4 Sep
1996
4 Sep
'96
3:53 p.m.
In message <xoienkiiahn.fsf@chops.icp.net>, Sean Doran writes:
Curtis Villamizar <curtis@ans.net> writes:
There are routers that handle {LS}SRR at near full speed. Stuffing it on the broken path in is a bug. Fast-path/broken-path is just a bug.
I'm not sure I agree, particularly in the case where "broken-path" involves delegating work to another processor.
Whenever you discover that in your design you risk sending 5,000-10,000 little delegations of work per second from one processor to another and taking the entire box down in the process it is time to seriously consider whether you've made the right design decisions. That's all. Curtis