To compute the number of pixels available in the top-right and
bottom-left edges, get_intra_edges() received frame_w_in_b (MiCols) and
frame_h_in_b (MiRows) as parameters, initialized as follow:

    MiCols = 2 * ( ( FrameWidth + 7 ) >> 3 )
    MiRows = 2 * ( ( FrameHeight + 7 ) >> 3 )

<https://aomediacodec.github.io/av1-spec/#compute-image-size-function>

The sizes computed by get_intra_edges() were basically the frame
dimensions rounded up to the next multiple of 8, decimated:

    (MI_SIZE >> plane_cfg.xdec) * frame_w_in_b
    (MI_SIZE >> plane_cfg.ydec) * frame_h_in_b

But in Frame::new(), the luma plane dimensions are also initialized with
the frame dimensions rounded up to the next multiple of 8. Therefore, it
is equivalent to directly use the plane dimensions.

The function get_sad() was called with block width and block height
parameters swapped.

As a consequence, in tests, associate the precomputed SAD values to the
transposed block size.

This function is not needed in rust and it is mostly a convenience for
other languages. Instead move this chunk in the appropriate bindings.

* Extract archives in parallel with download
* Fetch sccache binary release
* Use sccache for C and C++ dependencies
* Limit sccache size to 500M
* Use CI generic cache to store compiler cache

Romain Vimont authored 6 years ago and Thomas Daede committed 6 years ago

<https://github.com/xiph/rav1e/pull/1092#discussion_r264275780

>

Suggested-by: David Michael Barr <b@rr-dav.id.au>

Romain Vimont authored 6 years ago and Thomas Daede committed 6 years ago

The functions sad_ssse3() and sad_sse2() only support u16 and u8
respectively, so they are not generic.

Make the caller pass the expected type.

This doesn't actually call the assembly functdions yet.

It acts a aligned memory wrapper.

Fixes #1101

Preliminary to use a different backing storage.

Apparently cbindgen has problems parsing them in the former rendition.

This reduces the odds of a Travis timeout.

Also, check out the v1.0.0-errata1 tag of libaom.

Many of the settings change nothing at the default speed.

Make cargo doc happier and make crav1e not depend on compiler
unstable feature.

Vladimir Kazakov authored 6 years ago and Thomas Daede committed 6 years ago

* https://rust-lang.github.io/rust-clippy/master/index.html#erasing_op
* https://rust-lang.github.io/rust-clippy/master/index.html#identity_op
* https://rust-lang.github.io/rust-clippy/master/index.html#neg_multiply

Since commit 0a35174f, the function
convert_slice_2d() is not used anymore.

These functions return a subslice that exceed its parent slice. We want
to avoid this behavior for tiling.

The function get_intra_edges() acceded 1 pixel above and 1 pixel on the
left of the given PlaneSlice. This behavior will be problematic for
tiling.

Instead, use EdgedPlaneSlice to pass the extended region, so that a
subslice never exceeds its parent slice.

For tiling, we want subslices to never exceed their parent slice.

However, several functions need to read few pixels on top and on the
left of a given PlaneSlice.

For that purpose, introduce EdgedPlaneSlice, containing a PlaneSlice
which includes edges and the edges size.

The current threads usage is limited to motion estimation and
on machines with a large number of cores the default thread pool
dimension is causing performance degradation.

Fixes #1064

The feature is already provided by the Index and IndexMut traits (i.e.
operator[]).

Concretely, we always use:

    plane_slice[index]

instead of:

    plane_slice.row(index)