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
|
/*
Copyright (c) 2007-2010 iMatix Corporation
This file is part of 0MQ.
0MQ is free software; you can redistribute it and/or modify it under
the terms of the Lesser GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
0MQ is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Lesser GNU General Public License for more details.
You should have received a copy of the Lesser GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef __ZMQ_CONFIG_HPP_INCLUDED__
#define __ZMQ_CONFIG_HPP_INCLUDED__
namespace zmq
{
// Compile-time settings.
enum
{
// Maximal number of OS threads that can own 0MQ sockets
// at the same time.
max_app_threads = 512,
// Number of new messages in message pipe needed to trigger new memory
// allocation. Setting this parameter to 256 decreases the impact of
// memory allocation by approximately 99.6%
message_pipe_granularity = 256,
// Number of signals that can be read by the signaler
// using a single system call.
signal_buffer_size = 8,
// Determines how often does socket poll for new commands when it
// still has unprocessed messages to handle. Thus, if it is set to 100,
// socket will process 100 inbound messages before doing the poll.
// If there are no unprocessed messages available, poll is done
// immediately. Decreasing the value trades overall latency for more
// real-time behaviour (less latency peaks).
inbound_poll_rate = 100,
// Maximal batching size for engines with receiving functionality.
// So, if there are 10 messages that fit into the batch size, all of
// them may be read by a single 'recv' system call, thus avoiding
// unnecessary network stack traversals.
in_batch_size = 8192,
// Maximal batching size for engines with sending functionality.
// So, if there are 10 messages that fit into the batch size, all of
// them may be written by a single 'send' system call, thus avoiding
// unnecessary network stack traversals.
out_batch_size = 8192,
// Maximal delta between high and low watermark.
max_wm_delta = 1024,
// Maximum number of events the I/O thread can process in one go.
max_io_events = 256,
// Maximal wait time for a timer (milliseconds).
max_timer_period = 100,
// Maximal delay to process command in API thread (in CPU ticks).
// 3,000,000 ticks equals to 1 - 2 milliseconds on current CPUs.
max_command_delay = 3000000,
// Maximal number of non-accepted connections that can be held by
// TCP listener object.
tcp_connection_backlog = 10,
// Maximum transport data unit size for PGM (TPDU).
pgm_max_tpdu = 1500
};
}
#endif
|