Re: [Openocd-development] OpenOCD libftdi and ftd2xx benchmark

[Xiaofan C. <xia...@gm...>]

On Sat, Jul 16, 2011 at 10:49 AM, Peter Stuge <pe...@st...> wrote:
> Many thanks for making these tests! Awesome!
>
> Xiaofan Chen wrote:
>> Actually the result is pretty close for the LPC-P2148 based test.
>> jtag_khz = 1500 KHz, 38.927 KiB/s (ftd2xx) versus 38.754 KiB/s.
>
> So in conclusion there is almost no difference in performance between
> OpenOCD using libftdi-0.19 and ftd2xx, or did I overlook something?

Yes ftd2xx is only slightly faster based on my limited tests.
Freddie, Laurant and others may have more inputs

> If not, I would suggest to only use libftdi in OpenOCD.

I think it is still good to have both options.

> Uwe, do you see a possibility to port libftdi-0 over to use the
> libusb-1 API? If only synchronous is used it should be straight
> forward. I'm not saying that *you* must do it, but am rather asking
> how much effort it might be.

libftdi-1.0 is based on libusb-1.0 and uses both sync API
and a bit of async API. It is right now API compatible with libftdi-0.19.
http://developer.intra2net.com/git/?p=libftdi-1.0

> The benefit for OpenOCD would be that it can use libusb-1 on Windows
> without first having to rewrite the ft2232 driver.

That is right. It is actually not difficult to switch to libftdi-1.0 since
the current 1.0 API is compatible with libftdi-0.1x.

> That rewrite is
> absolutely desirable, but it seems that everyone agrees that the
> driver deserves and needs a thorough job, so would be nice to be able
> to punt on that and still be able to start using WinUSB and libusb-1.
>

I did some tests for libftdi-1.0 last time and it did not
offer any speed improvement for OpenOCD since OpenOCD
has not taken the advantage of the libftdi-1.0 async API.

http://lists.berlios.de/pipermail/openocd-development/2010-February/014895.html
http://lists.berlios.de/pipermail/openocd-development/2010-February/014896.html

-- 
Xiaofan