Re: [Linux-gpib-general] patch proposal

[Michael S. <mi...@sc...>]

On 29.07.24 00:24, marcello.carla wrote:
> It is a possibility and has been considered. The reason that led to
> prefer edge interupts to level ones is the unavoidable overhead that
> comes from the doubling of NRFD and NDAC interrupts. Not huge, but not
> negligible. In any case, hardware problems like a chattering interrupt
> or repeated interrupts due to line reflections cannot be handled in
> software. 

They can be handled in software, and IMHO they should be - that is part 
of designing robust software.

When dealing with an asynchronous bus mechanism like GPIB, such stuff 
can happen and software needs to cope with it.

The only overhead I see is that using level-triggered interrupts, I need 
to enable/disable the interrupts and re-program the level for each 
transfer. I don't know how costly those operations are, but since we are 
talking about a 800MHz clock CPU pushing data at something like 1MB/s, 
this should not be a huge problem.

cu

Michael