Re: InDataOrder (or not InDataOrder)

[julian_satran@il.ibm.com]

Definitely - many high performance controllers, when they have cache
misses, return data out of order for operations that involve many blocks.

Julo

"Somesh Gupta" <somesh_gupta@worldnet.att.net> on 17/01/2001 06:43:22

Please respond to somesh_gupta@silverbacksystems.com

To:   "IPS" <ips@ece.cmu.edu>
cc:
Subject:  InDataOrder (or not InDataOrder)





Looking at the Appendix C of the latest draft, and the InDataOrder
key, it seems that both sides can decide that they want data in order
(for a single command) - or each side can independently agree to
accept data in order.

I have the following questions:

1. Does any target return data for a read out of order? Althought the
potential benefit is there, in practice is it used?

2. For a write, does this restriction apply to the entire
write operation or only to individual r2t's (I assume the r2t) although
in practice there seems little reason for the r2t's to be requested
out of order?

Thanks,
Somesh