SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Politics : Formerly About Advanced Micro Devices
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Paul Engel who wrote (116222)6/18/2000 3:55:00 PM
From: EricRR  Read Replies (1) of 1572139
 
Foster will be 100% compatible with existing x86 software - so any existing x86 scripts - PERL, etc., - & JVM should run on it quite nicely - and at impressive speeds !

I don't think you've addressed my question- After all, Itanium is also 100% x86 compatible. My point point is that run-time linked programs tend to generate substantially more mispredicts, so any resulting penalty is critical in evaluating server performance.

Lets take a simple example: A processor with a 10 stage pipeline runs at 1GHz. It has a 95% branch predict success rate (for non scripting code). Thus: 5% * 10= 50% = The processor runs ~33% slower because of those mispredicts.

Compare to a 2GHz processor with a 20 stage pipeline (willy's "in trace" penalty) and a 95% success rate. 5% * 20 = 100% = the processor runs ~50% slower.

Thus the second processor runs only 50% faster than the first, despite being clocked twice as fast.

If the second processor had a mispredict of 40 cycles (all I know is that willy's non trace is greater than 30), it would provide an equivalent performance to the first processor.

Now Willy is supposed to have a better prediction algorithm than coppermine's 95%, but how much better? 97? 98? And will such schemes help so much on run time linked languages?

The only thing I can say for sure is that the branch mispredict problem won't show up in SPECint!
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  

HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2025 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News