Rockbox Development > New Ports

Cowon D2

<< < (168/194) > >>

shotofadds:
I've run test_codec several times over the full set of sample tracks with IRAM enabled and haven't run into any problems, so things appear to be stable. Improvements against the latest test on the CodecPerformanceComparison wiki (r25447, no IRAM) are:

AAC-HE +6%
Cook +14..16%
FLAC +7%
MP3 +7%
MPC +12%
AAC-LC +10..14%
Vorbis +13..17%
WMA +16%

APE/ALAC/WavPack are all more-or-less unchanged. I've attached the latest test_codec log to this post (tested using r26288 + IRAM).

Regarding battery life, I doubt even these improvements will make a lot of difference. There has been very little experimentation with power management / clocking changes / turning off outputs etc on this device, so I think codec performance is the least of our worries.  ;)

Also, while doing the above tests I did (re)discover that both the core and the MPA codec (at least) will attempt to put things in IRAM sections even if USE_IRAM is undefined. This is the reason for the current .lds hacks for this CPU, and should really be put right as it causes undefined instruction errors in the case of the MPA codec.

EDIT: Looks like some hackery in mad_iram.h is responsible for the MPA codec putting things in .icode/.ibss when it shouldn't. Not sure I'm the best person to tackle that, though.

longfeng328:
shotofadds,your attached file is empty...  :-[

shotofadds:
So it was. I've fixed it now.

saratoga:
http://www.rockbox.org/wiki/CodecPerformanceComparison#Telechips_TCC7801_ARM926EJ_S

Updated.

mbk1969:
Are binaries of "current build" built with that "IRAM"-optimization ?

Navigation

[0] Message Index

[#] Next page

[*] Previous page