mpv/filters/f_hwtransfer.c

425 lines
12 KiB
C
Raw Normal View History

video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
#include <libavutil/buffer.h>
#include <libavutil/hwcontext.h>
#include <libavutil/mem.h>
#include "video/fmt-conversion.h"
#include "video/hwdec.h"
#include "video/mp_image.h"
#include "video/mp_image_pool.h"
#include "f_hwtransfer.h"
#include "filter_internal.h"
struct priv {
AVBufferRef *av_device_ctx;
AVBufferRef *hw_pool;
int last_input_fmt;
int last_upload_fmt;
int last_sw_fmt;
// Hardware wrapper format, e.g. IMGFMT_VAAPI.
int hw_imgfmt;
// List of supported underlying surface formats.
int *fmts;
int num_fmts;
// List of supported upload image formats. May contain duplicate entries
// (which should be ignored).
int *upload_fmts;
int num_upload_fmts;
// For fmts[n], fmt_upload_index[n] gives the index of the first supported
// upload format in upload_fmts[], and fmt_upload_num[n] gives the number
// of formats at this position.
int *fmt_upload_index;
int *fmt_upload_num;
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
struct mp_hwupload public;
};
struct ffmpeg_and_other_bugs {
int imgfmt; // hw format
const int *const whitelist_formats; // if non-NULL, allow only these
// sw formats
bool force_same_upload_fmt; // force upload fmt == sw fmt
};
// This garbage is so complex and buggy. Hardcoding knowledge makes it work,
// trying to use the dynamic information returned by the API does not. So fuck
// this shit, I'll just whitelist the cases that work, what the fuck.
static const struct ffmpeg_and_other_bugs shitlist[] = {
{
.imgfmt = IMGFMT_VAAPI,
.whitelist_formats = (const int[]){IMGFMT_NV12, IMGFMT_P010, IMGFMT_BGRA,
IMGFMT_ABGR, IMGFMT_RGB0, 0},
.force_same_upload_fmt = true,
},
{0}
};
static bool select_format(struct priv *p, int input_fmt, int *out_sw_fmt,
int *out_upload_fmt)
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
{
if (!input_fmt)
return false;
// First find the closest sw fmt. Some hwdec APIs return crazy lists of
// "supported" formats, which then are not supported or crash (???), so
// the this is a good way to avoid problems.
// (Actually we should just have hardcoded everything instead of relying on
// this fragile bullshit FFmpeg API and the fragile bullshit hwdec drivers.)
int sw_fmt = mp_imgfmt_select_best_list(p->fmts, p->num_fmts, input_fmt);
if (!sw_fmt)
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
return false;
// Dumb, but find index for p->fmts[index]==sw_fmt.
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
int index = -1;
for (int n = 0; n < p->num_fmts; n++) {
if (p->fmts[n] == sw_fmt)
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
index = n;
}
if (index < 0)
return false;
// Now check the available upload formats. This is the format our sw frame
// has to be in, and which the upload API will take (probably).
int *upload_fmts = &p->upload_fmts[p->fmt_upload_index[index]];
int num_upload_fmts = p->fmt_upload_num[index];
int up_fmt = mp_imgfmt_select_best_list(upload_fmts, num_upload_fmts,
input_fmt);
if (!up_fmt)
return false;
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
*out_sw_fmt = sw_fmt;
*out_upload_fmt = up_fmt;
return true;
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
}
int mp_hwupload_find_upload_format(struct mp_hwupload *u, int imgfmt)
{
struct priv *p = u->f->priv;
int sw = 0, up = 0;
select_format(p, imgfmt, &sw, &up);
return up;
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
}
static void process(struct mp_filter *f)
{
struct priv *p = f->priv;
if (!mp_pin_can_transfer_data(f->ppins[1], f->ppins[0]))
return;
struct mp_frame frame = mp_pin_out_read(f->ppins[0]);
if (mp_frame_is_signaling(frame)) {
mp_pin_in_write(f->ppins[1], frame);
return;
}
if (frame.type != MP_FRAME_VIDEO) {
MP_ERR(f, "unsupported frame type\n");
goto error;
}
struct mp_image *src = frame.data;
// As documented, just pass though HW frames.
if (IMGFMT_IS_HWACCEL(src->imgfmt)) {
mp_pin_in_write(f->ppins[1], frame);
return;
}
if (src->w % 2 || src->h % 2) {
MP_ERR(f, "non-mod 2 input frames unsupported\n");
goto error;
}
if (src->imgfmt != p->last_input_fmt) {
if (!select_format(p, src->imgfmt, &p->last_sw_fmt, &p->last_upload_fmt))
{
MP_ERR(f, "no hw upload format found\n");
goto error;
}
if (src->imgfmt != p->last_upload_fmt) {
// Should not fail; if it does, mp_hwupload_find_upload_format()
// does not return the src->imgfmt format.
MP_ERR(f, "input format not an upload format\n");
goto error;
}
p->last_input_fmt = src->imgfmt;
MP_INFO(f, "upload %s -> %s[%s]\n",
mp_imgfmt_to_name(p->last_input_fmt),
mp_imgfmt_to_name(p->hw_imgfmt),
mp_imgfmt_to_name(p->last_sw_fmt));
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
}
if (!mp_update_av_hw_frames_pool(&p->hw_pool, p->av_device_ctx, p->hw_imgfmt,
p->last_sw_fmt, src->w, src->h))
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
{
MP_ERR(f, "failed to create frame pool\n");
goto error;
}
struct mp_image *dst = mp_av_pool_image_hw_upload(p->hw_pool, src);
if (!dst)
goto error;
mp_frame_unref(&frame);
mp_pin_in_write(f->ppins[1], MAKE_FRAME(MP_FRAME_VIDEO, dst));
return;
error:
mp_frame_unref(&frame);
MP_ERR(f, "failed to upload frame\n");
mp_filter_internal_mark_failed(f);
}
static void destroy(struct mp_filter *f)
{
struct priv *p = f->priv;
av_buffer_unref(&p->hw_pool);
av_buffer_unref(&p->av_device_ctx);
}
static const struct mp_filter_info hwupload_filter = {
.name = "hwupload",
.priv_size = sizeof(struct priv),
.process = process,
.destroy = destroy,
};
// The VO layer might have restricted format support. It might actually
// work if this is input to a conversion filter anyway, but our format
// negotiation is too stupid and non-existent to detect this.
// So filter out all not explicitly supported formats.
static bool vo_supports(struct mp_hwdec_ctx *ctx, int hw_fmt, int sw_fmt)
{
if (!ctx->hw_imgfmt)
return true; // if unset, all formats are allowed
if (ctx->hw_imgfmt != hw_fmt)
return false;
for (int i = 0; ctx->supported_formats && ctx->supported_formats[i]; i++) {
if (ctx->supported_formats[i] == sw_fmt)
return true;
}
return false;
}
static bool probe_formats(struct mp_hwupload *u, int hw_imgfmt)
{
struct priv *p = u->f->priv;
p->hw_imgfmt = hw_imgfmt;
p->num_fmts = 0;
p->num_upload_fmts = 0;
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
struct mp_stream_info *info = mp_filter_find_stream_info(u->f);
if (!info || !info->hwdec_devs) {
MP_ERR(u->f, "no hw context\n");
return false;
}
struct mp_hwdec_ctx *ctx = NULL;
AVHWFramesConstraints *cstr = NULL;
for (int n = 0; ; n++) {
struct mp_hwdec_ctx *cur = hwdec_devices_get_n(info->hwdec_devs, n);
if (!cur)
break;
if (!cur->av_device_ref)
continue;
cstr = av_hwdevice_get_hwframe_constraints(cur->av_device_ref, NULL);
if (!cstr)
continue;
bool found = false;
for (int i = 0; cstr->valid_hw_formats &&
cstr->valid_hw_formats[i] != AV_PIX_FMT_NONE; i++)
{
found |= cstr->valid_hw_formats[i] == imgfmt2pixfmt(hw_imgfmt);
}
if (found && (!cur->hw_imgfmt || cur->hw_imgfmt == hw_imgfmt)) {
ctx = cur;
break;
}
av_hwframe_constraints_free(&cstr);
}
if (!ctx) {
MP_ERR(u->f, "no support for this hw format\n");
return false;
}
// Probe for supported formats. This is very roundabout, because the
// hwcontext API does not give us this information directly. We resort to
// creating temporary AVHWFramesContexts in order to retrieve the list of
// supported formats. This should be relatively cheap as we don't create
// any real frames (although some backends do for probing info).
const struct ffmpeg_and_other_bugs *bugs = NULL;
for (int n = 0; shitlist[n].imgfmt; n++) {
if (shitlist[n].imgfmt == hw_imgfmt) {
bugs = &shitlist[n];
break;
}
}
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
for (int n = 0; cstr->valid_sw_formats &&
cstr->valid_sw_formats[n] != AV_PIX_FMT_NONE; n++)
{
int imgfmt = pixfmt2imgfmt(cstr->valid_sw_formats[n]);
if (!imgfmt)
continue;
MP_VERBOSE(u->f, "looking at format %s/%s\n",
mp_imgfmt_to_name(hw_imgfmt),
mp_imgfmt_to_name(imgfmt));
if (bugs && bugs->whitelist_formats) {
bool found = false;
for (int i = 0; bugs->whitelist_formats[i]; i++) {
if (bugs->whitelist_formats[i] == imgfmt) {
found = true;
break;
}
}
if (!found) {
MP_VERBOSE(u->f, "... skipping blacklisted format\n");
continue;
}
}
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
// Creates an AVHWFramesContexts with the given parameters.
AVBufferRef *frames = NULL;
if (!mp_update_av_hw_frames_pool(&frames, ctx->av_device_ref,
hw_imgfmt, imgfmt, 128, 128))
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
{
MP_WARN(u->f, "failed to allocate pool\n");
continue;
}
enum AVPixelFormat *fmts;
if (av_hwframe_transfer_get_formats(frames,
AV_HWFRAME_TRANSFER_DIRECTION_TO, &fmts, 0) >= 0)
{
int index = p->num_fmts;
MP_TARRAY_APPEND(p, p->fmts, p->num_fmts, imgfmt);
MP_TARRAY_GROW(p, p->fmt_upload_index, index);
MP_TARRAY_GROW(p, p->fmt_upload_num, index);
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
p->fmt_upload_index[index] = p->num_upload_fmts;
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
for (int i = 0; fmts[i] != AV_PIX_FMT_NONE; i++) {
int fmt = pixfmt2imgfmt(fmts[i]);
if (!fmt)
continue;
MP_VERBOSE(u->f, " supports %s\n", mp_imgfmt_to_name(fmt));
if (bugs && bugs->force_same_upload_fmt && imgfmt != fmt) {
MP_VERBOSE(u->f, " ... skipping blacklisted format\n");
continue;
}
if (!vo_supports(ctx, hw_imgfmt, fmt)) {
MP_VERBOSE(u->f, " ... not supported by VO\n");
continue;
}
MP_TARRAY_APPEND(p, p->upload_fmts, p->num_upload_fmts, fmt);
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
}
p->fmt_upload_num[index] =
p->num_upload_fmts - p->fmt_upload_index[index];
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
av_free(fmts);
}
av_buffer_unref(&frames);
}
p->av_device_ctx = av_buffer_ref(ctx->av_device_ref);
if (!p->av_device_ctx)
return false;
return p->num_upload_fmts > 0;
video: rewrite filtering glue code Get rid of the old vf.c code. Replace it with a generic filtering framework, which can potentially handle more than just --vf. At least reimplementing --af with this code is planned. This changes some --vf semantics (including runtime behavior and the "vf" command). The most important ones are listed in interface-changes. vf_convert.c is renamed to f_swscale.c. It is now an internal filter that can not be inserted by the user manually. f_lavfi.c is a refactor of player/lavfi.c. The latter will be removed once --lavfi-complex is reimplemented on top of f_lavfi.c. (which is conceptually easy, but a big mess due to the data flow changes). The existing filters are all changed heavily. The data flow of the new filter framework is different. Especially EOF handling changes - EOF is now a "frame" rather than a state, and must be passed through exactly once. Another major thing is that all filters must support dynamic format changes. The filter reconfig() function goes away. (This sounds complex, but since all filters need to handle EOF draining anyway, they can use the same code, and it removes the mess with reconfig() having to predict the output format, which completely breaks with libavfilter anyway.) In addition, there is no automatic format negotiation or conversion. libavfilter's primitive and insufficient API simply doesn't allow us to do this in a reasonable way. Instead, filters can use f_autoconvert as sub-filter, and tell it which formats they support. This filter will in turn add actual conversion filters, such as f_swscale, to perform necessary format changes. vf_vapoursynth.c uses the same basic principle of operation as before, but with worryingly different details in data flow. Still appears to work. The hardware deint filters (vf_vavpp.c, vf_d3d11vpp.c, vf_vdpaupp.c) are heavily changed. Fortunately, they all used refqueue.c, which is for sharing the data flow logic (especially for managing future/past surfaces and such). It turns out it can be used to factor out most of the data flow. Some of these filters accepted software input. Instead of having ad-hoc upload code in each filter, surface upload is now delegated to f_autoconvert, which can use f_hwupload to perform this. Exporting VO capabilities is still a big mess (mp_stream_info stuff). The D3D11 code drops the redundant image formats, and all code uses the hw_subfmt (sw_format in FFmpeg) instead. Although that too seems to be a big mess for now. f_async_queue is unused.
2018-01-16 10:53:44 +00:00
}
struct mp_hwupload *mp_hwupload_create(struct mp_filter *parent, int hw_imgfmt)
{
struct mp_filter *f = mp_filter_create(parent, &hwupload_filter);
if (!f)
return NULL;
struct priv *p = f->priv;
struct mp_hwupload *u = &p->public;
u->f = f;
mp_filter_add_pin(f, MP_PIN_IN, "in");
mp_filter_add_pin(f, MP_PIN_OUT, "out");
if (!probe_formats(u, hw_imgfmt)) {
MP_ERR(f, "hardware format not supported\n");
goto error;
}
return u;
error:
talloc_free(f);
return NULL;
}
static void hwdownload_process(struct mp_filter *f)
{
struct mp_hwdownload *d = f->priv;
if (!mp_pin_can_transfer_data(f->ppins[1], f->ppins[0]))
return;
struct mp_frame frame = mp_pin_out_read(f->ppins[0]);
if (frame.type != MP_FRAME_VIDEO)
goto passthrough;
struct mp_image *src = frame.data;
if (!src->hwctx)
goto passthrough;
struct mp_image *dst = mp_image_hw_download(src, d->pool);
if (!dst) {
MP_ERR(f, "Could not copy hardware frame to CPU memory.\n");
goto passthrough;
}
mp_frame_unref(&frame);
mp_pin_in_write(f->ppins[1], MAKE_FRAME(MP_FRAME_VIDEO, dst));
return;
passthrough:
mp_pin_in_write(f->ppins[1], frame);
return;
}
static const struct mp_filter_info hwdownload_filter = {
.name = "hwdownload",
.priv_size = sizeof(struct mp_hwdownload),
.process = hwdownload_process,
};
struct mp_hwdownload *mp_hwdownload_create(struct mp_filter *parent)
{
struct mp_filter *f = mp_filter_create(parent, &hwdownload_filter);
if (!f)
return NULL;
struct mp_hwdownload *d = f->priv;
d->f = f;
d->pool = mp_image_pool_new(d);
mp_filter_add_pin(f, MP_PIN_IN, "in");
mp_filter_add_pin(f, MP_PIN_OUT, "out");
return d;
}