GitLab

nasm should infer .text but does not for windows:
https://bugzilla.nasm.us/show_bug.cgi?id=3392451

Change-Id: Ib195465e5f33405f5ff61c4cf88aa2a72640cacb

Change-Id: Ie60381a0c6ee01f828cd364a43f01517f4cb03e9

Change-Id: I9f95f47bc7ecbb7980f21cbc3a91f699624141af

Change-Id: If8b91aaa883c01107f0ea3468139fa24cfb301d2

Fixes a build issue when relocation is not allowed:
relocation R_X86_64_32 against '.rodata' can not be used when making a shared object

Change-Id: Ica3e90c926847bc384e818d7854f0030f4d69aa0

SSE2 instrinsic vs AVX2 intrinsic speed gains:
blocksize   16: ~1.33
blocksize   64: ~1.51
blocksize  256: ~3.03
blocksize 1024: ~3.71

Change-Id: I79b28cba82d21f9dd765e79881aa16d24fd0cb58

Change-Id: I7364a157de39eb7137b599808474b8d46d19d376

The added AVX-512 support requires the subset of AVX-512 added in Skylake-X.

Change-Id: I39666b00d10bf96d06c709823663eb09b89265b7

Eliminates the following instruction for the x86 (64 bit)
intrinsic code:

movslq %esi,%rax

Change-Id: I8f5ebd40726f998708a668b0f52ea7a0576befae

Changed the intrinsics to perform summation similiar to the way the assembly does.

The new code diverges from the assembly by preferring unsaturated additions.

Results for haswell

SSSE3
Horiz/Vert  Size  Speedup
Horiz       x4    ~32%
Horiz       x8    ~6%
Vert        x8    ~4%

AVX2
Horiz/Vert  Size  Speedup
Horiz       x16   ~16%
Vert        x16   ~14%

BUG=webm:1471

Change-Id: I7ad98ea688c904b1ba324adf8eb977873c8b8668

Use an intermediate buffer before storing to coeffs when
highbitdepth is enabled.

Change-Id: I101981a1995f1108ad107c55c37d6e09eadb404b

~10% performance gain.  Fixed the cosmetics noted in the
previous commit.

Change-Id: Iddf475f34d0d0a3e356b2143682aeabac459ed13

This version is ~1.91x faster than the sse2 version.  When
highbitdepth is enabled, it is ~1.74x.

Change-Id: I2b0e92ede9f55c6259ca07bf1f8c8a5d0d0955bd

Change-Id: I6539111dfb35a43028e9755785b2e9ea31854305

Note this change will trigger the different C version on SSSE3 and
generate different scaled output.

Its speed is 2x compared with the version calling vpx_scaled_2d_ssse3().

Change-Id: I17fff122cd0a5ac8aa451d84daa606582da8e194

Change-Id: I51c190f0a88685867df36912522e67bdae58a673

Also adds vpx_convolve8_avg_horiz_avx2.

Change-Id: I38783d972ac26bec77610e9e15a0a058ed498cbf

vpx_convolve8_avg works by first running a normal horizontal filter then a
vertical filter averages at the end.

The added vpx_convolve8_avg_avx2 calls pre-existing AVX2 code for the
horizontal step.

vpx_convolve8_avg_vert_avx2 is also added, but only uses ssse3 code.

Change-Id: If5160c0c8e778e10de61ee9bf42ee4be5975c983

Change-Id: I882da3a04884d5fabd4cd591c28682cbb2d76aa5

BUG=webm:1462,766721

Change-Id: Icfa536a8e38623636b96c396e3c94889bfde7a98

Change-Id: Id6a8c549709a3c516ed5d7b719b05117c5ef8bac

Add some load and store sse2 inline functions.

Change-Id: Ib1e0650b5a3d8e2b3736ab7c7642d6e384354222

Change-Id: Ic369dd86b3b81686f68fbc13ad34ab8ea8846878

C vs SSE2 speed gains:
_4x4 : ~1.81x

C vs SSSE3 speed gains:
_8x8 : ~1.96x
_16x16 : ~1.88x
_32x32 : ~2.02x

BUG=webm:1411

Change-Id: Iefaf8b39afbbfe34c1ad1d21e3a003b20f1f61e0

C vs SSE2 speed gains:
_4x4 : ~2.04x

C vs SSSE3 speed gains:
_8x8 : ~2.82x
_16x16 : ~5.93x
_32x32 : ~2.79x

BUG=webm:1411

Change-Id: I31d949695991c067dac89d91e0bed3e666c94993

Exposed by fuzz test in high bitdepth.
The bug is introduced in commit 64653fa1.

BUG=webm:1466

Change-Id: Idd77d5c6a60efb9241471611ce1aba0646cb6ff5

C vs SSE2 speed gains:
_4x4 : ~1.95x

C vs SSSE3 speed gains:
_8x8 : ~3.30x
_16x16 : ~5.67x
_32x32 : ~3.87x

BUG=webm:1411

Change-Id: Ib483989b25614aa89b635e8c087d0879a5d71904

C vs SSSE3 speed gains:
_4x4 : ~2.45x
_8x8 : ~10.61x
_16x16 : ~11.34x
_32x32 : ~6.36x

BUG=webm:1411

Change-Id: Ic91389a4f1a8ad093f498afe53765b897fb9be09

BUG=webm:1450

Change-Id: If59743aafe99226e0ec67ab5d20678ce25f53ab8

BUG=webm:1450

Change-Id: Ib046fe28caec5b9ebdc9d0152df7c54ff4266858

The unnecessary upcast to (int) will be cleaned later.

BUG=webm:1450

Change-Id: Ia234575206d5a74540526924b06ed3939322d063

Extract a couple of static functions into their caller functions.

Change-Id: If8d8a0e217fba6b402d2a79ede13b5b444ff08a0

C vs SSE2 speed gains:
_4x4 : ~2.94x

C vs SSSE3 speed gains:
_8x8 : ~8.69x
_16x16 : ~6.32x
_32x32 : ~5.33x

BUG=webm:1411

Change-Id: I2c35b527eac2229f17aaa9d118fb601e7195efe4

This reverts commit 8c42237b.

Because ssse3 code is used for the reference, the qcoeff and dqcoeff
reference buffers must be aligned.

Original change's description:
> quantize avx: copy 32x32 implementation
>
> Ensure avx and ssse3 stay in sync by testing them against each other.
>
> Change-Id: I699f3b48785c83260825402d7826231f475f697c

Change-Id: Ieeef11b9406964194028b0d81d84bcb63296ae06

C vs SSE2 speed gains:
_4x4 : ~2.31x

C vs SSSE3 speed gains:
_8x8 : ~4.73x
_16x16 : ~10.88x
_32x32 : ~4.80x

BUG=webm:1411

Change-Id: I0bac29db261079181ddabc6814bd62c463109caf

So that 4 to 1 frame scaling can call them.

Change-Id: I9ec438aa63b923ba164ad3c59d7ecfa12789eab5

so that the convolve functions are independent of table alignment.

Change-Id: Ieab132a30d72c6e75bbe9473544fbe2cf51541ee

C vs SSE2 speed gains:
_4x4 : ~7.64x
_8x8 : ~16.60x
_16x16 : ~8.15x
_32x32 : ~5.05x

BUG=webm:1411

Change-Id: If165d419711cfda901bd428a05ca1560a009e62e

C vs SSE2 speed gains:
_4x4 : ~6.49x
_8x8 : ~10.82x
_16x16 : ~7.61x
_32x32 : ~5.29x

BUG=webm:1411

Change-Id: Ibc30c50cb7139049bf05298010803499e6ef949b

C vs SSE2 speed gains:
_4x4 : ~7.39x
_8x8 : ~11.36x
_16x16 : ~8.68x
_32x32 : ~4.33x

BUG=webm:1411

Change-Id: I7f1487cd1531d4e7f0fbb4596fed3bfb72a59d58