FFmpeg: libswscale/sparc/yuv2rgb

00001 /*
00002  * VIS optimized software YUV to RGB converter
00003  * Copyright (c) 2007 Denes Balatoni <dbalatoni@programozo.hu>
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00022 #include <inttypes.h>
00023 #include <stdlib.h>
00024 
00025 #include "libavutil/attributes.h"
00026 #include "libswscale/swscale.h"
00027 #include "libswscale/swscale_internal.h"
00028 
00029 #define YUV2RGB_INIT                               \
00030     "wr %%g0, 0x10, %%gsr \n\t"                    \
00031     "ldd [%5],      %%f32 \n\t"                    \
00032     "ldd [%5 +  8], %%f34 \n\t"                    \
00033     "ldd [%5 + 16], %%f36 \n\t"                    \
00034     "ldd [%5 + 24], %%f38 \n\t"                    \
00035     "ldd [%5 + 32], %%f40 \n\t"                    \
00036     "ldd [%5 + 40], %%f42 \n\t"                    \
00037     "ldd [%5 + 48], %%f44 \n\t"                    \
00038     "ldd [%5 + 56], %%f46 \n\t"                    \
00039     "ldd [%5 + 64], %%f48 \n\t"                    \
00040     "ldd [%5 + 72], %%f50 \n\t"
00041 
00042 #define YUV2RGB_KERNEL                             \
00043     /* ^^^^ f0=Y f3=u f5=v */                      \
00044     "fmul8x16 %%f3,  %%f48,  %%f6 \n\t"            \
00045     "fmul8x16 %%f19, %%f48, %%f22 \n\t"            \
00046     "fmul8x16 %%f5,  %%f44,  %%f8 \n\t"            \
00047     "fmul8x16 %%f21, %%f44, %%f24 \n\t"            \
00048     "fmul8x16 %%f0,  %%f42,  %%f0 \n\t"            \
00049     "fmul8x16 %%f16, %%f42, %%f16 \n\t"            \
00050     "fmul8x16 %%f3,  %%f50,  %%f2 \n\t"            \
00051     "fmul8x16 %%f19, %%f50, %%f18 \n\t"            \
00052     "fmul8x16 %%f5,  %%f46,  %%f4 \n\t"            \
00053     "fmul8x16 %%f21, %%f46, %%f20 \n\t"            \
00054                                                    \
00055     "fpsub16 %%f6,  %%f34,  %%f6 \n\t" /* 1 */     \
00056     "fpsub16 %%f22, %%f34, %%f22 \n\t" /* 1 */     \
00057     "fpsub16 %%f8,  %%f38,  %%f8 \n\t" /* 3 */     \
00058     "fpsub16 %%f24, %%f38, %%f24 \n\t" /* 3 */     \
00059     "fpsub16 %%f0,  %%f32,  %%f0 \n\t" /* 0 */     \
00060     "fpsub16 %%f16, %%f32, %%f16 \n\t" /* 0 */     \
00061     "fpsub16 %%f2,  %%f36,  %%f2 \n\t" /* 2 */     \
00062     "fpsub16 %%f18, %%f36, %%f18 \n\t" /* 2 */     \
00063     "fpsub16 %%f4,  %%f40,  %%f4 \n\t" /* 4 */     \
00064     "fpsub16 %%f20, %%f40, %%f20 \n\t" /* 4 */     \
00065                                                    \
00066     "fpadd16 %%f0,  %%f8,  %%f8  \n\t" /* Gt */    \
00067     "fpadd16 %%f16, %%f24, %%f24 \n\t" /* Gt */    \
00068     "fpadd16 %%f0,  %%f4,  %%f4  \n\t" /* R */     \
00069     "fpadd16 %%f16, %%f20, %%f20 \n\t" /* R */     \
00070     "fpadd16 %%f0,  %%f6,  %%f6  \n\t" /* B */     \
00071     "fpadd16 %%f16, %%f22, %%f22 \n\t" /* B */     \
00072     "fpadd16 %%f8,  %%f2,  %%f2  \n\t" /* G */     \
00073     "fpadd16 %%f24, %%f18, %%f18 \n\t" /* G */     \
00074                                                    \
00075     "fpack16 %%f4,  %%f4  \n\t"                    \
00076     "fpack16 %%f20, %%f20 \n\t"                    \
00077     "fpack16 %%f6,  %%f6  \n\t"                    \
00078     "fpack16 %%f22, %%f22 \n\t"                    \
00079     "fpack16 %%f2,  %%f2  \n\t"                    \
00080     "fpack16 %%f18, %%f18 \n\t"
00081 
00082 // FIXME: must be changed to set alpha to 255 instead of 0
00083 static int vis_420P_ARGB32(SwsContext *c, uint8_t *src[], int srcStride[],
00084                            int srcSliceY, int srcSliceH,
00085                            uint8_t *dst[], int dstStride[])
00086 {
00087     int y, out1, out2, out3, out4, out5, out6;
00088 
00089     for (y = 0; y < srcSliceH; ++y)
00090         __asm__ volatile (
00091             YUV2RGB_INIT
00092             "wr %%g0, 0xd2, %%asi        \n\t"  /* ASI_FL16_P */
00093             "1:                          \n\t"
00094             "ldda [%1]     %%asi, %%f2   \n\t"
00095             "ldda [%1 + 2] %%asi, %%f18  \n\t"
00096             "ldda [%2]     %%asi, %%f4   \n\t"
00097             "ldda [%2 + 2] %%asi, %%f20  \n\t"
00098             "ld [%0], %%f0               \n\t"
00099             "ld [%0+4], %%f16            \n\t"
00100             "fpmerge %%f3,  %%f3,  %%f2  \n\t"
00101             "fpmerge %%f19, %%f19, %%f18 \n\t"
00102             "fpmerge %%f5,  %%f5,  %%f4  \n\t"
00103             "fpmerge %%f21, %%f21, %%f20 \n\t"
00104             YUV2RGB_KERNEL
00105             "fzero %%f0                  \n\t"
00106             "fpmerge %%f4,  %%f6,  %%f8  \n\t"  // r, b, t1
00107             "fpmerge %%f20, %%f22, %%f24 \n\t"  // r, b, t1
00108             "fpmerge %%f0,  %%f2,  %%f10 \n\t"  // 0, g, t2
00109             "fpmerge %%f0,  %%f18, %%f26 \n\t"  // 0, g, t2
00110             "fpmerge %%f10, %%f8,  %%f4  \n\t"  // t2, t1, msb
00111             "fpmerge %%f26, %%f24, %%f20 \n\t"  // t2, t1, msb
00112             "fpmerge %%f11, %%f9,  %%f6  \n\t"  // t2, t1, lsb
00113             "fpmerge %%f27, %%f25, %%f22 \n\t"  // t2, t1, lsb
00114             "std %%f4,  [%3]             \n\t"
00115             "std %%f20, [%3 + 16]        \n\t"
00116             "std %%f6,  [%3 +  8]        \n\t"
00117             "std %%f22, [%3 + 24]        \n\t"
00118 
00119             "add %0, 8, %0   \n\t"
00120             "add %1, 4, %1   \n\t"
00121             "add %2, 4, %2   \n\t"
00122             "subcc %4, 8, %4 \n\t"
00123             "bne 1b          \n\t"
00124             "add %3, 32, %3  \n\t"              // delay slot
00125             : "=r" (out1), "=r" (out2), "=r" (out3), "=r" (out4), "=r" (out5), "=r" (out6)
00126             : "0" (src[0] + (y + srcSliceY) * srcStride[0]), "1" (src[1] + ((y + srcSliceY) >> 1) * srcStride[1]),
00127             "2" (src[2] + ((y + srcSliceY) >> 1) * srcStride[2]), "3" (dst[0] + (y + srcSliceY) * dstStride[0]),
00128             "4" (c->dstW),
00129             "5" (c->sparc_coeffs)
00130             );
00131 
00132     return srcSliceH;
00133 }
00134 
00135 // FIXME: must be changed to set alpha to 255 instead of 0
00136 static int vis_422P_ARGB32(SwsContext *c, uint8_t *src[], int srcStride[],
00137                            int srcSliceY, int srcSliceH,
00138                            uint8_t *dst[], int dstStride[])
00139 {
00140     int y, out1, out2, out3, out4, out5, out6;
00141 
00142     for (y = 0; y < srcSliceH; ++y)
00143         __asm__ volatile (
00144             YUV2RGB_INIT
00145             "wr %%g0, 0xd2, %%asi        \n\t" /* ASI_FL16_P */
00146             "1:                          \n\t"
00147             "ldda [%1]     %%asi, %%f2   \n\t"
00148             "ldda [%1 + 2] %%asi, %%f18  \n\t"
00149             "ldda [%2]     %%asi, %%f4   \n\t"
00150             "ldda [%2 + 2] %%asi, %%f20  \n\t"
00151             "ld [%0],     %%f0           \n\t"
00152             "ld [%0 + 4], %%f16          \n\t"
00153             "fpmerge %%f3,  %%f3,  %%f2  \n\t"
00154             "fpmerge %%f19, %%f19, %%f18 \n\t"
00155             "fpmerge %%f5,  %%f5,  %%f4  \n\t"
00156             "fpmerge %%f21, %%f21, %%f20 \n\t"
00157             YUV2RGB_KERNEL
00158             "fzero %%f0 \n\t"
00159             "fpmerge %%f4,  %%f6,  %%f8  \n\t"  // r,b,t1
00160             "fpmerge %%f20, %%f22, %%f24 \n\t"  // r,b,t1
00161             "fpmerge %%f0,  %%f2,  %%f10 \n\t"  // 0,g,t2
00162             "fpmerge %%f0,  %%f18, %%f26 \n\t"  // 0,g,t2
00163             "fpmerge %%f10, %%f8,  %%f4  \n\t"  // t2,t1,msb
00164             "fpmerge %%f26, %%f24, %%f20 \n\t"  // t2,t1,msb
00165             "fpmerge %%f11, %%f9,  %%f6  \n\t"  // t2,t1,lsb
00166             "fpmerge %%f27, %%f25, %%f22 \n\t"  // t2,t1,lsb
00167             "std %%f4,  [%3]             \n\t"
00168             "std %%f20, [%3 + 16]        \n\t"
00169             "std %%f6,  [%3 + 8]         \n\t"
00170             "std %%f22, [%3 + 24]        \n\t"
00171 
00172             "add %0, 8, %0   \n\t"
00173             "add %1, 4, %1   \n\t"
00174             "add %2, 4, %2   \n\t"
00175             "subcc %4, 8, %4 \n\t"
00176             "bne 1b          \n\t"
00177             "add %3, 32, %3  \n\t" //delay slot
00178             : "=r" (out1), "=r" (out2), "=r" (out3), "=r" (out4), "=r" (out5), "=r" (out6)
00179             : "0" (src[0] + (y + srcSliceY) * srcStride[0]), "1" (src[1] + (y + srcSliceY) * srcStride[1]),
00180             "2" (src[2] + (y + srcSliceY) * srcStride[2]), "3" (dst[0] + (y + srcSliceY) * dstStride[0]),
00181             "4" (c->dstW),
00182             "5" (c->sparc_coeffs)
00183             );
00184 
00185     return srcSliceH;
00186 }
00187 
00188 av_cold SwsFunc ff_yuv2rgb_init_vis(SwsContext *c)
00189 {
00190     c->sparc_coeffs[5] = c->yCoeff;
00191     c->sparc_coeffs[6] = c->vgCoeff;
00192     c->sparc_coeffs[7] = c->vrCoeff;
00193     c->sparc_coeffs[8] = c->ubCoeff;
00194     c->sparc_coeffs[9] = c->ugCoeff;
00195 
00196     c->sparc_coeffs[0] = (((int16_t)c->yOffset * (int16_t)c->yCoeff  >> 11) & 0xffff) * 0x0001000100010001ULL;
00197     c->sparc_coeffs[1] = (((int16_t)c->uOffset * (int16_t)c->ubCoeff >> 11) & 0xffff) * 0x0001000100010001ULL;
00198     c->sparc_coeffs[2] = (((int16_t)c->uOffset * (int16_t)c->ugCoeff >> 11) & 0xffff) * 0x0001000100010001ULL;
00199     c->sparc_coeffs[3] = (((int16_t)c->vOffset * (int16_t)c->vgCoeff >> 11) & 0xffff) * 0x0001000100010001ULL;
00200     c->sparc_coeffs[4] = (((int16_t)c->vOffset * (int16_t)c->vrCoeff >> 11) & 0xffff) * 0x0001000100010001ULL;
00201 
00202     if (c->dstFormat == PIX_FMT_RGB32 && c->srcFormat == PIX_FMT_YUV422P && (c->dstW & 7) == 0) {
00203         av_log(c, AV_LOG_INFO,
00204                "SPARC VIS accelerated YUV422P -> RGB32 (WARNING: alpha value is wrong)\n");
00205         return vis_422P_ARGB32;
00206     } else if (c->dstFormat == PIX_FMT_RGB32 && c->srcFormat == PIX_FMT_YUV420P && (c->dstW & 7) == 0) {
00207         av_log(c, AV_LOG_INFO,
00208                "SPARC VIS accelerated YUV420P -> RGB32 (WARNING: alpha value is wrong)\n");
00209         return vis_420P_ARGB32;
00210     }
00211     return NULL;
00212 }
libswscale/sparc/yuv2rgb_vis.c
