On Tue, 20 Jan 2004, David Brownell wrote:

> Date: Tue, 20 Jan 2004 19:38:13 -0800
> From: David Brownell <[EMAIL PROTECTED]>
> To: John Heil <[EMAIL PROTECTED]>
> Cc: [EMAIL PROTECTED], [EMAIL PROTECTED],
>      linux-usb-devel <[EMAIL PROTECTED]>
> Subject: Re: [linux-usb-devel] Re: Test of ehci iso patch
>
> Hi John,
>
> More later, but one quick comment:
>
> > Each urb at present has 2048 bytes so the mult bit is on and it
> > comes in 1K chunks. The production version will be about 2800+
> > so 3 chunks.
>
> Interesting ... so eight urbs per uframe, one IRQ each 125 usec.
>
> What happens when you make each URB larger, like maybe taking up
> a full frame with each one (16 KB)?  Or if you set URB_NO_INTERRUPT
> on some significant fraction (7/8 or better) of those urbs?
>

Oops. Sorry Dave!
I meant:  2048 bytes per transaction and 8 transaction per
URB/iTD and each read iTD is laterally chained to a write iTD and
the pair is anchored by a single periodic frame list slot so that
the controller re-reads the pair 8 times per slot.

There is a 1 to 1 mapping of urbs to iTDs.

Each iTD uses all 8 of its transactions so each uframe gets
2048 bytes inbound and 2048 outbound for 4096 bytes on the bus
per uframe, overall 4096 bytes * 8 uframes = 32K bytes on the bus,
in my testing mode.

The actual hardware will increaes that to about 2800 bytes
in + 2800 bytes out per uframe (Right now I'm on Cypress FX2 boards).

I do set the transactions URB_NO_INTERRUPT for 7 of the 8 transactions
per iTD. I will try setting it to 15 of the 16 (8 read + 8 write), thus
placing the interrupt on the last of the write transactions (which always
must follow the read).

>
> > It seems tho that the shorter our interrupt path the better especially
> > given that we have preemption capability and deadline sched'ng.
>
> Getting 1/8 as many interrupts is for many purposes the same
> win as an interrupt path 1/8 as long ... :)
>
> - Dave
>

True, but really tight top half code is still beneficial...
for everyone :)
Also, I won't have the luxury of staying on fast Intel silicon in the
final and prod versions and at that point, every cycle may count.

johnh

>
> -------------------------------------------------------
> The SF.Net email is sponsored by EclipseCon 2004
> Premiere Conference on Open Tools Development and Integration
> See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
> http://www.eclipsecon.org/osdn
> _______________________________________________
> [EMAIL PROTECTED]
> To unsubscribe, use the last form field at:
> https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
>

-
-----------------------------------------------------------------
John Heil
South Coast Software
Custom systems software for UNIX and IBM MVS mainframes
1-714-774-6952
[EMAIL PROTECTED]
http://www.sc-software.com
-----------------------------------------------------------------


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to