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
|
//
// Whenever there is a lot to calculate, the question arises as to how
// tasks can be carried out simultaneously. We have already learned about
// one possibility, namely asynchronous processes, in Exercises 84-91.
//
// However, the computing power of the processor is only distributed to
// the started tasks, which always reaches its limits when pure computing
// power is called up.
//
// For example, in blockchains based on proof of work, the miners have
// to find a nonce for a certain character string so that the first m bits
// in the hash of the character string and the nonce are zeros.
// As the miner who can solve the task first receives the reward, everyone
// tries to complete the calculations as quickly as possible.
//
// This is where multithreading comes into play, where tasks are actually
// distributed across several cores of the CPU or GPU, which then really
// means a multiplication of performance.
//
// The following diagram roughly illustrates the difference between the
// various types of process execution.
// The 'Overall Time' column is intended to illustrate how the time is
// affected if, instead of one core as in synchronous and asynchronous
// processing, a second core now helps to complete the work in multithreading.
//
// In the ideal case shown, execution takes only half the time compared
// to the synchronous single thread. And even asynchronous processing
// is only slightly faster in comparison.
//
//
// Synchronous Asynchronous
// Processing Processing Multithreading
// ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐
// │ Thread 1 │ │ Thread 1 │ │ Thread 1 │ │ Thread 2 │
// ├──────────┤ ├──────────┤ ├──────────┤ ├──────────┤ Overall Time
// └──┼┼┼┼┼───┴─┴──┼┼┼┼┼───┴──┴──┼┼┼┼┼───┴─┴──┼┼┼┼┼───┴──┬───────┬───────┬──
// ├───┤ ├───┤ ├───┤ ├───┤ │ │ │
// │ T │ │ T │ │ T │ │ T │ │ │ │
// │ a │ │ a │ │ a │ │ a │ │ │ │
// │ s │ │ s │ │ s │ │ s │ │ │ │
// │ k │ │ k │ │ k │ │ k │ │ │ │
// │ │ │ │ │ │ │ │ │ │ │
// │ 1 │ │ 1 │ │ 1 │ │ 3 │ │ │ │
// └─┬─┘ └─┬─┘ └─┬─┘ └─┬─┘ │ │ │
// │ │ │ │ 5 Sec │ │
// ┌────┴───┐ ┌─┴─┐ ┌─┴─┐ ┌─┴─┐ │ │ │
// │Blocking│ │ T │ │ T │ │ T │ │ │ │
// └────┬───┘ │ a │ │ a │ │ a │ │ │ │
// │ │ s │ │ s │ │ s │ │ 8 Sec │
// ┌─┴─┐ │ k │ │ k │ │ k │ │ │ │
// │ T │ │ │ │ │ │ │ │ │ │
// │ a │ │ 2 │ │ 2 │ │ 4 │ │ │ │
// │ s │ └─┬─┘ ├───┤ ├───┤ │ │ │
// │ k │ │ │┼┼┼│ │┼┼┼│ ▼ │ 10 Sec
// │ │ ┌─┴─┐ └───┴────────┴───┴───────── │ │
// │ 1 │ │ T │ │ │
// └─┬─┘ │ a │ │ │
// │ │ s │ │ │
// ┌─┴─┐ │ k │ │ │
// │ T │ │ │ │ │
// │ a │ │ 1 │ │ │
// │ s │ ├───┤ │ │
// │ k │ │┼┼┼│ ▼ │
// │ │ └───┴──────────────────────────────────────────── │
// │ 2 │ │
// ├───┤ │
// │┼┼┼│ ▼
// └───┴────────────────────────────────────────────────────────────────
//
//
// The diagram was modeled on the one in a blog in which the differences
// between asynchronous processing and multithreading are explained in detail:
// https://blog.devgenius.io/multi-threading-vs-asynchronous-programming-what-is-the-difference-3ebfe1179a5
//
// Our exercise is essentially about clarifying the approach in Zig and
// therefore we try to keep it as simple as possible.
// Multithreading in itself is already difficult enough. ;-)
//
const std = @import("std");
pub fn main() !void {
// This is where the preparatory work takes place
// before the parallel processing begins.
std.debug.print("Starting work...\n", .{});
// These curly brackets are very important, they are necessary
// to enclose the area where the threads are called.
// Without these brackets, the program would not wait for the
// end of the threads and they would continue to run beyond the
// end of the program.
{
// Now we start the first thread, with the number as parameter
const handle = try std.Thread.spawn(.{}, thread_function, .{1});
// Waits for the thread to complete,
// then deallocates any resources created on `spawn()`.
defer handle.join();
// Second thread
const handle2 = try std.Thread.spawn(.{}, thread_function, .{-4}); // that can't be right?
defer handle2.join();
// Third thread
const handle3 = try std.Thread.spawn(.{}, thread_function, .{3});
defer ??? // <-- something is missing
// After the threads have been started,
// they run in parallel and we can still do some work in between.
std.time.sleep((1) * std.time.ns_per_s);
std.debug.print("Some weird stuff, after starting the threads.\n", .{});
}
// After we have left the closed area, we wait until
// the threads have run through, if this has not yet been the case.
std.debug.print("Zig is cool!\n", .{});
}
// This function is started with every thread that we set up.
// In our example, we pass the number of the thread as a parameter.
fn thread_function(num: usize) !void {
std.debug.print("thread {d}: {s}\n", .{ num, "started." });
std.time.sleep((5 - num % 3) * std.time.ns_per_s);
std.debug.print("thread {d}: {s}\n", .{ num, "finished." });
}
// This is the easiest way to run threads in parallel.
// In general, however, more management effort is required,
// e.g. by setting up a pool and allowing the threads to communicate
// with each other using semaphores.
//
// But that's a topic for another exercise.
|
