cpu-speed limitations

[Alex Kavanagh]

Jay Vaughan wrote, On 05/08/08 16:32:
>> I don't think that this is necessarily the case.  OM people have  
>> lots of
>> fish to fry and writing a acceleration code running inside the Glamo
>> probably isn't *the* most important thing to be working on.  Getting  
>> the
>> core kernel, device libraries and some application software working  
>> IMHO
>> *is* the most important thing.
>>
>>     
>
> The thing is though, these sorts of things can be done by people who  
> haven't signed an NDA to get access to the docs. 

I'm not so sure that they can.  The various issues do actually seem to
need OpenMoko staff to solve them.  Look at all the work Andy is putting
into solve GPS issues, kernel issues, etc.  And Joerg on the GSM
issues.  I don't think that OM have enough staff to do all of the that
and write Glamo code.
>  It makes no sense to  
> me that the only people who can do something about the pitiable Glamo  
> situation are instead worrying about other stuff that the rest of the  
> community, unencumbered by NDA's, can handle ..
>
>   
>> I think you might be shooting the wrong people here.  S.Media are the
>> 'bad' guys as they won't/can't provide the documentation.  However, if
>> this were always the case then maybe the Glamo was a poor choice.
>> However, I imagine that it *was* the right decision _at_the_time_ the
>> decision was made; everything is clearer with hindsight and everybody
>> *thinks* they would've made a better decision.
>>     
>
>
> Sure, sure.  I just wish that the most pressing issues of non- 
> operating hardware were being properly addressed.  I don't think its  
> cool to buy a phone that has features that go unused, just because  
> we've got an NDA situation going on.  We in the Open Source community  
> can route around that damage, and it seems the only way to handle that  
> right now is if the 'official' folks in OpenMoko address the issue.
>
> It would be a terrible shame if, after all the 'more important work'  
> gets done, Glamo gets ignored because 'now its time to build new  
> hardware'.  I think it'd be rotten, in fact.
>   

I'd agree with that, if that happens - but that hasn't happened yet. 
You're imagining the worst possible outcome here? Perhaps I'm too
naive.  Although I can imagine a scenario where the OM team would move
quickly onto the GTA03 once the GTA02 is 'good enough', the community
can keep the pressure up to 'do' something with the glamo chip when the
other issues on the GTA02 have been solved.

Is the glamo graphics stuff the 'most important' thing that should be
done?  i.e. in firmware on the glamo chip?  The qtopia image (which uses
a framebuffer?) seems to get reasonably good performance without
anything extra.  Maybe just a bit more work needs to be done in the
Xserver to stop it doing 'expensive' ops (as Raster suggested?) rather
than full on acceleration in the chip.  This is something that could be
done in the community.  Complex stuff, either way.

-- 
Alex.