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
|
/*
* Copyright © 2009 Intel Corporation
*
* 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.
*
* Authors:
* Eric Anholt <eric@anholt.net>
*
*/
/**
* Roughly simulates repeatedly uploading frames of images, by uploading
* the data all at once with pwrite, and then blitting it to another buffer.
*
* You might think of this like a movie player, but that wouldn't be entirely
* accurate, since the access patterns of the memory would be different
* (generally, smaller source image, upscaled, an thus different memory access
* pattern in both texel fetch for the stretching and the destination writes).
* However, some things like swfdec would be doing something like this since
* they compute their data in host memory and upload the full sw rendered
* frame.
*
* Additionally, those applications should be rendering at the screen refresh
* rate, while this test has no limits, and so can get itself into the
* working set larger than aperture size performance disaster.
*
* The current workload we have that does large drm_intel_bo_map()
* uploads is texture upload for OpenGL (as it frequently is doing
* reformatting as it uploads the user's data, making bo_subdata less
* suitable)
*/
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <assert.h>
#include <fcntl.h>
#include <inttypes.h>
#include <errno.h>
#include <sys/stat.h>
#include <sys/time.h>
#include "drm.h"
#include "i915_drm.h"
#include "drmtest.h"
#include "intel_bufmgr.h"
#include "intel_batchbuffer.h"
#include "intel_gpu_tools.h"
#define OBJECT_WIDTH 1280
#define OBJECT_HEIGHT 720
static double
get_time_in_secs(void)
{
struct timeval tv;
gettimeofday(&tv, NULL);
return (double)tv.tv_sec + tv.tv_usec / 1000000.0;
}
static void
do_render(drm_intel_bufmgr *bufmgr, struct intel_batchbuffer *batch,
drm_intel_bo *dst_bo, int width, int height)
{
uint32_t *data;
drm_intel_bo *src_bo;
int i;
static uint32_t seed = 1;
src_bo = drm_intel_bo_alloc(bufmgr, "src", width * height * 4, 4096);
drm_intel_bo_map(src_bo, 1);
data = src_bo->virtual;
for (i = 0; i < width * height; i++) {
data[i] = seed++;
}
drm_intel_bo_unmap(src_bo);
/* Render the junk to the dst. */
BLIT_COPY_BATCH_START(batch->devid, 0);
OUT_BATCH((3 << 24) | /* 32 bits */
(0xcc << 16) | /* copy ROP */
(width * 4) /* dst pitch */);
OUT_BATCH(0); /* dst x1,y1 */
OUT_BATCH((height << 16) | width); /* dst x2,y2 */
OUT_RELOC(dst_bo, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, 0);
BLIT_RELOC_UDW(batch->devid);
OUT_BATCH(0); /* src x1,y1 */
OUT_BATCH(width * 4); /* src pitch */
OUT_RELOC(src_bo, I915_GEM_DOMAIN_RENDER, 0, 0);
BLIT_RELOC_UDW(batch->devid);
ADVANCE_BATCH();
intel_batchbuffer_flush(batch);
drm_intel_bo_unreference(src_bo);
}
int main(int argc, char **argv)
{
int fd;
int object_size = OBJECT_WIDTH * OBJECT_HEIGHT * 4;
double start_time, end_time;
drm_intel_bo *dst_bo;
drm_intel_bufmgr *bufmgr;
struct intel_batchbuffer *batch;
int i;
fd = drm_open_any();
bufmgr = drm_intel_bufmgr_gem_init(fd, 4096);
drm_intel_bufmgr_gem_enable_reuse(bufmgr);
batch = intel_batchbuffer_alloc(bufmgr, intel_get_drm_devid(fd));
dst_bo = drm_intel_bo_alloc(bufmgr, "dst", object_size, 4096);
/* Prep loop to get us warmed up. */
for (i = 0; i < 60; i++) {
do_render(bufmgr, batch, dst_bo, OBJECT_WIDTH, OBJECT_HEIGHT);
}
drm_intel_bo_wait_rendering(dst_bo);
/* Do the actual timing. */
start_time = get_time_in_secs();
for (i = 0; i < 200; i++) {
do_render(bufmgr, batch, dst_bo, OBJECT_WIDTH, OBJECT_HEIGHT);
}
drm_intel_bo_wait_rendering(dst_bo);
end_time = get_time_in_secs();
printf("%d iterations in %.03f secs: %.01f MB/sec\n", i,
end_time - start_time,
(double)i * OBJECT_WIDTH * OBJECT_HEIGHT * 4 / 1024.0 / 1024.0 /
(end_time - start_time));
intel_batchbuffer_free(batch);
drm_intel_bufmgr_destroy(bufmgr);
close(fd);
return 0;
}
|
