On Mon, Jan 28, 2013 at 9:25 PM, Laurent Gauch
<[email protected]> wrote:
> Le 28.01.2013 13:19, Xiaofan Chen a écrit :
>>> 1. It's said to be faster and it was faster than ft2232.c @ libftdi. I
>>> don't have a comparison with ftdi.c though...
>> I have done some comparisons last time when testing mpsse
>> codes, mpsse is still faster than the default d2xx implenmentation
>> as of now.
>
>> On the other hand, ft2232.c is not optimized for libftdi-1.0
>> and ftd2xx's asynchronous API. So the speed could
>> be improved using ftd2xx's async API or libftdi-1.0's
>> async API.
> Asychronous io access is not the problem, the problem is how to
> implement using it, in the goal to accelerate the JTAG and or SPI SWD
> ... :-)

Yes I agree.

> Also, we do not need true asynchronous io. Non-blocking fonction is not
> so important here. But we have to think more about the access gestion ->
> pipelining the access -> and let the host computer and the dongle as the
> Amontec JTAGkey-2 to work as two prosess and not as one process, like now !
>
>> I've seen one private ftd2xx mod which achieve similar
>> speed to mpsse under Windows.
>
> Yes,   with only 3 more lines of code in  the ft2232.c  we get the same
> speed as the mpsse.c libusb based, and maybe faster when over the d2xx
> kernel driver :-)

Could you post the modification you make to ft2232.c? Thanks.

> But for the story, we ( Amontec) have tried to come with these new code
> lines long time ago .... we received only negative feedback since we
> where working on the ft2232.c ;-(  .

Hmm, I do not know that. Maybe it is before I joined the mailing
list. I have some interests in OpenOCD as a JTAG debugger
and programmer but I mainly use OpenOCD as a test bed for
libusb-win32, libusb/libusbx and libftdi.

>>> 2. If one uses the adapter with some other tools that use ftd2xx driver
>>> (99% of commercial tools) then there's no need to change the driver
>>> every time you need to use the other program. Yes, I know about the
>>> filter-driver, but not everyone uses it.
>>
>> The libusb-win32 filter driver will work for libftdi-0.x but not libftdi-1.x
>> which depends on libusb-1.0/libusbx. libusb.org does not yet support
>> libusb-win32 driver yet. libusbx 1.0.14 does support libusb-win32
>> device driver but there are issues with the filter driver so the filter
>> driver will not work for FT2232x.
>>
>>> 3. Most adapter vendors don't bother to provide any other driver than
>>> ftd2xx (sometimes even with a nice OpenOCD binary linked to ftd2xx.dll).
>>>
>>> The most important is the first one, so I'll have to browse the net
>>> searching for such binary or maybe I'll try to compile myself and check...
>> Yes it would be interesting to see how you come to. I tend to believe
>> that mpsse will be faster than the stock ftd2xx code since ft2232.c
>> does not fully use the potential of ftd2xx.
>>
> The ft2232.c has the advantage to include the layout specific things.
> Tomek will say this is bad. Again, I will say this is good to have this
> hardcoded for stability for the end users. In an other hand, the
> excellent mpsse.c has a better use of buffers but it is too much libusb
> specific - really bad point for mpsse.c actually. MPSSE (the engine) is
> the protocol we use to remote JTAG SPI or SWD, the libusb must be seen
> as the kernel driver to transport the MPSSE command sequence over USB.
> The mpsse.c should be called mpsse-libusb.c as it is actually.

I tend to agree with you here. Technically you should be able to
use libftdi-1.0 or ftd2xx to replace libusb-1.0 to achieve similar
mpsse-libftdi1 and mpsse-ftd2xx.

-- 
Xiaofan

------------------------------------------------------------------------------
Master Visual Studio, SharePoint, SQL, ASP.NET, C# 2012, HTML5, CSS,
MVC, Windows 8 Apps, JavaScript and much more. Keep your skills current
with LearnDevNow - 3,200 step-by-step video tutorials by Microsoft
MVPs and experts. ON SALE this month only -- learn more at:
http://p.sf.net/sfu/learnnow-d2d
_______________________________________________
OpenOCD-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/openocd-devel

Reply via email to