Instead of computing y + N with a hard-coded index offset, calculate the
relative offset as a 16-bit integer in C and add that to the pointer directly.
Since we no longer mask the resulting combined address, this may result in
overread, but that's fine since we over-provisioned the array in the previous
commit.
I want to change the way offsets are calculated inside the dither asm;
and the cleanest way to solve that problem is to just over-allocate the
entire dither matrix based on the maximum offset range we expected.
This is an exceptionally unlikely (in fact, currently impossible) case to
actually hit, and not worth micro-optimizing for. More specifically, having
this special case prevents me from easily adding per-row offsets.
This is required placement by standard [[maybe_unused]] attribute, works
the same for __attribute__((unused)).
Signed-off-by: Kacper Michajłow <kasper93@gmail.com>
This fixes the following compiler error, if compiling with MSVC
for ARM (32 bit):
src/libswscale/ops_chain.c(48): error C2719: 'priv': formal parameter with requested alignment of 16 won't be aligned
This change shouldn't affect the performance of this operation
(which in itself probably isn't relevant); instead of copying the
contents of the SwsOpPriv struct from the stack as parameter,
it gets copied straight from the caller function's stack frame
instead.
Separately from this issue, MSVC 17.8 and 17.9 end up in an
internal compiler error when compiling libswscale/ops.c, but
older and newer versions do compile it successfully.
This covers most 8-bit and 16-bit ops, and some 32-bit ops. It also covers all
floating point operations. While this is not yet 100% coverage, it's good
enough for the vast majority of formats out there.
Of special note is the packed shuffle fast path, which uses pshufb at vector
sizes up to AVX512.
The scalar loop is replaced with masked AVX512 instructions.
For extracting the Y from UYVY, vperm2b is used instead of
various AND and packuswb.
Instead of loading the vectors with interleaved lanes as done
in AVX2 version, normal load is used. At the end of packuswb,
for U and V, an extra permute operation is done to get the
required layout.
AMD 7950x Zen 4 benchmark data:
uyvytoyuv422_c: 29105.0 ( 1.00x)
uyvytoyuv422_sse2: 3888.0 ( 7.49x)
uyvytoyuv422_avx: 3374.2 ( 8.63x)
uyvytoyuv422_avx2: 2649.8 (10.98x)
uyvytoyuv422_avx512icl: 1615.0 (18.02x)
Signed-off-by: Shreesh Adiga <16567adigashreesh@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
ff_w1111 and ff_bgr2(Y|UV)Offset are only used there
(and only on x86-32 since caaec2ea95).
Also make them static.
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This setting can be used to infuence the type of tone and gamut mapping used
internally when color space conversions are required. As discussed at VDD'24,
the default was set to relative colorimetric clipping, which is approximately
associative, surjective and idempotent. As such, it roundtrips well, although
it is strictly speaking not associative on out-of-gamut colors.
There is an issue with the constants used in YUV to YUV range conversion,
where the upper bound is not respected when converting to mpeg range.
With this commit, the constants are calculated at runtime, depending on
the bit depth. This approach also allows us to more easily understand how
the constants are derived.
For bit depths <= 14, the number of fixed point bits has been set to 14
for all conversions, to simplify the code.
For bit depths > 14, the number of fixed points bits has been raised and
set to 18, to allow for the conversion to be accurate enough for the mpeg
range to be respected.
The convert functions now take the conversion constants (coeff and offset)
as function arguments.
For bit depths <= 14, coeff is unsigned 16-bit and offset is 32-bit.
For bit depths > 14, coeff is unsigned 32-bit and offset is 64-bit.
x86_64:
chrRangeFromJpeg8_1920_c: 2127.4 2125.0 (1.00x)
chrRangeFromJpeg16_1920_c: 2325.2 2127.2 (1.09x)
chrRangeToJpeg8_1920_c: 3166.9 3168.7 (1.00x)
chrRangeToJpeg16_1920_c: 2152.4 3164.8 (0.68x)
lumRangeFromJpeg8_1920_c: 1263.0 1302.5 (0.97x)
lumRangeFromJpeg16_1920_c: 1080.5 1299.2 (0.83x)
lumRangeToJpeg8_1920_c: 1886.8 2112.2 (0.89x)
lumRangeToJpeg16_1920_c: 1077.0 1906.5 (0.56x)
aarch64 A55:
chrRangeFromJpeg8_1920_c: 28835.2 28835.6 (1.00x)
chrRangeFromJpeg16_1920_c: 28839.8 32680.8 (0.88x)
chrRangeToJpeg8_1920_c: 23074.7 23075.4 (1.00x)
chrRangeToJpeg16_1920_c: 17318.9 24996.0 (0.69x)
lumRangeFromJpeg8_1920_c: 15389.7 15384.5 (1.00x)
lumRangeFromJpeg16_1920_c: 15388.2 17306.7 (0.89x)
lumRangeToJpeg8_1920_c: 19227.8 19226.6 (1.00x)
lumRangeToJpeg16_1920_c: 15387.0 21146.3 (0.73x)
aarch64 A76:
chrRangeFromJpeg8_1920_c: 6324.4 6268.1 (1.01x)
chrRangeFromJpeg16_1920_c: 6339.9 11521.5 (0.55x)
chrRangeToJpeg8_1920_c: 9656.0 9612.8 (1.00x)
chrRangeToJpeg16_1920_c: 6340.4 11651.8 (0.54x)
lumRangeFromJpeg8_1920_c: 4422.0 4420.8 (1.00x)
lumRangeFromJpeg16_1920_c: 4420.9 5762.0 (0.77x)
lumRangeToJpeg8_1920_c: 5949.1 5977.5 (1.00x)
lumRangeToJpeg16_1920_c: 4446.8 5946.2 (0.75x)
NOTE: all simd optimizations for range_convert have been disabled.
they will be re-enabled when they are fixed for each architecture.
NOTE2: the same issue still exists in rgb2yuv conversions, which is not
addressed in this commit.
As part of a larger, ongoing effort to modernize and partially rewrite
libswscale, it was decided and generally agreed upon to introduce a new
public API for libswscale. This API is designed to be less stateful, more
explicitly defined, and considerably easier to use than the existing one.
Most of the API work has been already accomplished in the previous commits,
this commit merely introduces the ability to use sws_scale_frame()
dynamically, without prior sws_init_context() calls. Instead, the new API
takes frame properties from the frames themselves, and the implementation is
based on the new SwsGraph API, which we simply reinitialize as needed.
This high-level wrapper also recreates the logic that used to live inside
vf_scale for scaling interlaced frames, enabling it to be reused more easily
by end users.
Finally, this function is designed to simply copy refs directly when nothing
needs to be done, substantially improving throughput of the noop fast path.
Sponsored-by: Sovereign Tech Fund
Signed-off-by: Niklas Haas <git@haasn.dev>
This is a preliminary step to separating these into a new struct. This
commit contains no functional changes, it is a pure search-and-replace.
Sponsored-by: Sovereign Tech Fund
Signed-off-by: Niklas Haas <git@haasn.dev>
This commit also fixes the issue that the call to ff_sws_init_range_convert()
from sws_init_swscale() was not setting up the arch-specific optimizations.
And preserve the public SwsContext as separate name. The motivation here
is that I want to turn SwsContext into a public struct, while keeping the
internal implementation hidden. Additionally, I also want to be able to
use multiple internal implementations, e.g. for GPU devices.
This commit does not include any functional changes. For the most part, it is
a simple rename. The only complications arise from the public facing API
functions, which preserve their current type (and hence require an additional
unwrapping step internally), and the checkasm test framework, which directly
accesses SwsInternal.
For consistency, the affected functions that need to maintain a distionction
have generally been changed to refer to the SwsContext as *sws, and the
SwsInternal as *c.
In an upcoming commit, I will provide a backing definition for the public
SwsContext, and update `sws_internal()` to dereference the internal struct
instead of merely casting it.
Sponsored-by: Sovereign Tech Fund
Signed-off-by: Niklas Haas <git@haasn.dev>
I want to move away from having random leaf processing functions mutate
plane pointers, and while we're at it, we might as well make the strides
and tables const as well.
Sponsored-by: Sovereign Tech Fund
Signed-off-by: Niklas Haas <git@haasn.dev>
