1 /*
2 * Copyright 2017 Google Inc.
3 *
4 * Use of this source code is governed by a BSD-style license that can be
5 * found in the LICENSE file.
6 */
7
8 #include "SkJumper.h"
9 #include "SkJumper_misc.h" // SI, unaligned_load(), bit_cast()
10 #include "SkJumper_vectors.h" // F, I32, U32, U16, U8, cast(), expand()
11
12 // Our fundamental vector depth is our pixel stride.
13 static const size_t kStride = sizeof(F) / sizeof(float);
14
15 // A reminder:
16 // Code guarded by defined(JUMPER) can assume that it will be compiled by Clang
17 // and that F, I32, etc. are kStride-deep ext_vector_types of the appropriate type.
18 // Otherwise, F, I32, etc. just alias the basic scalar types (and so kStride == 1).
19
20 // You can use most constants in this file, but in a few rare exceptions we read from this struct.
21 using K = const SkJumper_constants;
22
23 // A little wrapper macro to name Stages differently depending on the instruction set.
24 // That lets us link together several options.
25 #if !defined(JUMPER)
26 #define WRAP(name) sk_##name
27 #elif defined(__aarch64__)
28 #define WRAP(name) sk_##name##_aarch64
29 #elif defined(__arm__)
30 #define WRAP(name) sk_##name##_vfp4
31 #elif defined(__AVX2__)
32 #define WRAP(name) sk_##name##_hsw
33 #elif defined(__AVX__)
34 #define WRAP(name) sk_##name##_avx
35 #elif defined(__SSE4_1__)
36 #define WRAP(name) sk_##name##_sse41
37 #elif defined(__SSE2__)
38 #define WRAP(name) sk_##name##_sse2
39 #endif
40
41 // We're finally going to get to what a Stage function looks like!
42 // tail == 0 ~~> work on a full kStride pixels
43 // tail != 0 ~~> work on only the first tail pixels
44 // tail is always < kStride.
45 //
46 // We keep program the second argument, so that it's passed in rsi for load_and_inc().
47 using Stage = void(K* k, void** program, size_t x, size_t y, size_t tail, F,F,F,F, F,F,F,F);
48
49 #if defined(JUMPER) && defined(__AVX__)
50 // We really want to make sure all paths go through this function's (implicit) vzeroupper.
51 // If they don't, we'll experience severe slowdowns when we first use SSE instructions again.
52 __attribute__((disable_tail_calls))
53 #endif
54 MAYBE_MSABI
WRAP(start_pipeline)55 extern "C" void WRAP(start_pipeline)(size_t x, size_t y, size_t limit, void** program, K* k) {
56 #if defined(JUMPER)
57 F v;
58 #else
59 F v{};
60 #endif
61 auto start = (Stage*)load_and_inc(program);
62 while (x + kStride <= limit) {
63 start(k,program,x,y,0, v,v,v,v, v,v,v,v);
64 x += kStride;
65 }
66 if (size_t tail = limit - x) {
67 start(k,program,x,y,tail, v,v,v,v, v,v,v,v);
68 }
69 }
70
71 #if defined(JUMPER) && defined(__AVX__)
72 // We really want to make sure all paths go through this function's (implicit) vzeroupper.
73 // If they don't, we'll experience severe slowdowns when we first use SSE instructions again.
74 __attribute__((disable_tail_calls))
75 #endif
76 #if defined(JUMPER)
77 __attribute__((flatten)) // Force-inline the call to start_pipeline().
78 #endif
79 MAYBE_MSABI
WRAP(start_pipeline_2d)80 extern "C" void WRAP(start_pipeline_2d)(size_t x, size_t y, size_t xlimit, size_t ylimit,
81 void** program, K* k) {
82 for (; y < ylimit; y++) {
83 WRAP(start_pipeline)(x,y,xlimit, program, k);
84 }
85 }
86
87 #define STAGE(name) \
88 SI void name##_k(K* k, LazyCtx ctx, size_t x, size_t y, size_t tail, \
89 F& r, F& g, F& b, F& a, F& dr, F& dg, F& db, F& da); \
90 extern "C" void WRAP(name)(K* k, void** program, size_t x, size_t y, size_t tail, \
91 F r, F g, F b, F a, F dr, F dg, F db, F da) { \
92 LazyCtx ctx(program); \
93 name##_k(k,ctx,x,y,tail, r,g,b,a, dr,dg,db,da); \
94 auto next = (Stage*)load_and_inc(program); \
95 next(k,program,x,y,tail, r,g,b,a, dr,dg,db,da); \
96 } \
97 SI void name##_k(K* k, LazyCtx ctx, size_t x, size_t y, size_t tail, \
98 F& r, F& g, F& b, F& a, F& dr, F& dg, F& db, F& da)
99
100
101 // just_return() is a simple no-op stage that only exists to end the chain,
102 // returning back up to start_pipeline(), and from there to the caller.
WRAP(just_return)103 extern "C" void WRAP(just_return)(K*, void**, size_t,size_t,size_t, F,F,F,F, F,F,F,F) {}
104
105
106 // We could start defining normal Stages now. But first, some helper functions.
107
108 // These load() and store() methods are tail-aware,
109 // but focus mainly on keeping the at-stride tail==0 case fast.
110
111 template <typename V, typename T>
load(const T * src,size_t tail)112 SI V load(const T* src, size_t tail) {
113 #if defined(JUMPER)
114 __builtin_assume(tail < kStride);
115 if (__builtin_expect(tail, 0)) {
116 V v{}; // Any inactive lanes are zeroed.
117 switch (tail) {
118 case 7: v[6] = src[6];
119 case 6: v[5] = src[5];
120 case 5: v[4] = src[4];
121 case 4: memcpy(&v, src, 4*sizeof(T)); break;
122 case 3: v[2] = src[2];
123 case 2: memcpy(&v, src, 2*sizeof(T)); break;
124 case 1: memcpy(&v, src, 1*sizeof(T)); break;
125 }
126 return v;
127 }
128 #endif
129 return unaligned_load<V>(src);
130 }
131
132 template <typename V, typename T>
store(T * dst,V v,size_t tail)133 SI void store(T* dst, V v, size_t tail) {
134 #if defined(JUMPER)
135 __builtin_assume(tail < kStride);
136 if (__builtin_expect(tail, 0)) {
137 switch (tail) {
138 case 7: dst[6] = v[6];
139 case 6: dst[5] = v[5];
140 case 5: dst[4] = v[4];
141 case 4: memcpy(dst, &v, 4*sizeof(T)); break;
142 case 3: dst[2] = v[2];
143 case 2: memcpy(dst, &v, 2*sizeof(T)); break;
144 case 1: memcpy(dst, &v, 1*sizeof(T)); break;
145 }
146 return;
147 }
148 #endif
149 unaligned_store(dst, v);
150 }
151
152 // AVX adds some mask loads and stores that make for shorter, faster code.
153 #if defined(JUMPER) && defined(__AVX__)
mask(size_t tail)154 SI U32 mask(size_t tail) {
155 // We go a little out of our way to avoid needing large constant values here.
156
157 // It's easiest to build the mask as 8 8-bit values, either 0x00 or 0xff.
158 // Start fully on, then shift away lanes from the top until we've got our mask.
159 uint64_t mask = 0xffffffffffffffff >> 8*(kStride-tail);
160
161 // Sign-extend each mask lane to its full width, 0x00000000 or 0xffffffff.
162 using S8 = int8_t __attribute__((ext_vector_type(8)));
163 using S32 = int32_t __attribute__((ext_vector_type(8)));
164 return (U32)__builtin_convertvector(unaligned_load<S8>(&mask), S32);
165 }
166
167 template <>
load(const uint32_t * src,size_t tail)168 inline U32 load(const uint32_t* src, size_t tail) {
169 __builtin_assume(tail < kStride);
170 if (__builtin_expect(tail, 0)) {
171 return (U32)_mm256_maskload_ps((const float*)src, mask(tail));
172 }
173 return unaligned_load<U32>(src);
174 }
175
176 template <>
store(uint32_t * dst,U32 v,size_t tail)177 inline void store(uint32_t* dst, U32 v, size_t tail) {
178 __builtin_assume(tail < kStride);
179 if (__builtin_expect(tail, 0)) {
180 return _mm256_maskstore_ps((float*)dst, mask(tail), (F)v);
181 }
182 unaligned_store(dst, v);
183 }
184 #endif
185
from_byte(U8 b)186 SI F from_byte(U8 b) {
187 return cast(expand(b)) * (1/255.0f);
188 }
from_565(U16 _565,F * r,F * g,F * b)189 SI void from_565(U16 _565, F* r, F* g, F* b) {
190 U32 wide = expand(_565);
191 *r = cast(wide & (31<<11)) * (1.0f / (31<<11));
192 *g = cast(wide & (63<< 5)) * (1.0f / (63<< 5));
193 *b = cast(wide & (31<< 0)) * (1.0f / (31<< 0));
194 }
from_4444(U16 _4444,F * r,F * g,F * b,F * a)195 SI void from_4444(U16 _4444, F* r, F* g, F* b, F* a) {
196 U32 wide = expand(_4444);
197 *r = cast(wide & (15<<12)) * (1.0f / (15<<12));
198 *g = cast(wide & (15<< 8)) * (1.0f / (15<< 8));
199 *b = cast(wide & (15<< 4)) * (1.0f / (15<< 4));
200 *a = cast(wide & (15<< 0)) * (1.0f / (15<< 0));
201 }
from_8888(U32 _8888,F * r,F * g,F * b,F * a)202 SI void from_8888(U32 _8888, F* r, F* g, F* b, F* a) {
203 *r = cast((_8888 ) & 0xff) * (1/255.0f);
204 *g = cast((_8888 >> 8) & 0xff) * (1/255.0f);
205 *b = cast((_8888 >> 16) & 0xff) * (1/255.0f);
206 *a = cast((_8888 >> 24) ) * (1/255.0f);
207 }
208
209 template <typename T>
ix_and_ptr(T ** ptr,const SkJumper_MemoryCtx * ctx,F x,F y)210 SI U32 ix_and_ptr(T** ptr, const SkJumper_MemoryCtx* ctx, F x, F y) {
211 *ptr = (const T*)ctx->pixels;
212 return trunc_(y)*ctx->stride + trunc_(x);
213 }
214
215 // Now finally, normal Stages!
216
STAGE(seed_shader)217 STAGE(seed_shader) {
218 // It's important for speed to explicitly cast(x) and cast(y),
219 // which has the effect of splatting them to vectors before converting to floats.
220 // On Intel this breaks a data dependency on previous loop iterations' registers.
221 r = cast(x) + 0.5f + unaligned_load<F>(k->iota_F);
222 g = cast(y) + 0.5f;
223 b = 1.0f;
224 a = 0;
225 dr = dg = db = da = 0;
226 }
227
STAGE(dither)228 STAGE(dither) {
229 auto rate = *(const float*)ctx;
230
231 // Get [(x,y), (x+1,y), (x+2,y), ...] loaded up in integer vectors.
232 U32 X = x + unaligned_load<U32>(k->iota_U32),
233 Y = y;
234
235 // We're doing 8x8 ordered dithering, see https://en.wikipedia.org/wiki/Ordered_dithering.
236 // In this case n=8 and we're using the matrix that looks like 1/64 x [ 0 48 12 60 ... ].
237
238 // We only need X and X^Y from here on, so it's easier to just think of that as "Y".
239 Y ^= X;
240
241 // We'll mix the bottom 3 bits of each of X and Y to make 6 bits,
242 // for 2^6 == 64 == 8x8 matrix values. If X=abc and Y=def, we make fcebda.
243 U32 M = (Y & 1) << 5 | (X & 1) << 4
244 | (Y & 2) << 2 | (X & 2) << 1
245 | (Y & 4) >> 1 | (X & 4) >> 2;
246
247 // Scale that dither to [0,1), then (-0.5,+0.5), here using 63/128 = 0.4921875 as 0.5-epsilon.
248 // We want to make sure our dither is less than 0.5 in either direction to keep exact values
249 // like 0 and 1 unchanged after rounding.
250 F dither = cast(M) * (2/128.0f) - (63/128.0f);
251
252 r += rate*dither;
253 g += rate*dither;
254 b += rate*dither;
255
256 r = max(0, min(r, a));
257 g = max(0, min(g, a));
258 b = max(0, min(b, a));
259 }
260
261 // load 4 floats from memory, and splat them into r,g,b,a
STAGE(uniform_color)262 STAGE(uniform_color) {
263 auto rgba = (const float*)ctx;
264 r = rgba[0];
265 g = rgba[1];
266 b = rgba[2];
267 a = rgba[3];
268 }
269
270 // splats opaque-black into r,g,b,a
STAGE(black_color)271 STAGE(black_color) {
272 r = g = b = 0.0f;
273 a = 1.0f;
274 }
275
STAGE(white_color)276 STAGE(white_color) {
277 r = g = b = a = 1.0f;
278 }
279
280 // load registers r,g,b,a from context (mirrors store_rgba)
STAGE(load_rgba)281 STAGE(load_rgba) {
282 auto ptr = (const float*)ctx;
283 r = unaligned_load<F>(ptr + 0*kStride);
284 g = unaligned_load<F>(ptr + 1*kStride);
285 b = unaligned_load<F>(ptr + 2*kStride);
286 a = unaligned_load<F>(ptr + 3*kStride);
287 }
288
289 // store registers r,g,b,a into context (mirrors load_rgba)
STAGE(store_rgba)290 STAGE(store_rgba) {
291 auto ptr = (float*)ctx;
292 unaligned_store(ptr + 0*kStride, r);
293 unaligned_store(ptr + 1*kStride, g);
294 unaligned_store(ptr + 2*kStride, b);
295 unaligned_store(ptr + 3*kStride, a);
296 }
297
298 // Most blend modes apply the same logic to each channel.
299 #define BLEND_MODE(name) \
300 SI F name##_channel(F s, F d, F sa, F da); \
301 STAGE(name) { \
302 r = name##_channel(r,dr,a,da); \
303 g = name##_channel(g,dg,a,da); \
304 b = name##_channel(b,db,a,da); \
305 a = name##_channel(a,da,a,da); \
306 } \
307 SI F name##_channel(F s, F d, F sa, F da)
308
inv(F x)309 SI F inv(F x) { return 1.0f - x; }
two(F x)310 SI F two(F x) { return x + x; }
311
BLEND_MODE(clear)312 BLEND_MODE(clear) { return 0; }
BLEND_MODE(srcatop)313 BLEND_MODE(srcatop) { return s*da + d*inv(sa); }
BLEND_MODE(dstatop)314 BLEND_MODE(dstatop) { return d*sa + s*inv(da); }
BLEND_MODE(srcin)315 BLEND_MODE(srcin) { return s * da; }
BLEND_MODE(dstin)316 BLEND_MODE(dstin) { return d * sa; }
BLEND_MODE(srcout)317 BLEND_MODE(srcout) { return s * inv(da); }
BLEND_MODE(dstout)318 BLEND_MODE(dstout) { return d * inv(sa); }
BLEND_MODE(srcover)319 BLEND_MODE(srcover) { return mad(d, inv(sa), s); }
BLEND_MODE(dstover)320 BLEND_MODE(dstover) { return mad(s, inv(da), d); }
321
BLEND_MODE(modulate)322 BLEND_MODE(modulate) { return s*d; }
BLEND_MODE(multiply)323 BLEND_MODE(multiply) { return s*inv(da) + d*inv(sa) + s*d; }
BLEND_MODE(plus_)324 BLEND_MODE(plus_) { return s + d; }
BLEND_MODE(screen)325 BLEND_MODE(screen) { return s + d - s*d; }
BLEND_MODE(xor_)326 BLEND_MODE(xor_) { return s*inv(da) + d*inv(sa); }
327 #undef BLEND_MODE
328
329 // Most other blend modes apply the same logic to colors, and srcover to alpha.
330 #define BLEND_MODE(name) \
331 SI F name##_channel(F s, F d, F sa, F da); \
332 STAGE(name) { \
333 r = name##_channel(r,dr,a,da); \
334 g = name##_channel(g,dg,a,da); \
335 b = name##_channel(b,db,a,da); \
336 a = mad(da, inv(a), a); \
337 } \
338 SI F name##_channel(F s, F d, F sa, F da)
339
BLEND_MODE(darken)340 BLEND_MODE(darken) { return s + d - max(s*da, d*sa) ; }
BLEND_MODE(lighten)341 BLEND_MODE(lighten) { return s + d - min(s*da, d*sa) ; }
BLEND_MODE(difference)342 BLEND_MODE(difference) { return s + d - two(min(s*da, d*sa)); }
BLEND_MODE(exclusion)343 BLEND_MODE(exclusion) { return s + d - two(s*d); }
344
BLEND_MODE(colorburn)345 BLEND_MODE(colorburn) {
346 return if_then_else(d == da, d + s*inv(da),
347 if_then_else(s == 0, s + d*inv(sa),
348 sa*(da - min(da, (da-d)*sa/s)) + s*inv(da) + d*inv(sa)));
349 }
BLEND_MODE(colordodge)350 BLEND_MODE(colordodge) {
351 return if_then_else(d == 0, d + s*inv(da),
352 if_then_else(s == sa, s + d*inv(sa),
353 sa*min(da, (d*sa)/(sa - s)) + s*inv(da) + d*inv(sa)));
354 }
BLEND_MODE(hardlight)355 BLEND_MODE(hardlight) {
356 return s*inv(da) + d*inv(sa)
357 + if_then_else(two(s) <= sa, two(s*d), sa*da - two((da-d)*(sa-s)));
358 }
BLEND_MODE(overlay)359 BLEND_MODE(overlay) {
360 return s*inv(da) + d*inv(sa)
361 + if_then_else(two(d) <= da, two(s*d), sa*da - two((da-d)*(sa-s)));
362 }
363
BLEND_MODE(softlight)364 BLEND_MODE(softlight) {
365 F m = if_then_else(da > 0, d / da, 0),
366 s2 = two(s),
367 m4 = two(two(m));
368
369 // The logic forks three ways:
370 // 1. dark src?
371 // 2. light src, dark dst?
372 // 3. light src, light dst?
373 F darkSrc = d*(sa + (s2 - sa)*(1.0f - m)), // Used in case 1.
374 darkDst = (m4*m4 + m4)*(m - 1.0f) + 7.0f*m, // Used in case 2.
375 liteDst = rcp(rsqrt(m)) - m, // Used in case 3.
376 liteSrc = d*sa + da*(s2 - sa) * if_then_else(two(two(d)) <= da, darkDst, liteDst); // 2 or 3?
377 return s*inv(da) + d*inv(sa) + if_then_else(s2 <= sa, darkSrc, liteSrc); // 1 or (2 or 3)?
378 }
379 #undef BLEND_MODE
380
381 // We're basing our implemenation of non-separable blend modes on
382 // https://www.w3.org/TR/compositing-1/#blendingnonseparable.
383 // and
384 // https://www.khronos.org/registry/OpenGL/specs/es/3.2/es_spec_3.2.pdf
385 // They're equivalent, but ES' math has been better simplified.
386 //
387 // Anything extra we add beyond that is to make the math work with premul inputs.
388
max(F r,F g,F b)389 SI F max(F r, F g, F b) { return max(r, max(g, b)); }
min(F r,F g,F b)390 SI F min(F r, F g, F b) { return min(r, min(g, b)); }
391
sat(F r,F g,F b)392 SI F sat(F r, F g, F b) { return max(r,g,b) - min(r,g,b); }
lum(F r,F g,F b)393 SI F lum(F r, F g, F b) { return r*0.30f + g*0.59f + b*0.11f; }
394
set_sat(F * r,F * g,F * b,F s)395 SI void set_sat(F* r, F* g, F* b, F s) {
396 F mn = min(*r,*g,*b),
397 mx = max(*r,*g,*b),
398 sat = mx - mn;
399
400 // Map min channel to 0, max channel to s, and scale the middle proportionally.
401 auto scale = [=](F c) {
402 return if_then_else(sat == 0, 0, (c - mn) * s / sat);
403 };
404 *r = scale(*r);
405 *g = scale(*g);
406 *b = scale(*b);
407 }
set_lum(F * r,F * g,F * b,F l)408 SI void set_lum(F* r, F* g, F* b, F l) {
409 F diff = l - lum(*r, *g, *b);
410 *r += diff;
411 *g += diff;
412 *b += diff;
413 }
clip_color(F * r,F * g,F * b,F a)414 SI void clip_color(F* r, F* g, F* b, F a) {
415 F mn = min(*r, *g, *b),
416 mx = max(*r, *g, *b),
417 l = lum(*r, *g, *b);
418
419 auto clip = [=](F c) {
420 c = if_then_else(mn >= 0, c, l + (c - l) * ( l) / (l - mn) );
421 c = if_then_else(mx > a, l + (c - l) * (a - l) / (mx - l), c);
422 c = max(c, 0); // Sometimes without this we may dip just a little negative.
423 return c;
424 };
425 *r = clip(*r);
426 *g = clip(*g);
427 *b = clip(*b);
428 }
429
STAGE(hue)430 STAGE(hue) {
431 F R = r*a,
432 G = g*a,
433 B = b*a;
434
435 set_sat(&R, &G, &B, sat(dr,dg,db)*a);
436 set_lum(&R, &G, &B, lum(dr,dg,db)*a);
437 clip_color(&R,&G,&B, a*da);
438
439 r = r*inv(da) + dr*inv(a) + R;
440 g = g*inv(da) + dg*inv(a) + G;
441 b = b*inv(da) + db*inv(a) + B;
442 a = a + da - a*da;
443 }
STAGE(saturation)444 STAGE(saturation) {
445 F R = dr*a,
446 G = dg*a,
447 B = db*a;
448
449 set_sat(&R, &G, &B, sat( r, g, b)*da);
450 set_lum(&R, &G, &B, lum(dr,dg,db)* a); // (This is not redundant.)
451 clip_color(&R,&G,&B, a*da);
452
453 r = r*inv(da) + dr*inv(a) + R;
454 g = g*inv(da) + dg*inv(a) + G;
455 b = b*inv(da) + db*inv(a) + B;
456 a = a + da - a*da;
457 }
STAGE(color)458 STAGE(color) {
459 F R = r*da,
460 G = g*da,
461 B = b*da;
462
463 set_lum(&R, &G, &B, lum(dr,dg,db)*a);
464 clip_color(&R,&G,&B, a*da);
465
466 r = r*inv(da) + dr*inv(a) + R;
467 g = g*inv(da) + dg*inv(a) + G;
468 b = b*inv(da) + db*inv(a) + B;
469 a = a + da - a*da;
470 }
STAGE(luminosity)471 STAGE(luminosity) {
472 F R = dr*a,
473 G = dg*a,
474 B = db*a;
475
476 set_lum(&R, &G, &B, lum(r,g,b)*da);
477 clip_color(&R,&G,&B, a*da);
478
479 r = r*inv(da) + dr*inv(a) + R;
480 g = g*inv(da) + dg*inv(a) + G;
481 b = b*inv(da) + db*inv(a) + B;
482 a = a + da - a*da;
483 }
484
STAGE(srcover_rgba_8888)485 STAGE(srcover_rgba_8888) {
486 auto ptr = *(uint32_t**)ctx + x;
487
488 U32 dst = load<U32>(ptr, tail);
489 dr = cast((dst ) & 0xff);
490 dg = cast((dst >> 8) & 0xff);
491 db = cast((dst >> 16) & 0xff);
492 da = cast((dst >> 24) );
493 // {dr,dg,db,da} are in [0,255]
494 // { r, g, b, a} are in [0, 1]
495
496 r = mad(dr, inv(a), r*255.0f);
497 g = mad(dg, inv(a), g*255.0f);
498 b = mad(db, inv(a), b*255.0f);
499 a = mad(da, inv(a), a*255.0f);
500 // { r, g, b, a} are now in [0,255]
501
502 dst = round(r, 1.0f)
503 | round(g, 1.0f) << 8
504 | round(b, 1.0f) << 16
505 | round(a, 1.0f) << 24;
506 store(ptr, dst, tail);
507 }
508
STAGE(clamp_0)509 STAGE(clamp_0) {
510 r = max(r, 0);
511 g = max(g, 0);
512 b = max(b, 0);
513 a = max(a, 0);
514 }
515
STAGE(clamp_1)516 STAGE(clamp_1) {
517 r = min(r, 1.0f);
518 g = min(g, 1.0f);
519 b = min(b, 1.0f);
520 a = min(a, 1.0f);
521 }
522
STAGE(clamp_a)523 STAGE(clamp_a) {
524 a = min(a, 1.0f);
525 r = min(r, a);
526 g = min(g, a);
527 b = min(b, a);
528 }
529
STAGE(clamp_a_dst)530 STAGE(clamp_a_dst) {
531 da = min(da, 1.0f);
532 dr = min(dr, da);
533 dg = min(dg, da);
534 db = min(db, da);
535 }
536
STAGE(set_rgb)537 STAGE(set_rgb) {
538 auto rgb = (const float*)ctx;
539 r = rgb[0];
540 g = rgb[1];
541 b = rgb[2];
542 }
STAGE(swap_rb)543 STAGE(swap_rb) {
544 auto tmp = r;
545 r = b;
546 b = tmp;
547 }
548
STAGE(move_src_dst)549 STAGE(move_src_dst) {
550 dr = r;
551 dg = g;
552 db = b;
553 da = a;
554 }
STAGE(move_dst_src)555 STAGE(move_dst_src) {
556 r = dr;
557 g = dg;
558 b = db;
559 a = da;
560 }
561
STAGE(premul)562 STAGE(premul) {
563 r = r * a;
564 g = g * a;
565 b = b * a;
566 }
STAGE(premul_dst)567 STAGE(premul_dst) {
568 dr = dr * da;
569 dg = dg * da;
570 db = db * da;
571 }
STAGE(unpremul)572 STAGE(unpremul) {
573 auto scale = if_then_else(a == 0, 0, 1.0f / a);
574 r *= scale;
575 g *= scale;
576 b *= scale;
577 }
578
from_srgb(F s)579 SI F from_srgb(F s) {
580 auto lo = s * (1/12.92f);
581 auto hi = mad(s*s, mad(s, 0.3000f, 0.6975f), 0.0025f);
582 return if_then_else(s < 0.055f, lo, hi);
583 }
584
STAGE(from_srgb)585 STAGE(from_srgb) {
586 r = from_srgb(r);
587 g = from_srgb(g);
588 b = from_srgb(b);
589 }
STAGE(from_srgb_dst)590 STAGE(from_srgb_dst) {
591 dr = from_srgb(dr);
592 dg = from_srgb(dg);
593 db = from_srgb(db);
594 }
STAGE(to_srgb)595 STAGE(to_srgb) {
596 auto fn = [&](F l) {
597 // We tweak c and d for each instruction set to make sure fn(1) is exactly 1.
598 #if defined(JUMPER) && defined(__SSE2__)
599 const float c = 1.130048394203f,
600 d = 0.141357362270f;
601 #elif defined(JUMPER) && (defined(__aarch64__) || defined(__arm__))
602 const float c = 1.129999995232f,
603 d = 0.141381442547f;
604 #else
605 const float c = 1.129999995232f,
606 d = 0.141377761960f;
607 #endif
608 F t = rsqrt(l);
609 auto lo = l * 12.92f;
610 auto hi = mad(t, mad(t, -0.0024542345f, 0.013832027f), c)
611 * rcp(d + t);
612 return if_then_else(l < 0.00465985f, lo, hi);
613 };
614 r = fn(r);
615 g = fn(g);
616 b = fn(b);
617 }
618
STAGE(rgb_to_hsl)619 STAGE(rgb_to_hsl) {
620 F mx = max(max(r,g), b),
621 mn = min(min(r,g), b),
622 d = mx - mn,
623 d_rcp = 1.0f / d;
624
625 F h = (1/6.0f) *
626 if_then_else(mx == mn, 0,
627 if_then_else(mx == r, (g-b)*d_rcp + if_then_else(g < b, 6.0f, 0),
628 if_then_else(mx == g, (b-r)*d_rcp + 2.0f,
629 (r-g)*d_rcp + 4.0f)));
630
631 F l = (mx + mn) * 0.5f;
632 F s = if_then_else(mx == mn, 0,
633 d / if_then_else(l > 0.5f, 2.0f-mx-mn, mx+mn));
634
635 r = h;
636 g = s;
637 b = l;
638 }
STAGE(hsl_to_rgb)639 STAGE(hsl_to_rgb) {
640 F h = r,
641 s = g,
642 l = b;
643
644 F q = l + if_then_else(l >= 0.5f, s - l*s, l*s),
645 p = 2.0f*l - q;
646
647 auto hue_to_rgb = [&](F t) {
648 t = fract(t);
649
650 F r = p;
651 r = if_then_else(t >= 4/6.0f, r, p + (q-p)*(4.0f - 6.0f*t));
652 r = if_then_else(t >= 3/6.0f, r, q);
653 r = if_then_else(t >= 1/6.0f, r, p + (q-p)*( 6.0f*t));
654 return r;
655 };
656
657 r = if_then_else(s == 0, l, hue_to_rgb(h + (1/3.0f)));
658 g = if_then_else(s == 0, l, hue_to_rgb(h ));
659 b = if_then_else(s == 0, l, hue_to_rgb(h - (1/3.0f)));
660 }
661
STAGE(scale_1_float)662 STAGE(scale_1_float) {
663 auto c = *(const float*)ctx;
664
665 r = r * c;
666 g = g * c;
667 b = b * c;
668 a = a * c;
669 }
STAGE(scale_u8)670 STAGE(scale_u8) {
671 auto ptr = *(const uint8_t**)ctx + x;
672
673 auto scales = load<U8>(ptr, tail);
674 auto c = from_byte(scales);
675
676 r = r * c;
677 g = g * c;
678 b = b * c;
679 a = a * c;
680 }
681
lerp(F from,F to,F t)682 SI F lerp(F from, F to, F t) {
683 return mad(to-from, t, from);
684 }
685
STAGE(lerp_1_float)686 STAGE(lerp_1_float) {
687 auto c = *(const float*)ctx;
688
689 r = lerp(dr, r, c);
690 g = lerp(dg, g, c);
691 b = lerp(db, b, c);
692 a = lerp(da, a, c);
693 }
STAGE(lerp_u8)694 STAGE(lerp_u8) {
695 auto ptr = *(const uint8_t**)ctx + x;
696
697 auto scales = load<U8>(ptr, tail);
698 auto c = from_byte(scales);
699
700 r = lerp(dr, r, c);
701 g = lerp(dg, g, c);
702 b = lerp(db, b, c);
703 a = lerp(da, a, c);
704 }
STAGE(lerp_565)705 STAGE(lerp_565) {
706 auto ptr = *(const uint16_t**)ctx + x;
707
708 F cr,cg,cb;
709 from_565(load<U16>(ptr, tail), &cr, &cg, &cb);
710
711 r = lerp(dr, r, cr);
712 g = lerp(dg, g, cg);
713 b = lerp(db, b, cb);
714 a = max(lerp(da, a, cr), lerp(da, a, cg), lerp(da, a, cb));
715 }
716
STAGE(load_tables)717 STAGE(load_tables) {
718 auto c = (const SkJumper_LoadTablesCtx*)ctx;
719
720 auto px = load<U32>((const uint32_t*)c->src + x, tail);
721 r = gather(c->r, (px ) & 0xff);
722 g = gather(c->g, (px >> 8) & 0xff);
723 b = gather(c->b, (px >> 16) & 0xff);
724 a = cast( (px >> 24)) * (1/255.0f);
725 }
STAGE(load_tables_u16_be)726 STAGE(load_tables_u16_be) {
727 auto c = (const SkJumper_LoadTablesCtx*)ctx;
728 auto ptr = (const uint16_t*)c->src + 4*x;
729
730 U16 R,G,B,A;
731 load4(ptr, tail, &R,&G,&B,&A);
732
733 // c->src is big-endian, so & 0xff grabs the 8 most signficant bits.
734 r = gather(c->r, expand(R) & 0xff);
735 g = gather(c->g, expand(G) & 0xff);
736 b = gather(c->b, expand(B) & 0xff);
737 a = (1/65535.0f) * cast(expand(bswap(A)));
738 }
STAGE(load_tables_rgb_u16_be)739 STAGE(load_tables_rgb_u16_be) {
740 auto c = (const SkJumper_LoadTablesCtx*)ctx;
741 auto ptr = (const uint16_t*)c->src + 3*x;
742
743 U16 R,G,B;
744 load3(ptr, tail, &R,&G,&B);
745
746 // c->src is big-endian, so & 0xff grabs the 8 most signficant bits.
747 r = gather(c->r, expand(R) & 0xff);
748 g = gather(c->g, expand(G) & 0xff);
749 b = gather(c->b, expand(B) & 0xff);
750 a = 1.0f;
751 }
752
STAGE(byte_tables)753 STAGE(byte_tables) {
754 struct Tables { const uint8_t *r, *g, *b, *a; };
755 auto tables = (const Tables*)ctx;
756
757 r = from_byte(gather(tables->r, round(r, 255.0f)));
758 g = from_byte(gather(tables->g, round(g, 255.0f)));
759 b = from_byte(gather(tables->b, round(b, 255.0f)));
760 a = from_byte(gather(tables->a, round(a, 255.0f)));
761 }
762
STAGE(byte_tables_rgb)763 STAGE(byte_tables_rgb) {
764 struct Tables { const uint8_t *r, *g, *b; int n; };
765 auto tables = (const Tables*)ctx;
766
767 F scale = tables->n - 1;
768 r = from_byte(gather(tables->r, round(r, scale)));
769 g = from_byte(gather(tables->g, round(g, scale)));
770 b = from_byte(gather(tables->b, round(b, scale)));
771 }
772
table(F v,const SkJumper_TableCtx * ctx)773 SI F table(F v, const SkJumper_TableCtx* ctx) {
774 return gather(ctx->table, round(v, ctx->size - 1));
775 }
STAGE(table_r)776 STAGE(table_r) { r = table(r, ctx); }
STAGE(table_g)777 STAGE(table_g) { g = table(g, ctx); }
STAGE(table_b)778 STAGE(table_b) { b = table(b, ctx); }
STAGE(table_a)779 STAGE(table_a) { a = table(a, ctx); }
780
parametric(F v,const SkJumper_ParametricTransferFunction * ctx)781 SI F parametric(F v, const SkJumper_ParametricTransferFunction* ctx) {
782 F r = if_then_else(v <= ctx->D, mad(ctx->C, v, ctx->F)
783 , approx_powf(mad(ctx->A, v, ctx->B), ctx->G) + ctx->E);
784 return min(max(r, 0), 1.0f); // Clamp to [0,1], with argument order mattering to handle NaN.
785 }
STAGE(parametric_r)786 STAGE(parametric_r) { r = parametric(r, ctx); }
STAGE(parametric_g)787 STAGE(parametric_g) { g = parametric(g, ctx); }
STAGE(parametric_b)788 STAGE(parametric_b) { b = parametric(b, ctx); }
STAGE(parametric_a)789 STAGE(parametric_a) { a = parametric(a, ctx); }
790
STAGE(lab_to_xyz)791 STAGE(lab_to_xyz) {
792 F L = r * 100.0f,
793 A = g * 255.0f - 128.0f,
794 B = b * 255.0f - 128.0f;
795
796 F Y = (L + 16.0f) * (1/116.0f),
797 X = Y + A*(1/500.0f),
798 Z = Y - B*(1/200.0f);
799
800 X = if_then_else(X*X*X > 0.008856f, X*X*X, (X - (16/116.0f)) * (1/7.787f));
801 Y = if_then_else(Y*Y*Y > 0.008856f, Y*Y*Y, (Y - (16/116.0f)) * (1/7.787f));
802 Z = if_then_else(Z*Z*Z > 0.008856f, Z*Z*Z, (Z - (16/116.0f)) * (1/7.787f));
803
804 // Adjust to D50 illuminant.
805 r = X * 0.96422f;
806 g = Y ;
807 b = Z * 0.82521f;
808 }
809
STAGE(load_a8)810 STAGE(load_a8) {
811 auto ptr = *(const uint8_t**)ctx + x;
812
813 r = g = b = 0.0f;
814 a = from_byte(load<U8>(ptr, tail));
815 }
STAGE(load_a8_dst)816 STAGE(load_a8_dst) {
817 auto ptr = *(const uint8_t**)ctx + x;
818
819 dr = dg = db = 0.0f;
820 da = from_byte(load<U8>(ptr, tail));
821 }
STAGE(gather_a8)822 STAGE(gather_a8) {
823 const uint8_t* ptr;
824 U32 ix = ix_and_ptr(&ptr, ctx, r,g);
825 r = g = b = 0.0f;
826 a = from_byte(gather(ptr, ix));
827 }
STAGE(store_a8)828 STAGE(store_a8) {
829 auto ptr = *(uint8_t**)ctx + x;
830
831 U8 packed = pack(pack(round(a, 255.0f)));
832 store(ptr, packed, tail);
833 }
834
STAGE(load_g8)835 STAGE(load_g8) {
836 auto ptr = *(const uint8_t**)ctx + x;
837
838 r = g = b = from_byte(load<U8>(ptr, tail));
839 a = 1.0f;
840 }
STAGE(load_g8_dst)841 STAGE(load_g8_dst) {
842 auto ptr = *(const uint8_t**)ctx + x;
843
844 dr = dg = db = from_byte(load<U8>(ptr, tail));
845 da = 1.0f;
846 }
STAGE(gather_g8)847 STAGE(gather_g8) {
848 const uint8_t* ptr;
849 U32 ix = ix_and_ptr(&ptr, ctx, r,g);
850 r = g = b = from_byte(gather(ptr, ix));
851 a = 1.0f;
852 }
853
STAGE(load_565)854 STAGE(load_565) {
855 auto ptr = *(const uint16_t**)ctx + x;
856
857 from_565(load<U16>(ptr, tail), &r,&g,&b);
858 a = 1.0f;
859 }
STAGE(load_565_dst)860 STAGE(load_565_dst) {
861 auto ptr = *(const uint16_t**)ctx + x;
862
863 from_565(load<U16>(ptr, tail), &dr,&dg,&db);
864 da = 1.0f;
865 }
STAGE(gather_565)866 STAGE(gather_565) {
867 const uint16_t* ptr;
868 U32 ix = ix_and_ptr(&ptr, ctx, r,g);
869 from_565(gather(ptr, ix), &r,&g,&b);
870 a = 1.0f;
871 }
STAGE(store_565)872 STAGE(store_565) {
873 auto ptr = *(uint16_t**)ctx + x;
874
875 U16 px = pack( round(r, 31.0f) << 11
876 | round(g, 63.0f) << 5
877 | round(b, 31.0f) );
878 store(ptr, px, tail);
879 }
880
STAGE(load_4444)881 STAGE(load_4444) {
882 auto ptr = *(const uint16_t**)ctx + x;
883 from_4444(load<U16>(ptr, tail), &r,&g,&b,&a);
884 }
STAGE(load_4444_dst)885 STAGE(load_4444_dst) {
886 auto ptr = *(const uint16_t**)ctx + x;
887 from_4444(load<U16>(ptr, tail), &dr,&dg,&db,&da);
888 }
STAGE(gather_4444)889 STAGE(gather_4444) {
890 const uint16_t* ptr;
891 U32 ix = ix_and_ptr(&ptr, ctx, r,g);
892 from_4444(gather(ptr, ix), &r,&g,&b,&a);
893 }
STAGE(store_4444)894 STAGE(store_4444) {
895 auto ptr = *(uint16_t**)ctx + x;
896 U16 px = pack( round(r, 15.0f) << 12
897 | round(g, 15.0f) << 8
898 | round(b, 15.0f) << 4
899 | round(a, 15.0f) );
900 store(ptr, px, tail);
901 }
902
STAGE(load_8888)903 STAGE(load_8888) {
904 auto ptr = *(const uint32_t**)ctx + x;
905 from_8888(load<U32>(ptr, tail), &r,&g,&b,&a);
906 }
STAGE(load_8888_dst)907 STAGE(load_8888_dst) {
908 auto ptr = *(const uint32_t**)ctx + x;
909 from_8888(load<U32>(ptr, tail), &dr,&dg,&db,&da);
910 }
STAGE(gather_8888)911 STAGE(gather_8888) {
912 const uint32_t* ptr;
913 U32 ix = ix_and_ptr(&ptr, ctx, r,g);
914 from_8888(gather(ptr, ix), &r,&g,&b,&a);
915 }
STAGE(store_8888)916 STAGE(store_8888) {
917 auto ptr = *(uint32_t**)ctx + x;
918
919 U32 px = round(r, 255.0f)
920 | round(g, 255.0f) << 8
921 | round(b, 255.0f) << 16
922 | round(a, 255.0f) << 24;
923 store(ptr, px, tail);
924 }
925
STAGE(store_8888_2d)926 STAGE(store_8888_2d) {
927 auto c = (const SkJumper_MemoryCtx*)ctx;
928 auto ptr = (uint32_t*)c->pixels + y*c->stride + x;
929
930 U32 px = round(r, 255.0f)
931 | round(g, 255.0f) << 8
932 | round(b, 255.0f) << 16
933 | round(a, 255.0f) << 24;
934 store(ptr, px, tail);
935 }
936
STAGE(load_bgra)937 STAGE(load_bgra) {
938 auto ptr = *(const uint32_t**)ctx + x;
939 from_8888(load<U32>(ptr, tail), &b,&g,&r,&a);
940 }
STAGE(load_bgra_dst)941 STAGE(load_bgra_dst) {
942 auto ptr = *(const uint32_t**)ctx + x;
943 from_8888(load<U32>(ptr, tail), &db,&dg,&dr,&da);
944 }
STAGE(gather_bgra)945 STAGE(gather_bgra) {
946 const uint32_t* ptr;
947 U32 ix = ix_and_ptr(&ptr, ctx, r,g);
948 from_8888(gather(ptr, ix), &b,&g,&r,&a);
949 }
STAGE(store_bgra)950 STAGE(store_bgra) {
951 auto ptr = *(uint32_t**)ctx + x;
952
953 U32 px = round(b, 255.0f)
954 | round(g, 255.0f) << 8
955 | round(r, 255.0f) << 16
956 | round(a, 255.0f) << 24;
957 store(ptr, px, tail);
958 }
959
STAGE(load_f16)960 STAGE(load_f16) {
961 auto ptr = *(const uint64_t**)ctx + x;
962
963 U16 R,G,B,A;
964 load4((const uint16_t*)ptr,tail, &R,&G,&B,&A);
965 r = from_half(R);
966 g = from_half(G);
967 b = from_half(B);
968 a = from_half(A);
969 }
STAGE(load_f16_dst)970 STAGE(load_f16_dst) {
971 auto ptr = *(const uint64_t**)ctx + x;
972
973 U16 R,G,B,A;
974 load4((const uint16_t*)ptr,tail, &R,&G,&B,&A);
975 dr = from_half(R);
976 dg = from_half(G);
977 db = from_half(B);
978 da = from_half(A);
979 }
STAGE(gather_f16)980 STAGE(gather_f16) {
981 const uint64_t* ptr;
982 U32 ix = ix_and_ptr(&ptr, ctx, r,g);
983 auto px = gather(ptr, ix);
984
985 U16 R,G,B,A;
986 load4((const uint16_t*)&px,0, &R,&G,&B,&A);
987 r = from_half(R);
988 g = from_half(G);
989 b = from_half(B);
990 a = from_half(A);
991 }
STAGE(store_f16)992 STAGE(store_f16) {
993 auto ptr = *(uint64_t**)ctx + x;
994 store4((uint16_t*)ptr,tail, to_half(r)
995 , to_half(g)
996 , to_half(b)
997 , to_half(a));
998 }
999
STAGE(load_u16_be)1000 STAGE(load_u16_be) {
1001 auto ptr = *(const uint16_t**)ctx + 4*x;
1002
1003 U16 R,G,B,A;
1004 load4(ptr,tail, &R,&G,&B,&A);
1005
1006 r = (1/65535.0f) * cast(expand(bswap(R)));
1007 g = (1/65535.0f) * cast(expand(bswap(G)));
1008 b = (1/65535.0f) * cast(expand(bswap(B)));
1009 a = (1/65535.0f) * cast(expand(bswap(A)));
1010 }
STAGE(load_rgb_u16_be)1011 STAGE(load_rgb_u16_be) {
1012 auto ptr = *(const uint16_t**)ctx + 3*x;
1013
1014 U16 R,G,B;
1015 load3(ptr,tail, &R,&G,&B);
1016
1017 r = (1/65535.0f) * cast(expand(bswap(R)));
1018 g = (1/65535.0f) * cast(expand(bswap(G)));
1019 b = (1/65535.0f) * cast(expand(bswap(B)));
1020 a = 1.0f;
1021 }
STAGE(store_u16_be)1022 STAGE(store_u16_be) {
1023 auto ptr = *(uint16_t**)ctx + 4*x;
1024
1025 U16 R = bswap(pack(round(r, 65535.0f))),
1026 G = bswap(pack(round(g, 65535.0f))),
1027 B = bswap(pack(round(b, 65535.0f))),
1028 A = bswap(pack(round(a, 65535.0f)));
1029
1030 store4(ptr,tail, R,G,B,A);
1031 }
1032
STAGE(load_f32)1033 STAGE(load_f32) {
1034 auto ptr = *(const float**)ctx + 4*x;
1035 load4(ptr,tail, &r,&g,&b,&a);
1036 }
STAGE(load_f32_dst)1037 STAGE(load_f32_dst) {
1038 auto ptr = *(const float**)ctx + 4*x;
1039 load4(ptr,tail, &dr,&dg,&db,&da);
1040 }
STAGE(store_f32)1041 STAGE(store_f32) {
1042 auto ptr = *(float**)ctx + 4*x;
1043 store4(ptr,tail, r,g,b,a);
1044 }
1045
ulp_before(F f)1046 SI F ulp_before(F f) {
1047 U32 bits = -1 + unaligned_load<U32>(&f);
1048 return unaligned_load<F>(&bits);
1049 }
1050
exclusive_clamp(F v,const SkJumper_TileCtx * ctx)1051 SI F exclusive_clamp(F v, const SkJumper_TileCtx* ctx) {
1052 v = max(0,v);
1053 return min(v, ulp_before(ctx->scale));
1054 }
exclusive_repeat(F v,const SkJumper_TileCtx * ctx)1055 SI F exclusive_repeat(F v, const SkJumper_TileCtx* ctx) {
1056 v = v - floor_(v*ctx->invScale)*ctx->scale;
1057 return min(v, ulp_before(ctx->scale));
1058 }
exclusive_mirror(F v,const SkJumper_TileCtx * ctx)1059 SI F exclusive_mirror(F v, const SkJumper_TileCtx* ctx) {
1060 auto limit = ctx->scale;
1061 auto invLimit = ctx->invScale;
1062 v = abs_( (v-limit) - (limit+limit)*floor_((v-limit)*(invLimit*0.5f)) - limit );
1063 return min(v, ulp_before(limit));
1064 }
1065 // Clamp x or y to [0,limit) == [0,limit - 1 ulp] (think, sampling from images).
STAGE(clamp_x)1066 STAGE(clamp_x) { r = exclusive_clamp (r, (const SkJumper_TileCtx*)ctx); }
STAGE(clamp_y)1067 STAGE(clamp_y) { g = exclusive_clamp (g, (const SkJumper_TileCtx*)ctx); }
STAGE(repeat_x)1068 STAGE(repeat_x) { r = exclusive_repeat(r, (const SkJumper_TileCtx*)ctx); }
STAGE(repeat_y)1069 STAGE(repeat_y) { g = exclusive_repeat(g, (const SkJumper_TileCtx*)ctx); }
STAGE(mirror_x)1070 STAGE(mirror_x) { r = exclusive_mirror(r, (const SkJumper_TileCtx*)ctx); }
STAGE(mirror_y)1071 STAGE(mirror_y) { g = exclusive_mirror(g, (const SkJumper_TileCtx*)ctx); }
1072
1073 // Clamp x to [0,1], both sides exclusive (think, gradients).
STAGE(clamp_x_1)1074 STAGE( clamp_x_1) { r = min(max(0, r), 1.0f); }
STAGE(repeat_x_1)1075 STAGE(repeat_x_1) { r = r - floor_(r); }
STAGE(mirror_x_1)1076 STAGE(mirror_x_1) { r = abs_( (r-1.0f) - two(floor_((r-1.0f)*0.5f)) - 1.0f ); }
1077
STAGE(luminance_to_alpha)1078 STAGE(luminance_to_alpha) {
1079 a = r*0.2126f + g*0.7152f + b*0.0722f;
1080 r = g = b = 0;
1081 }
1082
STAGE(matrix_translate)1083 STAGE(matrix_translate) {
1084 auto m = (const float*)ctx;
1085
1086 r += m[0];
1087 g += m[1];
1088 }
STAGE(matrix_scale_translate)1089 STAGE(matrix_scale_translate) {
1090 auto m = (const float*)ctx;
1091
1092 r = mad(r,m[2], m[0]);
1093 g = mad(g,m[3], m[1]);
1094 }
STAGE(matrix_2x3)1095 STAGE(matrix_2x3) {
1096 auto m = (const float*)ctx;
1097
1098 auto R = mad(r,m[0], mad(g,m[2], m[4])),
1099 G = mad(r,m[1], mad(g,m[3], m[5]));
1100 r = R;
1101 g = G;
1102 }
STAGE(matrix_3x4)1103 STAGE(matrix_3x4) {
1104 auto m = (const float*)ctx;
1105
1106 auto R = mad(r,m[0], mad(g,m[3], mad(b,m[6], m[ 9]))),
1107 G = mad(r,m[1], mad(g,m[4], mad(b,m[7], m[10]))),
1108 B = mad(r,m[2], mad(g,m[5], mad(b,m[8], m[11])));
1109 r = R;
1110 g = G;
1111 b = B;
1112 }
STAGE(matrix_4x5)1113 STAGE(matrix_4x5) {
1114 auto m = (const float*)ctx;
1115
1116 auto R = mad(r,m[0], mad(g,m[4], mad(b,m[ 8], mad(a,m[12], m[16])))),
1117 G = mad(r,m[1], mad(g,m[5], mad(b,m[ 9], mad(a,m[13], m[17])))),
1118 B = mad(r,m[2], mad(g,m[6], mad(b,m[10], mad(a,m[14], m[18])))),
1119 A = mad(r,m[3], mad(g,m[7], mad(b,m[11], mad(a,m[15], m[19]))));
1120 r = R;
1121 g = G;
1122 b = B;
1123 a = A;
1124 }
STAGE(matrix_4x3)1125 STAGE(matrix_4x3) {
1126 auto m = (const float*)ctx;
1127 auto X = r,
1128 Y = g;
1129
1130 r = mad(X, m[0], mad(Y, m[4], m[ 8]));
1131 g = mad(X, m[1], mad(Y, m[5], m[ 9]));
1132 b = mad(X, m[2], mad(Y, m[6], m[10]));
1133 a = mad(X, m[3], mad(Y, m[7], m[11]));
1134 }
STAGE(matrix_perspective)1135 STAGE(matrix_perspective) {
1136 // N.B. Unlike the other matrix_ stages, this matrix is row-major.
1137 auto m = (const float*)ctx;
1138
1139 auto R = mad(r,m[0], mad(g,m[1], m[2])),
1140 G = mad(r,m[3], mad(g,m[4], m[5])),
1141 Z = mad(r,m[6], mad(g,m[7], m[8]));
1142 r = R * rcp(Z);
1143 g = G * rcp(Z);
1144 }
1145
gradient_lookup(const SkJumper_GradientCtx * c,U32 idx,F t,F * r,F * g,F * b,F * a)1146 SI void gradient_lookup(const SkJumper_GradientCtx* c, U32 idx, F t,
1147 F* r, F* g, F* b, F* a) {
1148 F fr, br, fg, bg, fb, bb, fa, ba;
1149 #if defined(JUMPER) && defined(__AVX2__)
1150 if (c->stopCount <=8) {
1151 fr = _mm256_permutevar8x32_ps(_mm256_loadu_ps(c->fs[0]), idx);
1152 br = _mm256_permutevar8x32_ps(_mm256_loadu_ps(c->bs[0]), idx);
1153 fg = _mm256_permutevar8x32_ps(_mm256_loadu_ps(c->fs[1]), idx);
1154 bg = _mm256_permutevar8x32_ps(_mm256_loadu_ps(c->bs[1]), idx);
1155 fb = _mm256_permutevar8x32_ps(_mm256_loadu_ps(c->fs[2]), idx);
1156 bb = _mm256_permutevar8x32_ps(_mm256_loadu_ps(c->bs[2]), idx);
1157 fa = _mm256_permutevar8x32_ps(_mm256_loadu_ps(c->fs[3]), idx);
1158 ba = _mm256_permutevar8x32_ps(_mm256_loadu_ps(c->bs[3]), idx);
1159 } else
1160 #endif
1161 {
1162 fr = gather(c->fs[0], idx);
1163 br = gather(c->bs[0], idx);
1164 fg = gather(c->fs[1], idx);
1165 bg = gather(c->bs[1], idx);
1166 fb = gather(c->fs[2], idx);
1167 bb = gather(c->bs[2], idx);
1168 fa = gather(c->fs[3], idx);
1169 ba = gather(c->bs[3], idx);
1170 }
1171
1172 *r = mad(t, fr, br);
1173 *g = mad(t, fg, bg);
1174 *b = mad(t, fb, bb);
1175 *a = mad(t, fa, ba);
1176 }
1177
STAGE(evenly_spaced_gradient)1178 STAGE(evenly_spaced_gradient) {
1179 auto c = (const SkJumper_GradientCtx*)ctx;
1180 auto t = r;
1181 auto idx = trunc_(t * (c->stopCount-1));
1182 gradient_lookup(c, idx, t, &r, &g, &b, &a);
1183 }
1184
STAGE(gauss_a_to_rgba)1185 STAGE(gauss_a_to_rgba) {
1186 // x = 1 - x;
1187 // exp(-x * x * 4) - 0.018f;
1188 // ... now approximate with quartic
1189 //
1190 const float c4 = -2.26661229133605957031f;
1191 const float c3 = 2.89795351028442382812f;
1192 const float c2 = 0.21345567703247070312f;
1193 const float c1 = 0.15489584207534790039f;
1194 const float c0 = 0.00030726194381713867f;
1195 a = mad(a, mad(a, mad(a, mad(a, c4, c3), c2), c1), c0);
1196 r = a;
1197 g = a;
1198 b = a;
1199 }
1200
STAGE(gradient)1201 STAGE(gradient) {
1202 auto c = (const SkJumper_GradientCtx*)ctx;
1203 auto t = r;
1204 U32 idx = 0;
1205
1206 // N.B. The loop starts at 1 because idx 0 is the color to use before the first stop.
1207 for (size_t i = 1; i < c->stopCount; i++) {
1208 idx += if_then_else(t >= c->ts[i], U32(1), U32(0));
1209 }
1210
1211 gradient_lookup(c, idx, t, &r, &g, &b, &a);
1212 }
1213
STAGE(evenly_spaced_2_stop_gradient)1214 STAGE(evenly_spaced_2_stop_gradient) {
1215 struct Ctx { float f[4], b[4]; };
1216 auto c = (const Ctx*)ctx;
1217
1218 auto t = r;
1219 r = mad(t, c->f[0], c->b[0]);
1220 g = mad(t, c->f[1], c->b[1]);
1221 b = mad(t, c->f[2], c->b[2]);
1222 a = mad(t, c->f[3], c->b[3]);
1223 }
1224
STAGE(xy_to_unit_angle)1225 STAGE(xy_to_unit_angle) {
1226 F X = r,
1227 Y = g;
1228 F xabs = abs_(X),
1229 yabs = abs_(Y);
1230
1231 F slope = min(xabs, yabs)/max(xabs, yabs);
1232 F s = slope * slope;
1233
1234 // Use a 7th degree polynomial to approximate atan.
1235 // This was generated using sollya.gforge.inria.fr.
1236 // A float optimized polynomial was generated using the following command.
1237 // P1 = fpminimax((1/(2*Pi))*atan(x),[|1,3,5,7|],[|24...|],[2^(-40),1],relative);
1238 F phi = slope
1239 * (0.15912117063999176025390625f + s
1240 * (-5.185396969318389892578125e-2f + s
1241 * (2.476101927459239959716796875e-2f + s
1242 * (-7.0547382347285747528076171875e-3f))));
1243
1244 phi = if_then_else(xabs < yabs, 1.0f/4.0f - phi, phi);
1245 phi = if_then_else(X < 0.0f , 1.0f/2.0f - phi, phi);
1246 phi = if_then_else(Y < 0.0f , 1.0f - phi , phi);
1247 phi = if_then_else(phi != phi , 0 , phi); // Check for NaN.
1248 r = phi;
1249 }
1250
STAGE(xy_to_radius)1251 STAGE(xy_to_radius) {
1252 F X2 = r * r,
1253 Y2 = g * g;
1254 r = sqrt_(X2 + Y2);
1255 }
1256
solve_2pt_conical_quadratic(const SkJumper_2PtConicalCtx * c,F x,F y,F (* select)(F,F))1257 SI F solve_2pt_conical_quadratic(const SkJumper_2PtConicalCtx* c, F x, F y, F (*select)(F, F)) {
1258 // At this point, (x, y) is mapped into a synthetic gradient space with
1259 // the start circle centerd on (0, 0), and the end circle centered on (1, 0)
1260 // (see the stage setup).
1261 //
1262 // We're searching along X-axis for x', such that
1263 //
1264 // 1) r(x') is a linear interpolation between r0 and r1
1265 // 2) (x, y) is on the circle C(x', 0, r(x'))
1266 //
1267 // Solving this system boils down to a quadratic equation with coefficients
1268 //
1269 // a = 1 - (r1 - r0)^2 <- constant, precomputed in ctx->fCoeffA)
1270 //
1271 // b = -2 * (x + (r1 - r0) * r0)
1272 //
1273 // c = x^2 + y^2 - r0^2
1274 //
1275 // Since the start/end circle centers are the extremes of the [0, 1] interval
1276 // on the X axis, the solution (x') is exactly the t we are looking for.
1277
1278 const F coeffA = c->fCoeffA,
1279 coeffB = -2 * (x + c->fDR*c->fR0),
1280 coeffC = x*x + y*y - c->fR0*c->fR0;
1281
1282 const F disc = mad(coeffB, coeffB, -4 * coeffA * coeffC);
1283 const F sqrt_disc = sqrt_(disc);
1284
1285 const F invCoeffA = c->fInvCoeffA;
1286 return select((-coeffB + sqrt_disc) * (invCoeffA * 0.5f),
1287 (-coeffB - sqrt_disc) * (invCoeffA * 0.5f));
1288 }
1289
STAGE(xy_to_2pt_conical_quadratic_max)1290 STAGE(xy_to_2pt_conical_quadratic_max) {
1291 r = solve_2pt_conical_quadratic(ctx, r, g, max);
1292 }
1293
STAGE(xy_to_2pt_conical_quadratic_min)1294 STAGE(xy_to_2pt_conical_quadratic_min) {
1295 r = solve_2pt_conical_quadratic(ctx, r, g, min);
1296 }
1297
STAGE(xy_to_2pt_conical_linear)1298 STAGE(xy_to_2pt_conical_linear) {
1299 auto* c = (const SkJumper_2PtConicalCtx*)ctx;
1300
1301 const F coeffB = -2 * (r + c->fDR*c->fR0),
1302 coeffC = r*r + g*g - c->fR0*c->fR0;
1303
1304 r = -coeffC / coeffB;
1305 }
1306
STAGE(mask_2pt_conical_degenerates)1307 STAGE(mask_2pt_conical_degenerates) {
1308 auto* c = (SkJumper_2PtConicalCtx*)ctx;
1309
1310 // Compute and save a mask for degenerate values.
1311 U32 mask = 0xffffffff;
1312
1313 // TODO: mtklein kindly volunteered to revisit this at some point.
1314 #if defined(JUMPER)
1315 // Vector comparisons set all bits, so we can use something like this.
1316 mask = mask & (mad(r, c->fDR, c->fR0) >= 0); // R(t) >= 0
1317 mask = mask & (r == r); // t != NaN
1318 #else
1319 // The portable version is more involved, 'cause we only get one bit back.
1320 mask = mask & if_then_else(mad(r, c->fDR, c->fR0) >= 0, U32(0xffffffff), U32(0)); // R(t) >= 0
1321 mask = mask & if_then_else(r == r, U32(0xffffffff), U32(0)); // t != NaN
1322 #endif
1323
1324 unaligned_store(&c->fMask, mask);
1325 }
1326
STAGE(apply_vector_mask)1327 STAGE(apply_vector_mask) {
1328 const U32 mask = unaligned_load<U32>((const uint32_t*)ctx);
1329 r = bit_cast<F>(bit_cast<U32>(r) & mask);
1330 g = bit_cast<F>(bit_cast<U32>(g) & mask);
1331 b = bit_cast<F>(bit_cast<U32>(b) & mask);
1332 a = bit_cast<F>(bit_cast<U32>(a) & mask);
1333 }
1334
STAGE(save_xy)1335 STAGE(save_xy) {
1336 auto c = (SkJumper_SamplerCtx*)ctx;
1337
1338 // Whether bilinear or bicubic, all sample points are at the same fractional offset (fx,fy).
1339 // They're either the 4 corners of a logical 1x1 pixel or the 16 corners of a 3x3 grid
1340 // surrounding (x,y) at (0.5,0.5) off-center.
1341 F fx = fract(r + 0.5f),
1342 fy = fract(g + 0.5f);
1343
1344 // Samplers will need to load x and fx, or y and fy.
1345 unaligned_store(c->x, r);
1346 unaligned_store(c->y, g);
1347 unaligned_store(c->fx, fx);
1348 unaligned_store(c->fy, fy);
1349 }
1350
STAGE(accumulate)1351 STAGE(accumulate) {
1352 auto c = (const SkJumper_SamplerCtx*)ctx;
1353
1354 // Bilinear and bicubic filters are both separable, so we produce independent contributions
1355 // from x and y, multiplying them together here to get each pixel's total scale factor.
1356 auto scale = unaligned_load<F>(c->scalex)
1357 * unaligned_load<F>(c->scaley);
1358 dr = mad(scale, r, dr);
1359 dg = mad(scale, g, dg);
1360 db = mad(scale, b, db);
1361 da = mad(scale, a, da);
1362 }
1363
1364 // In bilinear interpolation, the 4 pixels at +/- 0.5 offsets from the sample pixel center
1365 // are combined in direct proportion to their area overlapping that logical query pixel.
1366 // At positive offsets, the x-axis contribution to that rectangle is fx, or (1-fx) at negative x.
1367 // The y-axis is symmetric.
1368
1369 template <int kScale>
bilinear_x(SkJumper_SamplerCtx * ctx,F * x)1370 SI void bilinear_x(SkJumper_SamplerCtx* ctx, F* x) {
1371 *x = unaligned_load<F>(ctx->x) + (kScale * 0.5f);
1372 F fx = unaligned_load<F>(ctx->fx);
1373
1374 F scalex;
1375 if (kScale == -1) { scalex = 1.0f - fx; }
1376 if (kScale == +1) { scalex = fx; }
1377 unaligned_store(ctx->scalex, scalex);
1378 }
1379 template <int kScale>
bilinear_y(SkJumper_SamplerCtx * ctx,F * y)1380 SI void bilinear_y(SkJumper_SamplerCtx* ctx, F* y) {
1381 *y = unaligned_load<F>(ctx->y) + (kScale * 0.5f);
1382 F fy = unaligned_load<F>(ctx->fy);
1383
1384 F scaley;
1385 if (kScale == -1) { scaley = 1.0f - fy; }
1386 if (kScale == +1) { scaley = fy; }
1387 unaligned_store(ctx->scaley, scaley);
1388 }
1389
STAGE(bilinear_nx)1390 STAGE(bilinear_nx) { bilinear_x<-1>(ctx, &r); }
STAGE(bilinear_px)1391 STAGE(bilinear_px) { bilinear_x<+1>(ctx, &r); }
STAGE(bilinear_ny)1392 STAGE(bilinear_ny) { bilinear_y<-1>(ctx, &g); }
STAGE(bilinear_py)1393 STAGE(bilinear_py) { bilinear_y<+1>(ctx, &g); }
1394
1395
1396 // In bicubic interpolation, the 16 pixels and +/- 0.5 and +/- 1.5 offsets from the sample
1397 // pixel center are combined with a non-uniform cubic filter, with higher values near the center.
1398 //
1399 // We break this function into two parts, one for near 0.5 offsets and one for far 1.5 offsets.
1400 // See GrCubicEffect for details of this particular filter.
1401
bicubic_near(F t)1402 SI F bicubic_near(F t) {
1403 // 1/18 + 9/18t + 27/18t^2 - 21/18t^3 == t ( t ( -21/18t + 27/18) + 9/18) + 1/18
1404 return mad(t, mad(t, mad((-21/18.0f), t, (27/18.0f)), (9/18.0f)), (1/18.0f));
1405 }
bicubic_far(F t)1406 SI F bicubic_far(F t) {
1407 // 0/18 + 0/18*t - 6/18t^2 + 7/18t^3 == t^2 (7/18t - 6/18)
1408 return (t*t)*mad((7/18.0f), t, (-6/18.0f));
1409 }
1410
1411 template <int kScale>
bicubic_x(SkJumper_SamplerCtx * ctx,F * x)1412 SI void bicubic_x(SkJumper_SamplerCtx* ctx, F* x) {
1413 *x = unaligned_load<F>(ctx->x) + (kScale * 0.5f);
1414 F fx = unaligned_load<F>(ctx->fx);
1415
1416 F scalex;
1417 if (kScale == -3) { scalex = bicubic_far (1.0f - fx); }
1418 if (kScale == -1) { scalex = bicubic_near(1.0f - fx); }
1419 if (kScale == +1) { scalex = bicubic_near( fx); }
1420 if (kScale == +3) { scalex = bicubic_far ( fx); }
1421 unaligned_store(ctx->scalex, scalex);
1422 }
1423 template <int kScale>
bicubic_y(SkJumper_SamplerCtx * ctx,F * y)1424 SI void bicubic_y(SkJumper_SamplerCtx* ctx, F* y) {
1425 *y = unaligned_load<F>(ctx->y) + (kScale * 0.5f);
1426 F fy = unaligned_load<F>(ctx->fy);
1427
1428 F scaley;
1429 if (kScale == -3) { scaley = bicubic_far (1.0f - fy); }
1430 if (kScale == -1) { scaley = bicubic_near(1.0f - fy); }
1431 if (kScale == +1) { scaley = bicubic_near( fy); }
1432 if (kScale == +3) { scaley = bicubic_far ( fy); }
1433 unaligned_store(ctx->scaley, scaley);
1434 }
1435
STAGE(bicubic_n3x)1436 STAGE(bicubic_n3x) { bicubic_x<-3>(ctx, &r); }
STAGE(bicubic_n1x)1437 STAGE(bicubic_n1x) { bicubic_x<-1>(ctx, &r); }
STAGE(bicubic_p1x)1438 STAGE(bicubic_p1x) { bicubic_x<+1>(ctx, &r); }
STAGE(bicubic_p3x)1439 STAGE(bicubic_p3x) { bicubic_x<+3>(ctx, &r); }
1440
STAGE(bicubic_n3y)1441 STAGE(bicubic_n3y) { bicubic_y<-3>(ctx, &g); }
STAGE(bicubic_n1y)1442 STAGE(bicubic_n1y) { bicubic_y<-1>(ctx, &g); }
STAGE(bicubic_p1y)1443 STAGE(bicubic_p1y) { bicubic_y<+1>(ctx, &g); }
STAGE(bicubic_p3y)1444 STAGE(bicubic_p3y) { bicubic_y<+3>(ctx, &g); }
1445
STAGE(callback)1446 STAGE(callback) {
1447 auto c = (SkJumper_CallbackCtx*)ctx;
1448 store4(c->rgba,0, r,g,b,a);
1449 c->fn(c, tail ? tail : kStride);
1450 load4(c->read_from,0, &r,&g,&b,&a);
1451 }
1452