Commit ae8a270f614bd16b52e40bf74358dcad3d240b5d
2020-12-23T21:37:40
Add SDL_SoftStretchLowerLinear() (Bug 5313)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 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 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773
diff --git a/include/SDL_surface.h b/include/SDL_surface.h
index d3f8c81..a9975d7 100644
--- a/include/SDL_surface.h
+++ b/include/SDL_surface.h
@@ -519,6 +519,17 @@ extern DECLSPEC int SDLCALL SDL_SoftStretch(SDL_Surface * src,
SDL_Surface * dst,
const SDL_Rect * dstrect);
+/**
+ * \brief Perform a bilinear scaling between two surfaces of the
+ * same pixel format, 32BPP.
+ *
+ */
+extern DECLSPEC int SDLCALL SDL_SoftStretchLinear(SDL_Surface * src,
+ const SDL_Rect * srcrect,
+ SDL_Surface * dst,
+ const SDL_Rect * dstrect);
+
+
#define SDL_BlitScaled SDL_UpperBlitScaled
/**
diff --git a/src/video/SDL_stretch.c b/src/video/SDL_stretch.c
index bb988e9..5fdb317 100644
--- a/src/video/SDL_stretch.c
+++ b/src/video/SDL_stretch.c
@@ -198,6 +198,7 @@ copy_row3(Uint8 * src, int src_w, Uint8 * dst, int dst_w)
}
static int SDL_SoftStretchLowerNearest(SDL_Surface *src, const SDL_Rect *srcrect, SDL_Surface *dst, const SDL_Rect *dstrect);
+static int SDL_SoftStretchLowerLinear(SDL_Surface *src, const SDL_Rect *srcrect, SDL_Surface *dst, const SDL_Rect *dstrect);
static int SDL_UpperSoftStretch(SDL_Surface * src, const SDL_Rect * srcrect, SDL_Surface * dst, const SDL_Rect * dstrect, SDL_ScaleMode scaleMode);
/* Perform a stretch blit between two surfaces of the same format.
@@ -210,6 +211,13 @@ SDL_SoftStretch(SDL_Surface *src, const SDL_Rect *srcrect,
return SDL_UpperSoftStretch(src, srcrect, dst, dstrect, SDL_ScaleModeNearest);
}
+int
+SDL_SoftStretchLinear(SDL_Surface *src, const SDL_Rect *srcrect,
+ SDL_Surface *dst, const SDL_Rect *dstrect)
+{
+ return SDL_UpperSoftStretch(src, srcrect, dst, dstrect, SDL_ScaleModeLinear);
+}
+
static int
SDL_UpperSoftStretch(SDL_Surface * src, const SDL_Rect * srcrect,
SDL_Surface * dst, const SDL_Rect * dstrect, SDL_ScaleMode scaleMode)
@@ -273,7 +281,9 @@ SDL_UpperSoftStretch(SDL_Surface * src, const SDL_Rect * srcrect,
}
if (scaleMode == SDL_ScaleModeNearest) {
- ret = SDL_SoftStretchLowerNearest(src, srcrect, dst, dstrect);
+ ret = SDL_SoftStretchLowerNearest(src, srcrect, dst, dstrect);
+ } else {
+ ret = SDL_SoftStretchLowerLinear(src, srcrect, dst, dstrect);
}
/* We need to unlock the surfaces if they're locked */
@@ -376,4 +386,713 @@ SDL_SoftStretchLowerNearest(SDL_Surface *src, const SDL_Rect *srcrect,
return 0;
}
+
+/* bilinear interpolation precision must be < 8
+ Because with SSE: add-multiply: _mm_madd_epi16 works with signed int
+ so pixels 0xb1...... are negatives and false the result
+ same in NEON probably */
+#define PRECISION 7
+
+#define FIXED_POINT(i) ((uint32_t)(i) << 16)
+#define SRC_INDEX(fp) ((uint32_t)(fp) >> 16)
+#define INTEGER(fp) ((uint32_t)(fp) >> PRECISION)
+#define FRAC(fp) ((uint32_t)(fp >> (16 - PRECISION)) & ((1<<PRECISION) - 1))
+#define FRAC_ZERO 0
+#define FRAC_ONE (1 << PRECISION)
+#define FP_ONE FIXED_POINT(1)
+
+
+#define BILINEAR___START \
+ int i; \
+ int fp_sum_h, fp_step_h, left_pad_h, right_pad_h; \
+ int fp_sum_w, fp_step_w, left_pad_w, right_pad_w; \
+ int fp_sum_w_init, left_pad_w_init, right_pad_w_init, dst_gap, middle_init; \
+ get_scaler_datas(src_h, dst_h, &fp_sum_h, &fp_step_h, &left_pad_h, &right_pad_h); \
+ get_scaler_datas(src_w, dst_w, &fp_sum_w, &fp_step_w, &left_pad_w, &right_pad_w); \
+ fp_sum_w_init = fp_sum_w + left_pad_w * fp_step_w; \
+ left_pad_w_init = left_pad_w; \
+ right_pad_w_init = right_pad_w; \
+ dst_gap = dst_pitch - 4 * dst_w; \
+ middle_init = dst_w - left_pad_w - right_pad_w; \
+
+#define BILINEAR___HEIGHT \
+ int index_h, frac_h0, frac_h1, middle; \
+ const Uint32 *src_h0, *src_h1; \
+ int no_padding, incr_h0, incr_h1; \
+ \
+ no_padding = !(i < left_pad_h || i > dst_h - 1 - right_pad_h); \
+ index_h = SRC_INDEX(fp_sum_h); \
+ frac_h0 = FRAC(fp_sum_h); \
+ \
+ index_h = no_padding ? index_h : (i < left_pad_h ? 0 : src_h - 1); \
+ frac_h0 = no_padding ? frac_h0 : 0; \
+ incr_h1 = no_padding ? src_pitch : 0; \
+ incr_h0 = index_h * src_pitch; \
+ \
+ src_h0 = (const Uint32 *)((const Uint8 *)src + incr_h0); \
+ src_h1 = (const Uint32 *)((const Uint8 *)src_h0 + incr_h1); \
+ \
+ fp_sum_h += fp_step_h; \
+ \
+ frac_h1 = FRAC_ONE - frac_h0; \
+ fp_sum_w = fp_sum_w_init; \
+ right_pad_w = right_pad_w_init; \
+ left_pad_w = left_pad_w_init; \
+ middle = middle_init; \
+
+
+
+
+static void
+#if defined(__clang__)
+// Remove inlining of this function
+// Crash with clang 9.0.8 / android-ndk-r21d
+// Ok with clang 11.0.5 / android-ndk-22
+# if __clang_major__ == 9 && __clang_minor__ == 0 && __clang_patchlevel__ == 8
+__attribute__((noinline))
+# endif
+#endif
+get_scaler_datas(int src_nb, int dst_nb, int *fp_start, int *fp_step, int *left_pad, int *right_pad)
+{
+
+ int step = FIXED_POINT(src_nb) / (dst_nb); /* source step in fixed point */
+ int x0 = FP_ONE / 2; /* dst first pixel center at 0.5 in fixed point */
+ int fp_sum;
+ int i;
+#if 0
+ /* scale to source coordinates */
+ x0 *= src_nb;
+ x0 /= dst_nb; /* x0 == step / 2 */
+#else
+ /* Use this code for perfect match with pixman */
+ Sint64 tmp[2];
+ tmp[0] = (Sint64)step * (x0 >> 16);
+ tmp[1] = (Sint64)step * (x0 & 0xFFFF);
+ x0 = tmp[0] + ((tmp[1] + 0x8000) >> 16); /* x0 == (step + 1) / 2 */
+#endif
+ /* -= 0.5, get back the pixel origin, in source coordinates */
+ x0 -= FP_ONE / 2;
+
+ *fp_start = x0;
+ *fp_step = step;
+ *left_pad = 0;
+ *right_pad = 0;
+
+ fp_sum = x0;
+ for (i = 0; i < dst_nb; i++) {
+ if (fp_sum < 0) {
+ *left_pad += 1;
+ } else {
+ int index = SRC_INDEX(fp_sum);
+ if (index > src_nb - 2) {
+ *right_pad += 1;
+ }
+ }
+ fp_sum += step;
+ }
+// SDL_Log("%d -> %d x0=%d step=%d left_pad=%d right_pad=%d", src_nb, dst_nb, *fp_start, *fp_step, *left_pad, *right_pad);
+}
+
+typedef struct color_t {
+ Uint8 a;
+ Uint8 b;
+ Uint8 c;
+ Uint8 d;
+} color_t;
+
+#if 0
+static void
+printf_64(const char *str, void *var)
+{
+ uint8_t *val = (uint8_t*) var;
+ printf(" * %s: %02x %02x %02x %02x _ %02x %02x %02x %02x\n",
+ str, val[0], val[1], val[2], val[3], val[4], val[5], val[6], val[7]);
+}
+#endif
+
+/* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */
+
+static inline void
+INTERPOL(const Uint32 *src_x0, const Uint32 *src_x1, int frac0, int frac1, Uint32 *dst)
+{
+ const color_t *c0 = (const color_t *)src_x0;
+ const color_t *c1 = (const color_t *)src_x1;
+ color_t *cx = (color_t *)dst;
+#if 0
+ cx->a = c0->a + INTEGER(frac0 * (c1->a - c0->a));
+ cx->b = c0->b + INTEGER(frac0 * (c1->b - c0->b));
+ cx->c = c0->c + INTEGER(frac0 * (c1->c - c0->c));
+ cx->d = c0->d + INTEGER(frac0 * (c1->d - c0->d));
+#else
+ cx->a = INTEGER(frac1 * c0->a + frac0 * c1->a);
+ cx->b = INTEGER(frac1 * c0->b + frac0 * c1->b);
+ cx->c = INTEGER(frac1 * c0->c + frac0 * c1->c);
+ cx->d = INTEGER(frac1 * c0->d + frac0 * c1->d);
+#endif
+}
+
+static inline void
+INTERPOL_BILINEAR(const Uint32 *s0, const Uint32 *s1, int frac_w0, int frac_h0, int frac_h1, Uint32 *dst)
+{
+ Uint32 tmp[2];
+ unsigned int frac_w1 = FRAC_ONE - frac_w0;
+
+ /* Vertical first, store to 'tmp' */
+ INTERPOL(s0, s1, frac_h0, frac_h1, tmp);
+ INTERPOL(s0 + 1, s1 + 1, frac_h0, frac_h1, tmp + 1);
+
+ /* Horizontal, store to 'dst' */
+ INTERPOL(tmp, tmp + 1, frac_w0, frac_w1, dst);
+}
+
+static int
+scale_mat(const Uint32 *src, int src_w, int src_h, int src_pitch,
+ Uint32 *dst, int dst_w, int dst_h, int dst_pitch)
+{
+ BILINEAR___START
+
+ for (i = 0; i < dst_h; i++) {
+
+ BILINEAR___HEIGHT
+
+ while (left_pad_w--) {
+ INTERPOL_BILINEAR(src_h0, src_h1, FRAC_ZERO, frac_h0, frac_h1, dst);
+ dst += 1;
+ }
+
+ while (middle--) {
+ const Uint32 *s_00_01;
+ const Uint32 *s_10_11;
+ int index_w = 4 * SRC_INDEX(fp_sum_w);
+ int frac_w = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+
+/*
+ x00 ... x0_ ..... x01
+ . . .
+ . x .
+ . . .
+ . . .
+ x10 ... x1_ ..... x11
+*/
+ s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w);
+ s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w);
+
+ INTERPOL_BILINEAR(s_00_01, s_10_11, frac_w, frac_h0, frac_h1, dst);
+
+ dst += 1;
+ }
+
+ while (right_pad_w--) {
+ int index_w = 4 * (src_w - 2);
+ const Uint32 *s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w);
+ const Uint32 *s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w);
+ INTERPOL_BILINEAR(s_00_01, s_10_11, FRAC_ONE, frac_h0, frac_h1, dst);
+ dst += 1;
+ }
+ dst = (Uint32 *)((Uint8 *)dst + dst_gap);
+ }
+ return 0;
+}
+
+#if defined(__SSE2__)
+# define HAVE_SSE2_INTRINSICS 1
+#endif
+
+#if defined(__ARM_NEON)
+# define HAVE_NEON_INTRINSICS 1
+#endif
+
+/* TODO: this didn't compile on Window10 universal package last time I tried .. */
+#if defined(__WINRT__)
+# if defined(HAVE_NEON_INTRINSICS)
+# undef HAVE_NEON_INTRINSICS
+# endif
+#endif
+
+#if defined(HAVE_SSE2_INTRINSICS)
+
+#if 0
+static void
+printf_128(const char *str, __m128i var)
+{
+ uint16_t *val = (uint16_t*) &var;
+ printf(" * %s: %04x %04x %04x %04x _ %04x %04x %04x %04x\n",
+ str, val[0], val[1], val[2], val[3], val[4], val[5], val[6], val[7]);
+}
+#endif
+
+static inline int
+hasSSE2()
+{
+ static int val = -1;
+ if (val != -1) {
+ return val;
+ }
+ val = SDL_HasSSE2();
+ return val;
+}
+
+static inline void
+INTERPOL_BILINEAR_SSE(const Uint32 *s0, const Uint32 *s1, int frac_w, __m128i v_frac_h0, __m128i v_frac_h1, Uint32 *dst, __m128i zero)
+{
+ __m128i x_00_01, x_10_11; /* Pixels in 4*uint8 in row */
+ __m128i v_frac_w0, k0, l0, d0, e0;
+
+ int f, f2;
+ f = frac_w;
+ f2 = FRAC_ONE - frac_w;
+ v_frac_w0 = _mm_set_epi16(f, f2, f, f2, f, f2, f, f2);
+
+
+ x_00_01 = _mm_loadl_epi64((const __m128i *)s0); /* Load x00 and x01 */
+ x_10_11 = _mm_loadl_epi64((const __m128i *)s1);
+
+ /* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */
+
+ /* Interpolation vertical */
+ k0 = _mm_mullo_epi16(_mm_unpacklo_epi8(x_00_01, zero), v_frac_h1);
+ l0 = _mm_mullo_epi16(_mm_unpacklo_epi8(x_10_11, zero), v_frac_h0);
+ k0 = _mm_add_epi16(k0, l0);
+
+ /* For perfect match, clear the factionnal part eventually. */
+ /*
+ k0 = _mm_srli_epi16(k0, PRECISION);
+ k0 = _mm_slli_epi16(k0, PRECISION);
+ */
+
+ /* Interpolation horizontal */
+ l0 = _mm_unpacklo_epi64(/* unused */ l0, k0);
+ k0 = _mm_madd_epi16(_mm_unpackhi_epi16(l0, k0), v_frac_w0);
+
+ /* Store 1 pixel */
+ d0 = _mm_srli_epi32(k0, PRECISION * 2);
+ e0 = _mm_packs_epi32(d0, d0);
+ e0 = _mm_packus_epi16(e0, e0);
+ *dst = _mm_cvtsi128_si32(e0);
+}
+
+static int
+scale_mat_SSE(const Uint32 *src, int src_w, int src_h, int src_pitch, Uint32 *dst, int dst_w, int dst_h, int dst_pitch)
+{
+ BILINEAR___START
+
+ for (i = 0; i < dst_h; i++) {
+ int nb_block2;
+ __m128i v_frac_h0;
+ __m128i v_frac_h1;
+ __m128i zero;
+
+ BILINEAR___HEIGHT
+
+ nb_block2 = middle / 2;
+
+ v_frac_h0 = _mm_set_epi16(frac_h0, frac_h0, frac_h0, frac_h0, frac_h0, frac_h0, frac_h0, frac_h0);
+ v_frac_h1 = _mm_set_epi16(frac_h1, frac_h1, frac_h1, frac_h1, frac_h1, frac_h1, frac_h1, frac_h1);
+ zero = _mm_setzero_si128();
+
+ while (left_pad_w--) {
+ INTERPOL_BILINEAR_SSE(src_h0, src_h1, FRAC_ZERO, v_frac_h0, v_frac_h1, dst, zero);
+ dst += 1;
+ }
+
+ while (nb_block2--) {
+ int index_w_0, frac_w_0;
+ int index_w_1, frac_w_1;
+
+ const Uint32 *s_00_01, *s_02_03, *s_10_11, *s_12_13;
+
+ __m128i x_00_01, x_10_11, x_02_03, x_12_13;/* Pixels in 4*uint8 in row */
+ __m128i v_frac_w0, k0, l0, d0, e0;
+ __m128i v_frac_w1, k1, l1, d1, e1;
+
+ int f, f2;
+ index_w_0 = 4 * SRC_INDEX(fp_sum_w);
+ frac_w_0 = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+ index_w_1 = 4 * SRC_INDEX(fp_sum_w);
+ frac_w_1 = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+/*
+ x00............ x01 x02...........x03
+ . . . . . .
+ j0 f0 j1 j2 f1 j3
+ . . . . . .
+ . . . . . .
+ . . . . . .
+ x10............ x11 x12...........x13
+ */
+ s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w_0);
+ s_02_03 = (const Uint32 *)((const Uint8 *)src_h0 + index_w_1);
+ s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_0);
+ s_12_13 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_1);
+
+ f = frac_w_0;
+ f2 = FRAC_ONE - frac_w_0;
+ v_frac_w0 = _mm_set_epi16(f, f2, f, f2, f, f2, f, f2);
+
+ f = frac_w_1;
+ f2 = FRAC_ONE - frac_w_1;
+ v_frac_w1 = _mm_set_epi16(f, f2, f, f2, f, f2, f, f2);
+
+ x_00_01 = _mm_loadl_epi64((const __m128i *)s_00_01); /* Load x00 and x01 */
+ x_02_03 = _mm_loadl_epi64((const __m128i *)s_02_03);
+ x_10_11 = _mm_loadl_epi64((const __m128i *)s_10_11);
+ x_12_13 = _mm_loadl_epi64((const __m128i *)s_12_13);
+
+ /* Interpolation vertical */
+ k0 = _mm_mullo_epi16(_mm_unpacklo_epi8(x_00_01, zero), v_frac_h1);
+ l0 = _mm_mullo_epi16(_mm_unpacklo_epi8(x_10_11, zero), v_frac_h0);
+ k0 = _mm_add_epi16(k0, l0);
+ k1 = _mm_mullo_epi16(_mm_unpacklo_epi8(x_02_03, zero), v_frac_h1);
+ l1 = _mm_mullo_epi16(_mm_unpacklo_epi8(x_12_13, zero), v_frac_h0);
+ k1 = _mm_add_epi16(k1, l1);
+
+ /* Interpolation horizontal */
+ l0 = _mm_unpacklo_epi64(/* unused */ l0, k0);
+ k0 = _mm_madd_epi16(_mm_unpackhi_epi16(l0, k0), v_frac_w0);
+ l1 = _mm_unpacklo_epi64(/* unused */ l1, k1);
+ k1 = _mm_madd_epi16(_mm_unpackhi_epi16(l1, k1), v_frac_w1);
+
+ /* Store 1 pixel */
+ d0 = _mm_srli_epi32(k0, PRECISION * 2);
+ e0 = _mm_packs_epi32(d0, d0);
+ e0 = _mm_packus_epi16(e0, e0);
+ *dst++ = _mm_cvtsi128_si32(e0);
+
+ /* Store 1 pixel */
+ d1 = _mm_srli_epi32(k1, PRECISION * 2);
+ e1 = _mm_packs_epi32(d1, d1);
+ e1 = _mm_packus_epi16(e1, e1);
+ *dst++ = _mm_cvtsi128_si32(e1);
+ }
+
+ /* Last point */
+ if (middle & 0x1) {
+ const Uint32 *s_00_01;
+ const Uint32 *s_10_11;
+ int index_w = 4 * SRC_INDEX(fp_sum_w);
+ int frac_w = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+ s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w);
+ s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w);
+ INTERPOL_BILINEAR_SSE(s_00_01, s_10_11, frac_w, v_frac_h0, v_frac_h1, dst, zero);
+ dst += 1;
+ }
+
+ while (right_pad_w--) {
+ int index_w = 4 * (src_w - 2);
+ const Uint32 *s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w);
+ const Uint32 *s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w);
+ INTERPOL_BILINEAR_SSE(s_00_01, s_10_11, FRAC_ONE, v_frac_h0, v_frac_h1, dst, zero);
+ dst += 1;
+ }
+ dst = (Uint32 *)((Uint8 *)dst + dst_gap);
+ }
+ return 0;
+}
+#endif
+
+#if defined(HAVE_NEON_INTRINSICS)
+
+static inline int
+hasNEON()
+{
+ static int val = -1;
+ if (val != -1) {
+ return val;
+ }
+ val = SDL_HasNEON();
+ return val;
+}
+
+static inline void
+INTERPOL_BILINEAR_NEON(const Uint32 *s0, const Uint32 *s1, int frac_w, uint8x8_t v_frac_h0, uint8x8_t v_frac_h1, Uint32 *dst)
+{
+ uint8x8_t x_00_01, x_10_11; /* Pixels in 4*uint8 in row */
+ uint16x8_t k0;
+ uint32x4_t l0;
+ uint16x8_t d0;
+ uint8x8_t e0;
+
+ x_00_01 = (uint8x8_t)vld1_u32(s0); /* Load 2 pixels */
+ x_10_11 = (uint8x8_t)vld1_u32(s1);
+
+ /* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */
+ k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */
+ k0 = vmlal_u8(k0, x_10_11, v_frac_h0); /* k0 += x1 * frac */
+
+ /* k0 now contains 2 interpolated pixels { j0, j1 } */
+ l0 = vshll_n_u16(vget_low_u16(k0), PRECISION);
+ l0 = vmlsl_n_u16(l0, vget_low_u16(k0), frac_w);
+ l0 = vmlal_n_u16(l0, vget_high_u16(k0), frac_w);
+
+ /* Shift and narrow */
+ d0 = vcombine_u16(
+ /* uint16x4_t */ vshrn_n_u32(l0, 2 * PRECISION),
+ /* uint16x4_t */ vshrn_n_u32(l0, 2 * PRECISION)
+ );
+
+ /* Narrow again */
+ e0 = vmovn_u16(d0);
+
+ /* Store 1 pixel */
+ *dst = vget_lane_u32((uint32x2_t)e0, 0);
+}
+
+ static int
+scale_mat_NEON(const Uint32 *src, int src_w, int src_h, int src_pitch, Uint32 *dst, int dst_w, int dst_h, int dst_pitch)
+{
+ BILINEAR___START
+
+ for (i = 0; i < dst_h; i++) {
+ int nb_block4;
+ uint8x8_t v_frac_h0, v_frac_h1;
+
+ BILINEAR___HEIGHT
+
+ nb_block4 = middle / 4;
+
+ v_frac_h0 = vmov_n_u8(frac_h0);
+ v_frac_h1 = vmov_n_u8(frac_h1);
+
+ while (left_pad_w--) {
+ INTERPOL_BILINEAR_NEON(src_h0, src_h1, FRAC_ZERO, v_frac_h0, v_frac_h1, dst);
+ dst += 1;
+ }
+
+ while (nb_block4--) {
+ int index_w_0, frac_w_0;
+ int index_w_1, frac_w_1;
+ int index_w_2, frac_w_2;
+ int index_w_3, frac_w_3;
+
+ const Uint32 *s_00_01, *s_02_03, *s_04_05, *s_06_07;
+ const Uint32 *s_10_11, *s_12_13, *s_14_15, *s_16_17;
+
+ uint8x8_t x_00_01, x_10_11, x_02_03, x_12_13;/* Pixels in 4*uint8 in row */
+ uint8x8_t x_04_05, x_14_15, x_06_07, x_16_17;
+
+ uint16x8_t k0, k1, k2, k3;
+ uint32x4_t l0, l1, l2, l3;
+ uint16x8_t d0, d1;
+ uint8x8_t e0, e1;
+ uint32x4_t f0;
+
+ index_w_0 = 4 * SRC_INDEX(fp_sum_w);
+ frac_w_0 = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+ index_w_1 = 4 * SRC_INDEX(fp_sum_w);
+ frac_w_1 = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+ index_w_2 = 4 * SRC_INDEX(fp_sum_w);
+ frac_w_2 = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+ index_w_3 = 4 * SRC_INDEX(fp_sum_w);
+ frac_w_3 = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+
+ s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w_0);
+ s_02_03 = (const Uint32 *)((const Uint8 *)src_h0 + index_w_1);
+ s_04_05 = (const Uint32 *)((const Uint8 *)src_h0 + index_w_2);
+ s_06_07 = (const Uint32 *)((const Uint8 *)src_h0 + index_w_3);
+ s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_0);
+ s_12_13 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_1);
+ s_14_15 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_2);
+ s_16_17 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_3);
+
+ /* Interpolation vertical */
+ x_00_01 = (uint8x8_t)vld1_u32(s_00_01); /* Load 2 pixels */
+ x_02_03 = (uint8x8_t)vld1_u32(s_02_03);
+ x_04_05 = (uint8x8_t)vld1_u32(s_04_05);
+ x_06_07 = (uint8x8_t)vld1_u32(s_06_07);
+ x_10_11 = (uint8x8_t)vld1_u32(s_10_11);
+ x_12_13 = (uint8x8_t)vld1_u32(s_12_13);
+ x_14_15 = (uint8x8_t)vld1_u32(s_14_15);
+ x_16_17 = (uint8x8_t)vld1_u32(s_16_17);
+
+ /* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */
+ k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */
+ k0 = vmlal_u8(k0, x_10_11, v_frac_h0); /* k0 += x1 * frac */
+
+ k1 = vmull_u8(x_02_03, v_frac_h1);
+ k1 = vmlal_u8(k1, x_12_13, v_frac_h0);
+
+ k2 = vmull_u8(x_04_05, v_frac_h1);
+ k2 = vmlal_u8(k2, x_14_15, v_frac_h0);
+
+ k3 = vmull_u8(x_06_07, v_frac_h1);
+ k3 = vmlal_u8(k3, x_16_17, v_frac_h0);
+
+ /* k0 now contains 2 interpolated pixels { j0, j1 } */
+ /* k1 now contains 2 interpolated pixels { j2, j3 } */
+ /* k2 now contains 2 interpolated pixels { j4, j5 } */
+ /* k3 now contains 2 interpolated pixels { j6, j7 } */
+
+ l0 = vshll_n_u16(vget_low_u16(k0), PRECISION);
+ l0 = vmlsl_n_u16(l0, vget_low_u16(k0), frac_w_0);
+ l0 = vmlal_n_u16(l0, vget_high_u16(k0), frac_w_0);
+
+ l1 = vshll_n_u16(vget_low_u16(k1), PRECISION);
+ l1 = vmlsl_n_u16(l1, vget_low_u16(k1), frac_w_1);
+ l1 = vmlal_n_u16(l1, vget_high_u16(k1), frac_w_1);
+
+ l2 = vshll_n_u16(vget_low_u16(k2), PRECISION);
+ l2 = vmlsl_n_u16(l2, vget_low_u16(k2), frac_w_2);
+ l2 = vmlal_n_u16(l2, vget_high_u16(k2), frac_w_2);
+
+ l3 = vshll_n_u16(vget_low_u16(k3), PRECISION);
+ l3 = vmlsl_n_u16(l3, vget_low_u16(k3), frac_w_3);
+ l3 = vmlal_n_u16(l3, vget_high_u16(k3), frac_w_3);
+
+ /* shift and narrow */
+ d0 = vcombine_u16(
+ /* uint16x4_t */ vshrn_n_u32(l0, 2 * PRECISION),
+ /* uint16x4_t */ vshrn_n_u32(l1, 2 * PRECISION)
+ );
+ /* narrow again */
+ e0 = vmovn_u16(d0);
+
+ /* Shift and narrow */
+ d1 = vcombine_u16(
+ /* uint16x4_t */ vshrn_n_u32(l2, 2 * PRECISION),
+ /* uint16x4_t */ vshrn_n_u32(l3, 2 * PRECISION)
+ );
+ /* Narrow again */
+ e1 = vmovn_u16(d1);
+
+ f0 = vcombine_u32((uint32x2_t)e0, (uint32x2_t)e1);
+ /* Store 4 pixels */
+ vst1q_u32(dst, f0);
+
+ dst += 4;
+ }
+
+ if (middle & 0x2) {
+ int index_w_0, frac_w_0;
+ int index_w_1, frac_w_1;
+ const Uint32 *s_00_01, *s_02_03;
+ const Uint32 *s_10_11, *s_12_13;
+ uint8x8_t x_00_01, x_10_11, x_02_03, x_12_13;/* Pixels in 4*uint8 in row */
+ uint16x8_t k0, k1;
+ uint32x4_t l0, l1;
+ uint16x8_t d0;
+ uint8x8_t e0;
+
+ index_w_0 = 4 * SRC_INDEX(fp_sum_w);
+ frac_w_0 = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+ index_w_1 = 4 * SRC_INDEX(fp_sum_w);
+ frac_w_1 = FRAC(fp_sum_w);
+ fp_sum_w += fp_step_w;
+/*
+ x00............ x01 x02...........x03
+ . . . . . .
+ j0 dest0 j1 j2 dest1 j3
+ . . . . . .
+ . . . . . .
+ . . . . . .
+ x10............ x11 x12...........x13
+*/
+ s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w_0);
+ s_02_03 = (const Uint32 *)((const Uint8 *)src_h0 + index_w_1);
+ s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_0);
+ s_12_13 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_1);
+
+ /* Interpolation vertical */
+ x_00_01 = (uint8x8_t)vld1_u32(s_00_01);/* Load 2 pixels */
+ x_02_03 = (uint8x8_t)vld1_u32(s_02_03);
+ x_10_11 = (uint8x8_t)vld1_u32(s_10_11);
+ x_12_13 = (uint8x8_t)vld1_u32(s_12_13);
+
+ /* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */
+ k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */
+ k0 = vmlal_u8(k0, x_10_11, v_frac_h0); /* k0 += x1 * frac */
+
+ k1 = vmull_u8(x_02_03, v_frac_h1);
+ k1 = vmlal_u8(k1, x_12_13, v_frac_h0);
+
+ /* k0 now contains 2 interpolated pixels { j0, j1 } */
+ /* k1 now contains 2 interpolated pixels { j2, j3 } */
+
+ l0 = vshll_n_u16(vget_low_u16(k0), PRECISION);
+ l0 = vmlsl_n_u16(l0, vget_low_u16(k0), frac_w_0);
+ l0 = vmlal_n_u16(l0, vget_high_u16(k0), frac_w_0);
+
+ l1 = vshll_n_u16(vget_low_u16(k1), PRECISION);
+ l1 = vmlsl_n_u16(l1, vget_low_u16(k1), frac_w_1);
+ l1 = vmlal_n_u16(l1, vget_high_u16(k1), frac_w_1);
+
+ /* Shift and narrow */
+
+ d0 = vcombine_u16(
+ /* uint16x4_t */ vshrn_n_u32(l0, 2 * PRECISION),
+ /* uint16x4_t */ vshrn_n_u32(l1, 2 * PRECISION)
+ );
+
+ /* Narrow again */
+ e0 = vmovn_u16(d0);
+
+ /* Store 2 pixels */
+ vst1_u32(dst, (uint32x2_t)e0);
+ dst += 2;
+ }
+
+ /* Last point */
+ if (middle & 0x1) {
+ int index_w = 4 * SRC_INDEX(fp_sum_w);
+ int frac_w = FRAC(fp_sum_w);
+ const Uint32 *s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w);
+ const Uint32 *s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w);
+ INTERPOL_BILINEAR_NEON(s_00_01, s_10_11, frac_w, v_frac_h0, v_frac_h1, dst);
+ dst += 1;
+ }
+
+ while (right_pad_w--) {
+ int index_w = 4 * (src_w - 2);
+ const Uint32 *s_00_01 = (const Uint32 *)((const Uint8 *)src_h0 + index_w);
+ const Uint32 *s_10_11 = (const Uint32 *)((const Uint8 *)src_h1 + index_w);
+ INTERPOL_BILINEAR_NEON(s_00_01, s_10_11, FRAC_ONE, v_frac_h0, v_frac_h1, dst);
+ dst += 1;
+ }
+
+ dst = (Uint32 *)((Uint8 *)dst + dst_gap);
+ }
+ return 0;
+}
+#endif
+
+int
+SDL_SoftStretchLowerLinear(SDL_Surface *s, const SDL_Rect *srcrect,
+ SDL_Surface *d, const SDL_Rect *dstrect)
+{
+ int ret = -1;
+ int src_w = srcrect->w;
+ int src_h = srcrect->h;
+ int dst_w = dstrect->w;
+ int dst_h = dstrect->h;
+ int src_pitch = s->pitch;
+ int dst_pitch = d->pitch;
+ Uint32 *src = (Uint32 *) ((Uint8 *)s->pixels + srcrect->x * 4 + srcrect->y * src_pitch);
+ Uint32 *dst = (Uint32 *) ((Uint8 *)d->pixels + dstrect->x * 4 + dstrect->y * dst_pitch);
+
+#if defined(HAVE_NEON_INTRINSICS)
+ if (ret == -1 && hasNEON()) {
+ ret = scale_mat_NEON(src, src_w, src_h, src_pitch, dst, dst_w, dst_h, dst_pitch);
+ }
+#endif
+
+#if defined(HAVE_SSE2_INTRINSICS)
+ if (ret == -1 && hasSSE2()) {
+ ret = scale_mat_SSE(src, src_w, src_h, src_pitch, dst, dst_w, dst_h, dst_pitch);
+ }
+#endif
+
+ if (ret == -1) {
+ scale_mat(src, src_w, src_h, src_pitch, dst, dst_w, dst_h, dst_pitch);
+ }
+
+ return ret;
+}
+
/* vi: set ts=4 sw=4 expandtab: */