Re: [Gems-users] questions about opal's branch prediction and ruby's fast path hit


Date: Wed, 29 Jun 2005 10:12:11 -0500 (CDT)
From: Luke Yen <lyen@xxxxxxxxxxx>
Subject: Re: [Gems-users] questions about opal's branch prediction and ruby's fast path hit
Ayse:

  1) There is no flag for perfect branch prediction.  However there is a
feature in Opal that I have never used, and that is reading/writing branch
traces.  I haven't looked into that code too closely but you might be able
to collect a branch trace for your workload and feed it back into Opal to
simulate perfect branch prediction.

  2) I don't believe setting fastpath hits to be > 1 cycle should affect
Ruby, and I think it should be independent of protocol.

  Luke

On Tue, 28 Jun 2005 yilmazer@xxxxxxxxxxx wrote:

> Hello,
> I have two questions.
> First one, Is there any perfect branch prediction in Opal?
> And, second, If I change L1 hit latency to 2 cycle, does ruby still work with
> fast path hit (for MOESI SMP directory protocol and MOSI SMP broadcast
> protocol) or not? If it doesn't, what happens?
> Thanks for your help,
> Ayse Yilmazer.
>
>
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
>
[← Prev in Thread] Current Thread [Next in Thread→]