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
|
#include <stdint.h>
#include <inttypes.h>
#include <math.h>
#include <stdlib.h>
#include "til.h"
#include "til_fb.h"
/* Copyright (C) 2016 Vito Caputo <vcaputo@pengaru.com> */
/* Some defines for the fixed-point stuff in render(). */
#define FIXED_TRIG_LUT_SIZE 4096 /* size of the cos/sin look-up tables */
#define FIXED_BITS 11 /* fractional bits */
#define FIXED_EXP (1 << FIXED_BITS) /* 2^FIXED_BITS */
#define FIXED_MASK (FIXED_EXP - 1) /* fractional part mask */
#define FIXED_COS(_rad) costab[(_rad) % FIXED_TRIG_LUT_SIZE]
#define FIXED_SIN(_rad) sintab[(_rad) % FIXED_TRIG_LUT_SIZE]
#define FIXED_MULT(_a, _b) (((_a) * (_b)) >> FIXED_BITS)
#define FIXED_NEW(_i) ((_i) << FIXED_BITS)
#define FIXED_TO_INT(_f) ((_f) >> FIXED_BITS)
typedef struct color_t {
int r, g, b;
} color_t;
typedef struct roto_context_t {
unsigned r, rr;
unsigned n_cpus;
} roto_context_t;
static int32_t costab[FIXED_TRIG_LUT_SIZE], sintab[FIXED_TRIG_LUT_SIZE];
static uint8_t texture[256][256];
static color_t palette[2];
static void * roto_create_context(unsigned ticks, unsigned num_cpus)
{
return calloc(1, sizeof(roto_context_t));
}
static void roto_destroy_context(void *context)
{
free(context);
}
/* linearly interpolate between two colors, alpha is fixed-point value 0-FIXED_EXP. */
static inline color_t lerp_color(color_t *a, color_t *b, int alpha)
{
/* TODO: This could be done without multiplies with a bit of effort,
* maybe a simple table mapping integer color deltas to shift values
* for shifting alpha which then gets simply added? A table may not even
* be necessary, use the order of the delta to derive how much to shift
* alpha?
*/
color_t c = {
.r = a->r + FIXED_MULT(alpha, b->r - a->r),
.g = a->g + FIXED_MULT(alpha, b->g - a->g),
.b = a->b + FIXED_MULT(alpha, b->b - a->b),
};
return c;
}
/* Return the bilinearly interpolated color palette[texture[ty][tx]] (Anti-Aliasing) */
/* tx, ty are fixed-point for fractions, palette colors are also in fixed-point format. */
static uint32_t bilerp_color(uint8_t texture[256][256], color_t *palette, int tx, int ty)
{
uint8_t itx = FIXED_TO_INT(tx), ity = FIXED_TO_INT(ty);
color_t n_color, s_color, color;
int x_alpha, y_alpha;
uint8_t nw, ne, sw, se;
/* We need the 4 texels constituting a 2x2 square pattern to interpolate.
* A point tx,ty can only intersect one texel; one corner of the 2x2 square.
* Where relative to the corner's center the intersection occurs determines which corner has been intersected,
* and the other corner texels may then be addressed relative to that corner.
* Alpha values must also be determined for both axis, these values describe the position between
* the 2x2 texel centers the intersection occurred, aka the weight or bias.
* Once the two alpha values are known, linear interpolation between the texel colors is trivial.
*/
if ((ty & FIXED_MASK) > (FIXED_EXP >> 1)) {
y_alpha = ty & (FIXED_MASK >> 1);
if ((tx & (FIXED_MASK)) > (FIXED_EXP >> 1)) {
nw = texture[ity][itx];
ne = texture[ity][(uint8_t)(itx + 1)];
sw = texture[(uint8_t)(ity + 1)][itx];
se = texture[(uint8_t)(ity + 1)][(uint8_t)(itx + 1)];
x_alpha = tx & (FIXED_MASK >> 1);
} else {
ne = texture[ity][itx];
nw = texture[ity][(uint8_t)(itx - 1)];
se = texture[(uint8_t)(ity + 1)][itx];
sw = texture[(uint8_t)(ity + 1)][(uint8_t)(itx - 1)];
x_alpha = (FIXED_EXP >> 1) + (tx & (FIXED_MASK >> 1));
}
} else {
y_alpha = (FIXED_EXP >> 1) + (ty & (FIXED_MASK >> 1));
if ((tx & (FIXED_MASK)) > (FIXED_EXP >> 1)) {
sw = texture[ity][itx];
se = texture[ity][(uint8_t)(itx + 1)];
nw = texture[(uint8_t)(ity - 1)][itx];
ne = texture[(uint8_t)(ity - 1)][(uint8_t)(itx + 1)];
x_alpha = tx & (FIXED_MASK >> 1);
} else {
se = texture[ity][itx];
sw = texture[ity][(uint8_t)(itx - 1)];
ne = texture[(uint8_t)(ity - 1)][itx];
nw = texture[(uint8_t)(ity - 1)][(uint8_t)(itx - 1)];
x_alpha = (FIXED_EXP >> 1) + (tx & (FIXED_MASK >> 1));
}
}
/* Skip interpolation of same colors, a substantial optimization with plain textures like the checker pattern */
if (nw == ne) {
if (ne == sw && sw == se) {
return (FIXED_TO_INT(palette[sw].r) << 16) | (FIXED_TO_INT(palette[sw].g) << 8) | FIXED_TO_INT(palette[sw].b);
}
n_color = palette[nw];
} else {
n_color = lerp_color(&palette[nw], &palette[ne], x_alpha);
}
if (sw == se) {
s_color = palette[sw];
} else {
s_color = lerp_color(&palette[sw], &palette[se], x_alpha);
}
color = lerp_color(&n_color, &s_color, y_alpha);
return (FIXED_TO_INT(color.r) << 16) | (FIXED_TO_INT(color.g) << 8) | FIXED_TO_INT(color.b);
}
static void init_roto(uint8_t texture[256][256], int32_t *costab, int32_t *sintab)
{
int x, y, i;
/* Generate simple checker pattern texture, nothing clever, feel free to play! */
/* If you modify texture on every frame instead of only @ initialization you can
* produce some neat output. These values are indexed into palette[] below. */
for (y = 0; y < 128; y++) {
for (x = 0; x < 128; x++)
texture[y][x] = 1;
for (; x < 256; x++)
texture[y][x] = 0;
}
for (; y < 256; y++) {
for (x = 0; x < 128; x++)
texture[y][x] = 0;
for (; x < 256; x++)
texture[y][x] = 1;
}
/* Generate fixed-point cos & sin LUTs. */
for (i = 0; i < FIXED_TRIG_LUT_SIZE; i++) {
costab[i] = ((cos((double)2*M_PI*i/FIXED_TRIG_LUT_SIZE))*FIXED_EXP);
sintab[i] = ((sin((double)2*M_PI*i/FIXED_TRIG_LUT_SIZE))*FIXED_EXP);
}
}
static int roto_fragmenter(void *context, const til_fb_fragment_t *fragment, unsigned number, til_fb_fragment_t *res_fragment)
{
roto_context_t *ctxt = context;
return til_fb_fragment_slice_single(fragment, ctxt->n_cpus, number, res_fragment);
}
/* prepare a frame for concurrent rendering */
static void roto_prepare_frame(void *context, unsigned ticks, unsigned n_cpus, til_fb_fragment_t *fragment, til_fragmenter_t *res_fragmenter)
{
roto_context_t *ctxt = context;
static int initialized;
if (!initialized) {
initialized = 1;
init_roto(texture, costab, sintab);
}
*res_fragmenter = roto_fragmenter;
ctxt->n_cpus = n_cpus;
// This governs the rotation and color cycle.
ctxt->r += FIXED_TO_INT(FIXED_MULT(FIXED_SIN(ctxt->rr), FIXED_NEW(16)));
ctxt->rr += 2;
}
/* Draw a rotating checkered 256x256 texture into fragment. */
static void roto_render_fragment(void *context, unsigned ticks, unsigned cpu, til_fb_fragment_t *fragment)
{
roto_context_t *ctxt = context;
int y_cos_r, y_sin_r, x_cos_r, x_sin_r, x_cos_r_init, x_sin_r_init, cos_r, sin_r;
int x, y, frame_width = fragment->frame_width, frame_height = fragment->frame_height;
uint32_t *buf = fragment->buf;
/* This is all done using fixed-point in the hopes of being faster, and yes assumptions
* are being made WRT the overflow of tx/ty as well, only tested on x86_64. */
cos_r = FIXED_COS(ctxt->r);
sin_r = FIXED_SIN(ctxt->r);
/* Vary the colors, this is just a mashup of sinusoidal rgb values. */
palette[0].r = (FIXED_MULT(FIXED_COS(ctxt->rr), FIXED_NEW(127)) + FIXED_NEW(128));
palette[0].g = (FIXED_MULT(FIXED_SIN(ctxt->rr / 2), FIXED_NEW(127)) + FIXED_NEW(128));
palette[0].b = (FIXED_MULT(FIXED_COS(ctxt->rr / 3), FIXED_NEW(127)) + FIXED_NEW(128));
palette[1].r = (FIXED_MULT(FIXED_SIN(ctxt->rr / 2), FIXED_NEW(127)) + FIXED_NEW(128));
palette[1].g = (FIXED_MULT(FIXED_COS(ctxt->rr / 2), FIXED_NEW(127)) + FIXED_NEW(128));
palette[1].b = (FIXED_MULT(FIXED_SIN(ctxt->rr), FIXED_NEW(127)) + FIXED_NEW(128));
/* The dimensions are cut in half and negated to center the rotation. */
/* The [xy]_{sin,cos}_r variables are accumulators to replace multiplication with addition. */
x_cos_r_init = FIXED_MULT(-FIXED_NEW(frame_width / 2) + FIXED_NEW(fragment->x), cos_r);
x_sin_r_init = FIXED_MULT(-FIXED_NEW(frame_width / 2) + FIXED_NEW(fragment->x), sin_r);
y_cos_r = FIXED_MULT(-FIXED_NEW(frame_height / 2) + FIXED_NEW(fragment->y), cos_r);
y_sin_r = FIXED_MULT(-FIXED_NEW(frame_height / 2) + FIXED_NEW(fragment->y), sin_r);
for (y = fragment->y; y < fragment->y + fragment->height; y++) {
x_cos_r = x_cos_r_init;
x_sin_r = x_sin_r_init;
for (x = fragment->x; x < fragment->x + fragment->width; x++, buf++) {
*buf = bilerp_color(texture, palette, x_sin_r - y_cos_r, y_sin_r + x_cos_r);
x_cos_r += cos_r;
x_sin_r += sin_r;
}
buf = ((void *)buf) + fragment->stride;
y_cos_r += cos_r;
y_sin_r += sin_r;
}
}
til_module_t roto_module = {
.create_context = roto_create_context,
.destroy_context = roto_destroy_context,
.prepare_frame = roto_prepare_frame,
.render_fragment = roto_render_fragment,
.name = "roto",
.description = "Anti-aliased tiled texture rotation (threaded)",
.author = "Vito Caputo <vcaputo@pengaru.com>",
};
|
