Rockbox

This is the bug/patch tracker for Rockbox. Click here for more information.

Quick links: Bugs · Patches · Rockbox frontpage

Tasklist

FS#6670 - Small dct32 routine for the MPA codec

Attached to Project: Rockbox
Opened by Tomasz Malesinski (tmal) - Sunday, 18 February 2007, 19:08 GMT+2
Task Type Patches
Category Codecs
Status Closed
Assigned To No-one
Player Type PortalPlayer-based
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 0%
Private No

Details

The dct32 routine in libmad/synth.c is pretty big and it causes lots of cache misses. I have written a smaller one in ARM assembler. It isn't as optimal in terms of number of arithmetic operations, but it's much more cache friendly because of smaller code size. It helps a lot on Iriver iFP where SRAM is really slow. I would like someone with other ARM targets to check if it helps on them as well.

The routine is currently compiled to SRAM, but I suppose it's worth trying to move it to IRAM.
   dct32.patch (8.6 KiB) 
 new-rockbox.0/apps/codecs/libmad/dct32_arm.S |  284 +++++++++++++++++++++++++++
 new-rockbox.0/apps/codecs/libmad/SOURCES     |    1 
 new-rockbox.0/apps/codecs/libmad/synth.c     |    9 
 3 files changed, 294 insertions(+)

This task depends upon

Closed by  Tomasz Malesinski (tmal)
Wednesday, 28 February 2007, 00:53 GMT+2
Reason for closing:  Duplicate
Additional comments about closing:  Moved to  FS#6705 
Comment by Chris (decayed.cell) - Saturday, 24 February 2007, 04:40 GMT+2
What are we looking for? I've tested this on a 5.5G 30GB Video iPod
Comment by Tomasz Malesinski (tmal) - Saturday, 24 February 2007, 21:17 GMT+2
As tested by Preglow on Feb 18th (see irc logs), it doesn't improve performance on Ipod (I'm not sure which model). We were looking for smaller boost ratio (info->debug->view audio thread).
Comment by Chris (decayed.cell) - Sunday, 25 February 2007, 01:13 GMT+2
Okay what about  FS#6357  does that help or is it an outdated patch?

Loading...