yuv2rgb_altivec.c 38.2 KB
Newer Older
1
/*
2 3 4 5
 * AltiVec acceleration for colorspace conversion
 *
 * copyright (C) 2004 Marc Hoffman <marc.hoffman@analog.com>
 *
6
 * This file is part of Libav.
7
 *
8
 * Libav is free software; you can redistribute it and/or
9 10 11
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
12
 *
13
 * Libav is distributed in the hope that it will be useful,
14
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 16
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
17
 *
18
 * You should have received a copy of the GNU Lesser General Public
19
 * License along with Libav; if not, write to the Free Software
20 21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */
22

23
/*
24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86
 * Convert I420 YV12 to RGB in various formats,
 * it rejects images that are not in 420 formats,
 * it rejects images that don't have widths of multiples of 16,
 * it rejects images that don't have heights of multiples of 2.
 * Reject defers to C simulation code.
 *
 * Lots of optimizations to be done here.
 *
 * 1. Need to fix saturation code. I just couldn't get it to fly with packs
 * and adds, so we currently use max/min to clip.
 *
 * 2. The inefficient use of chroma loading needs a bit of brushing up.
 *
 * 3. Analysis of pipeline stalls needs to be done. Use shark to identify
 * pipeline stalls.
 *
 *
 * MODIFIED to calculate coeffs from currently selected color space.
 * MODIFIED core to be a macro where you specify the output format.
 * ADDED UYVY conversion which is never called due to some thing in swscale.
 * CORRECTED algorithim selection to be strict on input formats.
 * ADDED runtime detection of AltiVec.
 *
 * ADDED altivec_yuv2packedX vertical scl + RGB converter
 *
 * March 27,2004
 * PERFORMANCE ANALYSIS
 *
 * The C version uses 25% of the processor or ~250Mips for D1 video rawvideo
 * used as test.
 * The AltiVec version uses 10% of the processor or ~100Mips for D1 video
 * same sequence.
 *
 * 720 * 480 * 30  ~10MPS
 *
 * so we have roughly 10 clocks per pixel. This is too high, something has
 * to be wrong.
 *
 * OPTIMIZED clip codes to utilize vec_max and vec_packs removing the
 * need for vec_min.
 *
 * OPTIMIZED DST OUTPUT cache/DMA controls. We are pretty much guaranteed to
 * have the input video frame, it was just decompressed so it probably resides
 * in L1 caches. However, we are creating the output video stream. This needs
 * to use the DSTST instruction to optimize for the cache. We couple this with
 * the fact that we are not going to be visiting the input buffer again so we
 * mark it Least Recently Used. This shaves 25% of the processor cycles off.
 *
 * Now memcpy is the largest mips consumer in the system, probably due
 * to the inefficient X11 stuff.
 *
 * GL libraries seem to be very slow on this machine 1.33Ghz PB running
 * Jaguar, this is not the case for my 1Ghz PB.  I thought it might be
 * a versioning issue, however I have libGL.1.2.dylib for both
 * machines. (We need to figure this out now.)
 *
 * GL2 libraries work now with patch for RGB32.
 *
 * NOTE: quartz vo driver ARGB32_to_RGB24 consumes 30% of the processor.
 *
 * Integrated luma prescaling adjustment for saturation/contrast/brightness
 * adjustment.
 */
87 88 89

#include <stdio.h>
#include <stdlib.h>
Alex Beregszaszi's avatar
Alex Beregszaszi committed
90
#include <string.h>
91 92
#include <inttypes.h>
#include <assert.h>
93

94
#include "config.h"
95 96 97
#include "libswscale/rgb2rgb.h"
#include "libswscale/swscale.h"
#include "libswscale/swscale_internal.h"
98
#include "libavutil/attributes.h"
99
#include "libavutil/cpu.h"
100
#include "yuv2rgb_altivec.h"
101

102 103
#if HAVE_ALTIVEC

104 105 106 107 108 109 110
#undef PROFILE_THE_BEAST
#undef INC_SCALING

typedef unsigned char ubyte;
typedef signed char   sbyte;

/* RGB interleaver, 16 planar pels 8-bit samples per channel in
111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
 * homogeneous vector registers x0,x1,x2 are interleaved with the
 * following technique:
 *
 *    o0 = vec_mergeh(x0, x1);
 *    o1 = vec_perm(o0, x2, perm_rgb_0);
 *    o2 = vec_perm(o0, x2, perm_rgb_1);
 *    o3 = vec_mergel(x0, x1);
 *    o4 = vec_perm(o3, o2, perm_rgb_2);
 *    o5 = vec_perm(o3, o2, perm_rgb_3);
 *
 * perm_rgb_0:   o0(RG).h v1(B) --> o1*
 *            0   1  2   3   4
 *           rgbr|gbrg|brgb|rgbr
 *           0010 0100 1001 0010
 *           0102 3145 2673 894A
 *
 * perm_rgb_1:   o0(RG).h v1(B) --> o2
 *            0   1  2   3   4
 *           gbrg|brgb|bbbb|bbbb
 *           0100 1001 1111 1111
 *           B5CD 6EF7 89AB CDEF
 *
 * perm_rgb_2:   o3(RG).l o2(rgbB.l) --> o4*
 *            0   1  2   3   4
 *           gbrg|brgb|rgbr|gbrg
 *           1111 1111 0010 0100
 *           89AB CDEF 0182 3945
 *
 * perm_rgb_2:   o3(RG).l o2(rgbB.l) ---> o5*
 *            0   1  2   3   4
 *           brgb|rgbr|gbrg|brgb
 *           1001 0010 0100 1001
 *           a67b 89cA BdCD eEFf
 *
 */
static const vector unsigned char
    perm_rgb_0 = { 0x00, 0x01, 0x10, 0x02, 0x03, 0x11, 0x04, 0x05,
                   0x12, 0x06, 0x07, 0x13, 0x08, 0x09, 0x14, 0x0a },
    perm_rgb_1 = { 0x0b, 0x15, 0x0c, 0x0d, 0x16, 0x0e, 0x0f, 0x17,
                   0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f },
    perm_rgb_2 = { 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
                   0x00, 0x01, 0x18, 0x02, 0x03, 0x19, 0x04, 0x05 },
    perm_rgb_3 = { 0x1a, 0x06, 0x07, 0x1b, 0x08, 0x09, 0x1c, 0x0a,
                   0x0b, 0x1d, 0x0c, 0x0d, 0x1e, 0x0e, 0x0f, 0x1f };

#define vec_merge3(x2, x1, x0, y0, y1, y2)     \
    do {                                       \
        __typeof__(x0) o0, o2, o3;             \
        o0 = vec_mergeh(x0, x1);               \
        y0 = vec_perm(o0, x2, perm_rgb_0);     \
        o2 = vec_perm(o0, x2, perm_rgb_1);     \
        o3 = vec_mergel(x0, x1);               \
        y1 = vec_perm(o3, o2, perm_rgb_2);     \
        y2 = vec_perm(o3, o2, perm_rgb_3);     \
    } while (0)

#define vec_mstbgr24(x0, x1, x2, ptr)          \
    do {                                       \
        __typeof__(x0) _0, _1, _2;             \
        vec_merge3(x0, x1, x2, _0, _1, _2);    \
        vec_st(_0, 0, ptr++);                  \
        vec_st(_1, 0, ptr++);                  \
        vec_st(_2, 0, ptr++);                  \
    } while (0)

#define vec_mstrgb24(x0, x1, x2, ptr)          \
    do {                                       \
        __typeof__(x0) _0, _1, _2;             \
        vec_merge3(x2, x1, x0, _0, _1, _2);    \
        vec_st(_0, 0, ptr++);                  \
        vec_st(_1, 0, ptr++);                  \
        vec_st(_2, 0, ptr++);                  \
    } while (0)
184 185

/* pack the pixels in rgb0 format
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
 * msb R
 * lsb 0
 */
#define vec_mstrgb32(T, x0, x1, x2, x3, ptr)                            \
    do {                                                                \
        T _0, _1, _2, _3;                                               \
        _0 = vec_mergeh(x0, x1);                                        \
        _1 = vec_mergeh(x2, x3);                                        \
        _2 = (T) vec_mergeh((vector unsigned short) _0,                 \
                            (vector unsigned short) _1);                \
        _3 = (T) vec_mergel((vector unsigned short) _0,                 \
                            (vector unsigned short) _1);                \
        vec_st(_2, 0 * 16, (T *) ptr);                                  \
        vec_st(_3, 1 * 16, (T *) ptr);                                  \
        _0 = vec_mergel(x0, x1);                                        \
        _1 = vec_mergel(x2, x3);                                        \
        _2 = (T) vec_mergeh((vector unsigned short) _0,                 \
                            (vector unsigned short) _1);                \
        _3 = (T) vec_mergel((vector unsigned short) _0,                 \
                            (vector unsigned short) _1);                \
        vec_st(_2, 2 * 16, (T *) ptr);                                  \
        vec_st(_3, 3 * 16, (T *) ptr);                                  \
        ptr += 4;                                                       \
    } while (0)
210 211

/*
212 213 214 215 216 217 218 219 220 221 222
 * 1     0       1.4021   | | Y |
 * 1    -0.3441 -0.7142   |x| Cb|
 * 1     1.7718  0        | | Cr|
 *
 *
 * Y:      [-128 127]
 * Cb/Cr : [-128 127]
 *
 * typical YUV conversion works on Y: 0-255 this version has been
 * optimized for JPEG decoding.
 */
223

224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253
#define vec_unh(x)                                                      \
    (vector signed short)                                               \
        vec_perm(x, (__typeof__(x)) { 0 },                              \
                 ((vector unsigned char) {                              \
                     0x10, 0x00, 0x10, 0x01, 0x10, 0x02, 0x10, 0x03,    \
                     0x10, 0x04, 0x10, 0x05, 0x10, 0x06, 0x10, 0x07 }))

#define vec_unl(x)                                                      \
    (vector signed short)                                               \
        vec_perm(x, (__typeof__(x)) { 0 },                              \
                 ((vector unsigned char) {                              \
                     0x10, 0x08, 0x10, 0x09, 0x10, 0x0A, 0x10, 0x0B,    \
                     0x10, 0x0C, 0x10, 0x0D, 0x10, 0x0E, 0x10, 0x0F }))

#define vec_clip_s16(x)                                                 \
    vec_max(vec_min(x, ((vector signed short) {                         \
                    235, 235, 235, 235, 235, 235, 235, 235 })),         \
            ((vector signed short) { 16, 16, 16, 16, 16, 16, 16, 16 }))

#define vec_packclp(x, y)                                               \
    (vector unsigned char)                                              \
        vec_packs((vector unsigned short)                               \
                      vec_max(x, ((vector signed short) { 0 })),        \
                  (vector unsigned short)                               \
                      vec_max(y, ((vector signed short) { 0 })))

static inline void cvtyuvtoRGB(SwsContext *c, vector signed short Y,
                               vector signed short U, vector signed short V,
                               vector signed short *R, vector signed short *G,
                               vector signed short *B)
254
{
255
    vector signed short vx, ux, uvx;
256

257 258 259 260 261
    Y = vec_mradds(Y, c->CY, c->OY);
    U = vec_sub(U, (vector signed short)
                       vec_splat((vector signed short) { 128 }, 0));
    V = vec_sub(V, (vector signed short)
                       vec_splat((vector signed short) { 128 }, 0));
262

263 264 265
    // ux  = (CBU * (u << c->CSHIFT) + 0x4000) >> 15;
    ux = vec_sl(U, c->CSHIFT);
    *B = vec_mradds(ux, c->CBU, Y);
266

267 268 269
    // vx  = (CRV * (v << c->CSHIFT) + 0x4000) >> 15;
    vx = vec_sl(V, c->CSHIFT);
    *R = vec_mradds(vx, c->CRV, Y);
270

271 272 273
    // uvx = ((CGU * u) + (CGV * v)) >> 15;
    uvx = vec_mradds(U, c->CGU, Y);
    *G  = vec_mradds(V, c->CGV, uvx);
274 275 276
}

/*
277 278 279 280
 * ------------------------------------------------------------------------------
 * CS converters
 * ------------------------------------------------------------------------------
 */
281

282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303
#define DEFCSP420_CVT(name, out_pixels)                                       \
static int altivec_ ## name(SwsContext *c, const unsigned char **in,          \
                            int *instrides, int srcSliceY, int srcSliceH,     \
                            unsigned char **oplanes, int *outstrides)         \
{                                                                             \
    int w = c->srcW;                                                          \
    int h = srcSliceH;                                                        \
    int i, j;                                                                 \
    int instrides_scl[3];                                                     \
    vector unsigned char y0, y1;                                              \
                                                                              \
    vector signed char u, v;                                                  \
                                                                              \
    vector signed short Y0, Y1, Y2, Y3;                                       \
    vector signed short U, V;                                                 \
    vector signed short vx, ux, uvx;                                          \
    vector signed short vx0, ux0, uvx0;                                       \
    vector signed short vx1, ux1, uvx1;                                       \
    vector signed short R0, G0, B0;                                           \
    vector signed short R1, G1, B1;                                           \
    vector unsigned char R, G, B;                                             \
                                                                              \
304
    const vector unsigned char *y1ivP, *y2ivP, *uivP, *vivP;                  \
305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
    vector unsigned char align_perm;                                          \
                                                                              \
    vector signed short lCY       = c->CY;                                    \
    vector signed short lOY       = c->OY;                                    \
    vector signed short lCRV      = c->CRV;                                   \
    vector signed short lCBU      = c->CBU;                                   \
    vector signed short lCGU      = c->CGU;                                   \
    vector signed short lCGV      = c->CGV;                                   \
    vector unsigned short lCSHIFT = c->CSHIFT;                                \
                                                                              \
    const ubyte *y1i = in[0];                                                 \
    const ubyte *y2i = in[0] + instrides[0];                                  \
    const ubyte *ui  = in[1];                                                 \
    const ubyte *vi  = in[2];                                                 \
                                                                              \
    vector unsigned char *oute =                                              \
        (vector unsigned char *)                                              \
            (oplanes[0] + srcSliceY * outstrides[0]);                         \
    vector unsigned char *outo =                                              \
        (vector unsigned char *)                                              \
            (oplanes[0] + srcSliceY * outstrides[0] + outstrides[0]);         \
                                                                              \
    /* loop moves y{1, 2}i by w */                                            \
    instrides_scl[0] = instrides[0] * 2 - w;                                  \
    /* loop moves ui by w / 2 */                                              \
    instrides_scl[1] = instrides[1] - w / 2;                                  \
    /* loop moves vi by w / 2 */                                              \
    instrides_scl[2] = instrides[2] - w / 2;                                  \
                                                                              \
    for (i = 0; i < h / 2; i++) {                                             \
        vec_dstst(outo, (0x02000002 | (((w * 3 + 32) / 32) << 16)), 0);       \
        vec_dstst(oute, (0x02000002 | (((w * 3 + 32) / 32) << 16)), 1);       \
                                                                              \
        for (j = 0; j < w / 16; j++) {                                        \
339 340 341 342
            y1ivP = (const vector unsigned char *) y1i;                       \
            y2ivP = (const vector unsigned char *) y2i;                       \
            uivP  = (const vector unsigned char *) ui;                        \
            vivP  = (const vector unsigned char *) vi;                        \
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440
                                                                              \
            align_perm = vec_lvsl(0, y1i);                                    \
            y0 = (vector unsigned char)                                       \
                     vec_perm(y1ivP[0], y1ivP[1], align_perm);                \
                                                                              \
            align_perm = vec_lvsl(0, y2i);                                    \
            y1 = (vector unsigned char)                                       \
                     vec_perm(y2ivP[0], y2ivP[1], align_perm);                \
                                                                              \
            align_perm = vec_lvsl(0, ui);                                     \
            u = (vector signed char)                                          \
                    vec_perm(uivP[0], uivP[1], align_perm);                   \
                                                                              \
            align_perm = vec_lvsl(0, vi);                                     \
            v = (vector signed char)                                          \
                    vec_perm(vivP[0], vivP[1], align_perm);                   \
                                                                              \
            u = (vector signed char)                                          \
                    vec_sub(u,                                                \
                            (vector signed char)                              \
                                vec_splat((vector signed char) { 128 }, 0));  \
            v = (vector signed char)                                          \
                    vec_sub(v,                                                \
                            (vector signed char)                              \
                                vec_splat((vector signed char) { 128 }, 0));  \
                                                                              \
            U = vec_unpackh(u);                                               \
            V = vec_unpackh(v);                                               \
                                                                              \
            Y0 = vec_unh(y0);                                                 \
            Y1 = vec_unl(y0);                                                 \
            Y2 = vec_unh(y1);                                                 \
            Y3 = vec_unl(y1);                                                 \
                                                                              \
            Y0 = vec_mradds(Y0, lCY, lOY);                                    \
            Y1 = vec_mradds(Y1, lCY, lOY);                                    \
            Y2 = vec_mradds(Y2, lCY, lOY);                                    \
            Y3 = vec_mradds(Y3, lCY, lOY);                                    \
                                                                              \
            /* ux  = (CBU * (u << CSHIFT) + 0x4000) >> 15 */                  \
            ux  = vec_sl(U, lCSHIFT);                                         \
            ux  = vec_mradds(ux, lCBU, (vector signed short) { 0 });          \
            ux0 = vec_mergeh(ux, ux);                                         \
            ux1 = vec_mergel(ux, ux);                                         \
                                                                              \
            /* vx  = (CRV * (v << CSHIFT) + 0x4000) >> 15; */                 \
            vx  = vec_sl(V, lCSHIFT);                                         \
            vx  = vec_mradds(vx, lCRV, (vector signed short) { 0 });          \
            vx0 = vec_mergeh(vx, vx);                                         \
            vx1 = vec_mergel(vx, vx);                                         \
                                                                              \
            /* uvx = ((CGU * u) + (CGV * v)) >> 15 */                         \
            uvx  = vec_mradds(U, lCGU, (vector signed short) { 0 });          \
            uvx  = vec_mradds(V, lCGV, uvx);                                  \
            uvx0 = vec_mergeh(uvx, uvx);                                      \
            uvx1 = vec_mergel(uvx, uvx);                                      \
                                                                              \
            R0 = vec_add(Y0, vx0);                                            \
            G0 = vec_add(Y0, uvx0);                                           \
            B0 = vec_add(Y0, ux0);                                            \
            R1 = vec_add(Y1, vx1);                                            \
            G1 = vec_add(Y1, uvx1);                                           \
            B1 = vec_add(Y1, ux1);                                            \
                                                                              \
            R = vec_packclp(R0, R1);                                          \
            G = vec_packclp(G0, G1);                                          \
            B = vec_packclp(B0, B1);                                          \
                                                                              \
            out_pixels(R, G, B, oute);                                        \
                                                                              \
            R0 = vec_add(Y2, vx0);                                            \
            G0 = vec_add(Y2, uvx0);                                           \
            B0 = vec_add(Y2, ux0);                                            \
            R1 = vec_add(Y3, vx1);                                            \
            G1 = vec_add(Y3, uvx1);                                           \
            B1 = vec_add(Y3, ux1);                                            \
            R  = vec_packclp(R0, R1);                                         \
            G  = vec_packclp(G0, G1);                                         \
            B  = vec_packclp(B0, B1);                                         \
                                                                              \
                                                                              \
            out_pixels(R, G, B, outo);                                        \
                                                                              \
            y1i += 16;                                                        \
            y2i += 16;                                                        \
            ui  += 8;                                                         \
            vi  += 8;                                                         \
        }                                                                     \
                                                                              \
        outo += (outstrides[0]) >> 4;                                         \
        oute += (outstrides[0]) >> 4;                                         \
                                                                              \
        ui  += instrides_scl[1];                                              \
        vi  += instrides_scl[2];                                              \
        y1i += instrides_scl[0];                                              \
        y2i += instrides_scl[0];                                              \
    }                                                                         \
    return srcSliceH;                                                         \
441 442
}

443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
#define out_abgr(a, b, c, ptr)                                          \
    vec_mstrgb32(__typeof__(a), ((__typeof__(a)) { 255 }), c, b, a, ptr)
#define out_bgra(a, b, c, ptr)                                          \
    vec_mstrgb32(__typeof__(a), c, b, a, ((__typeof__(a)) { 255 }), ptr)
#define out_rgba(a, b, c, ptr)                                          \
    vec_mstrgb32(__typeof__(a), a, b, c, ((__typeof__(a)) { 255 }), ptr)
#define out_argb(a, b, c, ptr)                                          \
    vec_mstrgb32(__typeof__(a), ((__typeof__(a)) { 255 }), a, b, c, ptr)
#define out_rgb24(a, b, c, ptr) vec_mstrgb24(a, b, c, ptr)
#define out_bgr24(a, b, c, ptr) vec_mstbgr24(a, b, c, ptr)

DEFCSP420_CVT(yuv2_abgr,  out_abgr)
DEFCSP420_CVT(yuv2_bgra,  out_bgra)
DEFCSP420_CVT(yuv2_rgba,  out_rgba)
DEFCSP420_CVT(yuv2_argb,  out_argb)
DEFCSP420_CVT(yuv2_rgb24, out_rgb24)
DEFCSP420_CVT(yuv2_bgr24, out_bgr24)
460 461 462

// uyvy|uyvy|uyvy|uyvy
// 0123 4567 89ab cdef
463 464 465 466 467 468 469 470 471 472 473 474 475
static const vector unsigned char
    demux_u = { 0x10, 0x00, 0x10, 0x00,
                0x10, 0x04, 0x10, 0x04,
                0x10, 0x08, 0x10, 0x08,
                0x10, 0x0c, 0x10, 0x0c },
    demux_v = { 0x10, 0x02, 0x10, 0x02,
                0x10, 0x06, 0x10, 0x06,
                0x10, 0x0A, 0x10, 0x0A,
                0x10, 0x0E, 0x10, 0x0E },
    demux_y = { 0x10, 0x01, 0x10, 0x03,
                0x10, 0x05, 0x10, 0x07,
                0x10, 0x09, 0x10, 0x0B,
                0x10, 0x0D, 0x10, 0x0F };
476 477

/*
478 479 480 481 482
 * this is so I can play live CCIR raw video
 */
static int altivec_uyvy_rgb32(SwsContext *c, const unsigned char **in,
                              int *instrides, int srcSliceY, int srcSliceH,
                              unsigned char **oplanes, int *outstrides)
483
{
484 485
    int w = c->srcW;
    int h = srcSliceH;
486
    int i, j;
487
    vector unsigned char uyvy;
488 489 490
    vector signed short Y, U, V;
    vector signed short R0, G0, B0, R1, G1, B1;
    vector unsigned char R, G, B;
491
    vector unsigned char *out;
492
    const ubyte *img;
493

494
    img = in[0];
495
    out = (vector unsigned char *) (oplanes[0] + srcSliceY * outstrides[0]);
496

497 498 499
    for (i = 0; i < h; i++)
        for (j = 0; j < w / 16; j++) {
            uyvy = vec_ld(0, img);
500

501 502
            U = (vector signed short)
                    vec_perm(uyvy, (vector unsigned char) { 0 }, demux_u);
503
            V = (vector signed short)
504
                    vec_perm(uyvy, (vector unsigned char) { 0 }, demux_v);
505
            Y = (vector signed short)
506
                    vec_perm(uyvy, (vector unsigned char) { 0 }, demux_y);
507

508
            cvtyuvtoRGB(c, Y, U, V, &R0, &G0, &B0);
509

510
            uyvy = vec_ld(16, img);
511

512 513
            U = (vector signed short)
                    vec_perm(uyvy, (vector unsigned char) { 0 }, demux_u);
514
            V = (vector signed short)
515
                    vec_perm(uyvy, (vector unsigned char) { 0 }, demux_v);
516
            Y = (vector signed short)
517
                    vec_perm(uyvy, (vector unsigned char) { 0 }, demux_y);
518

519
            cvtyuvtoRGB(c, Y, U, V, &R1, &G1, &B1);
520

521 522 523
            R = vec_packclp(R0, R1);
            G = vec_packclp(G0, G1);
            B = vec_packclp(B0, B1);
524

525 526
            // vec_mstbgr24 (R,G,B, out);
            out_rgba(R, G, B, out);
527

528 529 530
            img += 32;
        }
    return srcSliceH;
531 532
}

533 534
#endif /* HAVE_ALTIVEC */

535
/* Ok currently the acceleration routine only supports
536 537 538 539 540
 * inputs of widths a multiple of 16
 * and heights a multiple 2
 *
 * So we just fall back to the C codes for this.
 */
541
av_cold SwsFunc ff_yuv2rgb_init_ppc(SwsContext *c)
542
{
543
#if HAVE_ALTIVEC
544
    if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
545
        return NULL;
546

547
    /*
548 549 550 551 552 553 554 555
     * and this seems not to matter too much I tried a bunch of
     * videos with abnormal widths and MPlayer crashes elsewhere.
     * mplayer -vo x11 -rawvideo on:w=350:h=240 raw-350x240.eyuv
     * boom with X11 bad match.
     *
     */
    if ((c->srcW & 0xf) != 0)
        return NULL;
556 557

    switch (c->srcFormat) {
558 559
    case AV_PIX_FMT_YUV410P:
    case AV_PIX_FMT_YUV420P:
560
    /*case IMGFMT_CLPL:        ??? */
561 562 563
    case AV_PIX_FMT_GRAY8:
    case AV_PIX_FMT_NV12:
    case AV_PIX_FMT_NV21:
564 565 566
        if ((c->srcH & 0x1) != 0)
            return NULL;

567
        switch (c->dstFormat) {
568
        case AV_PIX_FMT_RGB24:
569 570
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space RGB24\n");
            return altivec_yuv2_rgb24;
571
        case AV_PIX_FMT_BGR24:
572 573
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space BGR24\n");
            return altivec_yuv2_bgr24;
574
        case AV_PIX_FMT_ARGB:
575 576
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space ARGB\n");
            return altivec_yuv2_argb;
577
        case AV_PIX_FMT_ABGR:
578 579
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space ABGR\n");
            return altivec_yuv2_abgr;
580
        case AV_PIX_FMT_RGBA:
581 582
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space RGBA\n");
            return altivec_yuv2_rgba;
583
        case AV_PIX_FMT_BGRA:
584 585 586 587 588 589
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space BGRA\n");
            return altivec_yuv2_bgra;
        default: return NULL;
        }
        break;

590
    case AV_PIX_FMT_UYVY422:
591
        switch (c->dstFormat) {
592
        case AV_PIX_FMT_BGR32:
593 594 595 596 597 598
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space UYVY -> RGB32\n");
            return altivec_uyvy_rgb32;
        default: return NULL;
        }
        break;
    }
599 600
#endif /* HAVE_ALTIVEC */

601
    return NULL;
602 603
}

604 605 606 607 608
av_cold void ff_yuv2rgb_init_tables_ppc(SwsContext *c,
                                        const int inv_table[4],
                                        int brightness,
                                        int contrast,
                                        int saturation)
609
{
610
#if HAVE_ALTIVEC
611
    union {
612
        DECLARE_ALIGNED(16, signed short, tmp)[8];
613 614 615
        vector signed short vec;
    } buf;

616 617 618
    if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
        return;

619 620 621 622 623 624 625 626 627 628 629 630 631 632
    buf.tmp[0] = ((0xffffLL) * contrast >> 8) >> 9;                               // cy
    buf.tmp[1] = -256 * brightness;                                               // oy
    buf.tmp[2] =   (inv_table[0] >> 3) * (contrast >> 16) * (saturation >> 16);   // crv
    buf.tmp[3] =   (inv_table[1] >> 3) * (contrast >> 16) * (saturation >> 16);   // cbu
    buf.tmp[4] = -((inv_table[2] >> 1) * (contrast >> 16) * (saturation >> 16));  // cgu
    buf.tmp[5] = -((inv_table[3] >> 1) * (contrast >> 16) * (saturation >> 16));  // cgv

    c->CSHIFT = (vector unsigned short) vec_splat_u16(2);
    c->CY     = vec_splat((vector signed short) buf.vec, 0);
    c->OY     = vec_splat((vector signed short) buf.vec, 1);
    c->CRV    = vec_splat((vector signed short) buf.vec, 2);
    c->CBU    = vec_splat((vector signed short) buf.vec, 3);
    c->CGU    = vec_splat((vector signed short) buf.vec, 4);
    c->CGV    = vec_splat((vector signed short) buf.vec, 5);
633
    return;
634
#endif /* HAVE_ALTIVEC */
635 636
}

637 638
#if HAVE_ALTIVEC

639 640 641 642 643 644 645 646 647 648 649 650
static av_always_inline void yuv2packedX_altivec(SwsContext *c,
                                                 const int16_t *lumFilter,
                                                 const int16_t **lumSrc,
                                                 int lumFilterSize,
                                                 const int16_t *chrFilter,
                                                 const int16_t **chrUSrc,
                                                 const int16_t **chrVSrc,
                                                 int chrFilterSize,
                                                 const int16_t **alpSrc,
                                                 uint8_t *dest,
                                                 int dstW, int dstY,
                                                 enum AVPixelFormat target)
651
{
652 653 654
    int i, j;
    vector signed short X, X0, X1, Y0, U0, V0, Y1, U1, V1, U, V;
    vector signed short R0, G0, B0, R1, G1, B1;
655

656 657
    vector unsigned char R, G, B;
    vector unsigned char *out, *nout;
658

659
    vector signed short RND   = vec_splat_s16(1 << 3);
660
    vector unsigned short SCL = vec_splat_u16(4);
661
    DECLARE_ALIGNED(16, unsigned int, scratch)[16];
662

663
    vector signed short *YCoeffs, *CCoeffs;
664

665 666
    YCoeffs = c->vYCoeffsBank + dstY * lumFilterSize;
    CCoeffs = c->vCCoeffsBank + dstY * chrFilterSize;
667

668
    out = (vector unsigned char *) dest;
669

670
    for (i = 0; i < dstW; i += 16) {
671 672 673
        Y0 = RND;
        Y1 = RND;
        /* extract 16 coeffs from lumSrc */
674 675 676 677 678
        for (j = 0; j < lumFilterSize; j++) {
            X0 = vec_ld(0, &lumSrc[j][i]);
            X1 = vec_ld(16, &lumSrc[j][i]);
            Y0 = vec_mradds(X0, YCoeffs[j], Y0);
            Y1 = vec_mradds(X1, YCoeffs[j], Y1);
679
        }
680

681 682 683
        U = RND;
        V = RND;
        /* extract 8 coeffs from U,V */
684 685 686 687 688
        for (j = 0; j < chrFilterSize; j++) {
            X = vec_ld(0, &chrUSrc[j][i / 2]);
            U = vec_mradds(X, CCoeffs[j], U);
            X = vec_ld(0, &chrVSrc[j][i / 2]);
            V = vec_mradds(X, CCoeffs[j], V);
689
        }
690

691
        /* scale and clip signals */
692 693 694 695
        Y0 = vec_sra(Y0, SCL);
        Y1 = vec_sra(Y1, SCL);
        U  = vec_sra(U, SCL);
        V  = vec_sra(V, SCL);
696

697 698 699 700
        Y0 = vec_clip_s16(Y0);
        Y1 = vec_clip_s16(Y1);
        U  = vec_clip_s16(U);
        V  = vec_clip_s16(V);
701 702

        /* now we have
703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724
         * Y0 = y0 y1 y2 y3 y4 y5 y6 y7    Y1 = y8 y9 y10 y11 y12 y13 y14 y15
         * U  = u0 u1 u2 u3 u4 u5 u6 u7    V  = v0 v1 v2 v3 v4 v5 v6 v7
         *
         * Y0 = y0 y1 y2 y3 y4 y5 y6 y7    Y1 = y8 y9 y10 y11 y12 y13 y14 y15
         * U0 = u0 u0 u1 u1 u2 u2 u3 u3    U1 = u4 u4 u5 u5 u6 u6 u7 u7
         * V0 = v0 v0 v1 v1 v2 v2 v3 v3    V1 = v4 v4 v5 v5 v6 v6 v7 v7
         */

        U0 = vec_mergeh(U, U);
        V0 = vec_mergeh(V, V);

        U1 = vec_mergel(U, U);
        V1 = vec_mergel(V, V);

        cvtyuvtoRGB(c, Y0, U0, V0, &R0, &G0, &B0);
        cvtyuvtoRGB(c, Y1, U1, V1, &R1, &G1, &B1);

        R = vec_packclp(R0, R1);
        G = vec_packclp(G0, G1);
        B = vec_packclp(B0, B1);

        switch (target) {
725
        case AV_PIX_FMT_ABGR:
726 727
            out_abgr(R, G, B, out);
            break;
728
        case AV_PIX_FMT_BGRA:
729 730
            out_bgra(R, G, B, out);
            break;
731
        case AV_PIX_FMT_RGBA:
732 733
            out_rgba(R, G, B, out);
            break;
734
        case AV_PIX_FMT_ARGB:
735 736
            out_argb(R, G, B, out);
            break;
737
        case AV_PIX_FMT_RGB24:
738 739
            out_rgb24(R, G, B, out);
            break;
740
        case AV_PIX_FMT_BGR24:
741 742
            out_bgr24(R, G, B, out);
            break;
Ramiro Polla's avatar
Ramiro Polla committed
743
        default:
744 745 746 747 748 749 750 751 752
        {
            /* If this is reached, the caller should have called yuv2packedXinC
             * instead. */
            static int printed_error_message;
            if (!printed_error_message) {
                av_log(c, AV_LOG_ERROR,
                       "altivec_yuv2packedX doesn't support %s output\n",
                       sws_format_name(c->dstFormat));
                printed_error_message = 1;
753
            }
754 755
            return;
        }
756
        }
757 758
    }

759 760 761 762 763 764
    if (i < dstW) {
        i -= 16;

        Y0 = RND;
        Y1 = RND;
        /* extract 16 coeffs from lumSrc */
765 766 767 768 769
        for (j = 0; j < lumFilterSize; j++) {
            X0 = vec_ld(0, &lumSrc[j][i]);
            X1 = vec_ld(16, &lumSrc[j][i]);
            Y0 = vec_mradds(X0, YCoeffs[j], Y0);
            Y1 = vec_mradds(X1, YCoeffs[j], Y1);
770
        }
771

772 773 774
        U = RND;
        V = RND;
        /* extract 8 coeffs from U,V */
775 776 777 778 779
        for (j = 0; j < chrFilterSize; j++) {
            X = vec_ld(0, &chrUSrc[j][i / 2]);
            U = vec_mradds(X, CCoeffs[j], U);
            X = vec_ld(0, &chrVSrc[j][i / 2]);
            V = vec_mradds(X, CCoeffs[j], V);
780
        }
781

782
        /* scale and clip signals */
783 784 785 786
        Y0 = vec_sra(Y0, SCL);
        Y1 = vec_sra(Y1, SCL);
        U  = vec_sra(U, SCL);
        V  = vec_sra(V, SCL);
787

788 789 790 791
        Y0 = vec_clip_s16(Y0);
        Y1 = vec_clip_s16(Y1);
        U  = vec_clip_s16(U);
        V  = vec_clip_s16(V);
792 793

        /* now we have
794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816
         * Y0 = y0 y1 y2 y3 y4 y5 y6 y7    Y1 = y8 y9 y10 y11 y12 y13 y14 y15
         * U  = u0 u1 u2 u3 u4 u5 u6 u7    V  = v0 v1 v2 v3 v4 v5 v6 v7
         *
         * Y0 = y0 y1 y2 y3 y4 y5 y6 y7    Y1 = y8 y9 y10 y11 y12 y13 y14 y15
         * U0 = u0 u0 u1 u1 u2 u2 u3 u3    U1 = u4 u4 u5 u5 u6 u6 u7 u7
         * V0 = v0 v0 v1 v1 v2 v2 v3 v3    V1 = v4 v4 v5 v5 v6 v6 v7 v7
         */

        U0 = vec_mergeh(U, U);
        V0 = vec_mergeh(V, V);

        U1 = vec_mergel(U, U);
        V1 = vec_mergel(V, V);

        cvtyuvtoRGB(c, Y0, U0, V0, &R0, &G0, &B0);
        cvtyuvtoRGB(c, Y1, U1, V1, &R1, &G1, &B1);

        R = vec_packclp(R0, R1);
        G = vec_packclp(G0, G1);
        B = vec_packclp(B0, B1);

        nout = (vector unsigned char *) scratch;
        switch (target) {
817
        case AV_PIX_FMT_ABGR:
818 819
            out_abgr(R, G, B, nout);
            break;
820
        case AV_PIX_FMT_BGRA:
821 822
            out_bgra(R, G, B, nout);
            break;
823
        case AV_PIX_FMT_RGBA:
824 825
            out_rgba(R, G, B, nout);
            break;
826
        case AV_PIX_FMT_ARGB:
827 828
            out_argb(R, G, B, nout);
            break;
829
        case AV_PIX_FMT_RGB24:
830 831
            out_rgb24(R, G, B, nout);
            break;
832
        case AV_PIX_FMT_BGR24:
833 834
            out_bgr24(R, G, B, nout);
            break;
Ramiro Polla's avatar
Ramiro Polla committed
835 836
        default:
            /* Unreachable, I think. */
837 838
            av_log(c, AV_LOG_ERROR,
                   "altivec_yuv2packedX doesn't support %s output\n",
Ramiro Polla's avatar
Ramiro Polla committed
839 840
                   sws_format_name(c->dstFormat));
            return;
841
        }
842

843
        memcpy(&((uint32_t *) dest)[i], scratch, (dstW - i) / 4);
844
    }
845
}
846

847 848 849 850 851 852 853 854 855 856 857 858
#define YUV2PACKEDX_WRAPPER(suffix, pixfmt)                             \
void ff_yuv2 ## suffix ## _X_altivec(SwsContext *c,                     \
                                     const int16_t *lumFilter,          \
                                     const int16_t **lumSrc,            \
                                     int lumFilterSize,                 \
                                     const int16_t *chrFilter,          \
                                     const int16_t **chrUSrc,           \
                                     const int16_t **chrVSrc,           \
                                     int chrFilterSize,                 \
                                     const int16_t **alpSrc,            \
                                     uint8_t *dest, int dstW, int dstY) \
{                                                                       \
859 860 861 862
    yuv2packedX_altivec(c, lumFilter, lumSrc, lumFilterSize,            \
                        chrFilter, chrUSrc, chrVSrc,                    \
                        chrFilterSize, alpSrc,                          \
                        dest, dstW, dstY, pixfmt);                      \
863 864
}

865 866 867 868 869 870
YUV2PACKEDX_WRAPPER(abgr,  AV_PIX_FMT_ABGR);
YUV2PACKEDX_WRAPPER(bgra,  AV_PIX_FMT_BGRA);
YUV2PACKEDX_WRAPPER(argb,  AV_PIX_FMT_ARGB);
YUV2PACKEDX_WRAPPER(rgba,  AV_PIX_FMT_RGBA);
YUV2PACKEDX_WRAPPER(rgb24, AV_PIX_FMT_RGB24);
YUV2PACKEDX_WRAPPER(bgr24, AV_PIX_FMT_BGR24);
871 872

#endif /* HAVE_ALTIVEC */