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
extern crate ocl;
mod kernel;
use ocl::{Platform, Device, ProQue, Buffer};
use ocl::core;
use ocl::core::{DeviceInfo};
pub struct Context{
pub compute_units: u32,
pro_que: ocl::ProQue,
}
pub struct Matrix {
pub rows: usize,
pub cols: usize,
pub data: Vec<f32>,
}
impl Context {
pub fn mul_matrix_scalar(&mut self, matrix: &Matrix, scalar: f32) -> Matrix {
let ref mut ocl_pq = self.pro_que;
ocl_pq.set_dims([matrix.rows*matrix.cols]);
let source_buffer = Buffer::new(
&ocl_pq.queue().clone(),
Some(core::MEM_READ_WRITE | core::MEM_COPY_HOST_PTR),
ocl_pq.dims().clone(),
Some(&matrix.data)).unwrap();
let mut result = vec![0.0f32; matrix.cols*matrix.rows];
let result_buffer: Buffer<f32> = ocl_pq.create_buffer().unwrap();
let kernel = ocl_pq.create_kernel("mul_matrix_scalar").unwrap()
.arg_scl(scalar)
.arg_buf(&source_buffer)
.arg_buf(&result_buffer);
kernel.enq().unwrap();
result_buffer.read(&mut result).enq().unwrap();
Matrix{rows: matrix.rows,cols: matrix.cols, data: result}
}
pub fn mul_matrix_matrix(&mut self, matrix_a: &Matrix, matrix_b: &Matrix) -> Matrix {
let ref mut ocl_pq = self.pro_que;
ocl_pq.set_dims([matrix_a.rows,matrix_a.cols]);
let matrix_a_buffer = Buffer::new(
&ocl_pq.queue().clone(),
Some(core::MEM_READ_WRITE | core::MEM_COPY_HOST_PTR),
ocl_pq.dims().clone(),
Some(&matrix_a.data)).unwrap();
ocl_pq.set_dims([matrix_b.rows,matrix_b.cols]);
let matrix_b_buffer = Buffer::new(
&ocl_pq.queue().clone(),
Some(core::MEM_READ_WRITE | core::MEM_COPY_HOST_PTR),
ocl_pq.dims().clone(),
Some(&matrix_b.data)).unwrap();
ocl_pq.set_dims([matrix_a.rows,matrix_b.cols]);
let mut result = vec![0.0f32; matrix_a.rows*matrix_b.cols];
let result_buffer: Buffer<f32> = ocl_pq.create_buffer().unwrap();
let kernel = ocl_pq.create_kernel("mul_matrix_matrix").unwrap()
.arg_buf(&matrix_a_buffer)
.arg_buf(&matrix_b_buffer)
.arg_buf(&result_buffer)
.arg_scl(matrix_a.rows as i32)
.arg_scl(matrix_a.cols as i32)
.arg_scl(matrix_b.cols as i32);
kernel.enq().unwrap();
result_buffer.read(&mut result).enq().unwrap();
Matrix{rows: matrix_a.rows, cols: matrix_b.cols, data: result}
}
}
pub fn new() -> Option<Context> {
let mut compute_units = 0;
let mut ocl_device = None;
let platforms = Platform::list();
for p_idx in 0..platforms.len() {
let platform = &platforms[p_idx];
let devices = Device::list_all(platform);
for d_idx in 0..devices.len() {
let device = devices[d_idx];
let deviceinforesult = core::get_device_info(&device, DeviceInfo::MaxComputeUnits);
let units = deviceinforesult.to_string().parse().unwrap();
if units > compute_units {
ocl_device = Some(device);
compute_units = units;
}
}
}
if compute_units == 0 {
return None
}
let que = ProQue::builder()
.device(ocl_device.unwrap())
.src(kernel::OCL_KERNEL)
.build().expect("Build ProQue");
Some(Context{
compute_units: compute_units,
pro_que : que,
})
}
#[test]
fn single_test() {
let eps = 1.0e-6;
let mut c = new().unwrap();
let m0 = Matrix{rows: 2, cols: 2, data: vec![1.0, 2.0, 3.0, 4.0]};
let m1 = Matrix{rows: 2, cols: 1, data: vec![4.0, 5.0]};
let m = c.mul_matrix_matrix(&m0,&m1);
assert!((m.data[0] - 14.0f32).abs() < eps);
assert!((m.data[1] - 32.0f32).abs() < eps);
let m = c.mul_matrix_scalar(&m0,1.5);
assert!( (m.data[0] - 1.5f32) < eps);
assert!( (m.data[1] - 3.0f32) < eps);
assert!( (m.data[2] - 4.5f32) < eps);
assert!( (m.data[3] - 6.0f32) < eps);
}