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
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
|
/* oclnv12toyuv.c
Copyright (c) 2003-2012 HandBrake Team
This file is part of the HandBrake source code
Homepage: <http://handbrake.fr/>.
It may be used under the terms of the GNU General Public License v2.
For full terms see the file COPYING file or visit http://www.gnu.org/licenses/gpl-2.0.html
Authors: Peng Gao <peng@multicorewareinc.com> <http://www.multicorewareinc.com/>
Li Cao <li@multicorewareinc.com> <http://www.multicorewareinc.com/>
*/
#ifdef USE_OPENCL
#ifdef USE_HWD
#include "vadxva2.h"
#include "oclnv12toyuv.h"
/**
* It creates are opencl bufs w is input frame width, h is input frame height
*/
static int hb_nv12toyuv_create_cl_buf( KernelEnv *kenv, int w, int h, hb_va_dxva2_t *dxva2 );
/**
* It creates are opencl kernel. kernel name is nv12toyuv
*/
static int hb_nv12toyuv_create_cl_kernel( KernelEnv *kenv, hb_va_dxva2_t *dxva2 );
/**
* It set opencl arg, input data,output data, input width, output height
*/
static int hb_nv12toyuv_setkernelarg( KernelEnv *kenv, int w, int h, hb_va_dxva2_t *dxva2 );
/**
* It initialize nv12 to yuv kernel.
*/
static int hb_init_nv12toyuv_ocl( KernelEnv *kenv, int w, int h, hb_va_dxva2_t *dxva2 );
/**
* Run nv12 to yuv kernel.
*/
static int hb_nv12toyuv( void **userdata, KernelEnv *kenv );
/**
* register nv12 to yuv kernel.
*/
static int hb_nv12toyuv_reg_kernel( void );
/**
* It creates are opencl bufs w is input frame width, h is input frame height
*/
static int hb_nv12toyuv_create_cl_buf( KernelEnv *kenv, int w, int h, hb_va_dxva2_t *dxva2 )
{
cl_int status = CL_SUCCESS;
int in_bytes = w*h*3/2;
CREATEBUF( dxva2->cl_mem_nv12, CL_MEM_READ_ONLY | CL_MEM_ALLOC_HOST_PTR, in_bytes );
CREATEBUF( dxva2->cl_mem_yuv, CL_MEM_READ_WRITE | CL_MEM_ALLOC_HOST_PTR, in_bytes );
return 0;
}
/**
* It creates are opencl kernel. kernel name is nv12toyuv
*/
static int hb_nv12toyuv_create_cl_kernel( KernelEnv *kenv, hb_va_dxva2_t *dxva2 )
{
int ret;
dxva2->nv12toyuv = clCreateKernel( kenv->program, "nv12toyuv", &ret );
return ret;
}
/**
* It set opencl arg, input data,output data, input width, output height
*/
static int hb_nv12toyuv_setkernelarg( KernelEnv *kenv, int w, int h, hb_va_dxva2_t *dxva2 )
{
int arg = 0, status;
kenv->kernel = dxva2->nv12toyuv;
OCLCHECK( clSetKernelArg, kenv->kernel, arg++, sizeof(cl_mem), &dxva2->cl_mem_nv12 );
OCLCHECK( clSetKernelArg, kenv->kernel, arg++, sizeof(cl_mem), &dxva2->cl_mem_yuv );
OCLCHECK( clSetKernelArg, kenv->kernel, arg++, sizeof(int), &w );
OCLCHECK( clSetKernelArg, kenv->kernel, arg++, sizeof(int), &h );
return 0;
}
/**
* It initialize nv12 to yuv kernel.
*/
static int hb_init_nv12toyuv_ocl( KernelEnv *kenv, int w, int h, hb_va_dxva2_t *dxva2 )
{
if( !dxva2->nv12toyuv )
{
if( hb_nv12toyuv_create_cl_buf( kenv, w, h, dxva2 ) )
{
hb_log( "nv12toyuv_create_cl_buf fial" );
return -1;
}
if (!dxva2->nv12toyuv_tmp_in)
dxva2->nv12toyuv_tmp_in = malloc (w*h*3/2);
if (!dxva2->nv12toyuv_tmp_out)
dxva2->nv12toyuv_tmp_out = malloc (w*h*3/2);
hb_nv12toyuv_create_cl_kernel( kenv, dxva2 );
}
return 0;
}
/**
* Run nv12 to yuv kernel.
*/
static int hb_nv12toyuv( void **userdata, KernelEnv *kenv )
{
int status;
int w = (int)userdata[0];
int h = (int)userdata[1];
uint8_t *bufi1 = userdata[2];
int *crop = userdata[3];
hb_va_dxva2_t *dxva2 = userdata[4];
uint8_t *bufi2 = userdata[5];
int p = (int)userdata[6];
int i;
if( hb_init_nv12toyuv_ocl( kenv, w, h, dxva2 ) )
return -1;
if( hb_nv12toyuv_setkernelarg( kenv, w, h, dxva2 ) )
return -1;
int in_bytes = w*h*3/2;
if( kenv->isAMD )
{
void *data = clEnqueueMapBuffer( kenv->command_queue, dxva2->cl_mem_nv12, CL_MAP_WRITE_INVALIDATE_REGION, CL_TRUE, 0, in_bytes, 0, NULL, NULL, NULL );
//memcpy( data, bufi, in_bytes );
for ( i = 0; i < dxva2->height; i++ )
{
memcpy( data+i*dxva2->width, bufi1+i*p, dxva2->width );
if ( i<dxva2->height>>1 )
memcpy( data+(dxva2->width*dxva2->height)+i*dxva2->width, bufi2+i*p, dxva2->width );
}
clEnqueueUnmapMemObject( kenv->command_queue, dxva2->cl_mem_nv12, data, 0, NULL, NULL );
}
else
{
uint8_t *tmp = (uint8_t*)malloc( dxva2->width*dxva2->height*3/2 );
for( i = 0; i < dxva2->height; i++ )
{
memcpy( tmp+i*dxva2->width, bufi1+i*p, dxva2->width );
if( i<dxva2->height>>1 )
memcpy( tmp+(dxva2->width*dxva2->height)+i*dxva2->width, bufi2+i*p, dxva2->width );
}
OCLCHECK( clEnqueueWriteBuffer, kenv->command_queue, dxva2->cl_mem_nv12, CL_TRUE, 0, in_bytes, tmp, 0, NULL, NULL );
free( tmp );
}
size_t gdim[2] = {w>>1, h>>1};
OCLCHECK( clEnqueueNDRangeKernel, kenv->command_queue, kenv->kernel, 2, NULL, gdim, NULL, 0, NULL, NULL );
if( crop[0] || crop[1] || crop[2] || crop[3] )
{
AVPicture pic_in;
AVPicture pic_crop;
clEnqueueReadBuffer( kenv->command_queue, dxva2->cl_mem_yuv, CL_TRUE, 0, in_bytes, dxva2->nv12toyuv_tmp_out, 0, NULL, NULL );
hb_buffer_t *in = hb_video_buffer_init( w, h );
memcpy( in->plane[0].data, dxva2->nv12toyuv_tmp_out, w * h );
memcpy( in->plane[1].data, dxva2->nv12toyuv_tmp_out + w * h, ( w * h )>>2 );
memcpy( in->plane[2].data, dxva2->nv12toyuv_tmp_out + w * h + ( ( w * h )>>2 ), ( w * h )>>2 );
hb_avpicture_fill( &pic_in, in );
av_picture_crop( &pic_crop, &pic_in, in->f.fmt, crop[0], crop[2] );
int i, ww = w - ( crop[2] + crop[3] ), hh = h - ( crop[0] + crop[1] );
for( i = 0; i< hh >> 1; i++ )
{
memcpy( dxva2->nv12toyuv_tmp_in + ( ( i<<1 ) + 0 ) * ww, pic_crop.data[0]+ ( ( i<<1 ) + 0 ) * pic_crop.linesize[0], ww );
memcpy( dxva2->nv12toyuv_tmp_in + ( ( i<<1 ) + 1 ) * ww, pic_crop.data[0]+ ( ( i<<1 ) + 1 ) * pic_crop.linesize[0], ww );
memcpy( dxva2->nv12toyuv_tmp_in + ( ww * hh ) + i * ( ww>>1 ), pic_crop.data[1] + i * pic_crop.linesize[1], ww >> 1 );
memcpy( dxva2->nv12toyuv_tmp_in + ( ww * hh ) + ( ( ww * hh )>>2 ) + i * ( ww>>1 ), pic_crop.data[2] + i * pic_crop.linesize[2], ww >> 1 );
}
if( kenv->isAMD )
{
void *data = clEnqueueMapBuffer( kenv->command_queue, dxva2->cl_mem_yuv, CL_MAP_WRITE_INVALIDATE_REGION, CL_TRUE, 0, ww * hh * 3 / 2, 0, NULL, NULL, NULL );
memcpy( data, dxva2->nv12toyuv_tmp_in, ww * hh * 3 / 2 );
clEnqueueUnmapMemObject( kenv->command_queue, dxva2->cl_mem_yuv, data, 0, NULL, NULL );
}
else
{
OCLCHECK( clEnqueueWriteBuffer, kenv->command_queue, dxva2->cl_mem_yuv, CL_TRUE, 0, in_bytes, dxva2->nv12toyuv_tmp_in, 0, NULL, NULL );
}
hb_buffer_close( &in );
}
return 0;
}
/**
* register nv12 to yuv kernel.
*/
static int hb_nv12toyuv_reg_kernel( void )
{
int st = hb_register_kernel_wrapper( "nv12toyuv", hb_nv12toyuv );
if( !st )
{
hb_log( "register kernel[%s] faild\n", "nv12toyuv" );
return -1;
}
return 0;
}
/**
* nv12 to yuv interface
* bufi is input frame of nv12, w is input frame width, h is input frame height
*/
int hb_ocl_nv12toyuv( uint8_t *bufi[], int p, int w, int h, int *crop, hb_va_dxva2_t *dxva2 )
//int hb_ocl_nv12toyuv( uint8_t *bufi, int w, int h, int *crop, hb_va_dxva2_t *dxva2 )
{
void *userdata[7];
userdata[0] = (void*)w;
userdata[1] = (void*)h;
userdata[2] = bufi[0];
userdata[3] = crop;
userdata[4] = dxva2;
userdata[5] = bufi[1];
userdata[6] = (void*)p;
if( hb_nv12toyuv_reg_kernel() )
return -1;
if( hb_run_kernel( "nv12toyuv", userdata ) )
{
printf( "run kernel[nv12toyuv] faild\n" );
return -1;
}
return 0;
}
#endif
#endif
|
