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.

1 files changed, 155 insertions, 163 deletions

diff --git a/video/filter/vf_d3d11vpp.c b/video/filter/vf_d3d11vpp.c
index 3be49ede80..fb96a44e65 100644
--- a/video/filter/vf_d3d11vpp.c
+++ b/video/filter/vf_d3d11vpp.c
@@ -25,9 +25,13 @@
 #include "common/common.h"
 #include "osdep/timer.h"
 #include "osdep/windows_utils.h"
-#include "vf.h"
+#include "filters/f_autoconvert.h"
+#include "filters/filter.h"
+#include "filters/filter_internal.h"
+#include "filters/user_filters.h"
 #include "refqueue.h"
 #include "video/hwdec.h"
+#include "video/mp_image.h"
 #include "video/mp_image_pool.h"
 
 // missing in MinGW
@@ -38,8 +42,17 @@
 #define D3D11_VIDEO_PROCESSOR_PROCESSOR_CAPS_INVERSE_TELECINE 0x10
 #define D3D11_VIDEO_PROCESSOR_PROCESSOR_CAPS_FRAME_RATE_CONVERSION 0x20
 
-struct vf_priv_s {
+struct opts {
+    int deint_enabled;
+    int interlaced_only;
+    int mode;
+};
+
+struct priv {
+    struct opts *opts;
+
     ID3D11Device *vo_dev;
+    const int *vo_formats;
 
     ID3D11DeviceContext *device_ctx;
     ID3D11VideoDevice *video_dev;
@@ -61,10 +74,6 @@ struct vf_priv_s {
     struct mp_image_pool *pool;
 
     struct mp_refqueue *queue;
-
-    int deint_enabled;
-    int interlaced_only;
-    int mode;
 };
 
 static void release_tex(void *arg)
@@ -76,8 +85,8 @@ static void release_tex(void *arg)
 
 static struct mp_image *alloc_pool(void *pctx, int fmt, int w, int h)
 {
-    struct vf_instance *vf = pctx;
-    struct vf_priv_s *p = vf->priv;
+    struct mp_filter *vf = pctx;
+    struct priv *p = vf->priv;
     HRESULT hr;
 
     ID3D11Texture2D *texture = NULL;
@@ -100,7 +109,7 @@ static struct mp_image *alloc_pool(void *pctx, int fmt, int w, int h)
     if (!mpi)
         abort();
 
-    mp_image_setfmt(mpi, p->out_params.imgfmt);
+    mp_image_setfmt(mpi, IMGFMT_D3D11);
     mp_image_set_size(mpi, w, h);
     mpi->params.hw_subfmt = p->out_params.hw_subfmt;
 
@@ -110,29 +119,15 @@ static struct mp_image *alloc_pool(void *pctx, int fmt, int w, int h)
     return mpi;
 }
 
-static void flush_frames(struct vf_instance *vf)
+static void flush_frames(struct mp_filter *vf)
 {
-    struct vf_priv_s *p = vf->priv;
+    struct priv *p = vf->priv;
     mp_refqueue_flush(p->queue);
 }
 
-static int filter_ext(struct vf_instance *vf, struct mp_image *in)
-{
-    struct vf_priv_s *p = vf->priv;
-
-    mp_refqueue_set_refs(p->queue, 0, 0);
-    mp_refqueue_set_mode(p->queue,
-        (p->deint_enabled ? MP_MODE_DEINT : 0) |
-        MP_MODE_OUTPUT_FIELDS |
-        (p->interlaced_only ? MP_MODE_INTERLACED_ONLY : 0));
-
-    mp_refqueue_add_input(p->queue, in);
-    return 0;
-}
-
-static void destroy_video_proc(struct vf_instance *vf)
+static void destroy_video_proc(struct mp_filter *vf)
 {
-    struct vf_priv_s *p = vf->priv;
+    struct priv *p = vf->priv;
 
     if (p->video_proc)
         ID3D11VideoProcessor_Release(p->video_proc);
@@ -143,9 +138,9 @@ static void destroy_video_proc(struct vf_instance *vf)
     p->vp_enum = NULL;
 }
 
-static int recreate_video_proc(struct vf_instance *vf)
+static int recreate_video_proc(struct mp_filter *vf)
 {
-    struct vf_priv_s *p = vf->priv;
+    struct priv *p = vf->priv;
     HRESULT hr;
 
     destroy_video_proc(vf);
@@ -168,7 +163,7 @@ static int recreate_video_proc(struct vf_instance *vf)
         goto fail;
 
     MP_VERBOSE(vf, "Found %d rate conversion caps. Looking for caps=0x%x.\n",
-               (int)caps.RateConversionCapsCount, p->mode);
+               (int)caps.RateConversionCapsCount, p->opts->mode);
 
     int rindex = -1;
     for (int n = 0; n < caps.RateConversionCapsCount; n++) {
@@ -178,7 +173,7 @@ static int recreate_video_proc(struct vf_instance *vf)
         if (FAILED(hr))
             goto fail;
         MP_VERBOSE(vf, "  - %d: 0x%08x\n", n, (unsigned)rcaps.ProcessorCaps);
-        if (rcaps.ProcessorCaps & p->mode) {
+        if (rcaps.ProcessorCaps & p->opts->mode) {
             MP_VERBOSE(vf, "       (matching)\n");
             if (rindex < 0)
                 rindex = n;
@@ -248,17 +243,19 @@ fail:
     return -1;
 }
 
-static int render(struct vf_instance *vf)
+static struct mp_image *render(struct mp_filter *vf)
 {
-    struct vf_priv_s *p = vf->priv;
+    struct priv *p = vf->priv;
     int res = -1;
     HRESULT hr;
     ID3D11VideoProcessorInputView *in_view = NULL;
     ID3D11VideoProcessorOutputView *out_view = NULL;
     struct mp_image *in = NULL, *out = NULL;
-    out = mp_image_pool_get(p->pool, p->out_params.imgfmt, p->params.w, p->params.h);
-    if (!out)
+    out = mp_image_pool_get(p->pool, IMGFMT_D3D11, p->params.w, p->params.h);
+    if (!out) {
+        MP_WARN(vf, "failed to allocate frame\n");
         goto cleanup;
+    }
 
     ID3D11Texture2D *d3d_out_tex = (void *)out->planes[0];
 
@@ -325,8 +322,10 @@ static int render(struct vf_instance *vf)
                                                           (ID3D11Resource *)d3d_out_tex,
                                                           p->vp_enum, &outdesc,
                                                           &out_view);
-    if (FAILED(hr))
+    if (FAILED(hr)) {
+        MP_ERR(vf, "Could not create ID3D11VideoProcessorOutputView\n");
         goto cleanup;
+    }
 
     D3D11_VIDEO_PROCESSOR_STREAM stream = {
         .Enable = TRUE,
@@ -346,87 +345,73 @@ cleanup:
         ID3D11VideoProcessorInputView_Release(in_view);
     if (out_view)
         ID3D11VideoProcessorOutputView_Release(out_view);
-    if (res >= 0) {
-        vf_add_output_frame(vf, out);
-    } else {
-        talloc_free(out);
-    }
-    mp_refqueue_next_field(p->queue);
-    return res;
+    if (res < 0)
+        TA_FREEP(&out);
+    return out;
 }
 
-static int filter_out(struct vf_instance *vf)
+static bool vo_supports(struct priv *p, int subfmt)
 {
-    struct vf_priv_s *p = vf->priv;
-
-    if (!mp_refqueue_has_output(p->queue))
-        return 0;
-
-    // no filtering
-    if (!mp_refqueue_should_deint(p->queue) && !p->require_filtering) {
-        struct mp_image *in = mp_image_new_ref(mp_refqueue_get(p->queue, 0));
-        if (!in)
-            return -1;
-        mp_image_set_params(in, &p->out_params);
-        vf_add_output_frame(vf, in);
-        mp_refqueue_next(p->queue);
-        return 0;
+    for (int n = 0; p->vo_formats && p->vo_formats[n]; n++) {
+        if (p->vo_formats[n] == subfmt)
+            return true;
     }
-
-    return render(vf);
+    return false;
 }
 
-static int reconfig(struct vf_instance *vf, struct mp_image_params *in,
-                    struct mp_image_params *out)
+static void vf_d3d11vpp_process(struct mp_filter *vf)
 {
-    struct vf_priv_s *p = vf->priv;
-
-    flush_frames(vf);
-    talloc_free(p->pool);
-    p->pool = NULL;
+    struct priv *p = vf->priv;
+
+    struct mp_image *in_fmt = mp_refqueue_execute_reinit(p->queue);
+    if (in_fmt) {
+        mp_image_pool_clear(p->pool);
+
+        destroy_video_proc(vf);
+
+        p->params = in_fmt->params;
+        p->out_params = p->params;
+
+        if (vo_supports(p, IMGFMT_NV12)) {
+            p->out_params.hw_subfmt = IMGFMT_NV12;
+            p->out_format = DXGI_FORMAT_NV12;
+            p->out_shared = false;
+            p->out_rgb = false;
+        } else {
+            p->out_params.hw_subfmt = IMGFMT_RGB0;
+            p->out_format = DXGI_FORMAT_B8G8R8A8_UNORM;
+            p->out_shared = true;
+            p->out_rgb = true;
+        }
+        p->out_params.hw_flags = 0;
 
-    destroy_video_proc(vf);
+        p->require_filtering = p->params.hw_subfmt != p->out_params.hw_subfmt;
+    }
 
-    *out = *in;
+    if (!mp_refqueue_can_output(p->queue))
+        return;
 
-    if (vf_next_query_format(vf, IMGFMT_D3D11VA) ||
-        vf_next_query_format(vf, IMGFMT_D3D11NV12))
-    {
-        out->imgfmt = vf_next_query_format(vf, IMGFMT_D3D11VA)
-                    ? IMGFMT_D3D11VA : IMGFMT_D3D11NV12;
-        out->hw_subfmt = IMGFMT_NV12;
-        p->out_format = DXGI_FORMAT_NV12;
-        p->out_shared = false;
-        p->out_rgb = false;
+    if (!mp_refqueue_should_deint(p->queue) && !p->require_filtering) {
+        // no filtering
+        struct mp_image *in = mp_image_new_ref(mp_refqueue_get(p->queue, 0));
+        if (!in) {
+            mp_filter_internal_mark_failed(vf);
+            return;
+        }
+        mp_refqueue_write_out_pin(p->queue, in);
     } else {
-        out->imgfmt = IMGFMT_D3D11RGB;
-        out->hw_subfmt = IMGFMT_RGB0;
-        p->out_format = DXGI_FORMAT_B8G8R8A8_UNORM;
-        p->out_shared = true;
-        p->out_rgb = true;
+        mp_refqueue_write_out_pin(p->queue, render(vf));
     }
-    out->hw_flags = 0;
-
-    p->require_filtering = in->hw_subfmt != out->hw_subfmt;
-
-    p->params = *in;
-    p->out_params = *out;
-
-    p->pool = mp_image_pool_new(vf);
-    mp_image_pool_set_allocator(p->pool, alloc_pool, vf);
-    mp_image_pool_set_lru(p->pool);
-
-    return 0;
 }
 
-static void uninit(struct vf_instance *vf)
+static void uninit(struct mp_filter *vf)
 {
-    struct vf_priv_s *p = vf->priv;
+    struct priv *p = vf->priv;
 
     destroy_video_proc(vf);
 
     flush_frames(vf);
-    mp_refqueue_free(p->queue);
+    talloc_free(p->queue);
     talloc_free(p->pool);
 
     if (p->video_ctx)
@@ -442,69 +427,55 @@ static void uninit(struct vf_instance *vf)
         ID3D11Device_Release(p->vo_dev);
 }
 
-static int query_format(struct vf_instance *vf, unsigned int imgfmt)
+static const struct mp_filter_info vf_d3d11vpp_filter = {
+    .name = "d3d11vpp",
+    .process = vf_d3d11vpp_process,
+    .reset = flush_frames,
+    .destroy = uninit,
+    .priv_size = sizeof(struct priv),
+};
+
+static struct mp_filter *vf_d3d11vpp_create(struct mp_filter *parent,
+                                            void *options)
 {
-    if (imgfmt == IMGFMT_D3D11VA ||
-        imgfmt == IMGFMT_D3D11NV12 ||
-        imgfmt == IMGFMT_D3D11RGB)
-    {
-        return vf_next_query_format(vf, IMGFMT_D3D11VA) ||
-               vf_next_query_format(vf, IMGFMT_D3D11NV12) ||
-               vf_next_query_format(vf, IMGFMT_D3D11RGB);
+    struct mp_filter *f = mp_filter_create(parent, &vf_d3d11vpp_filter);
+    if (!f) {
+        talloc_free(options);
+        return NULL;
     }
-    return 0;
-}
 
-static bool test_conversion(int in, int out)
-{
-    return (in == IMGFMT_D3D11VA ||
-            in == IMGFMT_D3D11NV12 ||
-            in == IMGFMT_D3D11RGB) &&
-           (out == IMGFMT_D3D11VA ||
-            out == IMGFMT_D3D11NV12 ||
-            out == IMGFMT_D3D11RGB);
-}
+    mp_filter_add_pin(f, MP_PIN_IN, "in");
+    mp_filter_add_pin(f, MP_PIN_OUT, "out");
 
-static int control(struct vf_instance *vf, int request, void* data)
-{
-    switch (request){
-    case VFCTRL_SEEK_RESET:
-        flush_frames(vf);
-        return true;
-    default:
-        return CONTROL_UNKNOWN;
-    }
-}
+    struct priv *p = f->priv;
+    p->opts = talloc_steal(p, options);
 
-static int vf_open(vf_instance_t *vf)
-{
-    struct vf_priv_s *p = vf->priv;
+    // Special path for vf_d3d11_create_outconv(): disable all processing except
+    // possibly surface format conversions.
+    if (!p->opts) {
+        static const struct opts opts = {0};
+        p->opts = (struct opts *)&opts;
+    }
 
-    if (!vf->hwdec_devs)
-        return 0;
+    p->queue = mp_refqueue_alloc(f);
 
-    vf->reconfig = reconfig;
-    vf->filter_ext = filter_ext;
-    vf->filter_out = filter_out;
-    vf->query_format = query_format;
-    vf->uninit = uninit;
-    vf->control = control;
+    struct mp_stream_info *info = mp_filter_find_stream_info(f);
+    if (!info || !info->hwdec_devs)
+        goto fail;
 
-    hwdec_devices_request_all(vf->hwdec_devs);
-    AVBufferRef *ref =
-        hwdec_devices_get_lavc(vf->hwdec_devs, AV_HWDEVICE_TYPE_D3D11VA);
-    if (!ref)
-        return 0;
+    hwdec_devices_request_all(info->hwdec_devs);
 
-    AVHWDeviceContext *hwctx = (void *)ref->data;
-    AVD3D11VADeviceContext *d3dctx = hwctx->hwctx;
+    struct mp_hwdec_ctx *hwctx =
+        hwdec_devices_get_by_lavc(info->hwdec_devs, AV_HWDEVICE_TYPE_D3D11VA);
+    if (!hwctx || !hwctx->av_device_ref)
+        goto fail;
+    AVHWDeviceContext *avhwctx = (void *)hwctx->av_device_ref->data;
+    AVD3D11VADeviceContext *d3dctx = avhwctx->hwctx;
 
     p->vo_dev = d3dctx->device;
     ID3D11Device_AddRef(p->vo_dev);
 
-    av_buffer_unref(&ref);
-
-    p->queue = mp_refqueue_alloc();
+    p->vo_formats = hwctx->supported_formats;
 
     HRESULT hr;
 
@@ -521,14 +492,26 @@ static int vf_open(vf_instance_t *vf)
     if (FAILED(hr))
         goto fail;
 
-    return 1;
+    p->pool = mp_image_pool_new(f);
+    mp_image_pool_set_allocator(p->pool, alloc_pool, f);
+    mp_image_pool_set_lru(p->pool);
+
+    mp_refqueue_add_in_format(p->queue, IMGFMT_D3D11, 0);
+
+    mp_refqueue_set_refs(p->queue, 0, 0);
+    mp_refqueue_set_mode(p->queue,
+        (p->opts->deint_enabled ? MP_MODE_DEINT : 0) |
+        MP_MODE_OUTPUT_FIELDS |
+        (p->opts->interlaced_only ? MP_MODE_INTERLACED_ONLY : 0));
+
+    return f;
 
 fail:
-    uninit(vf);
-    return 0;
+    talloc_free(f);
+    return NULL;
 }
 
-#define OPT_BASE_STRUCT struct vf_priv_s
+#define OPT_BASE_STRUCT struct opts
 static const m_option_t vf_opts_fields[] = {
     OPT_FLAG("deint", deint_enabled, 0),
     OPT_FLAG("interlaced-only", interlaced_only, 0),
@@ -542,16 +525,25 @@ static const m_option_t vf_opts_fields[] = {
     {0}
 };
 
-const vf_info_t vf_info_d3d11vpp = {
-    .description = "D3D11 Video Post-Process Filter",
-    .name = "d3d11vpp",
-    .test_conversion = test_conversion,
-    .open = vf_open,
-    .priv_size = sizeof(struct vf_priv_s),
-    .priv_defaults = &(const struct vf_priv_s) {
-        .deint_enabled = 1,
-        .interlaced_only = 1,
-        .mode = D3D11_VIDEO_PROCESSOR_PROCESSOR_CAPS_DEINTERLACE_BOB,
+const struct mp_user_filter_entry vf_d3d11vpp = {
+    .desc = {
+        .description = "D3D11 Video Post-Process Filter",
+        .name = "d3d11vpp",
+        .priv_size = sizeof(OPT_BASE_STRUCT),
+        .priv_defaults = &(const OPT_BASE_STRUCT) {
+            .deint_enabled = 1,
+            .interlaced_only = 1,
+            .mode = D3D11_VIDEO_PROCESSOR_PROCESSOR_CAPS_DEINTERLACE_BOB,
+        },
+        .options = vf_opts_fields,
     },
-    .options = vf_opts_fields,
+    .create = vf_d3d11vpp_create,
 };
+
+// Create a filter for the purpose of converting the sub-format for hwdec
+// interops which are incapable of handling some formats (ANGLE).
+struct mp_filter *vf_d3d11_create_outconv(struct mp_filter *parent)
+{
+    // options==NULL is normally not allowed, and specially handled.
+    return vf_d3d11vpp_create(parent, NULL);
+}