Re: [ECOS] Best structure for multiple devices on one interrupt?

[Gary Thomas <gary at mlbassoc dot com>]

Brett Delmage wrote:
> I am running eCos on the LPC2xxx ARM, that has 3 external interrupt pins. 
>
> There are 17 devices that can generate interrupts, at rates ranging from 
> once every 2 ms (with a max allowed service time of 1 ms) to infrequently: 
> every few seconds or maybe hours or days, when active.
>
> Most of these interrupts exist so that the CPU can maximally sleep to save 
> power and be woken up when something happens, instead of polling when 
> there will be little activity. (the high rate interrupt will often not be 
> active.)
>
> I am wondering about the optimal way to set up the interrupt servicing. 
> Most of these interrupts come in on one external interrupt pin (EINT0); 
> two high-frequency ones and power-down request come in on the other input 
> (EINT2), such that these might receive higher priority or treated as an 
> FIQ instead of an IRQ -- at a later date. Mixing IRQs and FIQs on an ARM 
> seems problematic, so I will avoid this for now.
>
> To determine the source of the external interrupt, bit flags in an 
> external register can be read. Nothing unusual here.
>
> As per the NOTE in section 3.2.3 of the ECOS User Manual, would it be best 
> to have a single ISR triggerred by the external interrupt, which would 
> explictly post the correct device-specific DSR after examining the 
> interrupt status register?  e.g. by calling cyg_interrupt_post_dsr
>
> As I understand this, that way, the specific device interrupt can be 
> masked immediately, allowing other devices to interrupt. Also, If the ISR 
> examines the interrupt status bits in priority order, then DSRs can be 
> posted in priority order.
>
> Does this approach make sense, and are there any caveats ?
> Are there any examples of this in the repository that I could look at?

I'd extend the notion of HAL interrupts on the platform to explicitly
list all 17 sources.  Presumably, each source can be individually
tested/enabled/disabled in various platform registers.  Then it's just
a matter of extending the variant routines (in hal/arm/lpc2xxx/var) to
recognize them.

A framework for this exists on the iPAQ platform (sa11x0 ARM variant).
There are no details, but you can see how it can be done.


--
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

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