This will be useful for other video codecs, but putting all codecs in
one file will be a little messy.
(cherry picked from commit 48a21d5d6a203563eab60922f8c24202d0ed7e10)
This is necessary to give H.264 decoders the data they need to
initialize, including frame size and profile.
(cherry picked from commit 457a69786b788d3158813156d11e70980f50f7b7)
The function is separated from the string-to-enum function it uses, and
the order is also inconsistent with header.
(cherry picked from commit bf1e0fac94e0d8599c0c540c24883ca5f5ea4131)
These aren't particularly small objects, but we were still copying them
around all over the place. When TrackEntry contains data buffers, they
won't need to be copied as well.
(cherry picked from commit 55fda2068b7334acaba2673c80c01c019aaf7075)
This should halve the size of frames in memory for frames with 8-bit
color components, which is the majority of videos.
Calculation of the size of subsampled planes has also been consolidated
into a struct. There are likely some places that will still need to
change over to this, but it should prevent issues due to differing
handling of rounding/ceiling.
(cherry picked from commit 40fe0cb9d5c40a5ee568a3196bf19452ea8fed2b)
BT.2020 will mainly be used with bit depths greater than 8, so having
this specialization is mostly pointless until we use fixed-point math
for higher bit depths.
(cherry picked from commit fe2a63d485bc9d494b06a780043eae21adfd6854)
Apparently I forgot to put read the value for this field, though this
generally doesn't matter since video bitstreams usually specify CICP as
well.
(cherry picked from commit d3f88b4987730e396bc41c2674edddb53845cb37)
This change is in preparation for implementing audio codecs into the
library and using audio as timing for video playback.
(cherry picked from commit 7c10e1a08d7a109b63c9258578eb98aa9dcc1425)
No effect on sunset-retro.png since that's not animated.
wow.gif (nee giphy.gif) (184k):
1.4M -> 255K
74.0 ms ± 1.1 ms -> 86.9 ms ± 3.3 ms
(from 7.6x as big as the gif input to 1.4x as big.
About 82% smaller, for a 16% slowdown.)
7z7c.gif (11K):
8.4K -> 8.6K
12.9 ms ± 0.5 ms -> 12.7 ms ± 0.5 ms
(2.4% bigger, so the transform makes things a bit worse for this
image.)
AnimationWriter already only stores the smallest rect that contains
changing pixels between two frames. For example, when doing a screen
recording and only the mouse cursor moves, we already only encode
the pixels in the (single) rectangle containing old and new mouse cursor
positions.
Within that rectangle, there can still be many pixels that are identical
over the two frames. When possible, we now replace all identical pixels
with transparent black. This has two advantages:
1. It can reduce the number of colors in the image. In particular,
for wow.gif (and likely many other gifs), new frames had more
than 256 colors before, and have fewer than 256 colors after this
change.
2. Long run of identical pixels compress better.
In some cases, this transform might make things slighly worse,
for example if the input image already consists of long runs of
a single color. We'll now add another color to it (transparent black),
without it helping much. And the decoder now must do some blending,
slowing down decoding a bit.
But most of the time this should be a pretty big win. We can tweak
the heuristic when to do it later.
This transform is possible when:
* The new frame doesn't already have transparent pixels (which are
different from the old frame)
* The encoder/decoder can handle frames with transparent pixels
For the latter reason, encoders currently have to opt in to this.
This is an attempt to fix the hanging CI on macOS caused by some
screenshot requests being stuck unprocessed. With this change, we at
least make sure that the HTML event loop processing, which triggers
repainting, will happen as long as there are navigables that need to be
repainted.
(cherry picked from commit 72b4d44d07e12cc04bde90872e2f31aaab64aa00)
By Setting setBordered propperty on header buttons to `Yes` this
path makes the whole button clickable. Previously the only the
icon was clickable, now it's easy to click.
(cherry picked from commit af909784547c4f0ea8218ce464f45b3a1b0f4cb7)
There are a few instances where comments and documentation have minor
grammar issues likely resulting from English being the author's second
language.
This PR fixes several such cases, changing to idiomatic English and
resolving where it is unclear whether the user or program/code is
being referred to.
This avoids an unnecessary lossy conversion for the current time from
double to i32. And avoids an UBSAN failure on macOS that's dependent
on the current uptime.
(cherry picked from commit 55c1b5d1f4d7c82f0a68323260cb2e0f7de2faae,
amended to fix a typo in the commit message)
BSD `du` can also print apparent size, it just needs a different flag.
Makes `serenity.sh run` get a bit farther trying to build the disk
image on macOS without brew installed. No intended behavior change
elsewhere.
On several platforms, we don't yet have audio
support, so audio devices are missing. Instead of
having AudioServer crash repeatedly, and not
having the ability to open any app that relies on
it, we should instead handle missing devices
gracefully. This implementation is minimal, audio
device hotplugging support and such should be
implemented together with multi-device support.
AudioServer will start up and seem to function
normally without an audio device, but it will
pretend the device has a sample rate of 44.1 kHz
and all audio input is discarded.
The methods try_create_with_size() and try_create_purgeable_with_size()
on AnonymousVMObject are almost identical, other than one member
that gets set (m_purgeable). This patch makes
try_create_purgeable_with_size() call try_create_with_size() so that
both methods re-use the same code.
This avoids a segfault that would previously occur when middle clicking
to close a tab if only 1 tab was open.
(cherry picked from commit b95c05b6115c8eb84fe6fec55f5152db852fe743)
This implements the start of lossless webp's compression scheme,
which is almost, but not quite, entirely unlike deflate.
The green channel is now green-or-length, and has up to 280
entries, instead of up to 256 before. We now use the 40-entry
distance code (even though it only ever stores 1s now).
Due to this, a few places change to taking spans instead of
Array<256>s.
The spec only has the transform from prefix or distance code
to value. The inverse prefix_decompose() in this patch is
my own invention. I checked with a python script that it's
a true inverse (see PR for the script).
We now look for back-references with a distance of 1, which is
equivalent to run-length encoding. It's fairly fast to compute,
but leaves compression on the table. Full window-based
back references will be in a future PR.
We also still don't do color cache entries yet, but that should
be fairly straightforward to add. (It will make the green channel
larger than 280 entries.)
We still use a single global huffman table for the entire image.
Doing one per tile should be doable with the organization we now
have, and might also be in a future PR.
File sizes, and perf numbers on HEAD before this patch series (see
previous commit for perf comparison to previous commit):
sunset-retro.png (876K):
1.7M -> 1.6M,
25.3 ms ± 0.5 ms -> 27.5 ms ± 0.8 ms
(helps little; from 1.94x as input to 1.83x as large.
About 5% smaller, for about a 10% slowdown.)
wow.gif (nee giphy.gif) (184k):
3.9M -> 1.4M
105.7 ms ± 1.7 ms -> 74.0 ms ± 1.1 ms
(from 21.2x as big as the gif input to 7.6x as big.
About 64% smaller, for a 28% speed _up_.)
7z7c.gif (11K):
40K -> 8.4K
13.9 ms ± 0.6 ms -> 12.9 ms ± 0.5 ms
(from 3.6x as big as the gif input to 0.76x as big :^)
About 79% smaller, for a 7% speed _up_.)
We now do this in two passes instead of in one. This is virtually free
performance-wise, and allows nicer factoring.
Perf numbers after this change (see previous commit for perf numbers
before):
Benchmark 1: image -o sunset-retro.webp sunset-retro.bmp
Time (mean ± σ): 26.7 ms ± 0.8 ms
Benchmark 1: animation -o 7z7c.webp 7z7c.gif
Time (mean ± σ): 14.5 ms ± 0.6 ms
Benchmark 1: animation -o wow.webp wow.gif
Time (mean ± σ): 108.2 ms ± 2.2 ms
No behavior change yet, but this will allow us to emit distance/length
and color cache symbols in addition to literal symbols.
Not super expensive perf-wise. Before:
Benchmark 1: image -o sunset-retro.webp sunset-retro.bmp
Time (mean ± σ): 25.3 ms ± 0.5 ms
Benchmark 1: animation -o 7z7c.webp 7z7c.gif
Time (mean ± σ): 13.9 ms ± 0.6 ms
Benchmark 1: animation -o wow.webp wow.gif
Time (mean ± σ): 105.7 ms ± 1.7 ms
After:
Benchmark 1: image -o sunset-retro.webp sunset-retro.bmp
Time (mean ± σ): 26.1 ms ± 0.6 ms
Benchmark 1: animation -o 7z7c.webp 7z7c.gif
Time (mean ± σ): 14.4 ms ± 0.6 ms
Benchmark 1: animation -o wow.webp wow.gif
Time (mean ± σ): 106.5 ms ± 1.9 ms
Take Spans instead of Arrays. There's no need to have one copy of this
function for every possible array size passed to it.
Hardcode the inline size of the BinaryHeap to 288 for now. If this
becomes a performance issue in the future, we can make that size
an (optional) template parameter then.
No behavior change.
Works around https://gcc.gnu.org/PR70413. Without this, the next
commit would cause -Wsubobject-linkage warnings for AK/BinaryHeap.h
when used in LibCompress/Huffman.h.
We can undo this once we're on GCC 14 everywhere.
No behavior change.
VirtIO was probably broken on Windows versions of
QEMU, but this seems to no longer be the case.
Enabling VirtIO is currently required for RISC-V
mouse and keyboard support.
There are three compiler bugs that influence this decision:
- Clang writing to (validly) destroyed coroutine frame with -O0 and
-fsanitize=null,address under some conditions
(https://godbolt.org/z/17Efq5Ma5) (AK_COROUTINE_DESTRUCTION_BROKEN);
- GCC being unable to handle statement expressions in coroutines
(AK_COROUTINE_STATEMENT_EXPRS_BROKEN);
- GCC being unable to deduce template type parameter for TryAwaiter
with nested CO_TRYs (AK_COROUTINE_TYPE_DEDUCTION_BROKEN).
Instead of growing an ifdef soup in AK/Coroutine.h and
LibTest/AsyncTestCase.h, define three macros in AK/Platform.h that
correspond to these bugs and use them accordingly in the said files.
In case the user requests this, init can drop directly to a shell
without trying to spawn SystemServer.
To test this on x86-64, run:
```
Meta/serenity.sh run x86_64 GNU "init_args=emergency"
```
Also, init will drop to emergency shell if mounting filesystems with
`mount -a` failed for some reason.
This functionality can be useful in many cases.
For example, if the user needs to perform a command that must not alter
a corrupted filesystem state, then this mode is useful as the filesystem
should be mounted in read-only mode.
Another example is the ability to get a functioning system in case
SystemServer behaves badly or inconsistently, or the user specified a
wrong fstab entry, so proceeding to boot is probably a bad option.
Let's make SystemServer simpler by not involving it with the basic
system initialization sequence.
That initialization sequence can be done in another program that
theoretically can be put in another filesystem.
Co-authored-by: Tim Schumacher <timschumi@gmx.de>
