> I was always intrigued by that FX2 GPIF stuff -- that's what you're
> talking about?

Yes, they call it GPIF when the 8051 is the master (this is the more
complicated mode!), they call it Slave FIFO when an external device (the
CPLD) is the master. The CPLD just reads and writes to a 16/8 bit bus pretty
much like it was a memory device. The 8051 is only used to setup the USB
configuration, after that it sits in a loop and does nothing. The CPLD has a
direct pipe to the USB data and everything can be controlled through USB
traffic.

> That's why I mentioned Cortex-M3.  There are a lot of those around;
> and GCC handles them quite nicely.  So does OpenOCD, for that matter!

I do like the Cortex-M3 I have designed a board already for another project
using a Luminary Micro Cortex device which has Ethernet on it and the newer
ones have USB OTG, I believe they are currently sampling a device with both
Ethernet and USB.

I'm using the FX2 primarily because it was on my desk from my last project
but also because it's a fully "high speed" device.

Liam.

----- Original Message ----- 
From: "David Brownell" <davi...@pacbell.net>
To: <openocd-development@lists.berlios.de>
Cc: "Liam Redmond (Rock Software)" <i...@rock-software.com>
Sent: Monday, November 23, 2009 11:26 PM
Subject: Re: [Openocd-development] Openocd vrs Commercial jtag dongles


> On Monday 23 November 2009, Liam Redmond (Rock Software) wrote:
>> > Sadly, that also requires a lot of engineering expertise to write that
>>
>> Actually the hardware description is not as bad as you might first think,
>> I
>> have a prototype Xilinx CoolRunnerII CPLD hooked up to a Cypress high
>> speed
>> FX2, this effectively gives a direct pipe from the OpenOCD USB calls to
>> the
>> CPLD (via the FX2 slave interface).
>
> I was always intrigued by that FX2 GPIF stuff -- that's what you're
> talking about?  I think they did that to speed up IDE-to-FIFO data
> transfers, so the CPU wouldn't be involved.  On the other hand, any
> 8051 is a turn-off for me.  :(
>
> That's why I mentioned Cortex-M3.  There are a lot of those around;
> and GCC handles them quite nicely.  So does OpenOCD, for that matter!
>
> - Dave
>
>

_______________________________________________
Openocd-development mailing list
Openocd-development@lists.berlios.de
https://lists.berlios.de/mailman/listinfo/openocd-development

Reply via email to