Difference between revisions of "SoC 2011/Assembly optimizations in VLC"
(Created page with "=== Abstract === The main goal is to optimize video and audio filters of VLC, add SIMD x86 instruction using functions and clean their C code up. Next step is profiling of mai…") |
|||
| (10 intermediate revisions by the same user not shown) | |||
| Line 1: | Line 1: | ||
| − | === | + | {{SoCProject|year=2011|student=[[User:Bizon|Martin Briza]]|mentor=[[User:j-b|Jean-Baptiste Kempf]]}} |
| − | + | ||
| + | = Repository = | ||
| + | |||
| + | http://git.videolan.org/?p=vlc/vlc-bizon.git;a=summary | ||
| + | |||
| + | = Abstract = | ||
The main goal is to optimize video and audio filters of VLC, add SIMD x86 instruction using functions and clean their C code up. Next step is profiling of main VLC. | The main goal is to optimize video and audio filters of VLC, add SIMD x86 instruction using functions and clean their C code up. Next step is profiling of main VLC. | ||
| − | |||
Optimization of functions that process huge amount of data with the same algorithm at the same time with SIMD instructions, so better performance can be achieved. Main focus will be on video and audio filters of VLC and after these are done, profiling of other parts of the application will be the next step. | Optimization of functions that process huge amount of data with the same algorithm at the same time with SIMD instructions, so better performance can be achieved. Main focus will be on video and audio filters of VLC and after these are done, profiling of other parts of the application will be the next step. | ||
| − | |||
Side effect of the work will be clean up and improvement of algorithms used (same as in sepia.c - change from RGB to YUV, to not lose time converting the format). | Side effect of the work will be clean up and improvement of algorithms used (same as in sepia.c - change from RGB to YUV, to not lose time converting the format). | ||
| − | |||
| − | === Additional information | + | = Progress = |
| + | |||
| + | |||
| + | === Weeks 1 and 2 (May 23 - June 5) === | ||
| + | |||
| + | Exams, no activity on the project | ||
| + | |||
| + | === Week 3 (June 6 - 12) === | ||
| + | |||
| + | Acclimatization, starting work on YUV to RGB converting function from filter_picture.h to use it in plugins using it. I experienced few minor problems, which unfortunately ended up in me searching for bug where it was not. | ||
| + | |||
| + | At the moment (as of 13th of June), I'm working on posterization filter. As SSE doesn't provide integer division, I have to find solution which will really speed up the computation. | ||
| + | |||
| + | === Week 4 (June 13 - 19) === | ||
| + | |||
| + | Working on posterization filter, YUV <-> RGB function were written to be used in other modules. Code isn't very clear, definitely not to be commited to the main repo. | ||
| + | |||
| + | === Week 5 (June 20 - 26) === | ||
| + | |||
| + | Started working on adjust filter. Work was slowed down by the fact, I had to leave home to arrange some stuff about school (college room etc.). I have cleared my work that has been done in sepia filter in spring to be more clear, MUCH faster and mainly - using SSE2 instead of SSE4.1 | ||
| + | |||
| + | === Week 6 (June 27 - July 3) === | ||
| + | |||
| + | Finished working on adjust and cleaned posterize. After showing the code on the IRC channel and discussion with j-b and Dark_Shikari (which was really helping), I decided to rewrite them in a much better way, which I couldn't see before. | ||
| + | |||
| + | === Week 7 (July 4 - 10) === | ||
| + | |||
| + | Currently in progress. | ||
| + | |||
| + | = Additional information = | ||
May 9 - June 10 are exams on our school, so I won't start coding exactly at the start of the coding period. I'll keep you updated on when I have finished.<br> | May 9 - June 10 are exams on our school, so I won't start coding exactly at the start of the coding period. I'll keep you updated on when I have finished.<br> | ||
Latest revision as of 23:40, 5 July 2011
This project is part of Google Summer of Code 2011.
|
Contents
Repository
http://git.videolan.org/?p=vlc/vlc-bizon.git;a=summary
Abstract
The main goal is to optimize video and audio filters of VLC, add SIMD x86 instruction using functions and clean their C code up. Next step is profiling of main VLC.
Optimization of functions that process huge amount of data with the same algorithm at the same time with SIMD instructions, so better performance can be achieved. Main focus will be on video and audio filters of VLC and after these are done, profiling of other parts of the application will be the next step.
Side effect of the work will be clean up and improvement of algorithms used (same as in sepia.c - change from RGB to YUV, to not lose time converting the format).
Progress
Weeks 1 and 2 (May 23 - June 5)
Exams, no activity on the project
Week 3 (June 6 - 12)
Acclimatization, starting work on YUV to RGB converting function from filter_picture.h to use it in plugins using it. I experienced few minor problems, which unfortunately ended up in me searching for bug where it was not.
At the moment (as of 13th of June), I'm working on posterization filter. As SSE doesn't provide integer division, I have to find solution which will really speed up the computation.
Week 4 (June 13 - 19)
Working on posterization filter, YUV <-> RGB function were written to be used in other modules. Code isn't very clear, definitely not to be commited to the main repo.
Week 5 (June 20 - 26)
Started working on adjust filter. Work was slowed down by the fact, I had to leave home to arrange some stuff about school (college room etc.). I have cleared my work that has been done in sepia filter in spring to be more clear, MUCH faster and mainly - using SSE2 instead of SSE4.1
Week 6 (June 27 - July 3)
Finished working on adjust and cleaned posterize. After showing the code on the IRC channel and discussion with j-b and Dark_Shikari (which was really helping), I decided to rewrite them in a much better way, which I couldn't see before.
Week 7 (July 4 - 10)
Currently in progress.
Additional information
May 9 - June 10 are exams on our school, so I won't start coding exactly at the start of the coding period. I'll keep you updated on when I have finished.
