From Fedora Project Wiki
Line 23: Line 23:
* Last updated: 2010-06-04
* Last updated: 2010-06-04
* Percentage of completion: 60%
* Percentage of completion: 60%
libjpeg-turbo package is currently reviewed - https://bugzilla.redhat.com/show_bug.cgi?id=600243


== Detailed Description ==
== Detailed Description ==

Revision as of 09:19, 4 June 2010



Replace libjpeg by libjpeg-turbo

Summary

libjpeg-turbo is fork of the original libjpeg project. It contains numerous performance related enhancements and is at least twice faster in JPEG compression/decompression than original libjpeg on platforms with MMX/SSE instruction set. It has same API/ABI like original libjpeg and also runs on non-SSE platforms where is around 25% faster.

Owner

Current status

  • Targeted release: Fedora 14
  • Last updated: 2010-06-04
  • Percentage of completion: 60%

libjpeg-turbo package is currently reviewed - https://bugzilla.redhat.com/show_bug.cgi?id=600243

Detailed Description

Everything is written in summary.

Benefit to Fedora

Much faster JPEG compression/decompression library. JPEG codec is widely used and it is good to speed up core library which implements it. Additionally, libjpeg-turbo is developed in more open-source manner than original libjpeg.

Scope

libjpeg-turbo has same API/ABI as libjpeg so no package needs to be rebuilt

How To Test

You can use any program which uses libjpeg-turbo code. You should try to perform JPEG compression/decompression to and check if you don't see artefacts. You can also compare speed of the libjpeg-turbo against libjpeg, you should be able to visually check that libjpeg-turbo is really faster.

User Experience

A lot faster JPEG compression/decompression

Dependencies

462 packages (repoquery --alldeps --whatrequires libjpeg |wc -l) but no rebuild is needed due API/ABI compatibility

Contingency Plan

Revert to libjpeg when libjpeg-turbo hits some serious problems.

Documentation

Release Notes

libjpeg library has been replaced by libjpeg-turbo library which has same API/ABI but is at least twice faster on all primary architectures and about 25% faster on secondary architectures

Comments and Discussion