STM CircleOS forum / Choice of RLink as an integrated programmer

paulfertser · 2009-09-29 17:15:04

Hi,

After some studying of Primer2 capabilities and internals i'm still quite puzzled by the choice of Rlink programmer.

Comparing with a generic usb<->gpio based solutions (ft2232 and the like) i don't see much advantages but some disadvantages of RLink are clear:
requires special drivers that can be installed only on non-standard OSes;
requires admin privileges to install drivers;
requires proprietary tools that can work only on non-standard OSes;
requires buying license to be able to step through CircleOS code (since it's above 32k artificial make-them-pay-for-air limit).

So can anybody enlighten me and explain what advantages does Rlink have and why anybody thinks they're worth it?

paulfertser · 2009-10-04 07:49:37

Francis,

I see you're quite active on the forum and yet you seem to be ignoring my question. I very much hope you'll reconsider and post the answer.

I can assure you that i did my best to write the previous post in the clearest and most polite way, refraining from all the words that can be declared insulting here despite the fact that i consider 32k debugging limit to be a _personal_ offense.

So please, tell us about the other alternatives that were considered during r&d phase and what advantages RLink has over them (especially comparing to plain USB<->GPIO converters).

gussenhovenS · 2009-10-04 21:28:37

Hello,
In spite of criticisms, RLink is a highly optimized interface for programming and debugging.

This said, the most important advantage of buiding the Primers around RLink is the price of the final product. Raisonance could have opted for another interface technology, but cost of development would have been much higher. The trade off would have been a higher price to the end user or fewer hardware and software features.

By optimizing development costs the user gets an exceptional tool, with a great set of development features (color LCD, touch screen capability, codec-based audio, accelerometer, MicroSD card, ...) at a very reasonable price.

-s

paulfertser · 2009-10-04 21:59:10

Hello,

I'm sorry but i can not exactly understand how exactly using e.g. ft2232 can increase the cost of development. The usb-gpio converter solutions were already developed and tested by numerous companies and individuals, see e.g. http://www.hs-augsburg.de/~hhoegl/proj/ … bjtag.html . So my assumption is that using one of those well-known schemes can not really increase the development cost because the schematics are well known and (mostly) obvious. I'm sorry to sound so skeptical here but could you please provide me with more information so i will have no doubt that it would indeed cause an unreasonable price increase. And even considering quite conservative values of $2000 increase of development costs (which sounds rather high to me given the task at hand) and 10000 of produced devices we're having only $0.2 price increase per unit. I bet it's negligible comparing to the amount of money spent on hand-reworking every unit (for adding those ground traces to prevent damage to the voltage regulator) and on repairing those that got damaged.

Also you say that "RLink is a highly optimized interface" but how much is the actual difference? Do you have any convincing numbers?

But i fully agree, the tool would be really exceptional and exciting if not for the questionable RLink feature you seem to feel "ok" about. Currently, i have mixed feelings about this device as a whole.

Francis · 2009-10-05 06:21:22

The question was more related to the business model. Up to now, Raisonance makes its money (for 32 bit cores) with the debugging tool since IDE and compiler are free. Obviously, we cannot sell a complete solution for $0.2. Now the Primer concept could be considered apart from the standard tools, and I agree that we should make the Primers more opened. We are going to this direction.
For the next generation of Primer, we'll still keep a proprietary solution, but there is a also a good technical reason: we need SWD, not only JTAG (that was already true for Primer2), and we will add SWV (trace, uart-like debugging port,...). But we will try to make it more opened.

paulfertser · 2009-10-05 21:29:25

Hello,

Francis, i really appreciate your frank reply. And it is great to hear you understand frustration of (some) users and that future Primers are going to be more open.

It would be nice to see these development boards to be usable without proprietary software since you know that quite some amount of talented programmers prefer it that way. It'd also give more flexibility for those who want to use a decent standard operating system (e.g. GNU/Linux).

I hope you will find a way to satisfy your business requirements (to promote Raisonance tools) without harming those advanced users who prefer to use OpenOCD/gdb and Emacs or Eclipse or even Vim for development. Please think about that for a moment: there's a lot of very talented folks who can learn fast, read the sources and fix bugs almost instantly, who are really bright. Also judging by my experience free software developers are usually quite friendly and helpful. It always depresses me to see them suffer (or do useless work like reverse engineering) because of some (not even always sensible money-wise) business policy.

Your average John Doe schoolboy will not bother installing OpenOCD and other "complex" development tools on windows anyway so the majority of users will still be "hooked" on your tools.

As to the SWD and SWV i do not see why it can not be implemented with a generic converter. Moreover i've already seen an SWD implementation that works over ft2232. Probably RLink is a bit faster but how actually faster is it? It would be nice to know exactly if its use is justified or not.

BTW you perfectly know that RLink's proprietary protocol was reverse engineered relatively long time ago. Do you think it caused _any_ financial loss? I really doubt that, in fact i think it was a gain because some OpenOCD users choosed to buy RLink and not another programmer just because it was supported in OpenOCD. And if this was essentially harmless why not publish it officially in the first place?

Thanks again for your time.

Francis · 2009-10-06 07:03:11

I agree with you on the principle... For the choice "ft2232 vs RLink", yes RLink is faster because it can be optimized. We typically manage 250 kbyte/s (2MBit/s) of useful data for uploading/downloading. Perhaps the ft2232 will be able, in the future, to support SWD/SWV, but it's not the case right now. Conclusion: RLink is for us the best (technically) solution.
Now, we would like to control the 'Windows' driver using RLink. You are right when you say that OpenOCD does not hurt our business. Perhaps we could find a partnership to provide an OpenOCD driver...

# 1 2009-09-29 17:15:04 Choice of RLink as an integrated programmer

Choice of RLink as an integrated programmer

# 2 2009-10-04 07:49:37 Choice of RLink as an integrated programmer

Re: Choice of RLink as an integrated programmer

# 3 2009-10-04 21:28:37 Choice of RLink as an integrated programmer

Re: Choice of RLink as an integrated programmer

# 4 2009-10-04 21:59:10 Choice of RLink as an integrated programmer

Re: Choice of RLink as an integrated programmer

# 5 2009-10-05 06:21:22 Choice of RLink as an integrated programmer

Re: Choice of RLink as an integrated programmer

# 6 2009-10-05 21:29:25 Choice of RLink as an integrated programmer

Re: Choice of RLink as an integrated programmer

# 7 2009-10-06 07:03:11 Choice of RLink as an integrated programmer

Re: Choice of RLink as an integrated programmer

Board footer