#9694 NORM 1.5-fir: Memory speed slower in OFW than Phoenix
Zarro Boogs per Child
bugtracker at laptop.org
Fri Nov 20 23:06:15 EST 2009
#9694: Memory speed slower in OFW than Phoenix
-------------------------------------------+--------------------------------
Reporter: wmb at firmworks.com | Owner: wmb at firmworks.com
Type: defect | Status: assigned
Priority: normal | Milestone: 1.5-firmware
Component: ofw - open firmware | Version: Development firmware
Resolution: | Keywords:
Next_action: reproduce | Verified: 0
Deployment_affected: | Blockedby:
Blocking: |
-------------------------------------------+--------------------------------
Changes (by wmb at firmworks.com):
* status: new => assigned
Comment:
One difference between OFW's and Phoenix's memory controller settings is
the "interleave type" field that controls which CPU address bits go to the
BA[2:0] (bank address) bits on the memory chips. That could affect memory
speed on some workloads by changing the level of concurrency.
In order to adopt the Phoenix settings, it is necessary to know the
mapping from CPU address bits to memory chip multiplexed address pins for
different interleave types. The reason that is needed is because it
affects the value you must write to the memory chips during the "MRS" and
"EMRS" mode setting steps.
That mapping is not documented in the Vx855 Programmer's Manual. The
mapping for the interleave type that OFW currently uses can be inferred
from the coreboot source code which uses that same setting (which is why
OFW uses that setting), but I don't have any similar way to infer the
mapping for Phoenix's interleave type. I have a question out to Via,
asking them to provide the mapping information.
--
Ticket URL: <http://dev.laptop.org/ticket/9694#comment:3>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system
More information about the Bugs
mailing list