michael@0: int x; \ michael@0: for (x = 0; x < width - 1; x += 2) { \ michael@0: uint8 ab = AVGB(AVGB(src_rgb0[B], src_rgb1[B]), \ michael@0: AVGB(src_rgb0[B + BPP], src_rgb1[B + BPP])); \ michael@0: uint8 ag = AVGB(AVGB(src_rgb0[G], src_rgb1[G]), \ michael@0: AVGB(src_rgb0[G + BPP], src_rgb1[G + BPP])); \ michael@0: uint8 ar = AVGB(AVGB(src_rgb0[R], src_rgb1[R]), \ michael@0: AVGB(src_rgb0[R + BPP], src_rgb1[R + BPP])); \ michael@0: dst_u[0] = RGBToUJ(ar, ag, ab); \ michael@0: dst_v[0] = RGBToVJ(ar, ag, ab); \ michael@0: src_rgb0 += BPP * 2; \ michael@0: src_rgb1 += BPP * 2; \ michael@0: dst_u += 1; \ michael@0: dst_v += 1; \ michael@0: } \ michael@0: if (width & 1) { \ michael@0: uint8 ab = AVGB(src_rgb0[B], src_rgb1[B]); \ michael@0: uint8 ag = AVGB(src_rgb0[G], src_rgb1[G]); \ michael@0: uint8 ar = AVGB(src_rgb0[R], src_rgb1[R]); \ michael@0: dst_u[0] = RGBToUJ(ar, ag, ab); \ michael@0: dst_v[0] = RGBToVJ(ar, ag, ab); \ michael@0: } \ michael@0: } michael@0: michael@0: MAKEROWYJ(ARGB, 2, 1, 0, 4) michael@0: #undef MAKEROWYJ michael@0: michael@0: void RGB565ToYRow_C(const uint8* src_rgb565, uint8* dst_y, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_rgb565[0] & 0x1f; michael@0: uint8 g = (src_rgb565[0] >> 5) | ((src_rgb565[1] & 0x07) << 3); michael@0: uint8 r = src_rgb565[1] >> 3; michael@0: b = (b << 3) | (b >> 2); michael@0: g = (g << 2) | (g >> 4); michael@0: r = (r << 3) | (r >> 2); michael@0: dst_y[0] = RGBToY(r, g, b); michael@0: src_rgb565 += 2; michael@0: dst_y += 1; michael@0: } michael@0: } michael@0: michael@0: void ARGB1555ToYRow_C(const uint8* src_argb1555, uint8* dst_y, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_argb1555[0] & 0x1f; michael@0: uint8 g = (src_argb1555[0] >> 5) | ((src_argb1555[1] & 0x03) << 3); michael@0: uint8 r = (src_argb1555[1] & 0x7c) >> 2; michael@0: b = (b << 3) | (b >> 2); michael@0: g = (g << 3) | (g >> 2); michael@0: r = (r << 3) | (r >> 2); michael@0: dst_y[0] = RGBToY(r, g, b); michael@0: src_argb1555 += 2; michael@0: dst_y += 1; michael@0: } michael@0: } michael@0: michael@0: void ARGB4444ToYRow_C(const uint8* src_argb4444, uint8* dst_y, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 b = src_argb4444[0] & 0x0f; michael@0: uint8 g = src_argb4444[0] >> 4; michael@0: uint8 r = src_argb4444[1] & 0x0f; michael@0: b = (b << 4) | b; michael@0: g = (g << 4) | g; michael@0: r = (r << 4) | r; michael@0: dst_y[0] = RGBToY(r, g, b); michael@0: src_argb4444 += 2; michael@0: dst_y += 1; michael@0: } michael@0: } michael@0: michael@0: void RGB565ToUVRow_C(const uint8* src_rgb565, int src_stride_rgb565, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: const uint8* next_rgb565 = src_rgb565 + src_stride_rgb565; michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint8 b0 = src_rgb565[0] & 0x1f; michael@0: uint8 g0 = (src_rgb565[0] >> 5) | ((src_rgb565[1] & 0x07) << 3); michael@0: uint8 r0 = src_rgb565[1] >> 3; michael@0: uint8 b1 = src_rgb565[2] & 0x1f; michael@0: uint8 g1 = (src_rgb565[2] >> 5) | ((src_rgb565[3] & 0x07) << 3); michael@0: uint8 r1 = src_rgb565[3] >> 3; michael@0: uint8 b2 = next_rgb565[0] & 0x1f; michael@0: uint8 g2 = (next_rgb565[0] >> 5) | ((next_rgb565[1] & 0x07) << 3); michael@0: uint8 r2 = next_rgb565[1] >> 3; michael@0: uint8 b3 = next_rgb565[2] & 0x1f; michael@0: uint8 g3 = (next_rgb565[2] >> 5) | ((next_rgb565[3] & 0x07) << 3); michael@0: uint8 r3 = next_rgb565[3] >> 3; michael@0: uint8 b = (b0 + b1 + b2 + b3); // 565 * 4 = 787. michael@0: uint8 g = (g0 + g1 + g2 + g3); michael@0: uint8 r = (r0 + r1 + r2 + r3); michael@0: b = (b << 1) | (b >> 6); // 787 -> 888. michael@0: r = (r << 1) | (r >> 6); michael@0: dst_u[0] = RGBToU(r, g, b); michael@0: dst_v[0] = RGBToV(r, g, b); michael@0: src_rgb565 += 4; michael@0: next_rgb565 += 4; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: if (width & 1) { michael@0: uint8 b0 = src_rgb565[0] & 0x1f; michael@0: uint8 g0 = (src_rgb565[0] >> 5) | ((src_rgb565[1] & 0x07) << 3); michael@0: uint8 r0 = src_rgb565[1] >> 3; michael@0: uint8 b2 = next_rgb565[0] & 0x1f; michael@0: uint8 g2 = (next_rgb565[0] >> 5) | ((next_rgb565[1] & 0x07) << 3); michael@0: uint8 r2 = next_rgb565[1] >> 3; michael@0: uint8 b = (b0 + b2); // 565 * 2 = 676. michael@0: uint8 g = (g0 + g2); michael@0: uint8 r = (r0 + r2); michael@0: b = (b << 2) | (b >> 4); // 676 -> 888 michael@0: g = (g << 1) | (g >> 6); michael@0: r = (r << 2) | (r >> 4); michael@0: dst_u[0] = RGBToU(r, g, b); michael@0: dst_v[0] = RGBToV(r, g, b); michael@0: } michael@0: } michael@0: michael@0: void ARGB1555ToUVRow_C(const uint8* src_argb1555, int src_stride_argb1555, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: const uint8* next_argb1555 = src_argb1555 + src_stride_argb1555; michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint8 b0 = src_argb1555[0] & 0x1f; michael@0: uint8 g0 = (src_argb1555[0] >> 5) | ((src_argb1555[1] & 0x03) << 3); michael@0: uint8 r0 = (src_argb1555[1] & 0x7c) >> 2; michael@0: uint8 b1 = src_argb1555[2] & 0x1f; michael@0: uint8 g1 = (src_argb1555[2] >> 5) | ((src_argb1555[3] & 0x03) << 3); michael@0: uint8 r1 = (src_argb1555[3] & 0x7c) >> 2; michael@0: uint8 b2 = next_argb1555[0] & 0x1f; michael@0: uint8 g2 = (next_argb1555[0] >> 5) | ((next_argb1555[1] & 0x03) << 3); michael@0: uint8 r2 = (next_argb1555[1] & 0x7c) >> 2; michael@0: uint8 b3 = next_argb1555[2] & 0x1f; michael@0: uint8 g3 = (next_argb1555[2] >> 5) | ((next_argb1555[3] & 0x03) << 3); michael@0: uint8 r3 = (next_argb1555[3] & 0x7c) >> 2; michael@0: uint8 b = (b0 + b1 + b2 + b3); // 555 * 4 = 777. michael@0: uint8 g = (g0 + g1 + g2 + g3); michael@0: uint8 r = (r0 + r1 + r2 + r3); michael@0: b = (b << 1) | (b >> 6); // 777 -> 888. michael@0: g = (g << 1) | (g >> 6); michael@0: r = (r << 1) | (r >> 6); michael@0: dst_u[0] = RGBToU(r, g, b); michael@0: dst_v[0] = RGBToV(r, g, b); michael@0: src_argb1555 += 4; michael@0: next_argb1555 += 4; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: if (width & 1) { michael@0: uint8 b0 = src_argb1555[0] & 0x1f; michael@0: uint8 g0 = (src_argb1555[0] >> 5) | ((src_argb1555[1] & 0x03) << 3); michael@0: uint8 r0 = (src_argb1555[1] & 0x7c) >> 2; michael@0: uint8 b2 = next_argb1555[0] & 0x1f; michael@0: uint8 g2 = (next_argb1555[0] >> 5) | ((next_argb1555[1] & 0x03) << 3); michael@0: uint8 r2 = next_argb1555[1] >> 3; michael@0: uint8 b = (b0 + b2); // 555 * 2 = 666. michael@0: uint8 g = (g0 + g2); michael@0: uint8 r = (r0 + r2); michael@0: b = (b << 2) | (b >> 4); // 666 -> 888. michael@0: g = (g << 2) | (g >> 4); michael@0: r = (r << 2) | (r >> 4); michael@0: dst_u[0] = RGBToU(r, g, b); michael@0: dst_v[0] = RGBToV(r, g, b); michael@0: } michael@0: } michael@0: michael@0: void ARGB4444ToUVRow_C(const uint8* src_argb4444, int src_stride_argb4444, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: const uint8* next_argb4444 = src_argb4444 + src_stride_argb4444; michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint8 b0 = src_argb4444[0] & 0x0f; michael@0: uint8 g0 = src_argb4444[0] >> 4; michael@0: uint8 r0 = src_argb4444[1] & 0x0f; michael@0: uint8 b1 = src_argb4444[2] & 0x0f; michael@0: uint8 g1 = src_argb4444[2] >> 4; michael@0: uint8 r1 = src_argb4444[3] & 0x0f; michael@0: uint8 b2 = next_argb4444[0] & 0x0f; michael@0: uint8 g2 = next_argb4444[0] >> 4; michael@0: uint8 r2 = next_argb4444[1] & 0x0f; michael@0: uint8 b3 = next_argb4444[2] & 0x0f; michael@0: uint8 g3 = next_argb4444[2] >> 4; michael@0: uint8 r3 = next_argb4444[3] & 0x0f; michael@0: uint8 b = (b0 + b1 + b2 + b3); // 444 * 4 = 666. michael@0: uint8 g = (g0 + g1 + g2 + g3); michael@0: uint8 r = (r0 + r1 + r2 + r3); michael@0: b = (b << 2) | (b >> 4); // 666 -> 888. michael@0: g = (g << 2) | (g >> 4); michael@0: r = (r << 2) | (r >> 4); michael@0: dst_u[0] = RGBToU(r, g, b); michael@0: dst_v[0] = RGBToV(r, g, b); michael@0: src_argb4444 += 4; michael@0: next_argb4444 += 4; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: if (width & 1) { michael@0: uint8 b0 = src_argb4444[0] & 0x0f; michael@0: uint8 g0 = src_argb4444[0] >> 4; michael@0: uint8 r0 = src_argb4444[1] & 0x0f; michael@0: uint8 b2 = next_argb4444[0] & 0x0f; michael@0: uint8 g2 = next_argb4444[0] >> 4; michael@0: uint8 r2 = next_argb4444[1] & 0x0f; michael@0: uint8 b = (b0 + b2); // 444 * 2 = 555. michael@0: uint8 g = (g0 + g2); michael@0: uint8 r = (r0 + r2); michael@0: b = (b << 3) | (b >> 2); // 555 -> 888. michael@0: g = (g << 3) | (g >> 2); michael@0: r = (r << 3) | (r >> 2); michael@0: dst_u[0] = RGBToU(r, g, b); michael@0: dst_v[0] = RGBToV(r, g, b); michael@0: } michael@0: } michael@0: michael@0: void ARGBToUV444Row_C(const uint8* src_argb, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 ab = src_argb[0]; michael@0: uint8 ag = src_argb[1]; michael@0: uint8 ar = src_argb[2]; michael@0: dst_u[0] = RGBToU(ar, ag, ab); michael@0: dst_v[0] = RGBToV(ar, ag, ab); michael@0: src_argb += 4; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: } michael@0: michael@0: void ARGBToUV422Row_C(const uint8* src_argb, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint8 ab = (src_argb[0] + src_argb[4]) >> 1; michael@0: uint8 ag = (src_argb[1] + src_argb[5]) >> 1; michael@0: uint8 ar = (src_argb[2] + src_argb[6]) >> 1; michael@0: dst_u[0] = RGBToU(ar, ag, ab); michael@0: dst_v[0] = RGBToV(ar, ag, ab); michael@0: src_argb += 8; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: if (width & 1) { michael@0: uint8 ab = src_argb[0]; michael@0: uint8 ag = src_argb[1]; michael@0: uint8 ar = src_argb[2]; michael@0: dst_u[0] = RGBToU(ar, ag, ab); michael@0: dst_v[0] = RGBToV(ar, ag, ab); michael@0: } michael@0: } michael@0: michael@0: void ARGBToUV411Row_C(const uint8* src_argb, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 3; x += 4) { michael@0: uint8 ab = (src_argb[0] + src_argb[4] + src_argb[8] + src_argb[12]) >> 2; michael@0: uint8 ag = (src_argb[1] + src_argb[5] + src_argb[9] + src_argb[13]) >> 2; michael@0: uint8 ar = (src_argb[2] + src_argb[6] + src_argb[10] + src_argb[14]) >> 2; michael@0: dst_u[0] = RGBToU(ar, ag, ab); michael@0: dst_v[0] = RGBToV(ar, ag, ab); michael@0: src_argb += 16; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: if ((width & 3) == 3) { michael@0: uint8 ab = (src_argb[0] + src_argb[4] + src_argb[8]) / 3; michael@0: uint8 ag = (src_argb[1] + src_argb[5] + src_argb[9]) / 3; michael@0: uint8 ar = (src_argb[2] + src_argb[6] + src_argb[10]) / 3; michael@0: dst_u[0] = RGBToU(ar, ag, ab); michael@0: dst_v[0] = RGBToV(ar, ag, ab); michael@0: } else if ((width & 3) == 2) { michael@0: uint8 ab = (src_argb[0] + src_argb[4]) >> 1; michael@0: uint8 ag = (src_argb[1] + src_argb[5]) >> 1; michael@0: uint8 ar = (src_argb[2] + src_argb[6]) >> 1; michael@0: dst_u[0] = RGBToU(ar, ag, ab); michael@0: dst_v[0] = RGBToV(ar, ag, ab); michael@0: } else if ((width & 3) == 1) { michael@0: uint8 ab = src_argb[0]; michael@0: uint8 ag = src_argb[1]; michael@0: uint8 ar = src_argb[2]; michael@0: dst_u[0] = RGBToU(ar, ag, ab); michael@0: dst_v[0] = RGBToV(ar, ag, ab); michael@0: } michael@0: } michael@0: michael@0: void ARGBGrayRow_C(const uint8* src_argb, uint8* dst_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 y = RGBToYJ(src_argb[2], src_argb[1], src_argb[0]); michael@0: dst_argb[2] = dst_argb[1] = dst_argb[0] = y; michael@0: dst_argb[3] = src_argb[3]; michael@0: dst_argb += 4; michael@0: src_argb += 4; michael@0: } michael@0: } michael@0: michael@0: // Convert a row of image to Sepia tone. michael@0: void ARGBSepiaRow_C(uint8* dst_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: int b = dst_argb[0]; michael@0: int g = dst_argb[1]; michael@0: int r = dst_argb[2]; michael@0: int sb = (b * 17 + g * 68 + r * 35) >> 7; michael@0: int sg = (b * 22 + g * 88 + r * 45) >> 7; michael@0: int sr = (b * 24 + g * 98 + r * 50) >> 7; michael@0: // b does not over flow. a is preserved from original. michael@0: dst_argb[0] = sb; michael@0: dst_argb[1] = clamp255(sg); michael@0: dst_argb[2] = clamp255(sr); michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: // Apply color matrix to a row of image. Matrix is signed. michael@0: // TODO(fbarchard): Consider adding rounding (+32). michael@0: void ARGBColorMatrixRow_C(const uint8* src_argb, uint8* dst_argb, michael@0: const int8* matrix_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: int b = src_argb[0]; michael@0: int g = src_argb[1]; michael@0: int r = src_argb[2]; michael@0: int a = src_argb[3]; michael@0: int sb = (b * matrix_argb[0] + g * matrix_argb[1] + michael@0: r * matrix_argb[2] + a * matrix_argb[3]) >> 6; michael@0: int sg = (b * matrix_argb[4] + g * matrix_argb[5] + michael@0: r * matrix_argb[6] + a * matrix_argb[7]) >> 6; michael@0: int sr = (b * matrix_argb[8] + g * matrix_argb[9] + michael@0: r * matrix_argb[10] + a * matrix_argb[11]) >> 6; michael@0: int sa = (b * matrix_argb[12] + g * matrix_argb[13] + michael@0: r * matrix_argb[14] + a * matrix_argb[15]) >> 6; michael@0: dst_argb[0] = Clamp(sb); michael@0: dst_argb[1] = Clamp(sg); michael@0: dst_argb[2] = Clamp(sr); michael@0: dst_argb[3] = Clamp(sa); michael@0: src_argb += 4; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: // Apply color table to a row of image. michael@0: void ARGBColorTableRow_C(uint8* dst_argb, const uint8* table_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: int b = dst_argb[0]; michael@0: int g = dst_argb[1]; michael@0: int r = dst_argb[2]; michael@0: int a = dst_argb[3]; michael@0: dst_argb[0] = table_argb[b * 4 + 0]; michael@0: dst_argb[1] = table_argb[g * 4 + 1]; michael@0: dst_argb[2] = table_argb[r * 4 + 2]; michael@0: dst_argb[3] = table_argb[a * 4 + 3]; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: // Apply color table to a row of image. michael@0: void RGBColorTableRow_C(uint8* dst_argb, const uint8* table_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: int b = dst_argb[0]; michael@0: int g = dst_argb[1]; michael@0: int r = dst_argb[2]; michael@0: dst_argb[0] = table_argb[b * 4 + 0]; michael@0: dst_argb[1] = table_argb[g * 4 + 1]; michael@0: dst_argb[2] = table_argb[r * 4 + 2]; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: void ARGBQuantizeRow_C(uint8* dst_argb, int scale, int interval_size, michael@0: int interval_offset, int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: int b = dst_argb[0]; michael@0: int g = dst_argb[1]; michael@0: int r = dst_argb[2]; michael@0: dst_argb[0] = (b * scale >> 16) * interval_size + interval_offset; michael@0: dst_argb[1] = (g * scale >> 16) * interval_size + interval_offset; michael@0: dst_argb[2] = (r * scale >> 16) * interval_size + interval_offset; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: #define REPEAT8(v) (v) | ((v) << 8) michael@0: #define SHADE(f, v) v * f >> 24 michael@0: michael@0: void ARGBShadeRow_C(const uint8* src_argb, uint8* dst_argb, int width, michael@0: uint32 value) { michael@0: const uint32 b_scale = REPEAT8(value & 0xff); michael@0: const uint32 g_scale = REPEAT8((value >> 8) & 0xff); michael@0: const uint32 r_scale = REPEAT8((value >> 16) & 0xff); michael@0: const uint32 a_scale = REPEAT8(value >> 24); michael@0: michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: const uint32 b = REPEAT8(src_argb[0]); michael@0: const uint32 g = REPEAT8(src_argb[1]); michael@0: const uint32 r = REPEAT8(src_argb[2]); michael@0: const uint32 a = REPEAT8(src_argb[3]); michael@0: dst_argb[0] = SHADE(b, b_scale); michael@0: dst_argb[1] = SHADE(g, g_scale); michael@0: dst_argb[2] = SHADE(r, r_scale); michael@0: dst_argb[3] = SHADE(a, a_scale); michael@0: src_argb += 4; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: #undef REPEAT8 michael@0: #undef SHADE michael@0: michael@0: #define REPEAT8(v) (v) | ((v) << 8) michael@0: #define SHADE(f, v) v * f >> 16 michael@0: michael@0: void ARGBMultiplyRow_C(const uint8* src_argb0, const uint8* src_argb1, michael@0: uint8* dst_argb, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: const uint32 b = REPEAT8(src_argb0[0]); michael@0: const uint32 g = REPEAT8(src_argb0[1]); michael@0: const uint32 r = REPEAT8(src_argb0[2]); michael@0: const uint32 a = REPEAT8(src_argb0[3]); michael@0: const uint32 b_scale = src_argb1[0]; michael@0: const uint32 g_scale = src_argb1[1]; michael@0: const uint32 r_scale = src_argb1[2]; michael@0: const uint32 a_scale = src_argb1[3]; michael@0: dst_argb[0] = SHADE(b, b_scale); michael@0: dst_argb[1] = SHADE(g, g_scale); michael@0: dst_argb[2] = SHADE(r, r_scale); michael@0: dst_argb[3] = SHADE(a, a_scale); michael@0: src_argb0 += 4; michael@0: src_argb1 += 4; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: #undef REPEAT8 michael@0: #undef SHADE michael@0: michael@0: #define SHADE(f, v) clamp255(v + f) michael@0: michael@0: void ARGBAddRow_C(const uint8* src_argb0, const uint8* src_argb1, michael@0: uint8* dst_argb, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: const int b = src_argb0[0]; michael@0: const int g = src_argb0[1]; michael@0: const int r = src_argb0[2]; michael@0: const int a = src_argb0[3]; michael@0: const int b_add = src_argb1[0]; michael@0: const int g_add = src_argb1[1]; michael@0: const int r_add = src_argb1[2]; michael@0: const int a_add = src_argb1[3]; michael@0: dst_argb[0] = SHADE(b, b_add); michael@0: dst_argb[1] = SHADE(g, g_add); michael@0: dst_argb[2] = SHADE(r, r_add); michael@0: dst_argb[3] = SHADE(a, a_add); michael@0: src_argb0 += 4; michael@0: src_argb1 += 4; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: #undef SHADE michael@0: michael@0: #define SHADE(f, v) clamp0(f - v) michael@0: michael@0: void ARGBSubtractRow_C(const uint8* src_argb0, const uint8* src_argb1, michael@0: uint8* dst_argb, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: const int b = src_argb0[0]; michael@0: const int g = src_argb0[1]; michael@0: const int r = src_argb0[2]; michael@0: const int a = src_argb0[3]; michael@0: const int b_sub = src_argb1[0]; michael@0: const int g_sub = src_argb1[1]; michael@0: const int r_sub = src_argb1[2]; michael@0: const int a_sub = src_argb1[3]; michael@0: dst_argb[0] = SHADE(b, b_sub); michael@0: dst_argb[1] = SHADE(g, g_sub); michael@0: dst_argb[2] = SHADE(r, r_sub); michael@0: dst_argb[3] = SHADE(a, a_sub); michael@0: src_argb0 += 4; michael@0: src_argb1 += 4; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: #undef SHADE michael@0: michael@0: // Sobel functions which mimics SSSE3. michael@0: void SobelXRow_C(const uint8* src_y0, const uint8* src_y1, const uint8* src_y2, michael@0: uint8* dst_sobelx, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: int a = src_y0[i]; michael@0: int b = src_y1[i]; michael@0: int c = src_y2[i]; michael@0: int a_sub = src_y0[i + 2]; michael@0: int b_sub = src_y1[i + 2]; michael@0: int c_sub = src_y2[i + 2]; michael@0: int a_diff = a - a_sub; michael@0: int b_diff = b - b_sub; michael@0: int c_diff = c - c_sub; michael@0: int sobel = Abs(a_diff + b_diff * 2 + c_diff); michael@0: dst_sobelx[i] = (uint8)(clamp255(sobel)); michael@0: } michael@0: } michael@0: michael@0: void SobelYRow_C(const uint8* src_y0, const uint8* src_y1, michael@0: uint8* dst_sobely, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: int a = src_y0[i + 0]; michael@0: int b = src_y0[i + 1]; michael@0: int c = src_y0[i + 2]; michael@0: int a_sub = src_y1[i + 0]; michael@0: int b_sub = src_y1[i + 1]; michael@0: int c_sub = src_y1[i + 2]; michael@0: int a_diff = a - a_sub; michael@0: int b_diff = b - b_sub; michael@0: int c_diff = c - c_sub; michael@0: int sobel = Abs(a_diff + b_diff * 2 + c_diff); michael@0: dst_sobely[i] = (uint8)(clamp255(sobel)); michael@0: } michael@0: } michael@0: michael@0: void SobelRow_C(const uint8* src_sobelx, const uint8* src_sobely, michael@0: uint8* dst_argb, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: int r = src_sobelx[i]; michael@0: int b = src_sobely[i]; michael@0: int s = clamp255(r + b); michael@0: dst_argb[0] = (uint8)(s); michael@0: dst_argb[1] = (uint8)(s); michael@0: dst_argb[2] = (uint8)(s); michael@0: dst_argb[3] = (uint8)(255u); michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: void SobelToPlaneRow_C(const uint8* src_sobelx, const uint8* src_sobely, michael@0: uint8* dst_y, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: int r = src_sobelx[i]; michael@0: int b = src_sobely[i]; michael@0: int s = clamp255(r + b); michael@0: dst_y[i] = (uint8)(s); michael@0: } michael@0: } michael@0: michael@0: void SobelXYRow_C(const uint8* src_sobelx, const uint8* src_sobely, michael@0: uint8* dst_argb, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: int r = src_sobelx[i]; michael@0: int b = src_sobely[i]; michael@0: int g = clamp255(r + b); michael@0: dst_argb[0] = (uint8)(b); michael@0: dst_argb[1] = (uint8)(g); michael@0: dst_argb[2] = (uint8)(r); michael@0: dst_argb[3] = (uint8)(255u); michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) { michael@0: // Copy a Y to RGB. michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: uint8 y = src_y[0]; michael@0: dst_argb[2] = dst_argb[1] = dst_argb[0] = y; michael@0: dst_argb[3] = 255u; michael@0: dst_argb += 4; michael@0: ++src_y; michael@0: } michael@0: } michael@0: michael@0: // C reference code that mimics the YUV assembly. michael@0: michael@0: #define YG 74 /* (int8)(1.164 * 64 + 0.5) */ michael@0: michael@0: #define UB 127 /* min(63,(int8)(2.018 * 64)) */ michael@0: #define UG -25 /* (int8)(-0.391 * 64 - 0.5) */ michael@0: #define UR 0 michael@0: michael@0: #define VB 0 michael@0: #define VG -52 /* (int8)(-0.813 * 64 - 0.5) */ michael@0: #define VR 102 /* (int8)(1.596 * 64 + 0.5) */ michael@0: michael@0: // Bias michael@0: #define BB UB * 128 + VB * 128 michael@0: #define BG UG * 128 + VG * 128 michael@0: #define BR UR * 128 + VR * 128 michael@0: michael@0: static __inline void YuvPixel(uint8 y, uint8 u, uint8 v, michael@0: uint8* b, uint8* g, uint8* r) { michael@0: int32 y1 = ((int32)(y) - 16) * YG; michael@0: *b = Clamp((int32)((u * UB + v * VB) - (BB) + y1) >> 6); michael@0: *g = Clamp((int32)((u * UG + v * VG) - (BG) + y1) >> 6); michael@0: *r = Clamp((int32)((u * UR + v * VR) - (BR) + y1) >> 6); michael@0: } michael@0: michael@0: #if !defined(LIBYUV_DISABLE_NEON) && \ michael@0: (defined(__ARM_NEON__) || defined(LIBYUV_NEON)) michael@0: // C mimic assembly. michael@0: // TODO(fbarchard): Remove subsampling from Neon. michael@0: void I444ToARGBRow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint8 u = (src_u[0] + src_u[1] + 1) >> 1; michael@0: uint8 v = (src_v[0] + src_v[1] + 1) >> 1; michael@0: YuvPixel(src_y[0], u, v, rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_y[1], u, v, rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: src_y += 2; michael@0: src_u += 2; michael@0: src_v += 2; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: } michael@0: } michael@0: #else michael@0: void I444ToARGBRow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: src_y += 1; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: rgb_buf += 4; // Advance 1 pixel. michael@0: } michael@0: } michael@0: #endif michael@0: // Also used for 420 michael@0: void I422ToARGBRow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], michael@0: rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: } michael@0: } michael@0: michael@0: void I422ToRGB24Row_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], michael@0: rgb_buf + 3, rgb_buf + 4, rgb_buf + 5); michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: rgb_buf += 6; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: } michael@0: } michael@0: michael@0: void I422ToRAWRow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 2, rgb_buf + 1, rgb_buf + 0); michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], michael@0: rgb_buf + 5, rgb_buf + 4, rgb_buf + 3); michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: rgb_buf += 6; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 2, rgb_buf + 1, rgb_buf + 0); michael@0: } michael@0: } michael@0: michael@0: void I422ToARGB4444Row_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* dst_argb4444, michael@0: int width) { michael@0: uint8 b0; michael@0: uint8 g0; michael@0: uint8 r0; michael@0: uint8 b1; michael@0: uint8 g1; michael@0: uint8 r1; michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], &b0, &g0, &r0); michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], &b1, &g1, &r1); michael@0: b0 = b0 >> 4; michael@0: g0 = g0 >> 4; michael@0: r0 = r0 >> 4; michael@0: b1 = b1 >> 4; michael@0: g1 = g1 >> 4; michael@0: r1 = r1 >> 4; michael@0: *(uint32*)(dst_argb4444) = b0 | (g0 << 4) | (r0 << 8) | michael@0: (b1 << 16) | (g1 << 20) | (r1 << 24) | 0xf000f000; michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: dst_argb4444 += 4; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], &b0, &g0, &r0); michael@0: b0 = b0 >> 4; michael@0: g0 = g0 >> 4; michael@0: r0 = r0 >> 4; michael@0: *(uint16*)(dst_argb4444) = b0 | (g0 << 4) | (r0 << 8) | michael@0: 0xf000; michael@0: } michael@0: } michael@0: michael@0: void I422ToARGB1555Row_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* dst_argb1555, michael@0: int width) { michael@0: uint8 b0; michael@0: uint8 g0; michael@0: uint8 r0; michael@0: uint8 b1; michael@0: uint8 g1; michael@0: uint8 r1; michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], &b0, &g0, &r0); michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], &b1, &g1, &r1); michael@0: b0 = b0 >> 3; michael@0: g0 = g0 >> 3; michael@0: r0 = r0 >> 3; michael@0: b1 = b1 >> 3; michael@0: g1 = g1 >> 3; michael@0: r1 = r1 >> 3; michael@0: *(uint32*)(dst_argb1555) = b0 | (g0 << 5) | (r0 << 10) | michael@0: (b1 << 16) | (g1 << 21) | (r1 << 26) | 0x80008000; michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: dst_argb1555 += 4; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], &b0, &g0, &r0); michael@0: b0 = b0 >> 3; michael@0: g0 = g0 >> 3; michael@0: r0 = r0 >> 3; michael@0: *(uint16*)(dst_argb1555) = b0 | (g0 << 5) | (r0 << 10) | michael@0: 0x8000; michael@0: } michael@0: } michael@0: michael@0: void I422ToRGB565Row_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* dst_rgb565, michael@0: int width) { michael@0: uint8 b0; michael@0: uint8 g0; michael@0: uint8 r0; michael@0: uint8 b1; michael@0: uint8 g1; michael@0: uint8 r1; michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], &b0, &g0, &r0); michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], &b1, &g1, &r1); michael@0: b0 = b0 >> 3; michael@0: g0 = g0 >> 2; michael@0: r0 = r0 >> 3; michael@0: b1 = b1 >> 3; michael@0: g1 = g1 >> 2; michael@0: r1 = r1 >> 3; michael@0: *(uint32*)(dst_rgb565) = b0 | (g0 << 5) | (r0 << 11) | michael@0: (b1 << 16) | (g1 << 21) | (r1 << 27); michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: dst_rgb565 += 4; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], &b0, &g0, &r0); michael@0: b0 = b0 >> 3; michael@0: g0 = g0 >> 2; michael@0: r0 = r0 >> 3; michael@0: *(uint16*)(dst_rgb565) = b0 | (g0 << 5) | (r0 << 11); michael@0: } michael@0: } michael@0: michael@0: void I411ToARGBRow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 3; x += 4) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], michael@0: rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: YuvPixel(src_y[2], src_u[0], src_v[0], michael@0: rgb_buf + 8, rgb_buf + 9, rgb_buf + 10); michael@0: rgb_buf[11] = 255; michael@0: YuvPixel(src_y[3], src_u[0], src_v[0], michael@0: rgb_buf + 12, rgb_buf + 13, rgb_buf + 14); michael@0: rgb_buf[15] = 255; michael@0: src_y += 4; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: rgb_buf += 16; // Advance 4 pixels. michael@0: } michael@0: if (width & 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], michael@0: rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: src_y += 2; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: } michael@0: } michael@0: michael@0: void NV12ToARGBRow_C(const uint8* src_y, michael@0: const uint8* usrc_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], usrc_v[0], usrc_v[1], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_y[1], usrc_v[0], usrc_v[1], michael@0: rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: src_y += 2; michael@0: usrc_v += 2; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], usrc_v[0], usrc_v[1], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: } michael@0: } michael@0: michael@0: void NV21ToARGBRow_C(const uint8* src_y, michael@0: const uint8* src_vu, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_vu[1], src_vu[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: michael@0: YuvPixel(src_y[1], src_vu[1], src_vu[0], michael@0: rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: michael@0: src_y += 2; michael@0: src_vu += 2; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_vu[1], src_vu[0], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: } michael@0: } michael@0: michael@0: void NV12ToRGB565Row_C(const uint8* src_y, michael@0: const uint8* usrc_v, michael@0: uint8* dst_rgb565, michael@0: int width) { michael@0: uint8 b0; michael@0: uint8 g0; michael@0: uint8 r0; michael@0: uint8 b1; michael@0: uint8 g1; michael@0: uint8 r1; michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], usrc_v[0], usrc_v[1], &b0, &g0, &r0); michael@0: YuvPixel(src_y[1], usrc_v[0], usrc_v[1], &b1, &g1, &r1); michael@0: b0 = b0 >> 3; michael@0: g0 = g0 >> 2; michael@0: r0 = r0 >> 3; michael@0: b1 = b1 >> 3; michael@0: g1 = g1 >> 2; michael@0: r1 = r1 >> 3; michael@0: *(uint32*)(dst_rgb565) = b0 | (g0 << 5) | (r0 << 11) | michael@0: (b1 << 16) | (g1 << 21) | (r1 << 27); michael@0: src_y += 2; michael@0: usrc_v += 2; michael@0: dst_rgb565 += 4; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], usrc_v[0], usrc_v[1], &b0, &g0, &r0); michael@0: b0 = b0 >> 3; michael@0: g0 = g0 >> 2; michael@0: r0 = r0 >> 3; michael@0: *(uint16*)(dst_rgb565) = b0 | (g0 << 5) | (r0 << 11); michael@0: } michael@0: } michael@0: michael@0: void NV21ToRGB565Row_C(const uint8* src_y, michael@0: const uint8* vsrc_u, michael@0: uint8* dst_rgb565, michael@0: int width) { michael@0: uint8 b0; michael@0: uint8 g0; michael@0: uint8 r0; michael@0: uint8 b1; michael@0: uint8 g1; michael@0: uint8 r1; michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], vsrc_u[1], vsrc_u[0], &b0, &g0, &r0); michael@0: YuvPixel(src_y[1], vsrc_u[1], vsrc_u[0], &b1, &g1, &r1); michael@0: b0 = b0 >> 3; michael@0: g0 = g0 >> 2; michael@0: r0 = r0 >> 3; michael@0: b1 = b1 >> 3; michael@0: g1 = g1 >> 2; michael@0: r1 = r1 >> 3; michael@0: *(uint32*)(dst_rgb565) = b0 | (g0 << 5) | (r0 << 11) | michael@0: (b1 << 16) | (g1 << 21) | (r1 << 27); michael@0: src_y += 2; michael@0: vsrc_u += 2; michael@0: dst_rgb565 += 4; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], vsrc_u[1], vsrc_u[0], &b0, &g0, &r0); michael@0: b0 = b0 >> 3; michael@0: g0 = g0 >> 2; michael@0: r0 = r0 >> 3; michael@0: *(uint16*)(dst_rgb565) = b0 | (g0 << 5) | (r0 << 11); michael@0: } michael@0: } michael@0: michael@0: void YUY2ToARGBRow_C(const uint8* src_yuy2, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_yuy2[0], src_yuy2[1], src_yuy2[3], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_yuy2[2], src_yuy2[1], src_yuy2[3], michael@0: rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: src_yuy2 += 4; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_yuy2[0], src_yuy2[1], src_yuy2[3], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: } michael@0: } michael@0: michael@0: void UYVYToARGBRow_C(const uint8* src_uyvy, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_uyvy[1], src_uyvy[0], src_uyvy[2], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_uyvy[3], src_uyvy[0], src_uyvy[2], michael@0: rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: src_uyvy += 4; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_uyvy[1], src_uyvy[0], src_uyvy[2], michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: } michael@0: } michael@0: michael@0: void I422ToBGRARow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 3, rgb_buf + 2, rgb_buf + 1); michael@0: rgb_buf[0] = 255; michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], michael@0: rgb_buf + 7, rgb_buf + 6, rgb_buf + 5); michael@0: rgb_buf[4] = 255; michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 3, rgb_buf + 2, rgb_buf + 1); michael@0: rgb_buf[0] = 255; michael@0: } michael@0: } michael@0: michael@0: void I422ToABGRRow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 2, rgb_buf + 1, rgb_buf + 0); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], michael@0: rgb_buf + 6, rgb_buf + 5, rgb_buf + 4); michael@0: rgb_buf[7] = 255; michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 2, rgb_buf + 1, rgb_buf + 0); michael@0: rgb_buf[3] = 255; michael@0: } michael@0: } michael@0: michael@0: void I422ToRGBARow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 1, rgb_buf + 2, rgb_buf + 3); michael@0: rgb_buf[0] = 255; michael@0: YuvPixel(src_y[1], src_u[0], src_v[0], michael@0: rgb_buf + 5, rgb_buf + 6, rgb_buf + 7); michael@0: rgb_buf[4] = 255; michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], src_u[0], src_v[0], michael@0: rgb_buf + 1, rgb_buf + 2, rgb_buf + 3); michael@0: rgb_buf[0] = 255; michael@0: } michael@0: } michael@0: michael@0: void YToARGBRow_C(const uint8* src_y, uint8* rgb_buf, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: YuvPixel(src_y[0], 128, 128, michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: YuvPixel(src_y[1], 128, 128, michael@0: rgb_buf + 4, rgb_buf + 5, rgb_buf + 6); michael@0: rgb_buf[7] = 255; michael@0: src_y += 2; michael@0: rgb_buf += 8; // Advance 2 pixels. michael@0: } michael@0: if (width & 1) { michael@0: YuvPixel(src_y[0], 128, 128, michael@0: rgb_buf + 0, rgb_buf + 1, rgb_buf + 2); michael@0: rgb_buf[3] = 255; michael@0: } michael@0: } michael@0: michael@0: void MirrorRow_C(const uint8* src, uint8* dst, int width) { michael@0: int x; michael@0: src += width - 1; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst[x] = src[0]; michael@0: dst[x + 1] = src[-1]; michael@0: src -= 2; michael@0: } michael@0: if (width & 1) { michael@0: dst[width - 1] = src[0]; michael@0: } michael@0: } michael@0: michael@0: void MirrorUVRow_C(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) { michael@0: int x; michael@0: src_uv += (width - 1) << 1; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst_u[x] = src_uv[0]; michael@0: dst_u[x + 1] = src_uv[-2]; michael@0: dst_v[x] = src_uv[1]; michael@0: dst_v[x + 1] = src_uv[-2 + 1]; michael@0: src_uv -= 4; michael@0: } michael@0: if (width & 1) { michael@0: dst_u[width - 1] = src_uv[0]; michael@0: dst_v[width - 1] = src_uv[1]; michael@0: } michael@0: } michael@0: michael@0: void ARGBMirrorRow_C(const uint8* src, uint8* dst, int width) { michael@0: int x; michael@0: const uint32* src32 = (const uint32*)(src); michael@0: uint32* dst32 = (uint32*)(dst); michael@0: src32 += width - 1; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst32[x] = src32[0]; michael@0: dst32[x + 1] = src32[-1]; michael@0: src32 -= 2; michael@0: } michael@0: if (width & 1) { michael@0: dst32[width - 1] = src32[0]; michael@0: } michael@0: } michael@0: michael@0: void SplitUVRow_C(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst_u[x] = src_uv[0]; michael@0: dst_u[x + 1] = src_uv[2]; michael@0: dst_v[x] = src_uv[1]; michael@0: dst_v[x + 1] = src_uv[3]; michael@0: src_uv += 4; michael@0: } michael@0: if (width & 1) { michael@0: dst_u[width - 1] = src_uv[0]; michael@0: dst_v[width - 1] = src_uv[1]; michael@0: } michael@0: } michael@0: michael@0: void MergeUVRow_C(const uint8* src_u, const uint8* src_v, uint8* dst_uv, michael@0: int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst_uv[0] = src_u[x]; michael@0: dst_uv[1] = src_v[x]; michael@0: dst_uv[2] = src_u[x + 1]; michael@0: dst_uv[3] = src_v[x + 1]; michael@0: dst_uv += 4; michael@0: } michael@0: if (width & 1) { michael@0: dst_uv[0] = src_u[width - 1]; michael@0: dst_uv[1] = src_v[width - 1]; michael@0: } michael@0: } michael@0: michael@0: void CopyRow_C(const uint8* src, uint8* dst, int count) { michael@0: memcpy(dst, src, count); michael@0: } michael@0: michael@0: void SetRow_C(uint8* dst, uint32 v8, int count) { michael@0: #ifdef _MSC_VER michael@0: // VC will generate rep stosb. michael@0: int x; michael@0: for (x = 0; x < count; ++x) { michael@0: dst[x] = v8; michael@0: } michael@0: #else michael@0: memset(dst, v8, count); michael@0: #endif michael@0: } michael@0: michael@0: void ARGBSetRows_C(uint8* dst, uint32 v32, int width, michael@0: int dst_stride, int height) { michael@0: int y; michael@0: for (y = 0; y < height; ++y) { michael@0: uint32* d = (uint32*)(dst); michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: d[x] = v32; michael@0: } michael@0: dst += dst_stride; michael@0: } michael@0: } michael@0: michael@0: // Filter 2 rows of YUY2 UV's (422) into U and V (420). michael@0: void YUY2ToUVRow_C(const uint8* src_yuy2, int src_stride_yuy2, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: // Output a row of UV values, filtering 2 rows of YUY2. michael@0: int x; michael@0: for (x = 0; x < width; x += 2) { michael@0: dst_u[0] = (src_yuy2[1] + src_yuy2[src_stride_yuy2 + 1] + 1) >> 1; michael@0: dst_v[0] = (src_yuy2[3] + src_yuy2[src_stride_yuy2 + 3] + 1) >> 1; michael@0: src_yuy2 += 4; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: } michael@0: michael@0: // Copy row of YUY2 UV's (422) into U and V (422). michael@0: void YUY2ToUV422Row_C(const uint8* src_yuy2, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: // Output a row of UV values. michael@0: int x; michael@0: for (x = 0; x < width; x += 2) { michael@0: dst_u[0] = src_yuy2[1]; michael@0: dst_v[0] = src_yuy2[3]; michael@0: src_yuy2 += 4; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: } michael@0: michael@0: // Copy row of YUY2 Y's (422) into Y (420/422). michael@0: void YUY2ToYRow_C(const uint8* src_yuy2, uint8* dst_y, int width) { michael@0: // Output a row of Y values. michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst_y[x] = src_yuy2[0]; michael@0: dst_y[x + 1] = src_yuy2[2]; michael@0: src_yuy2 += 4; michael@0: } michael@0: if (width & 1) { michael@0: dst_y[width - 1] = src_yuy2[0]; michael@0: } michael@0: } michael@0: michael@0: // Filter 2 rows of UYVY UV's (422) into U and V (420). michael@0: void UYVYToUVRow_C(const uint8* src_uyvy, int src_stride_uyvy, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: // Output a row of UV values. michael@0: int x; michael@0: for (x = 0; x < width; x += 2) { michael@0: dst_u[0] = (src_uyvy[0] + src_uyvy[src_stride_uyvy + 0] + 1) >> 1; michael@0: dst_v[0] = (src_uyvy[2] + src_uyvy[src_stride_uyvy + 2] + 1) >> 1; michael@0: src_uyvy += 4; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: } michael@0: michael@0: // Copy row of UYVY UV's (422) into U and V (422). michael@0: void UYVYToUV422Row_C(const uint8* src_uyvy, michael@0: uint8* dst_u, uint8* dst_v, int width) { michael@0: // Output a row of UV values. michael@0: int x; michael@0: for (x = 0; x < width; x += 2) { michael@0: dst_u[0] = src_uyvy[0]; michael@0: dst_v[0] = src_uyvy[2]; michael@0: src_uyvy += 4; michael@0: dst_u += 1; michael@0: dst_v += 1; michael@0: } michael@0: } michael@0: michael@0: // Copy row of UYVY Y's (422) into Y (420/422). michael@0: void UYVYToYRow_C(const uint8* src_uyvy, uint8* dst_y, int width) { michael@0: // Output a row of Y values. michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst_y[x] = src_uyvy[1]; michael@0: dst_y[x + 1] = src_uyvy[3]; michael@0: src_uyvy += 4; michael@0: } michael@0: if (width & 1) { michael@0: dst_y[width - 1] = src_uyvy[1]; michael@0: } michael@0: } michael@0: michael@0: #define BLEND(f, b, a) (((256 - a) * b) >> 8) + f michael@0: michael@0: // Blend src_argb0 over src_argb1 and store to dst_argb. michael@0: // dst_argb may be src_argb0 or src_argb1. michael@0: // This code mimics the SSSE3 version for better testability. michael@0: void ARGBBlendRow_C(const uint8* src_argb0, const uint8* src_argb1, michael@0: uint8* dst_argb, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: uint32 fb = src_argb0[0]; michael@0: uint32 fg = src_argb0[1]; michael@0: uint32 fr = src_argb0[2]; michael@0: uint32 a = src_argb0[3]; michael@0: uint32 bb = src_argb1[0]; michael@0: uint32 bg = src_argb1[1]; michael@0: uint32 br = src_argb1[2]; michael@0: dst_argb[0] = BLEND(fb, bb, a); michael@0: dst_argb[1] = BLEND(fg, bg, a); michael@0: dst_argb[2] = BLEND(fr, br, a); michael@0: dst_argb[3] = 255u; michael@0: michael@0: fb = src_argb0[4 + 0]; michael@0: fg = src_argb0[4 + 1]; michael@0: fr = src_argb0[4 + 2]; michael@0: a = src_argb0[4 + 3]; michael@0: bb = src_argb1[4 + 0]; michael@0: bg = src_argb1[4 + 1]; michael@0: br = src_argb1[4 + 2]; michael@0: dst_argb[4 + 0] = BLEND(fb, bb, a); michael@0: dst_argb[4 + 1] = BLEND(fg, bg, a); michael@0: dst_argb[4 + 2] = BLEND(fr, br, a); michael@0: dst_argb[4 + 3] = 255u; michael@0: src_argb0 += 8; michael@0: src_argb1 += 8; michael@0: dst_argb += 8; michael@0: } michael@0: michael@0: if (width & 1) { michael@0: uint32 fb = src_argb0[0]; michael@0: uint32 fg = src_argb0[1]; michael@0: uint32 fr = src_argb0[2]; michael@0: uint32 a = src_argb0[3]; michael@0: uint32 bb = src_argb1[0]; michael@0: uint32 bg = src_argb1[1]; michael@0: uint32 br = src_argb1[2]; michael@0: dst_argb[0] = BLEND(fb, bb, a); michael@0: dst_argb[1] = BLEND(fg, bg, a); michael@0: dst_argb[2] = BLEND(fr, br, a); michael@0: dst_argb[3] = 255u; michael@0: } michael@0: } michael@0: #undef BLEND michael@0: #define ATTENUATE(f, a) (a | (a << 8)) * (f | (f << 8)) >> 24 michael@0: michael@0: // Multiply source RGB by alpha and store to destination. michael@0: // This code mimics the SSSE3 version for better testability. michael@0: void ARGBAttenuateRow_C(const uint8* src_argb, uint8* dst_argb, int width) { michael@0: int i; michael@0: for (i = 0; i < width - 1; i += 2) { michael@0: uint32 b = src_argb[0]; michael@0: uint32 g = src_argb[1]; michael@0: uint32 r = src_argb[2]; michael@0: uint32 a = src_argb[3]; michael@0: dst_argb[0] = ATTENUATE(b, a); michael@0: dst_argb[1] = ATTENUATE(g, a); michael@0: dst_argb[2] = ATTENUATE(r, a); michael@0: dst_argb[3] = a; michael@0: b = src_argb[4]; michael@0: g = src_argb[5]; michael@0: r = src_argb[6]; michael@0: a = src_argb[7]; michael@0: dst_argb[4] = ATTENUATE(b, a); michael@0: dst_argb[5] = ATTENUATE(g, a); michael@0: dst_argb[6] = ATTENUATE(r, a); michael@0: dst_argb[7] = a; michael@0: src_argb += 8; michael@0: dst_argb += 8; michael@0: } michael@0: michael@0: if (width & 1) { michael@0: const uint32 b = src_argb[0]; michael@0: const uint32 g = src_argb[1]; michael@0: const uint32 r = src_argb[2]; michael@0: const uint32 a = src_argb[3]; michael@0: dst_argb[0] = ATTENUATE(b, a); michael@0: dst_argb[1] = ATTENUATE(g, a); michael@0: dst_argb[2] = ATTENUATE(r, a); michael@0: dst_argb[3] = a; michael@0: } michael@0: } michael@0: #undef ATTENUATE michael@0: michael@0: // Divide source RGB by alpha and store to destination. michael@0: // b = (b * 255 + (a / 2)) / a; michael@0: // g = (g * 255 + (a / 2)) / a; michael@0: // r = (r * 255 + (a / 2)) / a; michael@0: // Reciprocal method is off by 1 on some values. ie 125 michael@0: // 8.8 fixed point inverse table with 1.0 in upper short and 1 / a in lower. michael@0: #define T(a) 0x01000000 + (0x10000 / a) michael@0: const uint32 fixed_invtbl8[256] = { michael@0: 0x01000000, 0x0100ffff, T(0x02), T(0x03), T(0x04), T(0x05), T(0x06), T(0x07), michael@0: T(0x08), T(0x09), T(0x0a), T(0x0b), T(0x0c), T(0x0d), T(0x0e), T(0x0f), michael@0: T(0x10), T(0x11), T(0x12), T(0x13), T(0x14), T(0x15), T(0x16), T(0x17), michael@0: T(0x18), T(0x19), T(0x1a), T(0x1b), T(0x1c), T(0x1d), T(0x1e), T(0x1f), michael@0: T(0x20), T(0x21), T(0x22), T(0x23), T(0x24), T(0x25), T(0x26), T(0x27), michael@0: T(0x28), T(0x29), T(0x2a), T(0x2b), T(0x2c), T(0x2d), T(0x2e), T(0x2f), michael@0: T(0x30), T(0x31), T(0x32), T(0x33), T(0x34), T(0x35), T(0x36), T(0x37), michael@0: T(0x38), T(0x39), T(0x3a), T(0x3b), T(0x3c), T(0x3d), T(0x3e), T(0x3f), michael@0: T(0x40), T(0x41), T(0x42), T(0x43), T(0x44), T(0x45), T(0x46), T(0x47), michael@0: T(0x48), T(0x49), T(0x4a), T(0x4b), T(0x4c), T(0x4d), T(0x4e), T(0x4f), michael@0: T(0x50), T(0x51), T(0x52), T(0x53), T(0x54), T(0x55), T(0x56), T(0x57), michael@0: T(0x58), T(0x59), T(0x5a), T(0x5b), T(0x5c), T(0x5d), T(0x5e), T(0x5f), michael@0: T(0x60), T(0x61), T(0x62), T(0x63), T(0x64), T(0x65), T(0x66), T(0x67), michael@0: T(0x68), T(0x69), T(0x6a), T(0x6b), T(0x6c), T(0x6d), T(0x6e), T(0x6f), michael@0: T(0x70), T(0x71), T(0x72), T(0x73), T(0x74), T(0x75), T(0x76), T(0x77), michael@0: T(0x78), T(0x79), T(0x7a), T(0x7b), T(0x7c), T(0x7d), T(0x7e), T(0x7f), michael@0: T(0x80), T(0x81), T(0x82), T(0x83), T(0x84), T(0x85), T(0x86), T(0x87), michael@0: T(0x88), T(0x89), T(0x8a), T(0x8b), T(0x8c), T(0x8d), T(0x8e), T(0x8f), michael@0: T(0x90), T(0x91), T(0x92), T(0x93), T(0x94), T(0x95), T(0x96), T(0x97), michael@0: T(0x98), T(0x99), T(0x9a), T(0x9b), T(0x9c), T(0x9d), T(0x9e), T(0x9f), michael@0: T(0xa0), T(0xa1), T(0xa2), T(0xa3), T(0xa4), T(0xa5), T(0xa6), T(0xa7), michael@0: T(0xa8), T(0xa9), T(0xaa), T(0xab), T(0xac), T(0xad), T(0xae), T(0xaf), michael@0: T(0xb0), T(0xb1), T(0xb2), T(0xb3), T(0xb4), T(0xb5), T(0xb6), T(0xb7), michael@0: T(0xb8), T(0xb9), T(0xba), T(0xbb), T(0xbc), T(0xbd), T(0xbe), T(0xbf), michael@0: T(0xc0), T(0xc1), T(0xc2), T(0xc3), T(0xc4), T(0xc5), T(0xc6), T(0xc7), michael@0: T(0xc8), T(0xc9), T(0xca), T(0xcb), T(0xcc), T(0xcd), T(0xce), T(0xcf), michael@0: T(0xd0), T(0xd1), T(0xd2), T(0xd3), T(0xd4), T(0xd5), T(0xd6), T(0xd7), michael@0: T(0xd8), T(0xd9), T(0xda), T(0xdb), T(0xdc), T(0xdd), T(0xde), T(0xdf), michael@0: T(0xe0), T(0xe1), T(0xe2), T(0xe3), T(0xe4), T(0xe5), T(0xe6), T(0xe7), michael@0: T(0xe8), T(0xe9), T(0xea), T(0xeb), T(0xec), T(0xed), T(0xee), T(0xef), michael@0: T(0xf0), T(0xf1), T(0xf2), T(0xf3), T(0xf4), T(0xf5), T(0xf6), T(0xf7), michael@0: T(0xf8), T(0xf9), T(0xfa), T(0xfb), T(0xfc), T(0xfd), T(0xfe), 0x01000100 }; michael@0: #undef T michael@0: michael@0: void ARGBUnattenuateRow_C(const uint8* src_argb, uint8* dst_argb, int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: uint32 b = src_argb[0]; michael@0: uint32 g = src_argb[1]; michael@0: uint32 r = src_argb[2]; michael@0: const uint32 a = src_argb[3]; michael@0: const uint32 ia = fixed_invtbl8[a] & 0xffff; // 8.8 fixed point michael@0: b = (b * ia) >> 8; michael@0: g = (g * ia) >> 8; michael@0: r = (r * ia) >> 8; michael@0: // Clamping should not be necessary but is free in assembly. michael@0: dst_argb[0] = clamp255(b); michael@0: dst_argb[1] = clamp255(g); michael@0: dst_argb[2] = clamp255(r); michael@0: dst_argb[3] = a; michael@0: src_argb += 4; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: void ComputeCumulativeSumRow_C(const uint8* row, int32* cumsum, michael@0: const int32* previous_cumsum, int width) { michael@0: int32 row_sum[4] = {0, 0, 0, 0}; michael@0: int x; michael@0: for (x = 0; x < width; ++x) { michael@0: row_sum[0] += row[x * 4 + 0]; michael@0: row_sum[1] += row[x * 4 + 1]; michael@0: row_sum[2] += row[x * 4 + 2]; michael@0: row_sum[3] += row[x * 4 + 3]; michael@0: cumsum[x * 4 + 0] = row_sum[0] + previous_cumsum[x * 4 + 0]; michael@0: cumsum[x * 4 + 1] = row_sum[1] + previous_cumsum[x * 4 + 1]; michael@0: cumsum[x * 4 + 2] = row_sum[2] + previous_cumsum[x * 4 + 2]; michael@0: cumsum[x * 4 + 3] = row_sum[3] + previous_cumsum[x * 4 + 3]; michael@0: } michael@0: } michael@0: michael@0: void CumulativeSumToAverageRow_C(const int32* tl, const int32* bl, michael@0: int w, int area, uint8* dst, int count) { michael@0: float ooa = 1.0f / area; michael@0: int i; michael@0: for (i = 0; i < count; ++i) { michael@0: dst[0] = (uint8)((bl[w + 0] + tl[0] - bl[0] - tl[w + 0]) * ooa); michael@0: dst[1] = (uint8)((bl[w + 1] + tl[1] - bl[1] - tl[w + 1]) * ooa); michael@0: dst[2] = (uint8)((bl[w + 2] + tl[2] - bl[2] - tl[w + 2]) * ooa); michael@0: dst[3] = (uint8)((bl[w + 3] + tl[3] - bl[3] - tl[w + 3]) * ooa); michael@0: dst += 4; michael@0: tl += 4; michael@0: bl += 4; michael@0: } michael@0: } michael@0: michael@0: // Copy pixels from rotated source to destination row with a slope. michael@0: LIBYUV_API michael@0: void ARGBAffineRow_C(const uint8* src_argb, int src_argb_stride, michael@0: uint8* dst_argb, const float* uv_dudv, int width) { michael@0: int i; michael@0: // Render a row of pixels from source into a buffer. michael@0: float uv[2]; michael@0: uv[0] = uv_dudv[0]; michael@0: uv[1] = uv_dudv[1]; michael@0: for (i = 0; i < width; ++i) { michael@0: int x = (int)(uv[0]); michael@0: int y = (int)(uv[1]); michael@0: *(uint32*)(dst_argb) = michael@0: *(const uint32*)(src_argb + y * src_argb_stride + michael@0: x * 4); michael@0: dst_argb += 4; michael@0: uv[0] += uv_dudv[2]; michael@0: uv[1] += uv_dudv[3]; michael@0: } michael@0: } michael@0: michael@0: // Blend 2 rows into 1 for conversions such as I422ToI420. michael@0: void HalfRow_C(const uint8* src_uv, int src_uv_stride, michael@0: uint8* dst_uv, int pix) { michael@0: int x; michael@0: for (x = 0; x < pix; ++x) { michael@0: dst_uv[x] = (src_uv[x] + src_uv[src_uv_stride + x] + 1) >> 1; michael@0: } michael@0: } michael@0: michael@0: // C version 2x2 -> 2x1. michael@0: void InterpolateRow_C(uint8* dst_ptr, const uint8* src_ptr, michael@0: ptrdiff_t src_stride, michael@0: int width, int source_y_fraction) { michael@0: int y1_fraction = source_y_fraction; michael@0: int y0_fraction = 256 - y1_fraction; michael@0: const uint8* src_ptr1 = src_ptr + src_stride; michael@0: int x; michael@0: if (source_y_fraction == 0) { michael@0: memcpy(dst_ptr, src_ptr, width); michael@0: return; michael@0: } michael@0: if (source_y_fraction == 128) { michael@0: HalfRow_C(src_ptr, (int)(src_stride), dst_ptr, width); michael@0: return; michael@0: } michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst_ptr[0] = (src_ptr[0] * y0_fraction + src_ptr1[0] * y1_fraction) >> 8; michael@0: dst_ptr[1] = (src_ptr[1] * y0_fraction + src_ptr1[1] * y1_fraction) >> 8; michael@0: src_ptr += 2; michael@0: src_ptr1 += 2; michael@0: dst_ptr += 2; michael@0: } michael@0: if (width & 1) { michael@0: dst_ptr[0] = (src_ptr[0] * y0_fraction + src_ptr1[0] * y1_fraction) >> 8; michael@0: } michael@0: } michael@0: michael@0: // Select 2 channels from ARGB on alternating pixels. e.g. BGBGBGBG michael@0: void ARGBToBayerRow_C(const uint8* src_argb, michael@0: uint8* dst_bayer, uint32 selector, int pix) { michael@0: int index0 = selector & 0xff; michael@0: int index1 = (selector >> 8) & 0xff; michael@0: // Copy a row of Bayer. michael@0: int x; michael@0: for (x = 0; x < pix - 1; x += 2) { michael@0: dst_bayer[0] = src_argb[index0]; michael@0: dst_bayer[1] = src_argb[index1]; michael@0: src_argb += 8; michael@0: dst_bayer += 2; michael@0: } michael@0: if (pix & 1) { michael@0: dst_bayer[0] = src_argb[index0]; michael@0: } michael@0: } michael@0: michael@0: // Select G channel from ARGB. e.g. GGGGGGGG michael@0: void ARGBToBayerGGRow_C(const uint8* src_argb, michael@0: uint8* dst_bayer, uint32 selector, int pix) { michael@0: // Copy a row of G. michael@0: int x; michael@0: for (x = 0; x < pix - 1; x += 2) { michael@0: dst_bayer[0] = src_argb[1]; michael@0: dst_bayer[1] = src_argb[5]; michael@0: src_argb += 8; michael@0: dst_bayer += 2; michael@0: } michael@0: if (pix & 1) { michael@0: dst_bayer[0] = src_argb[1]; michael@0: } michael@0: } michael@0: michael@0: // Use first 4 shuffler values to reorder ARGB channels. michael@0: void ARGBShuffleRow_C(const uint8* src_argb, uint8* dst_argb, michael@0: const uint8* shuffler, int pix) { michael@0: int index0 = shuffler[0]; michael@0: int index1 = shuffler[1]; michael@0: int index2 = shuffler[2]; michael@0: int index3 = shuffler[3]; michael@0: // Shuffle a row of ARGB. michael@0: int x; michael@0: for (x = 0; x < pix; ++x) { michael@0: // To support in-place conversion. michael@0: uint8 b = src_argb[index0]; michael@0: uint8 g = src_argb[index1]; michael@0: uint8 r = src_argb[index2]; michael@0: uint8 a = src_argb[index3]; michael@0: dst_argb[0] = b; michael@0: dst_argb[1] = g; michael@0: dst_argb[2] = r; michael@0: dst_argb[3] = a; michael@0: src_argb += 4; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: void I422ToYUY2Row_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* dst_frame, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst_frame[0] = src_y[0]; michael@0: dst_frame[1] = src_u[0]; michael@0: dst_frame[2] = src_y[1]; michael@0: dst_frame[3] = src_v[0]; michael@0: dst_frame += 4; michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: } michael@0: if (width & 1) { michael@0: dst_frame[0] = src_y[0]; michael@0: dst_frame[1] = src_u[0]; michael@0: dst_frame[2] = src_y[0]; // duplicate last y michael@0: dst_frame[3] = src_v[0]; michael@0: } michael@0: } michael@0: michael@0: void I422ToUYVYRow_C(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* dst_frame, int width) { michael@0: int x; michael@0: for (x = 0; x < width - 1; x += 2) { michael@0: dst_frame[0] = src_u[0]; michael@0: dst_frame[1] = src_y[0]; michael@0: dst_frame[2] = src_v[0]; michael@0: dst_frame[3] = src_y[1]; michael@0: dst_frame += 4; michael@0: src_y += 2; michael@0: src_u += 1; michael@0: src_v += 1; michael@0: } michael@0: if (width & 1) { michael@0: dst_frame[0] = src_u[0]; michael@0: dst_frame[1] = src_y[0]; michael@0: dst_frame[2] = src_v[0]; michael@0: dst_frame[3] = src_y[0]; // duplicate last y michael@0: } michael@0: } michael@0: michael@0: #if !defined(LIBYUV_DISABLE_X86) && defined(HAS_I422TOARGBROW_SSSE3) michael@0: // row_win.cc has asm version, but GCC uses 2 step wrapper. michael@0: #if defined(__x86_64__) || defined(__i386__) michael@0: void I422ToRGB565Row_SSSE3(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: // Allocate a row of ARGB. michael@0: align_buffer_64(row, width * 4); michael@0: I422ToARGBRow_SSSE3(src_y, src_u, src_v, row, width); michael@0: ARGBToRGB565Row_SSE2(row, rgb_buf, width); michael@0: free_aligned_buffer_64(row); michael@0: } michael@0: #endif // defined(__x86_64__) || defined(__i386__) michael@0: michael@0: #if defined(_M_IX86) || defined(__x86_64__) || defined(__i386__) michael@0: void I422ToARGB1555Row_SSSE3(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: // Allocate a row of ARGB. michael@0: align_buffer_64(row, width * 4); michael@0: I422ToARGBRow_SSSE3(src_y, src_u, src_v, row, width); michael@0: ARGBToARGB1555Row_SSE2(row, rgb_buf, width); michael@0: free_aligned_buffer_64(row); michael@0: } michael@0: michael@0: void I422ToARGB4444Row_SSSE3(const uint8* src_y, michael@0: const uint8* src_u, michael@0: const uint8* src_v, michael@0: uint8* rgb_buf, michael@0: int width) { michael@0: // Allocate a row of ARGB. michael@0: align_buffer_64(row, width * 4); michael@0: I422ToARGBRow_SSSE3(src_y, src_u, src_v, row, width); michael@0: ARGBToARGB4444Row_SSE2(row, rgb_buf, width); michael@0: free_aligned_buffer_64(row); michael@0: } michael@0: michael@0: void NV12ToRGB565Row_SSSE3(const uint8* src_y, michael@0: const uint8* src_uv, michael@0: uint8* dst_rgb565, michael@0: int width) { michael@0: // Allocate a row of ARGB. michael@0: align_buffer_64(row, width * 4); michael@0: NV12ToARGBRow_SSSE3(src_y, src_uv, row, width); michael@0: ARGBToRGB565Row_SSE2(row, dst_rgb565, width); michael@0: free_aligned_buffer_64(row); michael@0: } michael@0: michael@0: void NV21ToRGB565Row_SSSE3(const uint8* src_y, michael@0: const uint8* src_vu, michael@0: uint8* dst_rgb565, michael@0: int width) { michael@0: // Allocate a row of ARGB. michael@0: align_buffer_64(row, width * 4); michael@0: NV21ToARGBRow_SSSE3(src_y, src_vu, row, width); michael@0: ARGBToRGB565Row_SSE2(row, dst_rgb565, width); michael@0: free_aligned_buffer_64(row); michael@0: } michael@0: michael@0: void YUY2ToARGBRow_SSSE3(const uint8* src_yuy2, michael@0: uint8* dst_argb, michael@0: int width) { michael@0: // Allocate a rows of yuv. michael@0: align_buffer_64(row_y, ((width + 63) & ~63) * 2); michael@0: uint8* row_u = row_y + ((width + 63) & ~63); michael@0: uint8* row_v = row_u + ((width + 63) & ~63) / 2; michael@0: YUY2ToUV422Row_SSE2(src_yuy2, row_u, row_v, width); michael@0: YUY2ToYRow_SSE2(src_yuy2, row_y, width); michael@0: I422ToARGBRow_SSSE3(row_y, row_u, row_v, dst_argb, width); michael@0: free_aligned_buffer_64(row_y); michael@0: } michael@0: michael@0: void YUY2ToARGBRow_Unaligned_SSSE3(const uint8* src_yuy2, michael@0: uint8* dst_argb, michael@0: int width) { michael@0: // Allocate a rows of yuv. michael@0: align_buffer_64(row_y, ((width + 63) & ~63) * 2); michael@0: uint8* row_u = row_y + ((width + 63) & ~63); michael@0: uint8* row_v = row_u + ((width + 63) & ~63) / 2; michael@0: YUY2ToUV422Row_Unaligned_SSE2(src_yuy2, row_u, row_v, width); michael@0: YUY2ToYRow_Unaligned_SSE2(src_yuy2, row_y, width); michael@0: I422ToARGBRow_Unaligned_SSSE3(row_y, row_u, row_v, dst_argb, width); michael@0: free_aligned_buffer_64(row_y); michael@0: } michael@0: michael@0: void UYVYToARGBRow_SSSE3(const uint8* src_uyvy, michael@0: uint8* dst_argb, michael@0: int width) { michael@0: // Allocate a rows of yuv. michael@0: align_buffer_64(row_y, ((width + 63) & ~63) * 2); michael@0: uint8* row_u = row_y + ((width + 63) & ~63); michael@0: uint8* row_v = row_u + ((width + 63) & ~63) / 2; michael@0: UYVYToUV422Row_SSE2(src_uyvy, row_u, row_v, width); michael@0: UYVYToYRow_SSE2(src_uyvy, row_y, width); michael@0: I422ToARGBRow_SSSE3(row_y, row_u, row_v, dst_argb, width); michael@0: free_aligned_buffer_64(row_y); michael@0: } michael@0: michael@0: void UYVYToARGBRow_Unaligned_SSSE3(const uint8* src_uyvy, michael@0: uint8* dst_argb, michael@0: int width) { michael@0: // Allocate a rows of yuv. michael@0: align_buffer_64(row_y, ((width + 63) & ~63) * 2); michael@0: uint8* row_u = row_y + ((width + 63) & ~63); michael@0: uint8* row_v = row_u + ((width + 63) & ~63) / 2; michael@0: UYVYToUV422Row_Unaligned_SSE2(src_uyvy, row_u, row_v, width); michael@0: UYVYToYRow_Unaligned_SSE2(src_uyvy, row_y, width); michael@0: I422ToARGBRow_Unaligned_SSSE3(row_y, row_u, row_v, dst_argb, width); michael@0: free_aligned_buffer_64(row_y); michael@0: } michael@0: michael@0: #endif // defined(_M_IX86) || defined(__x86_64__) || defined(__i386__) michael@0: #endif // !defined(LIBYUV_DISABLE_X86) michael@0: michael@0: void ARGBPolynomialRow_C(const uint8* src_argb, michael@0: uint8* dst_argb, const float* poly, michael@0: int width) { michael@0: int i; michael@0: for (i = 0; i < width; ++i) { michael@0: float b = (float)(src_argb[0]); michael@0: float g = (float)(src_argb[1]); michael@0: float r = (float)(src_argb[2]); michael@0: float a = (float)(src_argb[3]); michael@0: float b2 = b * b; michael@0: float g2 = g * g; michael@0: float r2 = r * r; michael@0: float a2 = a * a; michael@0: float db = poly[0] + poly[4] * b; michael@0: float dg = poly[1] + poly[5] * g; michael@0: float dr = poly[2] + poly[6] * r; michael@0: float da = poly[3] + poly[7] * a; michael@0: float b3 = b2 * b; michael@0: float g3 = g2 * g; michael@0: float r3 = r2 * r; michael@0: float a3 = a2 * a; michael@0: db += poly[8] * b2; michael@0: dg += poly[9] * g2; michael@0: dr += poly[10] * r2; michael@0: da += poly[11] * a2; michael@0: db += poly[12] * b3; michael@0: dg += poly[13] * g3; michael@0: dr += poly[14] * r3; michael@0: da += poly[15] * a3; michael@0: michael@0: dst_argb[0] = Clamp((int32)(db)); michael@0: dst_argb[1] = Clamp((int32)(dg)); michael@0: dst_argb[2] = Clamp((int32)(dr)); michael@0: dst_argb[3] = Clamp((int32)(da)); michael@0: src_argb += 4; michael@0: dst_argb += 4; michael@0: } michael@0: } michael@0: michael@0: void ARGBLumaColorTableRow_C(const uint8* src_argb, uint8* dst_argb, int width, michael@0: const uint8* luma, uint32 lumacoeff) { michael@0: uint32 bc = lumacoeff & 0xff; michael@0: uint32 gc = (lumacoeff >> 8) & 0xff; michael@0: uint32 rc = (lumacoeff >> 16) & 0xff; michael@0: michael@0: int i; michael@0: for (i = 0; i < width - 1; i += 2) { michael@0: // Luminance in rows, color values in columns. michael@0: const uint8* luma0 = ((src_argb[0] * bc + src_argb[1] * gc + michael@0: src_argb[2] * rc) & 0x7F00u) + luma; michael@0: const uint8* luma1; michael@0: dst_argb[0] = luma0[src_argb[0]]; michael@0: dst_argb[1] = luma0[src_argb[1]]; michael@0: dst_argb[2] = luma0[src_argb[2]]; michael@0: dst_argb[3] = src_argb[3]; michael@0: luma1 = ((src_argb[4] * bc + src_argb[5] * gc + michael@0: src_argb[6] * rc) & 0x7F00u) + luma; michael@0: dst_argb[4] = luma1[src_argb[4]]; michael@0: dst_argb[5] = luma1[src_argb[5]]; michael@0: dst_argb[6] = luma1[src_argb[6]]; michael@0: dst_argb[7] = src_argb[7]; michael@0: src_argb += 8; michael@0: dst_argb += 8; michael@0: } michael@0: if (width & 1) { michael@0: // Luminance in rows, color values in columns. michael@0: const uint8* luma0 = ((src_argb[0] * bc + src_argb[1] * gc + michael@0: src_argb[2] * rc) & 0x7F00u) + luma; michael@0: dst_argb[0] = luma0[src_argb[0]]; michael@0: dst_argb[1] = luma0[src_argb[1]]; michael@0: dst_argb[2] = luma0[src_argb[2]]; michael@0: dst_argb[3] = src_argb[3]; michael@0: } michael@0: } michael@0: michael@0: void ARGBCopyAlphaRow_C(const uint8* src, uint8* dst, int width) { michael@0: int i; michael@0: for (i = 0; i < width - 1; i += 2) { michael@0: dst[3] = src[3]; michael@0: dst[7] = src[7]; michael@0: dst += 8; michael@0: src += 8; michael@0: } michael@0: if (width & 1) { michael@0: dst[3] = src[3]; michael@0: } michael@0: } michael@0: michael@0: void ARGBCopyYToAlphaRow_C(const uint8* src, uint8* dst, int width) { michael@0: int i; michael@0: for (i = 0; i < width - 1; i += 2) { michael@0: dst[3] = src[0]; michael@0: dst[7] = src[1]; michael@0: dst += 8; michael@0: src += 2; michael@0: } michael@0: if (width & 1) { michael@0: dst[3] = src[0]; michael@0: } michael@0: } michael@0: michael@0: #ifdef __cplusplus michael@0: } // extern "C" michael@0: } // namespace libyuv michael@0: #endif