We must reload error just before calling dav1d_decode_frame_exit, as
it may have become stale between the last load and that call.
This can result in crashes since we signal a seemingly successfully decoded
frame, when it's not.
Reloading error within the frame done condition's body ensures a non-stale
value, as we use 'f->task_thread.task_counter == 0' to ensure all other
threads / tasks have already completed when entering it. In other words, only
the last thread still working on this frame can execute this code, after
all other threads have returned to doing something else.

Pack the 5 bytes of data to improve memory and perf.

This fixes a regression from 7409a189

Co-authored-by: Henrik Gramner <gramner@twoorioles.com>

Henrik Gramner authored 2 years ago and Henrik Gramner committed 2 years ago

Improves readability.

Henrik Gramner authored 2 years ago and Henrik Gramner committed 2 years ago

It would previously print the full report() info for C functions
(with broken horizontal alignment as a side effect).

98b0c96d added an include of
src/ref.h in src/fg_apply_tmpl.c. That template source file is
included in tests/checkasm/filmgrain.c.

src/ref.h includes <stdatomic.h>. Including this file requires
declaring a dependency on stdatomic_dependencies in meson, which
provides the fallback implementation of stdatomic.h when building
with MSVC.

Create new references instead.

Signed-off-by: James Almer <jamrial@gmail.com>

Signed-off-by: James Almer <jamrial@gmail.com>

"From VideoLAN with love"

"only and except are not being actively developed. rules is the
preferred keyword to control when to add jobs to pipelines." [0]

[0] https://docs.gitlab.com/ee/ci/yaml/index.html#only--except

Calling meson with no command is deprecated since 0.64.0

The code in dav1d_drain_picture could result in a desync between
c->task_thread.first (oldest submitted frame) and c->frame_thread.next (first
frame to retrieve and/or next submit location).
As we loop through drain, we always increment next, but first only if the
frame has data. If the frame is visible we return. The problem arises when
encountering (an) invisible frame(s), and the next entries haven't been fed
yet, we then keep on looping increasing next but not first, as these have no
data.

We should always return when we encountered data (visible or
invisible decoded frame): for visible, the code already returns, for
invisible, we can store a boolean indicating we drained at least one frame,
whenever we reach an empty entry after that, we return (all subsequent
entries are guaranteed to be empty anyway), not incrementing next nor first.
This will have the effect to insert the next frame at the first free spot
(which is much better than the weird skips it's doing now).

So basically, c->frame_thread.next could skip some (empty) entries.
Now it's contiguous.

Fixes #416.

This reverts commit a51b6ce4.

We can't increment first when no data is there, otherwise we might do it
while the first frame was not yet decoded, messing up ordering: imagine
having a framedelay of 8, and a file with 7 frames. We feed 7 frames over 8
slots, now next points to [7] (empty entry), and we start draining cause EOF.
We do need next to be incremented to reach the first frame ([0]), so it can
be outputted, and only then first too.

Fixes #418.

Add an option for selecting the core where the single thread of
checkasm runs. This allows benchmarking on specific CPU cores on
heterogenous CPUs, like ARM big.LITTLE configurations.

On Linux, one can easily wrap an invocation of checkasm with
"taskset -c <n> [...]" - so this option isn't very essential
there - however it is quite useful on Windows.

On Windows, it is somewhat possible to do the same by launching
the tool with "start /B /affinity <hexmask> [...]", but that
doesn't work well with scripting ("start" returns before the
command has finished running, and it's not obvious how to
invoke "start" from within WSL).

Using "taskset" to launch processes on specific cores within WSL
on Windows doesn't work - regardless of the Linux level affinity,
the process ends up running on the performance cores anyway.

The implementation is a hybrid between two approaches; one generic
(but non-ideal) for cases with large max_base_y, which fills two
pixel columns at a time, i.e. looping over pixels first vertically,
then horizontally - i.e. in a non-optimal manner.

For cases with smaller max_base_y, it does two rows at a time, essentially
doing gathers with the TBX instruction.

Relative speedup over the C code:

                         Cortex A53    A55    A72    A73    A76   Apple M1
intra_pred_z3_w4_8bpc_neon:    3.32   2.89   2.78   3.52   2.52   9.67
intra_pred_z3_w8_8bpc_neon:    6.24   5.55   4.76   5.60   4.11   6.40
intra_pred_z3_w16_8bpc_neon:   7.64   7.07   4.37   6.23   4.18   8.60
intra_pred_z3_w32_8bpc_neon:   7.51   7.21   4.34   5.92   4.27   7.88
intra_pred_z3_w64_8bpc_neon:   6.82   6.25   4.08   5.83   3.52   7.31

Relative speedup over the C code:

                         Cortex A53    A55    A72    A73    A76  Apple M1
intra_pred_z1_w4_8bpc_neon:    4.09   3.15   3.63   4.16   3.27  13.00
intra_pred_z1_w8_8bpc_neon:    6.93   5.66   5.57   6.76   5.51   5.50
intra_pred_z1_w16_8bpc_neon:   7.81   6.85   6.24   7.78   6.59   9.00
intra_pred_z1_w32_8bpc_neon:  10.56   9.95   8.72  10.95   8.28  13.33
intra_pred_z1_w64_8bpc_neon:  11.00  11.38   9.11  11.62   8.65  14.61

(The speedup numbers for M1 are kinda noisy due to the very coarse
granularity of the timer used there.)

These functions contain a number of different codepaths; try to
make sure that we hit most codepaths for each size combination.

This both gives better test coverage in one single run of checkasm,
but also should give a better averaged runtime in benchmarks.

Fixes #416.

Henrik Gramner authored 2 years ago and Henrik Gramner committed 2 years ago

The intent was good, but in practice it results in a significant
amount of problems due to various compiler bugs for negligible gains.

Should be useful for scenarios like wanting only keyframes to quickly generate
a set of preview images of the whole stream.

Henrik Gramner authored 2 years ago and Henrik Gramner committed 2 years ago

bits_left could underflow after reaching EOB.

Credit to OSS-Fuzz.

Henrik Gramner authored 2 years ago and Henrik Gramner committed 2 years ago

A length of 1 is by far the most common case, and having a special case
for that is not only slightly faster but also reduces code size by a
decent amount due to not having to pass a length argument every time.