summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/radeon/r600_test_dma.c
blob: 1e60f6affa74e1c12290833ab98e786125c589aa (plain)
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
/*
 * Copyright 2016 Advanced Micro Devices, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 */

/* This file implements randomized SDMA texture blit tests. */

#include "r600_pipe_common.h"
#include "util/u_surface.h"

static uint64_t seed_xorshift128plus[2];

/* Super fast random number generator.
 *
 * This rand_xorshift128plus function by Sebastiano Vigna belongs
 * to the public domain.
 */
static uint64_t rand_xorshift128plus(void)
{
	uint64_t *s = seed_xorshift128plus;

	uint64_t s1 = s[0];
	const uint64_t s0 = s[1];
	s[0] = s0;
	s1 ^= s1 << 23;
	s[1] = s1 ^ s0 ^ (s1 >> 18) ^ (s0 >> 5);
	return s[1] + s0;
}

#define RAND_NUM_SIZE 8

/* The GPU blits are emulated on the CPU using these CPU textures. */

struct cpu_texture {
	uint8_t *ptr;
	uint64_t size;
	uint64_t layer_stride;
	unsigned stride;
};

static void alloc_cpu_texture(struct cpu_texture *tex,
			      struct pipe_resource *templ, int bpp)
{
	tex->stride = align(templ->width0 * bpp, RAND_NUM_SIZE);
	tex->layer_stride = (uint64_t)tex->stride * templ->height0;
	tex->size = tex->layer_stride * templ->array_size;
	tex->ptr = malloc(tex->size);
	assert(tex->ptr);
}

static void set_random_pixels(struct pipe_context *ctx,
			      struct pipe_resource *tex,
			      struct cpu_texture *cpu)
{
	struct pipe_transfer *t;
	uint8_t *map;
	int x,y,z;

	map = pipe_transfer_map_3d(ctx, tex, 0, PIPE_TRANSFER_WRITE,
				   0, 0, 0, tex->width0, tex->height0,
				   tex->array_size, &t);
	assert(map);

	for (z = 0; z < tex->array_size; z++) {
		for (y = 0; y < tex->height0; y++) {
			uint64_t *ptr = (uint64_t*)
				(map + t->layer_stride*z + t->stride*y);
			uint64_t *ptr_cpu = (uint64_t*)
				(cpu->ptr + cpu->layer_stride*z + cpu->stride*y);
			unsigned size = cpu->stride / RAND_NUM_SIZE;

			assert(t->stride % RAND_NUM_SIZE == 0);
			assert(cpu->stride % RAND_NUM_SIZE == 0);

			for (x = 0; x < size; x++)
				*ptr++ = *ptr_cpu++ = rand_xorshift128plus();
		}
	}

	pipe_transfer_unmap(ctx, t);
}

static bool compare_textures(struct pipe_context *ctx,
			     struct pipe_resource *tex,
			     struct cpu_texture *cpu, int bpp)
{
	struct pipe_transfer *t;
	uint8_t *map;
	int y,z;
	bool pass = true;

	map = pipe_transfer_map_3d(ctx, tex, 0, PIPE_TRANSFER_READ,
				   0, 0, 0, tex->width0, tex->height0,
				   tex->array_size, &t);
	assert(map);

	for (z = 0; z < tex->array_size; z++) {
		for (y = 0; y < tex->height0; y++) {
			uint8_t *ptr = map + t->layer_stride*z + t->stride*y;
			uint8_t *cpu_ptr = cpu->ptr +
					   cpu->layer_stride*z + cpu->stride*y;

			if (memcmp(ptr, cpu_ptr, tex->width0 * bpp)) {
				pass = false;
				goto done;
			}
		}
	}
done:
	pipe_transfer_unmap(ctx, t);
	return pass;
}

static enum pipe_format get_format_from_bpp(int bpp)
{
	switch (bpp) {
	case 1:
		return PIPE_FORMAT_R8_UINT;
	case 2:
		return PIPE_FORMAT_R16_UINT;
	case 4:
		return PIPE_FORMAT_R32_UINT;
	case 8:
		return PIPE_FORMAT_R32G32_UINT;
	case 16:
		return PIPE_FORMAT_R32G32B32A32_UINT;
	default:
		assert(0);
		return PIPE_FORMAT_NONE;
	}
}

static const char *array_mode_to_string(unsigned mode)
{
	switch (mode) {
	case RADEON_SURF_MODE_LINEAR_ALIGNED:
		return "LINEAR_ALIGNED";
	case RADEON_SURF_MODE_1D:
		return "1D_TILED_THIN1";
	case RADEON_SURF_MODE_2D:
		return "2D_TILED_THIN1";
	default:
		assert(0);
		return "       UNKNOWN";
	}
}

static unsigned generate_max_tex_side(unsigned max_tex_side)
{
	switch (rand() % 4) {
	case 0:
		/* Try to hit large sizes in 1/4 of the cases. */
		return max_tex_side;
	case 1:
		/* Try to hit 1D tiling in 1/4 of the cases. */
		return 128;
	default:
		/* Try to hit common sizes in 2/4 of the cases. */
		return 2048;
	}
}

void r600_test_dma(struct r600_common_screen *rscreen)
{
	struct pipe_screen *screen = &rscreen->b;
	struct pipe_context *ctx = screen->context_create(screen, NULL, 0);
	struct r600_common_context *rctx = (struct r600_common_context*)ctx;
	uint64_t max_alloc_size;
	unsigned i, iterations, num_partial_copies, max_levels, max_tex_side;
	unsigned num_pass = 0, num_fail = 0;

	max_levels = screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS);
	max_tex_side = 1 << (max_levels - 1);

	/* Max 128 MB allowed for both textures. */
	max_alloc_size = 128 * 1024 * 1024;

	/* the seed for random test parameters */
	srand(0x9b47d95b);
	/* the seed for random pixel data */
	seed_xorshift128plus[0] = 0x3bffb83978e24f88;
	seed_xorshift128plus[1] = 0x9238d5d56c71cd35;

	iterations = 1000000000; /* just kill it when you are bored */
	num_partial_copies = 30;

	/* These parameters are randomly generated per test:
	 * - whether to do one whole-surface copy or N partial copies per test
	 * - which tiling modes to use (LINEAR_ALIGNED, 1D, 2D)
	 * - which texture dimensions to use
	 * - whether to use VRAM (all tiling modes) and GTT (staging, linear
	 *   only) allocations
	 * - random initial pixels in src
	 * - generate random subrectangle copies for partial blits
	 */
	for (i = 0; i < iterations; i++) {
		struct pipe_resource tsrc = {}, tdst = {}, *src, *dst;
		struct r600_texture *rdst;
		struct r600_texture *rsrc;
		struct cpu_texture src_cpu, dst_cpu;
		unsigned bpp, max_width, max_height, max_depth, j, num;
		unsigned gfx_blits = 0, dma_blits = 0, max_tex_side_gen;
		unsigned max_tex_layers;
		bool pass;
		bool do_partial_copies = rand() & 1;

		/* generate a random test case */
		tsrc.target = tdst.target = PIPE_TEXTURE_2D_ARRAY;
		tsrc.depth0 = tdst.depth0 = 1;

		bpp = 1 << (rand() % 5);
		tsrc.format = tdst.format = get_format_from_bpp(bpp);

		max_tex_side_gen = generate_max_tex_side(max_tex_side);
		max_tex_layers = rand() % 4 ? 1 : 5;

		tsrc.width0 = (rand() % max_tex_side_gen) + 1;
		tsrc.height0 = (rand() % max_tex_side_gen) + 1;
		tsrc.array_size = (rand() % max_tex_layers) + 1;

		/* Have a 1/4 chance of getting power-of-two dimensions. */
		if (rand() % 4 == 0) {
			tsrc.width0 = util_next_power_of_two(tsrc.width0);
			tsrc.height0 = util_next_power_of_two(tsrc.height0);
		}

		if (!do_partial_copies) {
			/* whole-surface copies only, same dimensions */
			tdst = tsrc;
		} else {
			max_tex_side_gen = generate_max_tex_side(max_tex_side);
			max_tex_layers = rand() % 4 ? 1 : 5;

			/* many partial copies, dimensions can be different */
			tdst.width0 = (rand() % max_tex_side_gen) + 1;
			tdst.height0 = (rand() % max_tex_side_gen) + 1;
			tdst.array_size = (rand() % max_tex_layers) + 1;

			/* Have a 1/4 chance of getting power-of-two dimensions. */
			if (rand() % 4 == 0) {
				tdst.width0 = util_next_power_of_two(tdst.width0);
				tdst.height0 = util_next_power_of_two(tdst.height0);
			}
		}

		/* check texture sizes */
		if ((uint64_t)tsrc.width0 * tsrc.height0 * tsrc.array_size * bpp +
		    (uint64_t)tdst.width0 * tdst.height0 * tdst.array_size * bpp >
		    max_alloc_size) {
			/* too large, try again */
			i--;
			continue;
		}

		/* VRAM + the tiling mode depends on dimensions (3/4 of cases),
		 * or GTT + linear only (1/4 of cases)
		 */
		tsrc.usage = rand() % 4 ? PIPE_USAGE_DEFAULT : PIPE_USAGE_STAGING;
		tdst.usage = rand() % 4 ? PIPE_USAGE_DEFAULT : PIPE_USAGE_STAGING;

		/* Allocate textures (both the GPU and CPU copies).
		 * The CPU will emulate what the GPU should be doing.
		 */
		src = screen->resource_create(screen, &tsrc);
		dst = screen->resource_create(screen, &tdst);
		assert(src);
		assert(dst);
		rdst = (struct r600_texture*)dst;
		rsrc = (struct r600_texture*)src;
		alloc_cpu_texture(&src_cpu, &tsrc, bpp);
		alloc_cpu_texture(&dst_cpu, &tdst, bpp);

		printf("%4u: dst = (%5u x %5u x %u, %s), "
		       " src = (%5u x %5u x %u, %s), bpp = %2u, ",
		       i, tdst.width0, tdst.height0, tdst.array_size,
		       array_mode_to_string(rdst->surface.level[0].mode),
		       tsrc.width0, tsrc.height0, tsrc.array_size,
		       array_mode_to_string(rsrc->surface.level[0].mode), bpp);
		fflush(stdout);

		/* set src pixels */
		set_random_pixels(ctx, src, &src_cpu);

		/* clear dst pixels */
		rctx->clear_buffer(ctx, dst, 0, rdst->surface.bo_size, 0, true);
		memset(dst_cpu.ptr, 0, dst_cpu.layer_stride * tdst.array_size);

		/* preparation */
		max_width = MIN2(tsrc.width0, tdst.width0);
		max_height = MIN2(tsrc.height0, tdst.height0);
		max_depth = MIN2(tsrc.array_size, tdst.array_size);

		num = do_partial_copies ? num_partial_copies : 1;
		for (j = 0; j < num; j++) {
			int width, height, depth;
			int srcx, srcy, srcz, dstx, dsty, dstz;
			struct pipe_box box;
			unsigned old_num_draw_calls = rctx->num_draw_calls;
			unsigned old_num_dma_calls = rctx->num_dma_calls;

			if (!do_partial_copies) {
				/* copy whole src to dst */
				width = max_width;
				height = max_height;
				depth = max_depth;

				srcx = srcy = srcz = dstx = dsty = dstz = 0;
			} else {
				/* random sub-rectangle copies from src to dst */
				depth = (rand() % max_depth) + 1;
				srcz = rand() % (tsrc.array_size - depth + 1);
				dstz = rand() % (tdst.array_size - depth + 1);

				/* special code path to hit the tiled partial copies */
				if (rsrc->surface.level[0].mode >= RADEON_SURF_MODE_1D &&
				    rdst->surface.level[0].mode >= RADEON_SURF_MODE_1D &&
				    rand() & 1) {
					if (max_width < 8 || max_height < 8)
						continue;
					width = ((rand() % (max_width / 8)) + 1) * 8;
					height = ((rand() % (max_height / 8)) + 1) * 8;

					srcx = rand() % (tsrc.width0 - width + 1) & ~0x7;
					srcy = rand() % (tsrc.height0 - height + 1) & ~0x7;

					dstx = rand() % (tdst.width0 - width + 1) & ~0x7;
					dsty = rand() % (tdst.height0 - height + 1) & ~0x7;
				} else {
					/* just make sure that it doesn't divide by zero */
					assert(max_width > 0 && max_height > 0);

					width = (rand() % max_width) + 1;
					height = (rand() % max_height) + 1;

					srcx = rand() % (tsrc.width0 - width + 1);
					srcy = rand() % (tsrc.height0 - height + 1);

					dstx = rand() % (tdst.width0 - width + 1);
					dsty = rand() % (tdst.height0 - height + 1);
				}

				/* special code path to hit out-of-bounds reads in L2T */
				if (rsrc->surface.level[0].mode == RADEON_SURF_MODE_LINEAR_ALIGNED &&
				    rdst->surface.level[0].mode >= RADEON_SURF_MODE_1D &&
				    rand() % 4 == 0) {
					srcx = 0;
					srcy = 0;
					srcz = 0;
				}
			}

			/* GPU copy */
			u_box_3d(srcx, srcy, srcz, width, height, depth, &box);
			rctx->dma_copy(ctx, dst, 0, dstx, dsty, dstz, src, 0, &box);

			/* See which engine was used. */
			gfx_blits += rctx->num_draw_calls > old_num_draw_calls;
			dma_blits += rctx->num_dma_calls > old_num_dma_calls;

			/* CPU copy */
			util_copy_box(dst_cpu.ptr, tdst.format, dst_cpu.stride,
				      dst_cpu.layer_stride,
				      dstx, dsty, dstz, width, height, depth,
				      src_cpu.ptr, src_cpu.stride,
				      src_cpu.layer_stride,
				      srcx, srcy, srcz);
		}

		pass = compare_textures(ctx, dst, &dst_cpu, bpp);
		if (pass)
			num_pass++;
		else
			num_fail++;

		printf("BLITs: GFX = %2u, DMA = %2u, %s [%u/%u]\n",
		       gfx_blits, dma_blits, pass ? "pass" : "fail",
		       num_pass, num_pass+num_fail);

		/* cleanup */
		pipe_resource_reference(&src, NULL);
		pipe_resource_reference(&dst, NULL);
		free(src_cpu.ptr);
		free(dst_cpu.ptr);
	}

	ctx->destroy(ctx);
	exit(0);
}