// Copyright 2014 Google Inc. All Rights Reserved. // // Use of this source code is governed by a BSD-style license // that can be found in the COPYING file in the root of the source // tree. An additional intellectual property rights grant can be found // in the file PATENTS. All contributing project authors may // be found in the AUTHORS file in the root of the source tree. // ----------------------------------------------------------------------------- // // YUV->RGB conversion functions // // Author: Skal ([email protected]) #include "src/dsp/yuv.h" #if defined(WEBP_USE_SSE2) #include <stdlib.h> #include <emmintrin.h> #include "src/dsp/common_sse2.h" #include "src/utils/utils.h" //----------------------------------------------------------------------------- // Convert spans of 32 pixels to various RGB formats for the fancy upsampler. // These constants are 14b fixed-point version of ITU-R BT.601 constants. // R = (19077 * y + 26149 * v - 14234) >> 6 // G = (19077 * y - 6419 * u - 13320 * v + 8708) >> 6 // B = (19077 * y + 33050 * u - 17685) >> 6 static void ConvertYUV444ToRGB_SSE2(const __m128i* const Y0, const __m128i* const U0, const __m128i* const V0, __m128i* const R, __m128i* const G, __m128i* const B) { … } // Load the bytes into the *upper* part of 16b words. That's "<< 8", basically. static WEBP_INLINE __m128i Load_HI_16_SSE2(const uint8_t* src) { … } // Load and replicate the U/V samples static WEBP_INLINE __m128i Load_UV_HI_8_SSE2(const uint8_t* src) { … } // Convert 32 samples of YUV444 to R/G/B static void YUV444ToRGB_SSE2(const uint8_t* const y, const uint8_t* const u, const uint8_t* const v, __m128i* const R, __m128i* const G, __m128i* const B) { … } // Convert 32 samples of YUV420 to R/G/B static void YUV420ToRGB_SSE2(const uint8_t* const y, const uint8_t* const u, const uint8_t* const v, __m128i* const R, __m128i* const G, __m128i* const B) { … } // Pack R/G/B/A results into 32b output. static WEBP_INLINE void PackAndStore4_SSE2(const __m128i* const R, const __m128i* const G, const __m128i* const B, const __m128i* const A, uint8_t* const dst) { … } // Pack R/G/B/A results into 16b output. static WEBP_INLINE void PackAndStore4444_SSE2(const __m128i* const R, const __m128i* const G, const __m128i* const B, const __m128i* const A, uint8_t* const dst) { … } // Pack R/G/B results into 16b output. static WEBP_INLINE void PackAndStore565_SSE2(const __m128i* const R, const __m128i* const G, const __m128i* const B, uint8_t* const dst) { … } // Pack the planar buffers // rrrr... rrrr... gggg... gggg... bbbb... bbbb.... // triplet by triplet in the output buffer rgb as rgbrgbrgbrgb ... static WEBP_INLINE void PlanarTo24b_SSE2(__m128i* const in0, __m128i* const in1, __m128i* const in2, __m128i* const in3, __m128i* const in4, __m128i* const in5, uint8_t* const rgb) { … } void VP8YuvToRgba32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst) { … } void VP8YuvToBgra32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst) { … } void VP8YuvToArgb32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst) { … } void VP8YuvToRgba444432_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst) { … } void VP8YuvToRgb56532_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst) { … } void VP8YuvToRgb32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst) { … } void VP8YuvToBgr32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst) { … } //----------------------------------------------------------------------------- // Arbitrary-length row conversion functions static void YuvToRgbaRow_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst, int len) { … } static void YuvToBgraRow_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst, int len) { … } static void YuvToArgbRow_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst, int len) { … } static void YuvToRgbRow_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst, int len) { … } static void YuvToBgrRow_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v, uint8_t* dst, int len) { … } //------------------------------------------------------------------------------ // Entry point extern void WebPInitSamplersSSE2(void); WEBP_TSAN_IGNORE_FUNCTION void WebPInitSamplersSSE2(void) { … } //------------------------------------------------------------------------------ // RGB24/32 -> YUV converters // Load eight 16b-words from *src. #define LOAD_16(src) … // Store either 16b-words into *dst #define STORE_16(V, dst) … // Function that inserts a value of the second half of the in buffer in between // every two char of the first half. static WEBP_INLINE void RGB24PackedToPlanarHelper_SSE2( const __m128i* const in /*in[6]*/, __m128i* const out /*out[6]*/) { … } // Unpack the 8b input rgbrgbrgbrgb ... as contiguous registers: // rrrr... rrrr... gggg... gggg... bbbb... bbbb.... // Similar to PlanarTo24bHelper(), but in reverse order. static WEBP_INLINE void RGB24PackedToPlanar_SSE2( const uint8_t* const rgb, __m128i* const out /*out[6]*/) { … } // Convert 8 packed ARGB to r[], g[], b[] static WEBP_INLINE void RGB32PackedToPlanar_SSE2(const uint32_t* const argb, __m128i* const rgb /*in[6]*/) { … } // This macro computes (RG * MULT_RG + GB * MULT_GB + ROUNDER) >> DESCALE_FIX // It's a macro and not a function because we need to use immediate values with // srai_epi32, e.g. #define TRANSFORM … #define MK_CST_16 … static WEBP_INLINE void ConvertRGBToY_SSE2(const __m128i* const R, const __m128i* const G, const __m128i* const B, __m128i* const Y) { … } static WEBP_INLINE void ConvertRGBToUV_SSE2(const __m128i* const R, const __m128i* const G, const __m128i* const B, __m128i* const U, __m128i* const V) { … } #undef MK_CST_16 #undef TRANSFORM static void ConvertRGB24ToY_SSE2(const uint8_t* rgb, uint8_t* y, int width) { … } static void ConvertBGR24ToY_SSE2(const uint8_t* bgr, uint8_t* y, int width) { … } static void ConvertARGBToY_SSE2(const uint32_t* argb, uint8_t* y, int width) { … } // Horizontal add (doubled) of two 16b values, result is 16b. // in: A | B | C | D | ... -> out: 2*(A+B) | 2*(C+D) | ... static void HorizontalAddPack_SSE2(const __m128i* const A, const __m128i* const B, __m128i* const out) { … } static void ConvertARGBToUV_SSE2(const uint32_t* argb, uint8_t* u, uint8_t* v, int src_width, int do_store) { … } // Convert 16 packed ARGB 16b-values to r[], g[], b[] static WEBP_INLINE void RGBA32PackedToPlanar_16b_SSE2( const uint16_t* const rgbx, __m128i* const r, __m128i* const g, __m128i* const b) { … } static void ConvertRGBA32ToUV_SSE2(const uint16_t* rgb, uint8_t* u, uint8_t* v, int width) { … } //------------------------------------------------------------------------------ extern void WebPInitConvertARGBToYUVSSE2(void); WEBP_TSAN_IGNORE_FUNCTION void WebPInitConvertARGBToYUVSSE2(void) { … } #else // !WEBP_USE_SSE2 WEBP_DSP_INIT_STUB(WebPInitSamplersSSE2) WEBP_DSP_INIT_STUB(WebPInitConvertARGBToYUVSSE2) #endif // WEBP_USE_SSE2
Codebase Browser
godot