Input and output formats must be the same, the overlay format must be
the same as the input except possibly with an additional alpha component.
... | ... |
@@ -3239,6 +3239,7 @@ negate_filter_deps="lut_filter" |
3239 | 3239 |
nnedi_filter_deps="gpl" |
3240 | 3240 |
ocr_filter_deps="libtesseract" |
3241 | 3241 |
ocv_filter_deps="libopencv" |
3242 |
+overlay_opencl_filter_deps="opencl" |
|
3242 | 3243 |
overlay_qsv_filter_deps="libmfx" |
3243 | 3244 |
overlay_qsv_filter_select="qsvvpp" |
3244 | 3245 |
owdenoise_filter_deps="gpl" |
... | ... |
@@ -253,6 +253,8 @@ OBJS-$(CONFIG_OCV_FILTER) += vf_libopencv.o |
253 | 253 |
OBJS-$(CONFIG_OPENCL) += deshake_opencl.o unsharp_opencl.o |
254 | 254 |
OBJS-$(CONFIG_OSCILLOSCOPE_FILTER) += vf_datascope.o |
255 | 255 |
OBJS-$(CONFIG_OVERLAY_FILTER) += vf_overlay.o framesync.o |
256 |
+OBJS-$(CONFIG_OVERLAY_OPENCL_FILTER) += vf_overlay_opencl.o opencl.o \ |
|
257 |
+ opencl/overlay.o framesync.o |
|
256 | 258 |
OBJS-$(CONFIG_OVERLAY_QSV_FILTER) += vf_overlay_qsv.o |
257 | 259 |
OBJS-$(CONFIG_OWDENOISE_FILTER) += vf_owdenoise.o |
258 | 260 |
OBJS-$(CONFIG_PAD_FILTER) += vf_pad.o |
... | ... |
@@ -262,6 +262,7 @@ static void register_all(void) |
262 | 262 |
REGISTER_FILTER(OCV, ocv, vf); |
263 | 263 |
REGISTER_FILTER(OSCILLOSCOPE, oscilloscope, vf); |
264 | 264 |
REGISTER_FILTER(OVERLAY, overlay, vf); |
265 |
+ REGISTER_FILTER(OVERLAY_OPENCL, overlay_opencl, vf); |
|
265 | 266 |
REGISTER_FILTER(OVERLAY_QSV, overlay_qsv, vf); |
266 | 267 |
REGISTER_FILTER(OWDENOISE, owdenoise, vf); |
267 | 268 |
REGISTER_FILTER(PAD, pad, vf); |
268 | 269 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,104 @@ |
0 |
+/* |
|
1 |
+ * This file is part of FFmpeg. |
|
2 |
+ * |
|
3 |
+ * FFmpeg is free software; you can redistribute it and/or |
|
4 |
+ * modify it under the terms of the GNU Lesser General Public |
|
5 |
+ * License as published by the Free Software Foundation; either |
|
6 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
7 |
+ * |
|
8 |
+ * FFmpeg is distributed in the hope that it will be useful, |
|
9 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
10 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
11 |
+ * Lesser General Public License for more details. |
|
12 |
+ * |
|
13 |
+ * You should have received a copy of the GNU Lesser General Public |
|
14 |
+ * License along with FFmpeg; if not, write to the Free Software |
|
15 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
16 |
+ */ |
|
17 |
+ |
|
18 |
+__kernel void overlay_no_alpha(__write_only image2d_t dst, |
|
19 |
+ __read_only image2d_t main, |
|
20 |
+ __read_only image2d_t overlay, |
|
21 |
+ int x_position, |
|
22 |
+ int y_position) |
|
23 |
+{ |
|
24 |
+ const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE | |
|
25 |
+ CLK_FILTER_NEAREST); |
|
26 |
+ |
|
27 |
+ int2 overlay_size = get_image_dim(overlay); |
|
28 |
+ int2 loc = (int2)(get_global_id(0), get_global_id(1)); |
|
29 |
+ |
|
30 |
+ if (loc.x < x_position || |
|
31 |
+ loc.y < y_position || |
|
32 |
+ loc.x >= overlay_size.x + x_position || |
|
33 |
+ loc.y >= overlay_size.y + y_position) { |
|
34 |
+ float4 val = read_imagef(main, sampler, loc); |
|
35 |
+ write_imagef(dst, loc, val); |
|
36 |
+ } else { |
|
37 |
+ int2 loc_overlay = (int2)(x_position, y_position); |
|
38 |
+ float4 val = read_imagef(overlay, sampler, loc - loc_overlay); |
|
39 |
+ write_imagef(dst, loc, val); |
|
40 |
+ } |
|
41 |
+} |
|
42 |
+ |
|
43 |
+__kernel void overlay_internal_alpha(__write_only image2d_t dst, |
|
44 |
+ __read_only image2d_t main, |
|
45 |
+ __read_only image2d_t overlay, |
|
46 |
+ int x_position, |
|
47 |
+ int y_position) |
|
48 |
+{ |
|
49 |
+ const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE | |
|
50 |
+ CLK_FILTER_NEAREST); |
|
51 |
+ |
|
52 |
+ int2 overlay_size = get_image_dim(overlay); |
|
53 |
+ int2 loc = (int2)(get_global_id(0), get_global_id(1)); |
|
54 |
+ |
|
55 |
+ if (loc.x < x_position || |
|
56 |
+ loc.y < y_position || |
|
57 |
+ loc.x >= overlay_size.x + x_position || |
|
58 |
+ loc.y >= overlay_size.y + y_position) { |
|
59 |
+ float4 val = read_imagef(main, sampler, loc); |
|
60 |
+ write_imagef(dst, loc, val); |
|
61 |
+ } else { |
|
62 |
+ int2 loc_overlay = (int2)(x_position, y_position); |
|
63 |
+ float4 in_main = read_imagef(main, sampler, loc); |
|
64 |
+ float4 in_overlay = read_imagef(overlay, sampler, loc - loc_overlay); |
|
65 |
+ float4 val = in_overlay * in_overlay.w + in_main * (1.0f - in_overlay.w); |
|
66 |
+ write_imagef(dst, loc, val); |
|
67 |
+ } |
|
68 |
+} |
|
69 |
+ |
|
70 |
+__kernel void overlay_external_alpha(__write_only image2d_t dst, |
|
71 |
+ __read_only image2d_t main, |
|
72 |
+ __read_only image2d_t overlay, |
|
73 |
+ __read_only image2d_t alpha, |
|
74 |
+ int x_position, |
|
75 |
+ int y_position, |
|
76 |
+ int alpha_adj_x, |
|
77 |
+ int alpha_adj_y) |
|
78 |
+{ |
|
79 |
+ const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE | |
|
80 |
+ CLK_FILTER_NEAREST); |
|
81 |
+ |
|
82 |
+ int2 overlay_size = get_image_dim(overlay); |
|
83 |
+ int2 loc = (int2)(get_global_id(0), get_global_id(1)); |
|
84 |
+ |
|
85 |
+ if (loc.x < x_position || |
|
86 |
+ loc.y < y_position || |
|
87 |
+ loc.x >= overlay_size.x + x_position || |
|
88 |
+ loc.y >= overlay_size.y + y_position) { |
|
89 |
+ float4 val = read_imagef(main, sampler, loc); |
|
90 |
+ write_imagef(dst, loc, val); |
|
91 |
+ } else { |
|
92 |
+ int2 loc_overlay = (int2)(x_position, y_position); |
|
93 |
+ float4 in_main = read_imagef(main, sampler, loc); |
|
94 |
+ float4 in_overlay = read_imagef(overlay, sampler, loc - loc_overlay); |
|
95 |
+ |
|
96 |
+ int2 loc_alpha = (int2)(loc.x * alpha_adj_x, |
|
97 |
+ loc.y * alpha_adj_y) - loc_overlay; |
|
98 |
+ float4 in_alpha = read_imagef(alpha, sampler, loc_alpha); |
|
99 |
+ |
|
100 |
+ float4 val = in_overlay * in_alpha.x + in_main * (1.0f - in_alpha.x); |
|
101 |
+ write_imagef(dst, loc, val); |
|
102 |
+ } |
|
103 |
+} |
... | ... |
@@ -31,7 +31,7 @@ |
31 | 31 |
|
32 | 32 |
#define LIBAVFILTER_VERSION_MAJOR 7 |
33 | 33 |
#define LIBAVFILTER_VERSION_MINOR 2 |
34 |
-#define LIBAVFILTER_VERSION_MICRO 100 |
|
34 |
+#define LIBAVFILTER_VERSION_MICRO 101 |
|
35 | 35 |
|
36 | 36 |
#define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \ |
37 | 37 |
LIBAVFILTER_VERSION_MINOR, \ |
38 | 38 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,360 @@ |
0 |
+/* |
|
1 |
+ * This file is part of FFmpeg. |
|
2 |
+ * |
|
3 |
+ * FFmpeg is free software; you can redistribute it and/or |
|
4 |
+ * modify it under the terms of the GNU Lesser General Public |
|
5 |
+ * License as published by the Free Software Foundation; either |
|
6 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
7 |
+ * |
|
8 |
+ * FFmpeg is distributed in the hope that it will be useful, |
|
9 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
10 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
11 |
+ * Lesser General Public License for more details. |
|
12 |
+ * |
|
13 |
+ * You should have received a copy of the GNU Lesser General Public |
|
14 |
+ * License along with FFmpeg; if not, write to the Free Software |
|
15 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
16 |
+ */ |
|
17 |
+ |
|
18 |
+#include "libavutil/avassert.h" |
|
19 |
+#include "libavutil/buffer.h" |
|
20 |
+#include "libavutil/common.h" |
|
21 |
+#include "libavutil/hwcontext.h" |
|
22 |
+#include "libavutil/hwcontext_opencl.h" |
|
23 |
+#include "libavutil/log.h" |
|
24 |
+#include "libavutil/mathematics.h" |
|
25 |
+#include "libavutil/mem.h" |
|
26 |
+#include "libavutil/pixdesc.h" |
|
27 |
+#include "libavutil/opt.h" |
|
28 |
+ |
|
29 |
+#include "avfilter.h" |
|
30 |
+#include "framesync.h" |
|
31 |
+#include "internal.h" |
|
32 |
+#include "opencl.h" |
|
33 |
+#include "opencl_source.h" |
|
34 |
+#include "video.h" |
|
35 |
+ |
|
36 |
+typedef struct OverlayOpenCLContext { |
|
37 |
+ OpenCLFilterContext ocf; |
|
38 |
+ |
|
39 |
+ int initialised; |
|
40 |
+ cl_kernel kernel; |
|
41 |
+ cl_command_queue command_queue; |
|
42 |
+ |
|
43 |
+ FFFrameSync fs; |
|
44 |
+ |
|
45 |
+ int nb_planes; |
|
46 |
+ int x_subsample; |
|
47 |
+ int y_subsample; |
|
48 |
+ int alpha_separate; |
|
49 |
+ |
|
50 |
+ int x_position; |
|
51 |
+ int y_position; |
|
52 |
+} OverlayOpenCLContext; |
|
53 |
+ |
|
54 |
+static int overlay_opencl_load(AVFilterContext *avctx, |
|
55 |
+ enum AVPixelFormat main_format, |
|
56 |
+ enum AVPixelFormat overlay_format) |
|
57 |
+{ |
|
58 |
+ OverlayOpenCLContext *ctx = avctx->priv; |
|
59 |
+ cl_int cle; |
|
60 |
+ const char *source = ff_opencl_source_overlay; |
|
61 |
+ const char *kernel; |
|
62 |
+ const AVPixFmtDescriptor *main_desc, *overlay_desc; |
|
63 |
+ int err, i, main_planes, overlay_planes; |
|
64 |
+ |
|
65 |
+ main_desc = av_pix_fmt_desc_get(main_format); |
|
66 |
+ overlay_desc = av_pix_fmt_desc_get(overlay_format); |
|
67 |
+ |
|
68 |
+ main_planes = overlay_planes = 0; |
|
69 |
+ for (i = 0; i < main_desc->nb_components; i++) |
|
70 |
+ main_planes = FFMAX(main_planes, |
|
71 |
+ main_desc->comp[i].plane + 1); |
|
72 |
+ for (i = 0; i < overlay_desc->nb_components; i++) |
|
73 |
+ overlay_planes = FFMAX(overlay_planes, |
|
74 |
+ overlay_desc->comp[i].plane + 1); |
|
75 |
+ |
|
76 |
+ ctx->nb_planes = main_planes; |
|
77 |
+ ctx->x_subsample = 1 << main_desc->log2_chroma_w; |
|
78 |
+ ctx->y_subsample = 1 << main_desc->log2_chroma_h; |
|
79 |
+ |
|
80 |
+ if (ctx->x_position % ctx->x_subsample || |
|
81 |
+ ctx->y_position % ctx->y_subsample) { |
|
82 |
+ av_log(avctx, AV_LOG_WARNING, "Warning: overlay position (%d, %d) " |
|
83 |
+ "does not match subsampling (%d, %d).\n", |
|
84 |
+ ctx->x_position, ctx->y_position, |
|
85 |
+ ctx->x_subsample, ctx->y_subsample); |
|
86 |
+ } |
|
87 |
+ |
|
88 |
+ if (main_planes == overlay_planes) { |
|
89 |
+ if (main_desc->nb_components == overlay_desc->nb_components) |
|
90 |
+ kernel = "overlay_no_alpha"; |
|
91 |
+ else |
|
92 |
+ kernel = "overlay_internal_alpha"; |
|
93 |
+ ctx->alpha_separate = 0; |
|
94 |
+ } else { |
|
95 |
+ kernel = "overlay_external_alpha"; |
|
96 |
+ ctx->alpha_separate = 1; |
|
97 |
+ } |
|
98 |
+ |
|
99 |
+ av_log(avctx, AV_LOG_DEBUG, "Using kernel %s.\n", kernel); |
|
100 |
+ |
|
101 |
+ err = ff_opencl_filter_load_program(avctx, &source, 1); |
|
102 |
+ if (err < 0) |
|
103 |
+ goto fail; |
|
104 |
+ |
|
105 |
+ ctx->command_queue = clCreateCommandQueue(ctx->ocf.hwctx->context, |
|
106 |
+ ctx->ocf.hwctx->device_id, |
|
107 |
+ 0, &cle); |
|
108 |
+ if (!ctx->command_queue) { |
|
109 |
+ av_log(avctx, AV_LOG_ERROR, "Failed to create OpenCL " |
|
110 |
+ "command queue: %d.\n", cle); |
|
111 |
+ err = AVERROR(EIO); |
|
112 |
+ goto fail; |
|
113 |
+ } |
|
114 |
+ |
|
115 |
+ ctx->kernel = clCreateKernel(ctx->ocf.program, kernel, &cle); |
|
116 |
+ if (!ctx->kernel) { |
|
117 |
+ av_log(avctx, AV_LOG_ERROR, "Failed to create kernel: %d.\n", cle); |
|
118 |
+ err = AVERROR(EIO); |
|
119 |
+ goto fail; |
|
120 |
+ } |
|
121 |
+ |
|
122 |
+ ctx->initialised = 1; |
|
123 |
+ return 0; |
|
124 |
+ |
|
125 |
+fail: |
|
126 |
+ if (ctx->command_queue) |
|
127 |
+ clReleaseCommandQueue(ctx->command_queue); |
|
128 |
+ if (ctx->kernel) |
|
129 |
+ clReleaseKernel(ctx->kernel); |
|
130 |
+ return err; |
|
131 |
+} |
|
132 |
+ |
|
133 |
+static int overlay_opencl_blend(FFFrameSync *fs) |
|
134 |
+{ |
|
135 |
+ AVFilterContext *avctx = fs->parent; |
|
136 |
+ AVFilterLink *outlink = avctx->outputs[0]; |
|
137 |
+ OverlayOpenCLContext *ctx = avctx->priv; |
|
138 |
+ AVFrame *input_main, *input_overlay; |
|
139 |
+ AVFrame *output; |
|
140 |
+ cl_mem mem; |
|
141 |
+ cl_int cle, x, y; |
|
142 |
+ size_t global_work[2]; |
|
143 |
+ int kernel_arg = 0; |
|
144 |
+ int err, plane; |
|
145 |
+ |
|
146 |
+ err = ff_framesync_get_frame(fs, 0, &input_main, 0); |
|
147 |
+ if (err < 0) |
|
148 |
+ return err; |
|
149 |
+ err = ff_framesync_get_frame(fs, 1, &input_overlay, 0); |
|
150 |
+ if (err < 0) |
|
151 |
+ return err; |
|
152 |
+ |
|
153 |
+ if (!ctx->initialised) { |
|
154 |
+ AVHWFramesContext *main_fc = |
|
155 |
+ (AVHWFramesContext*)input_main->hw_frames_ctx->data; |
|
156 |
+ AVHWFramesContext *overlay_fc = |
|
157 |
+ (AVHWFramesContext*)input_overlay->hw_frames_ctx->data; |
|
158 |
+ |
|
159 |
+ err = overlay_opencl_load(avctx, main_fc->sw_format, |
|
160 |
+ overlay_fc->sw_format); |
|
161 |
+ if (err < 0) |
|
162 |
+ return err; |
|
163 |
+ } |
|
164 |
+ |
|
165 |
+ output = ff_get_video_buffer(outlink, outlink->w, outlink->h); |
|
166 |
+ if (!output) { |
|
167 |
+ err = AVERROR(ENOMEM); |
|
168 |
+ goto fail; |
|
169 |
+ } |
|
170 |
+ |
|
171 |
+ for (plane = 0; plane < ctx->nb_planes; plane++) { |
|
172 |
+ kernel_arg = 0; |
|
173 |
+ |
|
174 |
+ mem = (cl_mem)output->data[plane]; |
|
175 |
+ cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_mem), &mem); |
|
176 |
+ if (cle != CL_SUCCESS) |
|
177 |
+ goto fail_kernel_arg; |
|
178 |
+ |
|
179 |
+ mem = (cl_mem)input_main->data[plane]; |
|
180 |
+ cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_mem), &mem); |
|
181 |
+ if (cle != CL_SUCCESS) |
|
182 |
+ goto fail_kernel_arg; |
|
183 |
+ |
|
184 |
+ mem = (cl_mem)input_overlay->data[plane]; |
|
185 |
+ cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_mem), &mem); |
|
186 |
+ if (cle != CL_SUCCESS) |
|
187 |
+ goto fail_kernel_arg; |
|
188 |
+ |
|
189 |
+ if (ctx->alpha_separate) { |
|
190 |
+ mem = (cl_mem)input_overlay->data[ctx->nb_planes]; |
|
191 |
+ cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_mem), &mem); |
|
192 |
+ if (cle != CL_SUCCESS) |
|
193 |
+ goto fail_kernel_arg; |
|
194 |
+ } |
|
195 |
+ |
|
196 |
+ x = ctx->x_position / (plane == 0 ? 1 : ctx->x_subsample); |
|
197 |
+ y = ctx->y_position / (plane == 0 ? 1 : ctx->y_subsample); |
|
198 |
+ |
|
199 |
+ cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_int), &x); |
|
200 |
+ if (cle != CL_SUCCESS) |
|
201 |
+ goto fail_kernel_arg; |
|
202 |
+ cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_int), &y); |
|
203 |
+ if (cle != CL_SUCCESS) |
|
204 |
+ goto fail_kernel_arg; |
|
205 |
+ |
|
206 |
+ if (ctx->alpha_separate) { |
|
207 |
+ cl_int alpha_adj_x = plane == 0 ? 1 : ctx->x_subsample; |
|
208 |
+ cl_int alpha_adj_y = plane == 0 ? 1 : ctx->y_subsample; |
|
209 |
+ |
|
210 |
+ cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_int), &alpha_adj_x); |
|
211 |
+ if (cle != CL_SUCCESS) |
|
212 |
+ goto fail_kernel_arg; |
|
213 |
+ cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_int), &alpha_adj_y); |
|
214 |
+ if (cle != CL_SUCCESS) |
|
215 |
+ goto fail_kernel_arg; |
|
216 |
+ } |
|
217 |
+ |
|
218 |
+ global_work[0] = output->width; |
|
219 |
+ global_work[1] = output->height; |
|
220 |
+ |
|
221 |
+ cle = clEnqueueNDRangeKernel(ctx->command_queue, ctx->kernel, 2, NULL, |
|
222 |
+ global_work, NULL, 0, NULL, NULL); |
|
223 |
+ if (cle != CL_SUCCESS) { |
|
224 |
+ av_log(avctx, AV_LOG_ERROR, "Failed to enqueue " |
|
225 |
+ "overlay kernel for plane %d: %d.\n", cle, plane); |
|
226 |
+ err = AVERROR(EIO); |
|
227 |
+ goto fail; |
|
228 |
+ } |
|
229 |
+ } |
|
230 |
+ |
|
231 |
+ cle = clFinish(ctx->command_queue); |
|
232 |
+ if (cle != CL_SUCCESS) { |
|
233 |
+ av_log(avctx, AV_LOG_ERROR, "Failed to finish " |
|
234 |
+ "command queue: %d.\n", cle); |
|
235 |
+ err = AVERROR(EIO); |
|
236 |
+ goto fail; |
|
237 |
+ } |
|
238 |
+ |
|
239 |
+ err = av_frame_copy_props(output, input_main); |
|
240 |
+ |
|
241 |
+ av_log(avctx, AV_LOG_DEBUG, "Filter output: %s, %ux%u (%"PRId64").\n", |
|
242 |
+ av_get_pix_fmt_name(output->format), |
|
243 |
+ output->width, output->height, output->pts); |
|
244 |
+ |
|
245 |
+ return ff_filter_frame(outlink, output); |
|
246 |
+ |
|
247 |
+fail_kernel_arg: |
|
248 |
+ av_log(avctx, AV_LOG_ERROR, "Failed to set kernel arg %d: %d.\n", |
|
249 |
+ kernel_arg, cle); |
|
250 |
+ err = AVERROR(EIO); |
|
251 |
+fail: |
|
252 |
+ return err; |
|
253 |
+} |
|
254 |
+ |
|
255 |
+static int overlay_opencl_config_output(AVFilterLink *outlink) |
|
256 |
+{ |
|
257 |
+ AVFilterContext *avctx = outlink->src; |
|
258 |
+ OverlayOpenCLContext *ctx = avctx->priv; |
|
259 |
+ int err; |
|
260 |
+ |
|
261 |
+ err = ff_opencl_filter_config_output(outlink); |
|
262 |
+ if (err < 0) |
|
263 |
+ return err; |
|
264 |
+ |
|
265 |
+ err = ff_framesync_init_dualinput(&ctx->fs, avctx); |
|
266 |
+ if (err < 0) |
|
267 |
+ return err; |
|
268 |
+ |
|
269 |
+ return ff_framesync_configure(&ctx->fs); |
|
270 |
+} |
|
271 |
+ |
|
272 |
+static av_cold int overlay_opencl_init(AVFilterContext *avctx) |
|
273 |
+{ |
|
274 |
+ OverlayOpenCLContext *ctx = avctx->priv; |
|
275 |
+ |
|
276 |
+ ctx->fs.on_event = &overlay_opencl_blend; |
|
277 |
+ |
|
278 |
+ return ff_opencl_filter_init(avctx); |
|
279 |
+} |
|
280 |
+ |
|
281 |
+static int overlay_opencl_activate(AVFilterContext *avctx) |
|
282 |
+{ |
|
283 |
+ OverlayOpenCLContext *ctx = avctx->priv; |
|
284 |
+ |
|
285 |
+ return ff_framesync_activate(&ctx->fs); |
|
286 |
+} |
|
287 |
+ |
|
288 |
+static av_cold void overlay_opencl_uninit(AVFilterContext *avctx) |
|
289 |
+{ |
|
290 |
+ OverlayOpenCLContext *ctx = avctx->priv; |
|
291 |
+ cl_int cle; |
|
292 |
+ |
|
293 |
+ if (ctx->kernel) { |
|
294 |
+ cle = clReleaseKernel(ctx->kernel); |
|
295 |
+ if (cle != CL_SUCCESS) |
|
296 |
+ av_log(avctx, AV_LOG_ERROR, "Failed to release " |
|
297 |
+ "kernel: %d.\n", cle); |
|
298 |
+ } |
|
299 |
+ |
|
300 |
+ if (ctx->command_queue) { |
|
301 |
+ cle = clReleaseCommandQueue(ctx->command_queue); |
|
302 |
+ if (cle != CL_SUCCESS) |
|
303 |
+ av_log(avctx, AV_LOG_ERROR, "Failed to release " |
|
304 |
+ "command queue: %d.\n", cle); |
|
305 |
+ } |
|
306 |
+ |
|
307 |
+ ff_opencl_filter_uninit(avctx); |
|
308 |
+ |
|
309 |
+ ff_framesync_uninit(&ctx->fs); |
|
310 |
+} |
|
311 |
+ |
|
312 |
+#define OFFSET(x) offsetof(OverlayOpenCLContext, x) |
|
313 |
+#define FLAGS (AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_VIDEO_PARAM) |
|
314 |
+static const AVOption overlay_opencl_options[] = { |
|
315 |
+ { "x", "Overlay x position", |
|
316 |
+ OFFSET(x_position), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, .flags = FLAGS }, |
|
317 |
+ { "y", "Overlay y position", |
|
318 |
+ OFFSET(y_position), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, .flags = FLAGS }, |
|
319 |
+ { NULL }, |
|
320 |
+}; |
|
321 |
+ |
|
322 |
+AVFILTER_DEFINE_CLASS(overlay_opencl); |
|
323 |
+ |
|
324 |
+static const AVFilterPad overlay_opencl_inputs[] = { |
|
325 |
+ { |
|
326 |
+ .name = "main", |
|
327 |
+ .type = AVMEDIA_TYPE_VIDEO, |
|
328 |
+ .config_props = &ff_opencl_filter_config_input, |
|
329 |
+ }, |
|
330 |
+ { |
|
331 |
+ .name = "overlay", |
|
332 |
+ .type = AVMEDIA_TYPE_VIDEO, |
|
333 |
+ .config_props = &ff_opencl_filter_config_input, |
|
334 |
+ }, |
|
335 |
+ { NULL } |
|
336 |
+}; |
|
337 |
+ |
|
338 |
+static const AVFilterPad overlay_opencl_outputs[] = { |
|
339 |
+ { |
|
340 |
+ .name = "default", |
|
341 |
+ .type = AVMEDIA_TYPE_VIDEO, |
|
342 |
+ .config_props = &overlay_opencl_config_output, |
|
343 |
+ }, |
|
344 |
+ { NULL } |
|
345 |
+}; |
|
346 |
+ |
|
347 |
+AVFilter ff_vf_overlay_opencl = { |
|
348 |
+ .name = "overlay_opencl", |
|
349 |
+ .description = NULL_IF_CONFIG_SMALL("Overlay one video on top of another"), |
|
350 |
+ .priv_size = sizeof(OverlayOpenCLContext), |
|
351 |
+ .priv_class = &overlay_opencl_class, |
|
352 |
+ .init = &overlay_opencl_init, |
|
353 |
+ .uninit = &overlay_opencl_uninit, |
|
354 |
+ .query_formats = &ff_opencl_filter_query_formats, |
|
355 |
+ .activate = &overlay_opencl_activate, |
|
356 |
+ .inputs = overlay_opencl_inputs, |
|
357 |
+ .outputs = overlay_opencl_outputs, |
|
358 |
+ .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, |
|
359 |
+}; |