forked from Dadido3/blackcl
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathvector.go
123 lines (114 loc) · 3.39 KB
/
vector.go
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
package highCL
import (
"errors"
pure "github.com/opencl-pure/pureCL"
"reflect"
"unsafe"
)
// Vector is a memory buffer on device that holds []float32
type Vector struct {
buf *buffer
iSize, len int
typ reflect.Type
}
// Length the length of the vector
func (v *Vector) Length() int {
return v.len
}
// Release releases the buffer on the device
func (v *Vector) Release() error {
return v.buf.Release()
}
// NewVector want slice or array to create opencl vector in gpu
// I highly recommend primitive types such as int, uint, float32, uint8, ...,
// but you are free to experiment with GO structs, but you must keep in mind,
// that is there no guarantee how OpenCL will pass them
func (d *Device) NewVector(data interface{}) (*Vector, error) {
dataType := reflect.TypeOf(data)
if dataType.Kind() != reflect.Slice && dataType.Kind() != reflect.Array {
return nil, errors.New("data must be slice")
}
slice := reflect.ValueOf(data)
sliceLen := slice.Len()
if sliceLen == 0 {
return nil, errors.New("slice must have at least 1 item")
}
iSize := int(dataType.Elem().Size())
size := sliceLen * iSize
buf, err := newBuffer(d, size)
if err != nil {
return nil, err
}
v := &Vector{buf: buf, iSize: iSize, len: sliceLen, typ: dataType}
err = <-v.buf.copy(size, unsafe.Pointer(slice.Pointer()))
if err != nil {
_ = v.Release()
return nil, err
}
return v, nil
}
// Reset want equal data as NewVector was given (slice or array), it must have equal length as vector
// it is usefully for recall kernel with others data without locate new vector
// it's a non-blocking call, channel will return an error or nil if the data transfer is complete
func (v *Vector) Reset(data interface{}) <-chan error {
dataType := reflect.TypeOf(data)
if dataType != v.typ {
ch := make(chan error, 1)
ch <- errors.New("data must be slice equal type as been created")
return ch
}
slice := reflect.ValueOf(data)
l := slice.Len()
if v.Length() != l {
ch := make(chan error, 1)
ch <- errors.New("vector length not equal to data length")
return ch
}
return v.buf.copy(l*v.iSize, unsafe.Pointer(slice.Pointer()))
}
// Data gets data *reflect.Value in from device, it's a blocking call
// use v, err := Data(); elen := any(retrievedData.Index(i).Float()) ...
func (v *Vector) Data() (*reflect.Value, error) {
data := reflect.MakeSlice(v.typ, v.len, v.len)
err := pure.StatusToErr(pure.EnqueueReadBuffer(
v.buf.device.queue,
v.buf.memobj,
true,
0,
v.buf.size,
unsafe.Pointer(data.Pointer()),
0,
nil,
nil,
))
if err != nil {
return nil, err
}
return &data, nil
}
// DataArray gets data *reflect.Value in from device, it's a blocking call
// use v, err := DataArray(); array := *(v.Interface().(*[16]float32))
func (v *Vector) DataArray() (*reflect.Value, error) {
data := reflect.New(reflect.ArrayOf(v.len, v.typ.Elem()))
err := pure.StatusToErr(pure.EnqueueReadBuffer(
v.buf.device.queue,
v.buf.memobj,
true,
0,
v.buf.size,
unsafe.Pointer(data.Pointer()),
0,
nil,
nil,
))
if err != nil {
return nil, err
}
return &data, nil
}
// Map applies an map kernel on all elements of the vector
// It's a non-blocking call, so it can return an event object that you can wait on.
// The caller is responsible to release the returned event when it's not used anymore.
func (v *Vector) Map(k *Kernel, waitEvents []*Event) (*Event, error) {
return k.Global(v.Length()).Local(1).Run(waitEvents, v)
}