[FFmpeg-devel] avcodec/utvideodec : add x86 SIMD (SSSE3) for gradient prediction

Martin Vignali martin.vignali at gmail.com
Sun Nov 26 23:15:54 EET 2017


Patch in attach add SIMD (SSSE 3) for gradient prediction
and a checkasm test

Checkasm result (width = 1024) (kaby lake, macos 10.12)
add_gradient_pred_c: 1708.8
add_gradient_pred_ssse3: 533.0

Benchmark on a 3 min HD File in gradient (422)
without SIMD :
bench: utime=102.695s
bench: maxrss=102592512kB

with SIMD
bench: utime=91.712s
bench: maxrss=102543360kB

i will add AVX2 version later (need more clean before submitting, and will
conflict with another patch add_left_pred avx2 version)

This new dsp func, can probably also be use by magicyuv decoder

i'm not sure about the best asm way for load an uint8_t to all part of an

Comment welcome

Jokyo Images
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-avcodec-utvideodec-add-SIMD-SSSE3-for-gradient_pred.patch
Type: application/octet-stream
Size: 6180 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20171126/237dc19f/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0002-checkasm-llviddsp-add-test-for-add_gradient_pred.patch
Type: application/octet-stream
Size: 1892 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20171126/237dc19f/attachment-0001.obj>

More information about the ffmpeg-devel mailing list