mpv/video/mp_image.c

1217 lines
42 KiB
C
Raw Normal View History

/*
* This file is part of mpv.
*
* mpv is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* mpv is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with mpv. If not, see <http://www.gnu.org/licenses/>.
*/
#include <limits.h>
#include <pthread.h>
2012-12-11 23:43:36 +00:00
#include <assert.h>
#include <libavutil/mem.h>
#include <libavutil/common.h>
#include <libavutil/display.h>
#include <libavutil/bswap.h>
#include <libavutil/hwcontext.h>
#include <libavutil/intreadwrite.h>
#include <libavutil/rational.h>
#include <libavcodec/avcodec.h>
#include <libavutil/mastering_display_metadata.h>
#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(57, 16, 100)
# include <libavutil/dovi_meta.h>
#endif
#include "mpv_talloc.h"
#include "config.h"
#include "common/av_common.h"
#include "common/common.h"
#include "hwdec.h"
#include "mp_image.h"
#include "sws_utils.h"
#include "fmt-conversion.h"
// Determine strides, plane sizes, and total required size for an image
// allocation. Returns total size on success, <0 on error. Unused planes
// have out_stride/out_plane_size to 0, and out_plane_offset set to -1 up
// until MP_MAX_PLANES-1.
static int mp_image_layout(int imgfmt, int w, int h, int stride_align,
int out_stride[MP_MAX_PLANES],
int out_plane_offset[MP_MAX_PLANES],
int out_plane_size[MP_MAX_PLANES])
{
struct mp_imgfmt_desc desc = mp_imgfmt_get_desc(imgfmt);
w = MP_ALIGN_UP(w, desc.align_x);
h = MP_ALIGN_UP(h, desc.align_y);
struct mp_image_params params = {.imgfmt = imgfmt, .w = w, .h = h};
2012-12-11 23:43:36 +00:00
if (!mp_image_params_valid(&params) || desc.flags & MP_IMGFLAG_HWACCEL)
return -1;
// Note: for non-mod-2 4:2:0 YUV frames, we have to allocate an additional
// top/right border. This is needed for correct handling of such
// images in filter and VO code (e.g. vo_vdpau or vo_gpu).
for (int n = 0; n < MP_MAX_PLANES; n++) {
int alloc_w = mp_chroma_div_up(w, desc.xs[n]);
int alloc_h = MP_ALIGN_UP(h, 32) >> desc.ys[n];
int line_bytes = (alloc_w * desc.bpp[n] + 7) / 8;
out_stride[n] = MP_ALIGN_UP(line_bytes, stride_align);
out_plane_size[n] = out_stride[n] * alloc_h;
}
if (desc.flags & MP_IMGFLAG_PAL)
out_plane_size[1] = AVPALETTE_SIZE;
int sum = 0;
for (int n = 0; n < MP_MAX_PLANES; n++) {
out_plane_offset[n] = out_plane_size[n] ? sum : -1;
sum += out_plane_size[n];
}
return sum;
}
// Return the total size needed for an image allocation of the given
// configuration (imgfmt, w, h must be set). Returns -1 on error.
// Assumes the allocation is already aligned on stride_align (otherwise you
// need to add padding yourself).
int mp_image_get_alloc_size(int imgfmt, int w, int h, int stride_align)
{
int stride[MP_MAX_PLANES];
int plane_offset[MP_MAX_PLANES];
int plane_size[MP_MAX_PLANES];
return mp_image_layout(imgfmt, w, h, stride_align, stride, plane_offset,
plane_size);
}
// Fill the mpi->planes and mpi->stride fields of the given mpi with data
// from buffer according to the mpi's w/h/imgfmt fields. See mp_image_from_buffer
// aboud remarks how to allocate/use buffer/buffer_size.
// This does not free the data. You are expected to setup refcounting by
// setting mp_image.bufs before or after this function is called.
// Returns true on success, false on failure.
static bool mp_image_fill_alloc(struct mp_image *mpi, int stride_align,
void *buffer, int buffer_size)
{
int stride[MP_MAX_PLANES];
int plane_offset[MP_MAX_PLANES];
int plane_size[MP_MAX_PLANES];
int size = mp_image_layout(mpi->imgfmt, mpi->w, mpi->h, stride_align,
stride, plane_offset, plane_size);
if (size < 0 || size > buffer_size)
return false;
int align = MP_ALIGN_UP((uintptr_t)buffer, stride_align) - (uintptr_t)buffer;
if (buffer_size - size < align)
return false;
uint8_t *s = buffer;
s += align;
for (int n = 0; n < MP_MAX_PLANES; n++) {
mpi->planes[n] = plane_offset[n] >= 0 ? s + plane_offset[n] : NULL;
mpi->stride[n] = stride[n];
}
return true;
}
// Create a mp_image from the provided buffer. The mp_image is filled according
// to the imgfmt/w/h parameters, and respecting the stride_align parameter to
// align the plane start pointers and strides. Once the last reference to the
// returned image is destroyed, free(free_opaque, buffer) is called. (Be aware
// that this can happen from any thread.)
// The allocated size of buffer must be given by buffer_size. buffer_size should
// be at least the value returned by mp_image_get_alloc_size(). If buffer is not
// already aligned to stride_align, the function will attempt to align the
// pointer itself by incrementing the buffer pointer until ther alignment is
// achieved (if buffer_size is not large enough to allow aligning the buffer
// safely, the function fails). To be safe, you may want to overallocate the
// buffer by stride_align bytes, and include the overallocation in buffer_size.
// Returns NULL on failure. On failure, the free() callback is not called.
struct mp_image *mp_image_from_buffer(int imgfmt, int w, int h, int stride_align,
uint8_t *buffer, int buffer_size,
void *free_opaque,
void (*free)(void *opaque, uint8_t *data))
{
struct mp_image *mpi = mp_image_new_dummy_ref(NULL);
mp_image_setfmt(mpi, imgfmt);
mp_image_set_size(mpi, w, h);
if (!mp_image_fill_alloc(mpi, stride_align, buffer, buffer_size))
goto fail;
mpi->bufs[0] = av_buffer_create(buffer, buffer_size, free, free_opaque, 0);
if (!mpi->bufs[0])
goto fail;
return mpi;
fail:
talloc_free(mpi);
return NULL;
}
static bool mp_image_alloc_planes(struct mp_image *mpi)
{
assert(!mpi->planes[0]);
assert(!mpi->bufs[0]);
video: generally try to align image data on 64 bytes Generally, using x86 SIMD efficiently (or crash-free) requires aligning all data on boundaries of 16, 32, or 64 (depending on instruction set used). 64 bytes is needed or AVX-512, 32 for old AVX, 16 for SSE. Both FFmpeg and zimg usually require aligned data for this reason. FFmpeg is very unclear about alignment. Yes, it requires you to align data pointers and strides. No, it doesn't tell you how much, except sometimes (libavcodec has a legacy-looking avcodec_align_dimensions2() API function, that requires a heavy-weight AVCodecContext as argument). Sometimes, FFmpeg will take a shit on YOUR and ITS OWN alignment. For example, vf_crop will randomly reduce alignment of data pointers, depending on the crop parameters. On the other hand, some libavfilter filters or libavcodec encoders may randomly crash if they get the wrong alignment. I have no idea how this thing works at all. FFmpeg usually doesn't seem to signal alignment internal anywhere, and usually leaves it to av_malloc() etc. to allocate with proper alignment. libavutil/mem.c currently has a ALIGN define, which is set to 64 if FFmpeg is built with AVX-512 support, or as low as 16 if built without any AVX support. The really funny thing is that a normal FFmpeg build will e.g. align tiny string allocations to 64 bytes, even if the machine does not support AVX at all. For zimg use (in a later commit), we also want guaranteed alignment. Modern x86 should actually not be much slower at unaligned accesses, but that doesn't help. zimg's dumb intrinsic code apparently randomly chooses between aligned or unaligned accesses (depending on compiler, I guess), and on some CPUs these can even cause crashes. So just treat the requirement to align as a fact of life. All this means that we should probably make sure our own allocations are 64 bit aligned. This still doesn't guarantee alignment in all cases, but it's slightly better than before. This also makes me wonder whether we should always override libavcodec's buffer pool, just so we have a guaranteed alignment. Currently, we only do that if --vd-lavc-dr is used (and if that actually works). On the other hand, it always uses DR on my machine, so who cares.
2019-07-15 01:06:47 +00:00
int align = MP_IMAGE_BYTE_ALIGN;
int size = mp_image_get_alloc_size(mpi->imgfmt, mpi->w, mpi->h, align);
if (size < 0)
return false;
// Note: mp_image_pool assumes this creates only 1 AVBufferRef.
mpi->bufs[0] = av_buffer_alloc(size + align);
if (!mpi->bufs[0])
return false;
if (!mp_image_fill_alloc(mpi, align, mpi->bufs[0]->data, mpi->bufs[0]->size)) {
av_buffer_unref(&mpi->bufs[0]);
return false;
}
return true;
}
void mp_image_setfmt(struct mp_image *mpi, int out_fmt)
{
struct mp_image_params params = mpi->params;
struct mp_imgfmt_desc fmt = mp_imgfmt_get_desc(out_fmt);
params.imgfmt = fmt.id;
mpi->fmt = fmt;
mpi->imgfmt = fmt.id;
mpi->num_planes = fmt.num_planes;
mpi->params = params;
}
static void mp_image_destructor(void *ptr)
{
mp_image_t *mpi = ptr;
for (int p = 0; p < MP_MAX_PLANES; p++)
av_buffer_unref(&mpi->bufs[p]);
av_buffer_unref(&mpi->hwctx);
av_buffer_unref(&mpi->icc_profile);
av_buffer_unref(&mpi->a53_cc);
av_buffer_unref(&mpi->dovi);
av_buffer_unref(&mpi->film_grain);
for (int n = 0; n < mpi->num_ff_side_data; n++)
av_buffer_unref(&mpi->ff_side_data[n].buf);
talloc_free(mpi->ff_side_data);
}
int mp_chroma_div_up(int size, int shift)
2012-12-11 23:43:36 +00:00
{
return (size + (1 << shift) - 1) >> shift;
}
// Return the storage width in pixels of the given plane.
int mp_image_plane_w(struct mp_image *mpi, int plane)
{
return mp_chroma_div_up(mpi->w, mpi->fmt.xs[plane]);
}
// Return the storage height in pixels of the given plane.
int mp_image_plane_h(struct mp_image *mpi, int plane)
{
return mp_chroma_div_up(mpi->h, mpi->fmt.ys[plane]);
}
// Caller has to make sure this doesn't exceed the allocated plane data/strides.
void mp_image_set_size(struct mp_image *mpi, int w, int h)
{
assert(w >= 0 && h >= 0);
mpi->w = mpi->params.w = w;
mpi->h = mpi->params.h = h;
}
void mp_image_set_params(struct mp_image *image,
const struct mp_image_params *params)
{
// possibly initialize other stuff
mp_image_setfmt(image, params->imgfmt);
mp_image_set_size(image, params->w, params->h);
image->params = *params;
}
struct mp_image *mp_image_alloc(int imgfmt, int w, int h)
2012-12-11 23:43:36 +00:00
{
struct mp_image *mpi = talloc_zero(NULL, struct mp_image);
talloc_set_destructor(mpi, mp_image_destructor);
mp_image_set_size(mpi, w, h);
2012-12-11 23:43:36 +00:00
mp_image_setfmt(mpi, imgfmt);
if (!mp_image_alloc_planes(mpi)) {
talloc_free(mpi);
return NULL;
}
2012-12-11 23:43:36 +00:00
return mpi;
}
Implement backwards playback See manpage additions. This is a huge hack. You can bet there are shit tons of bugs. It's literally forcing square pegs into round holes. Hopefully, the manpage wall of text makes it clear enough that the whole shit can easily crash and burn. (Although it shouldn't literally crash. That would be a bug. It possibly _could_ start a fire by entering some sort of endless loop, not a literal one, just something where it tries to do work without making progress.) (Some obvious bugs I simply ignored for this initial version, but there's a number of potential bugs I can't even imagine. Normal playback should remain completely unaffected, though.) How this works is also described in the manpage. Basically, we demux in reverse, then we decode in reverse, then we render in reverse. The decoding part is the simplest: just reorder the decoder output. This weirdly integrates with the timeline/ordered chapter code, which also has special requirements on feeding the packets to the decoder in a non-straightforward way (it doesn't conflict, although a bugmessmass breaks correct slicing of segments, so EDL/ordered chapter playback is broken in backward direction). Backward demuxing is pretty involved. In theory, it could be much easier: simply iterating the usual demuxer output backward. But this just doesn't fit into our code, so there's a cthulhu nightmare of shit. To be specific, each stream (audio, video) is reversed separately. At least this means we can do backward playback within cached content (for example, you could play backwards in a live stream; on that note, it disables prefetching, which would lead to losing new live video, but this could be avoided). The fuckmess also meant that I didn't bother trying to support subtitles. Subtitles are a problem because they're "sparse" streams. They need to be "passively" demuxed: you don't try to read a subtitle packet, you demux audio and video, and then look whether there was a subtitle packet. This means to get subtitles for a time range, you need to know that you demuxed video and audio over this range, which becomes pretty messy when you demux audio and video backwards separately. Backward display is the most weird (and potentially buggy) part. To avoid that we need to touch a LOT of timing code, we negate all timestamps. The basic idea is that due to the navigation, all comparisons and subtractions of timestamps keep working, and you don't need to touch every single of them to "reverse" them. E.g.: bool before = pts_a < pts_b; would need to be: bool before = forward ? pts_a < pts_b : pts_a > pts_b; or: bool before = pts_a * dir < pts_b * dir; or if you, as it's implemented now, just do this after decoding: pts_a *= dir; pts_b *= dir; and then in the normal timing/renderer code: bool before = pts_a < pts_b; Consequently, we don't need many changes in the latter code. But some assumptions inhererently true for forward playback may have been broken anyway. What is mainly needed is fixing places where values are passed between positive and negative "domains". For example, seeking and timestamp user display always uses positive timestamps. The main mess is that it's not obvious which domain a given variable should or does use. Well, in my tests with a single file, it suddenly started to work when I did this. I'm honestly surprised that it did, and that I didn't have to change a single line in the timing code past decoder (just something minor to make external/cached text subtitles display). I committed it immediately while avoiding thinking about it. But there really likely are subtle problems of all sorts. As far as I'm aware, gstreamer also supports backward playback. When I looked at this years ago, I couldn't find a way to actually try this, and I didn't revisit it now. Back then I also read talk slides from the person who implemented it, and I'm not sure if and which ideas I might have taken from it. It's possible that the timestamp reversal is inspired by it, but I didn't check. (I think it claimed that it could avoid large changes by changing a sign?) VapourSynth has some sort of reverse function, which provides a backward view on a video. The function itself is trivial to implement, as VapourSynth aims to provide random access to video by frame numbers (so you just request decreasing frame numbers). From what I remember, it wasn't exactly fluid, but it worked. It's implemented by creating an index, and seeking to the target on demand, and a bunch of caching. mpv could use it, but it would either require using VapourSynth as demuxer and decoder for everything, or replacing the current file every time something is supposed to be played backwards. FFmpeg's libavfilter has reversal filters for audio and video. These require buffering the entire media data of the file, and don't really fit into mpv's architecture. It could be used by playing a libavfilter graph that also demuxes, but that's like VapourSynth but worse.
2019-05-18 00:10:51 +00:00
int mp_image_approx_byte_size(struct mp_image *img)
{
int total = sizeof(*img);
for (int n = 0; n < MP_MAX_PLANES; n++) {
struct AVBufferRef *buf = img->bufs[n];
if (buf)
total += buf->size;
}
return total;
}
2012-12-11 23:43:36 +00:00
struct mp_image *mp_image_new_copy(struct mp_image *img)
{
struct mp_image *new = mp_image_alloc(img->imgfmt, img->w, img->h);
if (!new)
return NULL;
2012-12-11 23:43:36 +00:00
mp_image_copy(new, img);
mp_image_copy_attributes(new, img);
return new;
}
// Make dst take over the image data of src, and free src.
// This is basically a safe version of *dst = *src; free(src);
// Only works with ref-counted images, and can't change image size/format.
void mp_image_steal_data(struct mp_image *dst, struct mp_image *src)
{
assert(dst->imgfmt == src->imgfmt && dst->w == src->w && dst->h == src->h);
assert(dst->bufs[0] && src->bufs[0]);
2012-12-11 23:43:36 +00:00
mp_image_destructor(dst); // unref old
talloc_free_children(dst);
2012-12-11 23:43:36 +00:00
*dst = *src;
*src = (struct mp_image){0};
2012-12-11 23:43:36 +00:00
talloc_free(src);
}
// Unref most data buffer (and clear the data array), but leave other fields
// allocated. In particular, mp_image.hwctx is preserved.
void mp_image_unref_data(struct mp_image *img)
{
for (int n = 0; n < MP_MAX_PLANES; n++) {
img->planes[n] = NULL;
img->stride[n] = 0;
av_buffer_unref(&img->bufs[n]);
}
}
static void ref_buffer(bool *ok, AVBufferRef **dst)
{
if (*dst) {
*dst = av_buffer_ref(*dst);
if (!*dst)
*ok = false;
}
}
2012-12-11 23:43:36 +00:00
// Return a new reference to img. The returned reference is owned by the caller,
// while img is left untouched.
struct mp_image *mp_image_new_ref(struct mp_image *img)
{
if (!img)
return NULL;
if (!img->bufs[0])
2012-12-11 23:43:36 +00:00
return mp_image_new_copy(img);
struct mp_image *new = talloc_ptrtype(NULL, new);
talloc_set_destructor(new, mp_image_destructor);
*new = *img;
bool ok = true;
for (int p = 0; p < MP_MAX_PLANES; p++)
ref_buffer(&ok, &new->bufs[p]);
ref_buffer(&ok, &new->hwctx);
ref_buffer(&ok, &new->icc_profile);
ref_buffer(&ok, &new->a53_cc);
ref_buffer(&ok, &new->dovi);
ref_buffer(&ok, &new->film_grain);
new->ff_side_data = talloc_memdup(NULL, new->ff_side_data,
new->num_ff_side_data * sizeof(new->ff_side_data[0]));
for (int n = 0; n < new->num_ff_side_data; n++)
ref_buffer(&ok, &new->ff_side_data[n].buf);
if (ok)
return new;
// Do this after _all_ bufs were changed; we don't want it to free bufs
// from the original image if this fails.
talloc_free(new);
return NULL;
2012-12-11 23:43:36 +00:00
}
struct free_args {
void *arg;
void (*free)(void *arg);
};
static void call_free(void *opaque, uint8_t *data)
{
struct free_args *args = opaque;
args->free(args->arg);
talloc_free(args);
}
// Create a new mp_image based on img, but don't set any buffers.
// Using this is only valid until the original img is unreferenced (including
// implicit unreferencing of the data by mp_image_make_writeable()), unless
// a new reference is set.
struct mp_image *mp_image_new_dummy_ref(struct mp_image *img)
2012-12-11 23:43:36 +00:00
{
struct mp_image *new = talloc_ptrtype(NULL, new);
talloc_set_destructor(new, mp_image_destructor);
*new = img ? *img : (struct mp_image){0};
for (int p = 0; p < MP_MAX_PLANES; p++)
new->bufs[p] = NULL;
new->hwctx = NULL;
new->icc_profile = NULL;
new->a53_cc = NULL;
new->dovi = NULL;
new->film_grain = NULL;
new->num_ff_side_data = 0;
new->ff_side_data = NULL;
2012-12-11 23:43:36 +00:00
return new;
}
// Return a reference counted reference to img. If the reference count reaches
// 0, call free(free_arg). The data passed by img must not be free'd before
// that. The new reference will be writeable.
// On allocation failure, unref the frame and return NULL.
// This is only used for hw decoding; this is important, because libav* expects
// all plane data to be accounted for by AVBufferRefs.
struct mp_image *mp_image_new_custom_ref(struct mp_image *img, void *free_arg,
void (*free)(void *arg))
{
struct mp_image *new = mp_image_new_dummy_ref(img);
struct free_args *args = talloc_ptrtype(NULL, args);
*args = (struct free_args){free_arg, free};
new->bufs[0] = av_buffer_create(NULL, 0, call_free, args,
AV_BUFFER_FLAG_READONLY);
if (new->bufs[0])
return new;
talloc_free(new);
return NULL;
}
2012-12-11 23:43:36 +00:00
bool mp_image_is_writeable(struct mp_image *img)
{
if (!img->bufs[0])
return true; // not ref-counted => always considered writeable
for (int p = 0; p < MP_MAX_PLANES; p++) {
if (!img->bufs[p])
break;
if (!av_buffer_is_writable(img->bufs[p]))
return false;
}
return true;
2012-12-11 23:43:36 +00:00
}
// Make the image data referenced by img writeable. This allocates new data
// if the data wasn't already writeable, and img->planes[] and img->stride[]
// will be set to the copy.
// Returns success; if false is returned, the image could not be made writeable.
bool mp_image_make_writeable(struct mp_image *img)
2012-12-11 23:43:36 +00:00
{
if (mp_image_is_writeable(img))
return true;
2012-12-11 23:43:36 +00:00
struct mp_image *new = mp_image_new_copy(img);
if (!new)
return false;
mp_image_steal_data(img, new);
2012-12-11 23:43:36 +00:00
assert(mp_image_is_writeable(img));
return true;
2012-12-11 23:43:36 +00:00
}
// Helper function: unrefs *p_img, and sets *p_img to a new ref of new_value.
// Only unrefs *p_img and sets it to NULL if out of memory.
2012-12-11 23:43:36 +00:00
void mp_image_setrefp(struct mp_image **p_img, struct mp_image *new_value)
{
if (*p_img != new_value) {
talloc_free(*p_img);
*p_img = new_value ? mp_image_new_ref(new_value) : NULL;
}
}
// Mere helper function (mp_image can be directly free'd with talloc_free)
void mp_image_unrefp(struct mp_image **p_img)
{
talloc_free(*p_img);
*p_img = NULL;
}
void memcpy_pic(void *dst, const void *src, int bytesPerLine, int height,
int dstStride, int srcStride)
{
if (bytesPerLine == dstStride && dstStride == srcStride && height) {
if (srcStride < 0) {
src = (uint8_t*)src + (height - 1) * srcStride;
dst = (uint8_t*)dst + (height - 1) * dstStride;
srcStride = -srcStride;
}
memcpy(dst, src, srcStride * (height - 1) + bytesPerLine);
} else {
for (int i = 0; i < height; i++) {
memcpy(dst, src, bytesPerLine);
src = (uint8_t*)src + srcStride;
dst = (uint8_t*)dst + dstStride;
}
}
}
void mp_image_copy(struct mp_image *dst, struct mp_image *src)
{
assert(dst->imgfmt == src->imgfmt);
assert(dst->w == src->w && dst->h == src->h);
assert(mp_image_is_writeable(dst));
for (int n = 0; n < dst->num_planes; n++) {
int line_bytes = (mp_image_plane_w(dst, n) * dst->fmt.bpp[n] + 7) / 8;
int plane_h = mp_image_plane_h(dst, n);
memcpy_pic(dst->planes[n], src->planes[n], line_bytes, plane_h,
dst->stride[n], src->stride[n]);
}
if (dst->fmt.flags & MP_IMGFLAG_PAL)
memcpy(dst->planes[1], src->planes[1], AVPALETTE_SIZE);
}
static enum mp_csp mp_image_params_get_forced_csp(struct mp_image_params *params)
{
int imgfmt = params->hw_subfmt ? params->hw_subfmt : params->imgfmt;
return mp_imgfmt_get_forced_csp(imgfmt);
}
static void assign_bufref(AVBufferRef **dst, AVBufferRef *new)
{
av_buffer_unref(dst);
if (new) {
*dst = av_buffer_ref(new);
MP_HANDLE_OOM(*dst);
}
}
void mp_image_copy_attributes(struct mp_image *dst, struct mp_image *src)
{
dst->pict_type = src->pict_type;
dst->fields = src->fields;
dst->pts = src->pts;
dst->dts = src->dts;
dst->pkt_duration = src->pkt_duration;
dst->params.rotate = src->params.rotate;
dst->params.stereo3d = src->params.stereo3d;
dst->params.p_w = src->params.p_w;
dst->params.p_h = src->params.p_h;
dst->params.color = src->params.color;
dst->params.chroma_location = src->params.chroma_location;
dst->params.alpha = src->params.alpha;
dst->nominal_fps = src->nominal_fps;
// ensure colorspace consistency
if (mp_image_params_get_forced_csp(&dst->params) !=
mp_image_params_get_forced_csp(&src->params))
dst->params.color = (struct mp_colorspace){0};
if ((dst->fmt.flags & MP_IMGFLAG_PAL) && (src->fmt.flags & MP_IMGFLAG_PAL)) {
if (dst->planes[1] && src->planes[1]) {
if (mp_image_make_writeable(dst))
memcpy(dst->planes[1], src->planes[1], AVPALETTE_SIZE);
}
}
assign_bufref(&dst->icc_profile, src->icc_profile);
assign_bufref(&dst->dovi, src->dovi);
assign_bufref(&dst->film_grain, src->film_grain);
assign_bufref(&dst->a53_cc, src->a53_cc);
}
// Crop the given image to (x0, y0)-(x1, y1) (bottom/right border exclusive)
// x0/y0 must be naturally aligned.
void mp_image_crop(struct mp_image *img, int x0, int y0, int x1, int y1)
{
assert(x0 >= 0 && y0 >= 0);
assert(x0 <= x1 && y0 <= y1);
assert(x1 <= img->w && y1 <= img->h);
assert(!(x0 & (img->fmt.align_x - 1)));
assert(!(y0 & (img->fmt.align_y - 1)));
for (int p = 0; p < img->num_planes; ++p) {
img->planes[p] += (y0 >> img->fmt.ys[p]) * img->stride[p] +
(x0 >> img->fmt.xs[p]) * img->fmt.bpp[p] / 8;
}
mp_image_set_size(img, x1 - x0, y1 - y0);
}
void mp_image_crop_rc(struct mp_image *img, struct mp_rect rc)
{
mp_image_crop(img, rc.x0, rc.y0, rc.x1, rc.y1);
}
// Repeatedly write count patterns of src[0..src_size] to p.
static void memset_pattern(void *p, size_t count, uint8_t *src, size_t src_size)
{
assert(src_size >= 1);
if (src_size == 1) {
memset(p, src[0], count);
} else if (src_size == 2) { // >8 bit YUV => common, be slightly less naive
uint16_t val;
memcpy(&val, src, 2);
uint16_t *p16 = p;
while (count--)
*p16++ = val;
} else {
while (count--) {
memcpy(p, src, src_size);
p = (char *)p + src_size;
}
}
}
static bool endian_swap_bytes(void *d, size_t bytes, size_t word_size)
{
if (word_size != 2 && word_size != 4)
return false;
size_t num_words = bytes / word_size;
uint8_t *ud = d;
switch (word_size) {
case 2:
for (size_t x = 0; x < num_words; x++)
AV_WL16(ud + x * 2, AV_RB16(ud + x * 2));
break;
case 4:
for (size_t x = 0; x < num_words; x++)
AV_WL32(ud + x * 2, AV_RB32(ud + x * 2));
break;
default:
MP_ASSERT_UNREACHABLE();
}
return true;
}
// Bottom/right border is allowed not to be aligned, but it might implicitly
// overwrite pixel data until the alignment (align_x/align_y) is reached.
// Alpha is cleared to 0 (fully transparent).
void mp_image_clear(struct mp_image *img, int x0, int y0, int x1, int y1)
{
assert(x0 >= 0 && y0 >= 0);
assert(x0 <= x1 && y0 <= y1);
assert(x1 <= img->w && y1 <= img->h);
assert(!(x0 & (img->fmt.align_x - 1)));
assert(!(y0 & (img->fmt.align_y - 1)));
struct mp_image area = *img;
struct mp_imgfmt_desc *fmt = &area.fmt;
mp_image_crop(&area, x0, y0, x1, y1);
// "Black" color for each plane.
uint8_t plane_clear[MP_MAX_PLANES][8] = {0};
int plane_size[MP_MAX_PLANES] = {0};
int misery = 1; // pixel group width
// YUV integer chroma needs special consideration, and technically luma is
// usually not 0 either.
if ((fmt->flags & (MP_IMGFLAG_HAS_COMPS | MP_IMGFLAG_PACKED_SS_YUV)) &&
(fmt->flags & MP_IMGFLAG_TYPE_MASK) == MP_IMGFLAG_TYPE_UINT &&
(fmt->flags & MP_IMGFLAG_COLOR_MASK) == MP_IMGFLAG_COLOR_YUV)
{
uint64_t plane_clear_i[MP_MAX_PLANES] = {0};
// Need to handle "multiple" pixels with packed YUV.
uint8_t luma_offsets[4] = {0};
if (fmt->flags & MP_IMGFLAG_PACKED_SS_YUV) {
misery = fmt->align_x;
if (misery <= MP_ARRAY_SIZE(luma_offsets)) // ignore if out of bounds
mp_imgfmt_get_packed_yuv_locations(fmt->id, luma_offsets);
}
for (int c = 0; c < 4; c++) {
struct mp_imgfmt_comp_desc *cd = &fmt->comps[c];
int plane_bits = fmt->bpp[cd->plane] * misery;
if (plane_bits <= 64 && plane_bits % 8u == 0 && cd->size) {
plane_size[cd->plane] = plane_bits / 8u;
int depth = cd->size + MPMIN(cd->pad, 0);
double m, o;
mp_get_csp_uint_mul(area.params.color.space,
area.params.color.levels,
depth, c + 1, &m, &o);
uint64_t val = MPCLAMP(lrint((0 - o) / m), 0, 1ull << depth);
plane_clear_i[cd->plane] |= val << cd->offset;
for (int x = 1; x < (c ? 0 : misery); x++)
plane_clear_i[cd->plane] |= val << luma_offsets[x];
}
}
for (int p = 0; p < MP_MAX_PLANES; p++) {
if (!plane_clear_i[p])
plane_size[p] = 0;
memcpy(&plane_clear[p][0], &plane_clear_i[p], 8); // endian dependent
if (fmt->endian_shift) {
endian_swap_bytes(&plane_clear[p][0], plane_size[p],
1 << fmt->endian_shift);
}
}
}
for (int p = 0; p < area.num_planes; p++) {
int p_h = mp_image_plane_h(&area, p);
int p_w = mp_image_plane_w(&area, p);
for (int y = 0; y < p_h; y++) {
void *ptr = area.planes[p] + (ptrdiff_t)area.stride[p] * y;
if (plane_size[p]) {
memset_pattern(ptr, p_w / misery, plane_clear[p], plane_size[p]);
} else {
memset(ptr, 0, mp_image_plane_bytes(&area, p, 0, area.w));
}
}
}
}
void mp_image_clear_rc(struct mp_image *mpi, struct mp_rect rc)
{
mp_image_clear(mpi, rc.x0, rc.y0, rc.x1, rc.y1);
}
// Clear the are of the image _not_ covered by rc.
void mp_image_clear_rc_inv(struct mp_image *mpi, struct mp_rect rc)
{
struct mp_rect clr[4];
int cnt = mp_rect_subtract(&(struct mp_rect){0, 0, mpi->w, mpi->h}, &rc, clr);
for (int n = 0; n < cnt; n++)
mp_image_clear_rc(mpi, clr[n]);
}
void mp_image_vflip(struct mp_image *img)
{
for (int p = 0; p < img->num_planes; p++) {
int plane_h = mp_image_plane_h(img, p);
img->planes[p] = img->planes[p] + img->stride[p] * (plane_h - 1);
img->stride[p] = -img->stride[p];
}
}
// Display size derived from image size and pixel aspect ratio.
void mp_image_params_get_dsize(const struct mp_image_params *p,
int *d_w, int *d_h)
{
*d_w = p->w;
*d_h = p->h;
if (p->p_w > p->p_h && p->p_h >= 1)
*d_w = MPCLAMP(*d_w * (int64_t)p->p_w / p->p_h, 1, INT_MAX);
if (p->p_h > p->p_w && p->p_w >= 1)
*d_h = MPCLAMP(*d_h * (int64_t)p->p_h / p->p_w, 1, INT_MAX);
}
void mp_image_params_set_dsize(struct mp_image_params *p, int d_w, int d_h)
{
AVRational ds = av_div_q((AVRational){d_w, d_h}, (AVRational){p->w, p->h});
p->p_w = ds.num;
p->p_h = ds.den;
}
char *mp_image_params_to_str_buf(char *b, size_t bs,
const struct mp_image_params *p)
{
if (p && p->imgfmt) {
snprintf(b, bs, "%dx%d", p->w, p->h);
if (p->p_w != p->p_h || !p->p_w)
mp_snprintf_cat(b, bs, " [%d:%d]", p->p_w, p->p_h);
mp_snprintf_cat(b, bs, " %s", mp_imgfmt_to_name(p->imgfmt));
if (p->hw_subfmt)
mp_snprintf_cat(b, bs, "[%s]", mp_imgfmt_to_name(p->hw_subfmt));
mp_snprintf_cat(b, bs, " %s/%s/%s/%s/%s",
m_opt_choice_str(mp_csp_names, p->color.space),
m_opt_choice_str(mp_csp_prim_names, p->color.primaries),
m_opt_choice_str(mp_csp_trc_names, p->color.gamma),
m_opt_choice_str(mp_csp_levels_names, p->color.levels),
m_opt_choice_str(mp_csp_light_names, p->color.light));
if (p->color.sig_peak)
mp_snprintf_cat(b, bs, " SP=%f", p->color.sig_peak);
mp_snprintf_cat(b, bs, " CL=%s",
m_opt_choice_str(mp_chroma_names, p->chroma_location));
if (p->rotate)
mp_snprintf_cat(b, bs, " rot=%d", p->rotate);
if (p->stereo3d > 0) {
mp_snprintf_cat(b, bs, " stereo=%s",
MP_STEREO3D_NAME_DEF(p->stereo3d, "?"));
}
if (p->alpha) {
mp_snprintf_cat(b, bs, " A=%s",
m_opt_choice_str(mp_alpha_names, p->alpha));
}
} else {
snprintf(b, bs, "???");
}
return b;
}
// Return whether the image parameters are valid.
// Some non-essential fields are allowed to be unset (like colorspace flags).
bool mp_image_params_valid(const struct mp_image_params *p)
{
// av_image_check_size has similar checks and triggers around 16000*16000
// It's mostly needed to deal with the fact that offsets are sometimes
// ints. We also should (for now) do the same as FFmpeg, to be sure large
// images don't crash with libswscale or when wrapping with AVFrame and
// passing the result to filters.
if (p->w <= 0 || p->h <= 0 || (p->w + 128LL) * (p->h + 128LL) >= INT_MAX / 8)
return false;
if (p->p_w < 0 || p->p_h < 0)
return false;
if (p->rotate < 0 || p->rotate >= 360)
return false;
struct mp_imgfmt_desc desc = mp_imgfmt_get_desc(p->imgfmt);
if (!desc.id)
return false;
if (p->hw_subfmt && !(desc.flags & MP_IMGFLAG_HWACCEL))
return false;
return true;
}
bool mp_image_params_equal(const struct mp_image_params *p1,
const struct mp_image_params *p2)
{
return p1->imgfmt == p2->imgfmt &&
p1->hw_subfmt == p2->hw_subfmt &&
p1->w == p2->w && p1->h == p2->h &&
p1->p_w == p2->p_w && p1->p_h == p2->p_h &&
mp_colorspace_equal(p1->color, p2->color) &&
2014-04-20 19:28:09 +00:00
p1->chroma_location == p2->chroma_location &&
p1->rotate == p2->rotate &&
p1->stereo3d == p2->stereo3d &&
p1->alpha == p2->alpha;
}
// Set most image parameters, but not image format or size.
// Display size is used to set the PAR.
void mp_image_set_attributes(struct mp_image *image,
const struct mp_image_params *params)
{
struct mp_image_params nparams = *params;
nparams.imgfmt = image->imgfmt;
nparams.w = image->w;
nparams.h = image->h;
if (nparams.imgfmt != params->imgfmt)
nparams.color = (struct mp_colorspace){0};
mp_image_set_params(image, &nparams);
}
// If details like params->colorspace/colorlevels are missing, guess them from
// the other settings. Also, even if they are set, make them consistent with
// the colorspace as implied by the pixel format.
void mp_image_params_guess_csp(struct mp_image_params *params)
{
enum mp_csp forced_csp = mp_image_params_get_forced_csp(params);
if (forced_csp == MP_CSP_AUTO) { // YUV/other
if (params->color.space != MP_CSP_BT_601 &&
params->color.space != MP_CSP_BT_709 &&
params->color.space != MP_CSP_BT_2020_NC &&
params->color.space != MP_CSP_BT_2020_C &&
params->color.space != MP_CSP_SMPTE_240M &&
params->color.space != MP_CSP_YCGCO)
{
// Makes no sense, so guess instead
// YCGCO should be separate, but libavcodec disagrees
params->color.space = MP_CSP_AUTO;
}
if (params->color.space == MP_CSP_AUTO)
params->color.space = mp_csp_guess_colorspace(params->w, params->h);
if (params->color.levels == MP_CSP_LEVELS_AUTO) {
if (params->color.gamma == MP_CSP_TRC_V_LOG) {
params->color.levels = MP_CSP_LEVELS_PC;
} else {
params->color.levels = MP_CSP_LEVELS_TV;
}
}
if (params->color.primaries == MP_CSP_PRIM_AUTO) {
// Guess based on the colormatrix as a first priority
if (params->color.space == MP_CSP_BT_2020_NC ||
params->color.space == MP_CSP_BT_2020_C) {
params->color.primaries = MP_CSP_PRIM_BT_2020;
} else if (params->color.space == MP_CSP_BT_709) {
params->color.primaries = MP_CSP_PRIM_BT_709;
} else {
// Ambiguous colormatrix for BT.601, guess based on res
params->color.primaries = mp_csp_guess_primaries(params->w, params->h);
}
}
if (params->color.gamma == MP_CSP_TRC_AUTO)
params->color.gamma = MP_CSP_TRC_BT_1886;
} else if (forced_csp == MP_CSP_RGB) {
params->color.space = MP_CSP_RGB;
params->color.levels = MP_CSP_LEVELS_PC;
// The majority of RGB content is either sRGB or (rarely) some other
// color space which we don't even handle, like AdobeRGB or
// ProPhotoRGB. The only reasonable thing we can do is assume it's
// sRGB and hope for the best, which should usually just work out fine.
// Note: sRGB primaries = BT.709 primaries
if (params->color.primaries == MP_CSP_PRIM_AUTO)
params->color.primaries = MP_CSP_PRIM_BT_709;
if (params->color.gamma == MP_CSP_TRC_AUTO)
params->color.gamma = MP_CSP_TRC_SRGB;
} else if (forced_csp == MP_CSP_XYZ) {
params->color.space = MP_CSP_XYZ;
params->color.levels = MP_CSP_LEVELS_PC;
// In theory, XYZ data does not really need a concept of 'primaries' to
// function, but this field can still be relevant for guiding gamut
// mapping optimizations, and it's also used by `mp_get_csp_matrix`
// when deciding what RGB space to map XYZ to for VOs that don't want
// to directly ingest XYZ into their color pipeline. BT.709 would be a
// sane default here, but it runs the risk of clipping any wide gamut
// content, so we pick BT.2020 instead to be on the safer side.
if (params->color.primaries == MP_CSP_PRIM_AUTO)
params->color.primaries = MP_CSP_PRIM_BT_2020;
if (params->color.gamma == MP_CSP_TRC_AUTO)
params->color.gamma = MP_CSP_TRC_LINEAR;
} else {
// We have no clue.
params->color.space = MP_CSP_AUTO;
params->color.levels = MP_CSP_LEVELS_AUTO;
params->color.primaries = MP_CSP_PRIM_AUTO;
params->color.gamma = MP_CSP_TRC_AUTO;
}
if (!params->color.sig_peak) {
if (params->color.gamma == MP_CSP_TRC_HLG) {
params->color.sig_peak = 1000 / MP_REF_WHITE; // reference display
} else {
// If the signal peak is unknown, we're forced to pick the TRC's
// nominal range as the signal peak to prevent clipping
params->color.sig_peak = mp_trc_nom_peak(params->color.gamma);
}
}
if (!mp_trc_is_hdr(params->color.gamma)) {
// Some clips have leftover HDR metadata after conversion to SDR, so to
// avoid blowing up the tone mapping code, strip/sanitize it
params->color.sig_peak = 1.0;
}
if (params->chroma_location == MP_CHROMA_AUTO) {
if (params->color.levels == MP_CSP_LEVELS_TV)
params->chroma_location = MP_CHROMA_LEFT;
if (params->color.levels == MP_CSP_LEVELS_PC)
params->chroma_location = MP_CHROMA_CENTER;
}
if (params->color.light == MP_CSP_LIGHT_AUTO) {
// HLG is always scene-referred (using its own OOTF), everything else
// we assume is display-refered by default.
if (params->color.gamma == MP_CSP_TRC_HLG) {
params->color.light = MP_CSP_LIGHT_SCENE_HLG;
} else {
params->color.light = MP_CSP_LIGHT_DISPLAY;
}
}
}
// Create a new mp_image reference to av_frame.
struct mp_image *mp_image_from_av_frame(struct AVFrame *src)
{
struct mp_image *dst = &(struct mp_image){0};
AVFrameSideData *sd;
for (int p = 0; p < MP_MAX_PLANES; p++)
dst->bufs[p] = src->buf[p];
dst->hwctx = src->hw_frames_ctx;
mp_image_setfmt(dst, pixfmt2imgfmt(src->format));
mp_image_set_size(dst, src->width, src->height);
dst->params.p_w = src->sample_aspect_ratio.num;
dst->params.p_h = src->sample_aspect_ratio.den;
for (int i = 0; i < 4; i++) {
dst->planes[i] = src->data[i];
dst->stride[i] = src->linesize[i];
}
dst->pict_type = src->pict_type;
dst->fields = 0;
if (src->interlaced_frame)
dst->fields |= MP_IMGFIELD_INTERLACED;
if (src->top_field_first)
dst->fields |= MP_IMGFIELD_TOP_FIRST;
if (src->repeat_pict == 1)
dst->fields |= MP_IMGFIELD_REPEAT_FIRST;
dst->params.color = (struct mp_colorspace){
.space = avcol_spc_to_mp_csp(src->colorspace),
.levels = avcol_range_to_mp_csp_levels(src->color_range),
.primaries = avcol_pri_to_mp_csp_prim(src->color_primaries),
.gamma = avcol_trc_to_mp_csp_trc(src->color_trc),
};
dst->params.chroma_location = avchroma_location_to_mp(src->chroma_location);
if (src->opaque_ref) {
struct mp_image_params *p = (void *)src->opaque_ref->data;
dst->params.rotate = p->rotate;
dst->params.stereo3d = p->stereo3d;
// Might be incorrect if colorspace changes.
dst->params.color.light = p->color.light;
dst->params.alpha = p->alpha;
}
sd = av_frame_get_side_data(src, AV_FRAME_DATA_DISPLAYMATRIX);
if (sd) {
double r = av_display_rotation_get((int32_t *)(sd->data));
if (!isnan(r))
dst->params.rotate = (((int)(-r) % 360) + 360) % 360;
}
sd = av_frame_get_side_data(src, AV_FRAME_DATA_ICC_PROFILE);
if (sd)
dst->icc_profile = sd->buf;
// Get the content light metadata if available
sd = av_frame_get_side_data(src, AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
if (sd) {
AVContentLightMetadata *clm = (AVContentLightMetadata *)sd->data;
dst->params.color.sig_peak = clm->MaxCLL / MP_REF_WHITE;
}
// Otherwise, try getting the mastering metadata if available
sd = av_frame_get_side_data(src, AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
if (!dst->params.color.sig_peak && sd) {
AVMasteringDisplayMetadata *mdm = (AVMasteringDisplayMetadata *)sd->data;
if (mdm->has_luminance)
dst->params.color.sig_peak = av_q2d(mdm->max_luminance) / MP_REF_WHITE;
}
sd = av_frame_get_side_data(src, AV_FRAME_DATA_A53_CC);
if (sd)
dst->a53_cc = sd->buf;
#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(57, 16, 100)
sd = av_frame_get_side_data(src, AV_FRAME_DATA_DOVI_METADATA);
if (sd) {
// Strip DoVi metadata that requires an EL, since it's near-impossible
// for us to support easily or sanely
const AVDOVIMetadata *metadata = (AVDOVIMetadata *) sd->buf->data;
const AVDOVIRpuDataHeader *rpu = av_dovi_get_header(metadata);
if (rpu->disable_residual_flag)
dst->dovi = sd->buf;
}
#endif
sd = av_frame_get_side_data(src, AV_FRAME_DATA_FILM_GRAIN_PARAMS);
if (sd)
dst->film_grain = sd->buf;
for (int n = 0; n < src->nb_side_data; n++) {
sd = src->side_data[n];
struct mp_ff_side_data mpsd = {
.type = sd->type,
.buf = sd->buf,
};
MP_TARRAY_APPEND(NULL, dst->ff_side_data, dst->num_ff_side_data, mpsd);
}
if (dst->hwctx) {
AVHWFramesContext *fctx = (void *)dst->hwctx->data;
dst->params.hw_subfmt = pixfmt2imgfmt(fctx->sw_format);
}
struct mp_image *res = mp_image_new_ref(dst);
// Allocated, but non-refcounted data.
talloc_free(dst->ff_side_data);
return res;
}
// Convert the mp_image reference to a AVFrame reference.
struct AVFrame *mp_image_to_av_frame(struct mp_image *src)
{
struct mp_image *new_ref = mp_image_new_ref(src);
AVFrame *dst = av_frame_alloc();
if (!dst || !new_ref) {
talloc_free(new_ref);
av_frame_free(&dst);
return NULL;
}
for (int p = 0; p < MP_MAX_PLANES; p++) {
dst->buf[p] = new_ref->bufs[p];
new_ref->bufs[p] = NULL;
}
dst->hw_frames_ctx = new_ref->hwctx;
new_ref->hwctx = NULL;
dst->format = imgfmt2pixfmt(src->imgfmt);
dst->width = src->w;
dst->height = src->h;
dst->sample_aspect_ratio.num = src->params.p_w;
dst->sample_aspect_ratio.den = src->params.p_h;
for (int i = 0; i < 4; i++) {
dst->data[i] = src->planes[i];
dst->linesize[i] = src->stride[i];
}
dst->extended_data = dst->data;
dst->pict_type = src->pict_type;
if (src->fields & MP_IMGFIELD_INTERLACED)
dst->interlaced_frame = 1;
if (src->fields & MP_IMGFIELD_TOP_FIRST)
dst->top_field_first = 1;
if (src->fields & MP_IMGFIELD_REPEAT_FIRST)
dst->repeat_pict = 1;
dst->colorspace = mp_csp_to_avcol_spc(src->params.color.space);
dst->color_range = mp_csp_levels_to_avcol_range(src->params.color.levels);
dst->color_primaries =
mp_csp_prim_to_avcol_pri(src->params.color.primaries);
dst->color_trc = mp_csp_trc_to_avcol_trc(src->params.color.gamma);
dst->chroma_location = mp_chroma_location_to_av(src->params.chroma_location);
dst->opaque_ref = av_buffer_alloc(sizeof(struct mp_image_params));
if (!dst->opaque_ref)
abort();
*(struct mp_image_params *)dst->opaque_ref->data = src->params;
if (src->icc_profile) {
AVFrameSideData *sd =
av_frame_new_side_data_from_buf(dst, AV_FRAME_DATA_ICC_PROFILE,
new_ref->icc_profile);
if (!sd)
abort();
new_ref->icc_profile = NULL;
}
if (src->params.color.sig_peak) {
AVContentLightMetadata *clm =
av_content_light_metadata_create_side_data(dst);
if (!clm)
abort();
clm->MaxCLL = src->params.color.sig_peak * MP_REF_WHITE;
}
// Add back side data, but only for types which are not specially handled
// above. Keep in mind that the types above will be out of sync anyway.
for (int n = 0; n < new_ref->num_ff_side_data; n++) {
struct mp_ff_side_data *mpsd = &new_ref->ff_side_data[n];
if (!av_frame_get_side_data(dst, mpsd->type)) {
AVFrameSideData *sd = av_frame_new_side_data_from_buf(dst, mpsd->type,
mpsd->buf);
if (!sd)
abort();
mpsd->buf = NULL;
}
}
talloc_free(new_ref);
if (dst->format == AV_PIX_FMT_NONE)
av_frame_free(&dst);
return dst;
}
// Same as mp_image_to_av_frame(), but unref img. (It does so even on failure.)
struct AVFrame *mp_image_to_av_frame_and_unref(struct mp_image *img)
{
AVFrame *frame = mp_image_to_av_frame(img);
talloc_free(img);
return frame;
}
void memset_pic(void *dst, int fill, int bytesPerLine, int height, int stride)
{
if (bytesPerLine == stride && height) {
memset(dst, fill, stride * (height - 1) + bytesPerLine);
} else {
for (int i = 0; i < height; i++) {
memset(dst, fill, bytesPerLine);
dst = (uint8_t *)dst + stride;
}
}
}
void memset16_pic(void *dst, int fill, int unitsPerLine, int height, int stride)
{
if (fill == 0) {
memset_pic(dst, 0, unitsPerLine * 2, height, stride);
} else {
for (int i = 0; i < height; i++) {
uint16_t *line = dst;
uint16_t *end = line + unitsPerLine;
while (line < end)
*line++ = fill;
dst = (uint8_t *)dst + stride;
}
}
}
// Pixel at the given luma position on the given plane. x/y always refer to
// non-subsampled coordinates (even if plane is chroma).
// The coordinates must be aligned to mp_imgfmt_desc.align_x/y (these are byte
// and chroma boundaries).
// You cannot access e.g. individual luma pixels on the luma plane with yuv420p.
void *mp_image_pixel_ptr(struct mp_image *img, int plane, int x, int y)
{
assert(MP_IS_ALIGNED(x, img->fmt.align_x));
assert(MP_IS_ALIGNED(y, img->fmt.align_y));
return mp_image_pixel_ptr_ny(img, plane, x, y);
}
// Like mp_image_pixel_ptr(), but do not require alignment on Y coordinates if
// the plane does not require it. Use with care.
// Useful for addressing luma rows.
void *mp_image_pixel_ptr_ny(struct mp_image *img, int plane, int x, int y)
{
assert(MP_IS_ALIGNED(x, img->fmt.align_x));
assert(MP_IS_ALIGNED(y, 1 << img->fmt.ys[plane]));
return img->planes[plane] +
img->stride[plane] * (ptrdiff_t)(y >> img->fmt.ys[plane]) +
(x >> img->fmt.xs[plane]) * (size_t)img->fmt.bpp[plane] / 8;
}
// Return size of pixels [x0, x0+w-1] in bytes. The coordinates refer to non-
// subsampled pixels (basically plane 0), and the size is rounded to chroma
// and byte alignment boundaries for the entire image, even if plane!=0.
// x0!=0 is useful for rounding (e.g. 8 bpp, x0=7, w=7 => 0..15 => 2 bytes).
size_t mp_image_plane_bytes(struct mp_image *img, int plane, int x0, int w)
{
int x1 = MP_ALIGN_UP(x0 + w, img->fmt.align_x);
x0 = MP_ALIGN_DOWN(x0, img->fmt.align_x);
size_t bpp = img->fmt.bpp[plane];
int xs = img->fmt.xs[plane];
return (x1 >> xs) * bpp / 8 - (x0 >> xs) * bpp / 8;
}