public inbox for ecos-discuss@sourceware.org
 help / color / mirror / Atom feed
* RE: [ECOS] What portions of eCos code can be locked in instruction cache
@ 2003-12-18 23:25 Michael Anburaj
  0 siblings, 0 replies; 2+ messages in thread
From: Michael Anburaj @ 2003-12-18 23:25 UTC (permalink / raw)
  To: sandeep, ecos-discuss

Sandeep,

Sometime back on an ARM920T with slow SRAM memory we had a problem with a 
certain ISR. It demanded a very quick response, latency lesser than 2us. And 
when we measured it using a logic analyzer it was randomly swinging between 
8-30us. Moving this to FIQ (ARMÂ’s special fast IRQ option) did not help 
much. So, I locked down the entire ISR path right from the HW vectors, 1st 
level handler & the ISR into the I-Cache & the latency dropped down to <1us 
(consistent).

This was one such situation; I guess different devices may have different 
cache lockdown requirements.

Hope this gives an idea,
-Mike.


>From: "sandeep" <sandeep@codito.com>
>To: <ecos-discuss@sources.redhat.com>
>Subject: [ECOS] What portions of eCos code can be locked in instruction 
>cache
>Date: Thu, 18 Dec 2003 19:46:38 +0530
>
>hi list,
>
>in case you happen to get eCos working on a chip that has support for
>instruction cache locking, what portions of eCos kernel would you suggest 
>should
>be locked there for optimal performance?
>
>* HAL code that gets called often viz. context switch code, 
>interrupt/exception
>handling code
>* roughly what section of network code?
>* what part of kernel code? apart from scheduler related code, unlock_inner
>function, isrs/dsrs
>
>to instruction-cache-lock certain non-HAL code some changes may have to be 
>made
>to non-HAL code in eCos, eg. may be marking some functions with special
>attributes. to what extents it is fine to modify the non-HAL eCos code? 
>surely
>coz of this syncing with global eCos CVS is going to be bit of hassles than
>otherwise.
>
>hoping to get your views.
>
>peace
>sandeep
>
>
>
>--
>Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
>and search the list archive: http://sources.redhat.com/ml/ecos-discuss
>

_________________________________________________________________
Make your home warm and cozy this winter with tips from MSN House & Home.  
http://special.msn.com/home/warmhome.armx


-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss

^ permalink raw reply	[flat|nested] 2+ messages in thread

* [ECOS] What portions of eCos code can be locked in instruction cache
@ 2003-12-18 14:13 sandeep
  0 siblings, 0 replies; 2+ messages in thread
From: sandeep @ 2003-12-18 14:13 UTC (permalink / raw)
  To: ecos-discuss

hi list,

in case you happen to get eCos working on a chip that has support for
instruction cache locking, what portions of eCos kernel would you suggest should
be locked there for optimal performance?

* HAL code that gets called often viz. context switch code, interrupt/exception
handling code
* roughly what section of network code?
* what part of kernel code? apart from scheduler related code, unlock_inner
function, isrs/dsrs

to instruction-cache-lock certain non-HAL code some changes may have to be made
to non-HAL code in eCos, eg. may be marking some functions with special
attributes. to what extents it is fine to modify the non-HAL eCos code? surely
coz of this syncing with global eCos CVS is going to be bit of hassles than
otherwise.

hoping to get your views.

peace
sandeep



-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2003-12-18 18:25 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-12-18 23:25 [ECOS] What portions of eCos code can be locked in instruction cache Michael Anburaj
  -- strict thread matches above, loose matches on Subject: below --
2003-12-18 14:13 sandeep

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).