/* * Colorspace conversion ( 64 - bit Arm Neon ) * * Copyright ( C ) 2020 , Arm Limited . All Rights Reserved . * * This software is provided ' as - is ' , without any express or implied * warranty . In no event will the authors be held liable for any damages * arising from the use of this software . * * Permission is granted to anyone to use this software for any purpose , * including commercial applications , and to alter it and redistribute it * freely , subject to the following restrictions : * * 1 . The origin of this software must not be misrepresented ; you must not * claim that you wrote the original software . If you use this software * in a product , an acknowledgment in the product documentation would be * appreciated but is not required . * 2 . Altered source versions must be plainly marked as such , and must not be * misrepresented as being the original software . * 3 . This notice may not be removed or altered from any source distribution . /* This file is included by jccolor-neon.c */ /* RGB -> YCbCr conversion is defined by the following equations: * Y = 0 . 29900 * R + 0 . 58700 * G + 0 . 11400 * B * Cb = - 0 . 16874 * R - 0 . 33126 * G + 0 . 50000 * B + 128 * Cr = 0 . 50000 * R - 0 . 41869 * G - 0 . 08131 * B + 128 * * Avoid floating point arithmetic by using shifted integer constants : * 0 . 29899597 = 19595 * 2 ^ - 16 * 0 . 58700561 = 38470 * 2 ^ - 16 * 0 . 11399841 = 7471 * 2 ^ - 16 * 0 . 16874695 = 11059 * 2 ^ - 16 * 0 . 33125305 = 21709 * 2 ^ - 16 * 0 . 50000000 = 32768 * 2 ^ - 16 * 0 . 41868592 = 27439 * 2 ^ - 16 * 0 . 08131409 = 5329 * 2 ^ - 16 * These constants are defined in jccolor - neon . c * * We add the fixed - point equivalent of 0 . 5 to Cb and Cr , which effectively * rounds up or down the result via integer truncation . void jsimd_rgb_ycc_convert_neon(JDIMENSION image_width, JSAMPARRAY input_buf,int num_rows)/* Pointer to RGB(X/A) input data */ /* Pointers to Y, Cb, and Cr output data */ /* Allocate temporary buffer for final (image_width % 16) pixels in row. */ 16 ) uint8_t tmp_buf[16 * RGB_PIXELSIZE];/* Set up conversion constants. */ const uint16x8_t consts = vld1q_u16(jsimd_rgb_ycc_neon_consts);const uint32x4_t scaled_128_5 = vdupq_n_u32((128 << 16 ) + 32767 );while (--num_rows >= 0 ) {0 ][output_row];1 ][output_row];2 ][output_row];int cols_remaining = image_width;for (; cols_remaining >= 16 ; cols_remaining -= 16 ) {#if RGB_PIXELSIZE == 4 #else #endif /* Compute Y = 0.29900 * R + 0.58700 * G + 0.11400 * B */ 0 );1 );2 );0 );1 );2 );0 );1 );2 );0 );1 );2 );/* Compute Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + 128 */ 3 );4 );5 );3 );4 );5 );3 );4 );5 );3 );4 );5 );/* Compute Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + 128 */ 5 );6 );7 );5 );6 );7 );5 );6 );7 );5 );6 );7 );/* Descale Y values (rounding right shift) and narrow to 16-bit. */ 16 ),16 ));16 ),16 ));/* Descale Cb values (right shift) and narrow to 16-bit. */ 16 ),16 ));16 ),16 ));/* Descale Cr values (right shift) and narrow to 16-bit. */ 16 ),16 ));16 ),16 ));/* Narrow Y, Cb, and Cr values to 8-bit and store to memory. Buffer * overwrite is permitted up to the next multiple of ALIGN_SIZE bytes . /* Increment pointers. */ 16 * RGB_PIXELSIZE);16 ;16 ;16 ;if (cols_remaining > 8 ) {/* To prevent buffer overread by the vector load instructions, the last * ( image_width % 16 ) columns of data are first memcopied to a temporary * buffer large enough to accommodate the vector load . #if RGB_PIXELSIZE == 4 #else #endif /* Compute Y = 0.29900 * R + 0.58700 * G + 0.11400 * B */ 0 );1 );2 );0 );1 );2 );0 );1 );2 );0 );1 );2 );/* Compute Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + 128 */ 3 );4 );5 );3 );4 );5 );3 );4 );5 );3 );4 );5 );/* Compute Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + 128 */ 5 );6 );7 );5 );6 );7 );5 );6 );7 );5 );6 );7 );/* Descale Y values (rounding right shift) and narrow to 16-bit. */ 16 ),16 ));16 ),16 ));/* Descale Cb values (right shift) and narrow to 16-bit. */ 16 ),16 ));16 ),16 ));/* Descale Cr values (right shift) and narrow to 16-bit. */ 16 ),16 ));16 ),16 ));/* Narrow Y, Cb, and Cr values to 8-bit and store to memory. Buffer * overwrite is permitted up to the next multiple of ALIGN_SIZE bytes . else if (cols_remaining > 0 ) {/* To prevent buffer overread by the vector load instructions, the last * ( image_width % 8 ) columns of data are first memcopied to a temporary * buffer large enough to accommodate the vector load . #if RGB_PIXELSIZE == 4 #else #endif /* Compute Y = 0.29900 * R + 0.58700 * G + 0.11400 * B */ 0 );1 );2 );0 );1 );2 );/* Compute Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + 128 */ 3 );4 );5 );3 );4 );5 );/* Compute Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + 128 */ 5 );6 );7 );5 );6 );7 );/* Descale Y values (rounding right shift) and narrow to 16-bit. */ 16 ),16 ));/* Descale Cb values (right shift) and narrow to 16-bit. */ 16 ),16 ));/* Descale Cr values (right shift) and narrow to 16-bit. */ 16 ),16 ));/* Narrow Y, Cb, and Cr values to 8-bit and store to memory. Buffer * overwrite is permitted up to the next multiple of ALIGN_SIZE bytes .
Messung V0.5 in Prozent C=94 H=99 G=96
¤ Dauer der Verarbeitung: 0.6 Sekunden
¤ *© Formatika GbR, Deutschland
