On Wednesday 12. October 2016 11.43.55 Luke Kenneth Casson Leighton wrote:
On 10/11/16, Paul Boddie paul@boddie.org.uk wrote:
Meanwhile, significant international partners should be doing their part, too: I noticed that Imagination had managed to prise the jz4780 SoC programming manual out of Ingenic for public dissemination around their Creator CI20
right. ingenic are an extremely small team: they simply haven't got time to mess about with "support" issues for people who aren't directly paying them. the amazing thing is that ingenic fully recognise this and just dump the entire source (albeit of really quite old kernels and u-boot but at least it works) directly onto their FTP server.
Your reports about the connectivity issues probably explain quite a lot. (Thanks a lot for posting entries from your travel diary, by the way!)
Ingenic has tried to push code into various places - SourceForge, GitHub - but it isn't a sustained effort, and the connectivity problem probably makes it even more of a chore to have to deal with such services. Best to just have one's own repositories and server, I think. Amusingly, the first obvious link I've seen to Kim Dotcom's Mega service appears on Ingenic's site. I guess that if you just want to get the stuff out there, that makes as much sense as Dropbox and all the "acceptable" sharing services. ;-)
i really really like ingenic, it's just that they've gone too specialist (and are sticking with ultra-low-power single and dual core), and their attempts to do 4-way or above SMP obviously didn't work out as well as they'd hoped (we had rumours of a jz4790 quad-core SMP design a couple years back, which never panned out).
They do have this focus on "Internet of Things", following on from smartwatches, and I suppose it makes sense for what they're offering. Similarly, those ICubeCorp people now seem to be focused on network equipment, if I understand the gist of their site.
now, here's the thing: what i might do is see if i can go meet them, and point them in the direction of RISC-V, particularly the work being done by the lowRISC team. they could either learn from that (and make themselves a decent SMP design) ooorr..... they could make a really decent RISC-V SoC, using some of the interface hard macros that they've already got.
I don't have much to say on whether their architecture choices make sense. I note that the jz4780 documentation mentions "MIPS" as opposed to just "XBurst", and maybe there is mileage for them in continuing along the same path, perhaps incorporating technologies from MIPS, perhaps evolving their own XBurst extensions.
Just a communication of interest could be useful, I suppose, given that Ingenic products aren't generally available through the obvious western channels (as far as I can tell) and thus no other convenient mechanism for communicating ongoing demand seems to exist.
Paul
P.S. On the subject of openness in general, for any small operation it does make sense to limit support to people who are genuine customers. One might argue that as long as they have customers who approach them directly and business is conducted on that basis thereafter, there is no need to cultivate a general audience for their products.
But at some point the lack of public documentation undermines general confidence and interest in the products being offered by any such company. Random engineers are less and less interested in having to enter (or having their managers or superiors enter) some kind of sales relationship to get access to materials, when it all could end in important materials being withheld once the supplier has gained what it wants from the relationship. (Think about how a lot of recruiters operate if you want another example!)
It's tempting to frame such differences in working as "east versus west" cultural differences, but they are not: "enter a relationship with us to get things done" is precisely the kind of thing that still permeates western academia, not helped by the rampant "intellectual property" opportunism infecting academic institutions. It may be the case that working together will bring about more rapid progress, but when one first has to evaluate who to work with, and when the basis of that evaluation is a limited amount of public information, a lot of time is wasted before any actual work can be done.