x in om2007.2

Andy Green andy at openmoko.com
Tue Jul 29 00:24:47 CEST 2008

Hash: SHA1

Somebody in the thread at some point said:
| On Sun, 27 Jul 2008 16:28:16 +0100 Andy Green <andy at openmoko.com> babbled:
|> Hash: SHA1
|> Somebody in the thread at some point said:
|> |> Why is this so (lesser memory bandwidth due to the glamo chip)?
|> |
|> | because that is the nature of the hardware on the freerunner. the glamo
|> | throttles write rates to its video ram (or read rates) and thus
brings the
|> | system into wait states waiting on the glamo (basically lock in your
|> cpu down to
|> | doing nothing WHILE you try and write data, and the writes do not go
|> as fast
|> | asht e cpu can manage - they hit about 7mb/sec - thats a fair bit less
|> than
|> | your average 100mbit ethernet can transport, and thats the video memory
|> | access from the cpu). it's just that is how it is, and it is a limit
|> you have to
|> | code around - read/write less data to/from video ram on updates.
|> Just a FYI, I have a pair of patches pending that claim to increase
|> throughput to Glamo without introducing new instability, and it is
|> possible to crank it up a bit more I think.
| reducing waitstates? i remember olv hand to increase them to stop
| instability... maybe it was a production issue at the time... any
increase will
| be very much welcomed - though i'd really like to see a 2x, 3x or 4x
speedup or
| more... but i doubt that's a happening thing. :(

No as I say without making new instability.  I've been running them
several days with a lot of SD Card traffic and no problems (hundreds of
kernel uploads and boots from SD), and I tested them when I did them
with the scrolling map app in ASU without any funnies.  It won't be 2X
but it should be something noticeable.

- -Andy
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org


More information about the devel mailing list