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
|
use super::{CheckRequest, Message, RowResult};
use std::sync::mpsc::{channel, Receiver, Sender};
use std::thread::JoinHandle;
#[derive(Debug)]
struct RequestBuffer {
mask_buff: Vec<u64>,
row_buff: Vec<Vec<u32>>,
pointer: usize,
}
impl RequestBuffer {
pub fn new(size: usize) -> Self {
RequestBuffer {
mask_buff: vec![0; size],
row_buff: vec![Vec::new(); size],
pointer: 0,
}
}
pub fn read(&mut self, request: CheckRequest) -> Option<(&[u64], &[Vec<u32>])> {
self.mask_buff[self.pointer] = request.bitmask;
self.row_buff[self.pointer] = request.rows;
self.pointer += 1;
if self.pointer == self.mask_buff.len() {
self.pointer = 0;
return Some((self.mask_buff.as_ref(), self.row_buff.as_ref()));
}
None
}
fn flush(&mut self) {
while self.read(CheckRequest::new(vec![], 0, 0)).is_none() {}
}
}
pub struct OclManager {
job_id: u64,
host_sender: Sender<Message>,
output_sender: Sender<Message>,
receiver: Receiver<Message>,
buffers: Vec<RequestBuffer>,
output_handle: JoinHandle<()>,
host_handle: Option<JoinHandle<()>>,
}
impl OclManager {
pub fn launch_sevice(
permutations: &[Vec<u32>],
permutations_mask: &[u64],
n: u32,
// Workgroup size, set to 0 for max
mut wg_size: usize,
result_output: Sender<Message>,
) -> (Sender<Message>, JoinHandle<()>) {
let (h, w) = crate::solvers::wall_stats(n);
let src = include_str!("check.cl");
let platform = ocl::Platform::default();
let device = ocl::Device::first(platform).expect("failed to create opencl device");
let max_wg_size = device.max_wg_size().expect("failed to query max_wg_size");
if wg_size == 0 {
wg_size = max_wg_size;
} else if wg_size > max_wg_size {
println!("invalid workgroup size");
}
let (output_sender, output_handle) =
super::output::Output::launch_sevice(permutations, permutations_mask, result_output);
let (host_sender, host_handle) = super::host::Host::launch_sevice(
permutations_mask,
n,
h,
w,
wg_size,
src,
output_sender.clone(),
)
.unwrap();
let (sender, receiver) = channel();
println!("wg {}", wg_size);
let mut buffers = Vec::with_capacity((n - h + 1) as usize);
for _ in 0..=(n - h) {
buffers.push(RequestBuffer::new(wg_size as usize));
}
let manager = Self {
job_id: 0,
host_sender,
output_sender,
receiver,
buffers,
output_handle,
host_handle: Some(host_handle),
};
(
sender,
std::thread::Builder::new()
.name("GPU Manager Deamon".into())
.spawn(move || {
manager.run();
})
.unwrap(),
)
}
fn run(mut self) {
loop {
match self.receiver.recv().expect("Channel to GPU Manager broke") {
Message::CheckRequest(request) => {
let queue = request.queue;
//println!("num: {:?} bit {:b}", request.rows, request.bitmask);
if let Some(buffer) = self.buffers[queue as usize].read(request) {
self.host_sender
.send(Message::HostMessage((self.job_id, queue, buffer.0.into())))
.unwrap();
self.output_sender
.send(Message::OutputMessage((self.job_id, buffer.1.into())))
.unwrap();
self.job_id += 1;
}
}
Message::CpuDone => {
for (i, b) in self.buffers.iter_mut().enumerate() {
b.flush();
self.host_sender
.send(Message::HostMessage((
self.job_id,
i as u32,
b.mask_buff.clone(),
)))
.unwrap();
self.output_sender
.send(Message::OutputMessage((self.job_id, b.row_buff.clone())))
.unwrap();
self.job_id += 1;
}
println!("flushing buffers");
self.host_sender.send(Message::CpuDone);
self.host_handle.take().unwrap().join();
}
Message::Terminate => {
self.output_sender.send(Message::Terminate);
self.output_handle.join();
return;
}
_ => println!("Invalid MessageType"),
}
}
}
}
|
