stds-1394: empirical character of SBP2 scatter/gather lists?

Peter Johansson PJohansson at ACM.org
Mon Jun 25 18:18:55 PDT 2001


At 05:34 PM 6/22/2001, Pat LaVarre wrote:

>Hi.  Can anyone tell me where the people hang out who are into the 
>empirical character of SBP2 scatter/gather lists?
>
>I've got a legacy i/o stack that likes to think of i/o in terms of a 
>series of filling a series of buffers expressed as pairs of virtual 
>address with length.  Since this stack uses a heap of virtual memory 
>divided into 4KB physical blocks, the physical translation of a random 
>virtual address with a random length usually turns out to be a leading 
>fragment, a series of 4K blocks, and then a trailing fragment.  (((The 
>addressing is a mess when the system is fragmented, but the series of 
>lengths are consistent in this way.)))
>
>But this is NOT what I see in my 1394 traces.  There I see a lot more 
>fragmentation than I knew to expect.  I'm not yet clear on Why so now I'm 
>looking for people who are into this enough to have done the empirical 
>characterisation of what is typical a long time before I got involved.

Pat, I suggest that you try T10 at T10.org (first) and SBP3 at ISG.Apple.com 
(second). This ordering may seem counter-intuitive---after all, you're 
asking an SBP question---but the SBP-3 reflector is focused on extensions 
to the protocol while on the T10 reflector you're likely to find someone 
who will help with your questions about today's devices and operating systems.

You might clarify what you mean by "... a lot more fragmentation ..." in 
your 1394 traces. Remember, even though a page table entry describes a 4 KB 
contiguous address space, Serial Bus devices may be constrained by maximum 
packet size considerations to transfer the data in chunks that range from 
512 to 2048 bytes.


Regards,

Peter Johansson

Congruent Software, Inc.
98 Colorado Avenue
Berkeley, CA  94707

(510) 527-3926
(510) 527-3856 FAX

PJohansson at ACM.org


******************************************************
SBP-3 protocol for FireWire Mailing List
Unsubscribing:
send email to requests at isg.apple.com with subject "unsubscribe sbp3"

Set to Digest Mode:
send email to requests at isg.apple.com with subject "subscribe digest sbp3"

Help?:
send email to requests at isg.apple.com with subject "help"
******************************************************




More information about the T10 mailing list