mesa/glthread: decrease the batch size for better perf scaling
[mesa.git] / src / mesa / main / glthread.h
1 /*
2 * Copyright © 2012 Intel Corporation
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21 * IN THE SOFTWARE.
22 */
23
24 #ifndef _GLTHREAD_H
25 #define _GLTHREAD_H
26
27 #include "main/mtypes.h"
28
29 /* The size of one batch and the maximum size of one call.
30 *
31 * This should be as low as possible, so that:
32 * - multiple synchronizations within a frame don't slow us down much
33 * - a smaller number of calls per frame can still get decent parallelism
34 * - the memory footprint of the queue is low, and with that comes a lower
35 * chance of experiencing CPU cache thrashing
36 * but it should be high enough so that u_queue overhead remains negligible.
37 */
38 #define MARSHAL_MAX_CMD_SIZE (8 * 1024)
39
40 /* The number of batch slots in memory.
41 *
42 * One batch is being executed, one batch is being filled, the rest are
43 * waiting batches. There must be at least 1 slot for a waiting batch,
44 * so the minimum number of batches is 3.
45 */
46 #define MARSHAL_MAX_BATCHES 8
47
48 #include <inttypes.h>
49 #include <stdbool.h>
50 #include <pthread.h>
51 #include "util/u_queue.h"
52
53 enum marshal_dispatch_cmd_id;
54
55 /** A single batch of commands queued up for execution. */
56 struct glthread_batch
57 {
58 /** Batch fence for waiting for the execution to finish. */
59 struct util_queue_fence fence;
60
61 /** The worker thread will access the context with this. */
62 struct gl_context *ctx;
63
64 /** Amount of data used by batch commands, in bytes. */
65 size_t used;
66
67 /** Data contained in the command buffer. */
68 uint8_t buffer[MARSHAL_MAX_CMD_SIZE];
69 };
70
71 struct glthread_state
72 {
73 /** Multithreaded queue. */
74 struct util_queue queue;
75
76 /** This is sent to the driver for framebuffer overlay / HUD. */
77 struct util_queue_monitoring stats;
78
79 /** The ring of batches in memory. */
80 struct glthread_batch batches[MARSHAL_MAX_BATCHES];
81
82 /** Index of the last submitted batch. */
83 unsigned last;
84
85 /** Index of the batch being filled and about to be submitted. */
86 unsigned next;
87
88 /**
89 * Tracks on the main thread side whether the current vertex array binding
90 * is in a VBO.
91 */
92 bool vertex_array_is_vbo;
93
94 /**
95 * Tracks on the main thread side whether the current element array (index
96 * buffer) binding is in a VBO.
97 */
98 bool element_array_is_vbo;
99 };
100
101 void _mesa_glthread_init(struct gl_context *ctx);
102 void _mesa_glthread_destroy(struct gl_context *ctx);
103
104 void _mesa_glthread_restore_dispatch(struct gl_context *ctx);
105 void _mesa_glthread_flush_batch(struct gl_context *ctx);
106 void _mesa_glthread_finish(struct gl_context *ctx);
107
108 #endif /* _GLTHREAD_H*/