This makes it similar to put_epel16_v6, and gives a 10-25%
speedup of this function.
Before: Cortex A7 A8 A9 A53 A72
vp8_put_epel16_h6v6_neon: 3058.0 2218.5 2459.8 2183.0 1572.2
After:
vp8_put_epel16_h6v6_neon: 2670.8 1934.2 2244.4 1729.4 1503.9
Signed-off-by: Martin Storsjö <martin@martin.st>
When targeting darwin, clang requires commas between arguments,
while the no-comma form is allowed for other targets.
Since Xcode 9.3, the bundled clang supports altmacro and doesn't
require using gas-preprocessor any longer.
Signed-off-by: Martin Storsjö <martin@martin.st>
Clang supports the macro expansion counter (used for making unique
labels within macro expansions), but not when targeting darwin.
Convert uses of the counter into normal local labels, as used
elsewhere.
Since Xcode 9.3, the bundled clang supports altmacro and doesn't
require using gas-preprocessor any longer.
Signed-off-by: Martin Storsjö <martin@martin.st>
When targeting darwin, clang requires commas between arguments,
while the no-comma form is allowed for other targets.
Since Xcode 9.3, the bundled clang supports altmacro and doesn't
require using gas-preprocessor any longer.
Signed-off-by: Martin Storsjö <martin@martin.st>
Compilation error "out of range" fixed for armeabi-v7a. Compilation failed
trying to build libvlc.aar for ARM7 android on ubuntu 16.04 host. Error
messages is "Offset out of range". The reason of the error is assembler LDR
directives in function "ff_hevc_transform_luma_4x4_neon_8" need local storage
in range <1k, but no such storage provided.
Based on a patch by Ihor Bobalo <bob@eleks.com>
Suggested-by: wbs
Signed-off-by: James Almer <jamrial@gmail.com>
In binutils 2.29, the behavior of the ADR instruction changed so that 1 is
added to the address of a Thumb function (previously nothing was added). This
allows the loaded address to be passed to a BLX instruction and the correct
mode change will occur.
See: https://sourceware.org/bugzilla/show_bug.cgi?id=21458
By using adr with a label that isn't annotated as a thumb function,
we avoid the new behaviour in binutils 2.29 and get the same behaviour
as in prior releases, and as in other assemblers (ms armasm.exe,
clang's built in assembler) - an idea that Janne Grunau came up with.
Signed-off-by: Martin Storsjö <martin@martin.st>
It is redundant with costable. The first half of sintable is
identical with the second half of costable. The second half
of sintable is negative value of the first half of sintable.
The computation is changed to handle sign of sin values, in
C code and ARM assembly code.
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
The code originally pre-multiply by 2 the steps, causing the running sum
of the h factors to drift away due to the lack of precision. It quickly
causes an inaccuracy > 0.01.
I tried diverse approaches such as multiply by 2.0 (instead of adding
the value itself) without success.
I'm unable to bench the impact of this change, feel free to compare.
This commit fixes the incoming aacpsdsp tests.
Following is an alternative simplified function (matching the incoming
AArch64 code) that may be used:
function ff_ps_stereo_interpolate_neon, export=1
vld1.32 {q0}, [r2]
vld1.32 {q1}, [r3]
ldr r12, [sp]
vmov.f32 q8, q0
vmov.f32 q9, q1
vzip.32 q8, q0
vzip.32 q9, q1
1:
vld1.32 {d4}, [r0,:64]
vld1.32 {d6}, [r1,:64]
vadd.f32 q8, q8, q9
vadd.f32 q0, q0, q1
vmov.f32 d5, d4
vmov.f32 d7, d6
vmul.f32 q2, q2, q8
vmla.f32 q2, q3, q0
vst1.32 {d4}, [r0,:64]!
vst1.32 {d5}, [r1,:64]!
subs r12, r12, #1
bgt 1b
bx lr
endfunc
clang now (in the upcoming 5.0 version) is capable of building our
arm assembly without relying on gas-preprocessor, although clang/LLVM
doesn't support .dn register aliases.
The VC1 MC assembly was only built and used if the chosen assembler
supported the .dn directives though. This was supported as long as
gas-preprocessor was used.
This means that VC1 decoding got a speed regression on clang 5.0,
unless the user manually chose using gas-preprocessor again.
By avoiding using the .dn register aliases, we can build the VC1 MC
assembly with the latest clang version.
Support for the .dn/.qn directives in clang/LLVM isn't actively planned,
see https://bugs.llvm.org/show_bug.cgi?id=18199.
This partially reverts 896a5bff64.
Signed-off-by: Martin Storsjö <martin@martin.st>
Instead, hardcode the use of the _arm implementation of add_pixels,
and use the C version for put_pixels (as no arm-optimized version
exists). Since there's separate implementations of idct{,_put,_add}
for neon, this has no practical impact on performance.
The main hevcdsp.c file calls this init function if HAVE_ARM is set,
regardless of whether neon support is available or not.
This fixes builds where neon isn't supported by the build tools at all.
Signed-off-by: Martin Storsjö <martin@martin.st>
This work is sponsored by, and copyright, Google.
This reduces the code size of libavcodec/arm/vp9itxfm_16bpp_neon.o from
17500 to 14516 bytes.
This gives a small slowdown of a couple tens of cycles, up to around
150 cycles for the full case of the largest transform, but makes
it more feasible to add more optimized versions of these transforms.
Before: Cortex A7 A8 A9 A53
vp9_inv_dct_dct_16x16_sub4_add_10_neon: 4237.4 3561.5 3971.8 2525.3
vp9_inv_dct_dct_16x16_sub16_add_10_neon: 6371.9 5452.0 5779.3 3910.5
vp9_inv_dct_dct_32x32_sub4_add_10_neon: 22068.8 17867.5 19555.2 13871.6
vp9_inv_dct_dct_32x32_sub32_add_10_neon: 37268.9 38684.2 32314.2 23969.0
After:
vp9_inv_dct_dct_16x16_sub4_add_10_neon: 4375.1 3571.9 4283.8 2567.2
vp9_inv_dct_dct_16x16_sub16_add_10_neon: 6415.6 5578.9 5844.6 3948.3
vp9_inv_dct_dct_32x32_sub4_add_10_neon: 22653.7 18079.7 19603.7 13905.3
vp9_inv_dct_dct_32x32_sub32_add_10_neon: 37593.2 38862.2 32235.8 24070.9
Signed-off-by: Martin Storsjö <martin@martin.st>
Keep the idct32 coefficients in narrow form in q6-q7, and idct16
coefficients in lengthened 32 bit form in q0-q3. Avoid clobbering
q0-q3 in the pass1 function, and squeeze the idct16 coefficients
into q0-q1 in the pass2 function to avoid reloading them.
The idct16 coefficients are clobbered and reloaded within idct32_odd
though, since that turns out to be faster than narrowing them and
swapping them into q6-q7.
Before: Cortex A7 A8 A9 A53
vp9_inv_dct_dct_32x32_sub4_add_10_neon: 22653.8 18268.4 19598.0 14079.0
vp9_inv_dct_dct_32x32_sub32_add_10_neon: 37699.0 38665.2 32542.3 24472.2
After:
vp9_inv_dct_dct_32x32_sub4_add_10_neon: 22270.8 18159.3 19531.0 13865.0
vp9_inv_dct_dct_32x32_sub32_add_10_neon: 37523.3 37731.6 32181.7 24071.2
Signed-off-by: Martin Storsjö <martin@martin.st>
Align the second/third operands as they usually are.
Due to the wildly varying sizes of the written out operands
in aarch64 assembly, the column alignment is usually not as clear
as in arm assembly.
This is cherrypicked from libav commit
7995ebfad1.
Signed-off-by: Martin Storsjö <martin@martin.st>
In the half/quarter cases where we don't use the min_eob array, defer
loading the pointer until we know it will be needed.
This is cherrypicked from libav commit
3a0d5e206d.
Signed-off-by: Martin Storsjö <martin@martin.st>
This reduces the number of lines and reduces the duplication.
Also simplify the eob check for the half case.
If we are in the half case, we know we at least will need to do the
first three slices, we only need to check eob for the fourth one,
so we can hardcode the value to check against instead of loading
from the min_eob array.
Since at most one slice can be skipped in the first pass, we can
unroll the loop for filling zeros completely, as it was done for
the quarter case before.
This allows skipping loading the min_eob pointer when using the
quarter/half cases.
This is cherrypicked from libav commit
98ee855ae0.
Signed-off-by: Martin Storsjö <martin@martin.st>
Align the second/third operands as they usually are.
Due to the wildly varying sizes of the written out operands
in aarch64 assembly, the column alignment is usually not as clear
as in arm assembly.
Signed-off-by: Martin Storsjö <martin@martin.st>
In the half/quarter cases where we don't use the min_eob array, defer
loading the pointer until we know it will be needed.
Signed-off-by: Martin Storsjö <martin@martin.st>
This reduces the number of lines and reduces the duplication.
Also simplify the eob check for the half case.
If we are in the half case, we know we at least will need to do the
first three slices, we only need to check eob for the fourth one,
so we can hardcode the value to check against instead of loading
from the min_eob array.
Since at most one slice can be skipped in the first pass, we can
unroll the loop for filling zeros completely, as it was done for
the quarter case before.
This allows skipping loading the min_eob pointer when using the
quarter/half cases.
Signed-off-by: Martin Storsjö <martin@martin.st>
This matches the order they are in the 16 bpp version.
There they are in this order, to make sure we access them in the
same order they are declared, easing loading only half of the
coefficients at a time.
This makes the 8 bpp version match the 16 bpp version better.
This is cherrypicked from libav commit
08074c092d.
Signed-off-by: Martin Storsjö <martin@martin.st>
All elements are used pairwise, except for the first one.
Previously, the 16th element was unused. Move the unused element
to the second slot, to make the later element pairs not split
across registers.
This simplifies loading only parts of the coefficients,
reducing the difference to the 16 bpp version.
This is cherrypicked from libav commit
de06bdfe6c.
Signed-off-by: Martin Storsjö <martin@martin.st>
The idct32x32 function actually pushed q4-q7 onto the stack even
though it didn't clobber them; there are plenty of registers that
can be used to allow keeping all the idct coefficients in registers
without having to reload different subsets of them at different
stages in the transform.
Since the idct16 core transform avoids clobbering q4-q7 (but clobbers
q2-q3 instead, to avoid needing to back up and restore q4-q7 at all
in the idct16 function), and the lanewise vmul needs a register in
the q0-q3 range, we move the stored coefficients from q2-q3 into q4-q5
while doing idct16.
While keeping these coefficients in registers, we still can skip pushing
q7.
Before: Cortex A7 A8 A9 A53
vp9_inv_dct_dct_32x32_sub32_add_neon: 18553.8 17182.7 14303.3 12089.7
After:
vp9_inv_dct_dct_32x32_sub32_add_neon: 18470.3 16717.7 14173.6 11860.8
This is cherrypicked from libav commit
402546a172.
Signed-off-by: Martin Storsjö <martin@martin.st>
For this case, with 8 inputs but only changing 4 of them, we can fit
all 16 input pixels into a q register, and still have enough temporary
registers for doing the loop filter.
The wd=8 filters would require too many temporary registers for
processing all 16 pixels at once though.
Before: Cortex A7 A8 A9 A53
vp9_loop_filter_mix2_v_44_16_neon: 289.7 256.2 237.5 181.2
After:
vp9_loop_filter_mix2_v_44_16_neon: 221.2 150.5 177.7 138.0
This is cherrypicked from libav commit
575e31e931.
Signed-off-by: Martin Storsjö <martin@martin.st>
This adds lots of extra .ifs, but speeds it up by a couple cycles,
by avoiding stalls.
This is cherrypicked from libav commit
e18c39005a.
Signed-off-by: Martin Storsjö <martin@martin.st>
Previously we first calculated hev, and then negated it.
Since we were able to schedule the negation in the middle
of another calculation, we don't see any gain in all cases.
Before: Cortex A7 A8 A9 A53 A53/AArch64
vp9_loop_filter_v_4_8_neon: 147.0 129.0 115.8 89.0 88.7
vp9_loop_filter_v_8_8_neon: 242.0 198.5 174.7 140.0 136.7
vp9_loop_filter_v_16_8_neon: 500.0 419.5 382.7 293.0 275.7
vp9_loop_filter_v_16_16_neon: 971.2 825.5 731.5 579.0 453.0
After:
vp9_loop_filter_v_4_8_neon: 143.0 127.7 114.8 88.0 87.7
vp9_loop_filter_v_8_8_neon: 241.0 197.2 173.7 140.0 136.7
vp9_loop_filter_v_16_8_neon: 497.0 419.5 379.7 293.0 275.7
vp9_loop_filter_v_16_16_neon: 965.2 818.7 731.4 579.0 452.0
This is cherrypicked from libav commit
e1f9de86f4.
Signed-off-by: Martin Storsjö <martin@martin.st>
James Almer
Martin Storsjö
Meng Wang
Aurelien Jacobs
Michael Niedermayer
Alexandra Hájková
Muhammad Faiz
Clément Bœsch
Seppo Tomperi
Ronald S. Bultje
Clément Bœsch