[uClinux-dev] 2.4.x questions
gerg at snapgear.com
Wed Nov 14 22:14:40 EST 2001
Travis Griggs wrote:
> Since we were able to get colilo working with the help from many from
> this group; we've got the nod to actually use it in our products. We
> dl'd the modified colilo we submitted a week or so ago, and found a
> couple of rough left overs; we'll submit the same cleaned up tomorrow.
> We also dl'd the new 2.4 based uClinux and begin putting together
> development environments and instructions for colleagues. It looks like
> it has a LOT more options. We had a number of questions though. Most of
> them had to do with options we encountered as we walked thru make xconfig.
> 1) There's this section at the bottom of the "Processor type and
> features" menu. We set the FLASH type to AMD. But we left the RAM/FLASH
> size/bit width set as AUTO. We actually know what the size of each is...
> but we wondered what the AUTO option does?
I made these options configurable by all ColdFire targets.
But they only do something if the code for that target actually
uses them. Not all currently do.
The idea of AUTO is for boards that can have multipe RAM
size configurations, but can detect it automatically.
It would also mean that for a board that only has 1 possible
RAM size to use that and ignore anything else.
(I know the Arnewsh 5206 board does autosizing).
The FLASH type defines also only apply to the blkmem driver.
(They are not related to the MTD driver config).
> 2) In the same menu, there is an option for the Kernel to execute from
> HIMEM. What is this? Is it the 4K on chip ram found in the 5272? Even
> more fundamentally, does (and how) uClinux use the on chip ram? Mybe for
> the stack?
This option is not used by ColdFire targets.
Currently the 4k internal RAM is not used by uClinux.
> 3) Another menu is the MTD one. If I've got the M5272C3 board, are any
> of these things going to be useful to me? It seemed like this might give
> me run time control over the flash ram, but I'm kind of new to this
> flash stuff.
You can configure MTD to use the FLASH on your board.
Be very carefull if you set this up, you don't want to
overwrite the standard dBUG ROM.
> 4) In the Application Configuration>>Core Applications menu, we enabled
> netflash and recover for grins. We found that when we then tried to
> 'make', it could build neither of these. Cursorily it seemed that it
> couldn't find zlib.h. Is this a known problem?
Yep, it is now. Fixed in next release.
> 5) Finally, we had built a little network client/server pair which timed
> how long it took to transmit 4MB of image data over the ethernet. With
> the older (2.0 based version) and a 2.2 kernel for the host box, we were
> seeing transmit times of ~1.9 seconds. One of the reasons we got excited
> about the 2.4 based uClinux, is we thought this time might improve a
> little. It did the opposite. The host box is running 2.4.(12 I think).
> With the 2.4 based uClinux, our xmit times go up to ~2.4 seconds. Is
> there a known reason for this? A setting we can flick somewhere which
> will improve performance at least back to the ~1.9?
I hadn't benchmarked both that closely. The FEC driver in each
kernel version is based on a different code base (with an acient
common heritage.) It might need some digging into to see why
the big difference.
Greg Ungerer -- Chief Software Wizard EMAIL: gerg at snapgear.com
SnapGear PHONE: +61 7 3435 2888
825 Stanley St, FAX: +61 7 3891 3630
Woolloongabba, QLD, 4102, Australia WEB: www.snapgear.com
This message resent by the uclinux-dev at uclinux.org list server http://www.uClinux.org/
More information about the uClinux-dev