[Info-vax] Poulson at hot-chips 2011

[JF Mezei]

John Wallace wrote:

> Someone at The Register wondered that too, went digging, and came up
> with a ReaWorld Tech article from a few months ago.
> 
> http://www.realworldtech.com/page.cfm?ArticleID=RWT051811113343&p=4

Intersting, So the replay buffer is more about out of order re-execution
when results changed the input of those execution rather than a RAS
feature to counter memory errors.

Interesting that Intel's powerpoint slides justified the replay buffer
as RAS instead of a way to deal with faulty out of order execution that
changed results.

Out of curiosity, does Alpha have the same problems, having to replay
instructions because it failed to see that previous instructions could
change the inputs to those instructions ?