We have two machines with AMD 64 processors, one with Asus motherboard & Via chipset. Another with Asus motherboard with Nvidia chipset. On the one with the Via chipset using WinXP, when an external interrupt is received from a stargen chip, it takes 8 microseconds for the response to the interrupt. On the Nvida chipset, it takes 40 microseconds. Both machines are running the same program (In house developed driver).
Is this a common occurance between different chipsets? Would LinuxBios help this situation any?
The software being developed on .NET will also be ported over to Linux using Mono in the near future (Suse Linux).
Art
On 10/11/05, Art Fore art.fore@comcast.net wrote:
What are you looking at to determine this? It may be a driver issue and not a hardware issue. Are you flipping a bit somewhere when the interrupt is acknowlegded in the driver or when your app responds?
-- Richard A. Smith
The PCI interrupt line was monitored with a scope. Driver was the same in both cases.
Art
Richard Smith wrote:
On 10/12/05, Art Fore art.fore@comcast.net wrote:
The PCI interrupt line was monitored with a scope. Driver was the same in both cases.
Ok so you are watching for the IRQ ack from the IRQ driver. I don't know a whole lot about the deep internals of the windows irq stack. Does the system ACK the IRQ and pass up up a notification or when you hook the IRQ is your driver responsible for all that.
By driver I meant the chipset driver so they can't be the same. Via vs nvidia. I would guess that the difference is in the way the chipset driver deals with interrupts.
Unless you can somehow track it down to a setting in the chipset I doubt linuxbios will be much help.
Are you using some type of RTOS extensions for XP? I wasn't aware that you could get response times that small with XP.
-- Richard A. Smith
No, we are not using any RTOS extensions yet, although we do have the embeded XP. I am trying to get them to test it under Suse Linux which also have installed on the machine, but the stargen PCI drivers need to be compiled for it and the software ported over. The user level software is in .net and we plan on using it on Mono. That is going to be interesting. We are getting a 2 microseccond latency in the startgen chips though. Thanks for the info though.
Art
Richard Smith wrote:
-
Art Fore -
Richard Smith