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.
Technology Stocks : Advanced Micro Devices - Moderated (AMD)
AMD 215.32-0.2%Dec 30 3:59 PM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Scot who wrote (6848)8/28/2000 10:21:47 PM
From: John EvansRead Replies (1) of 275872
 
RE:"Some software applications cause circuit on 1.13 GHz Pentiums to fail"

gcc is a good benchmark because it has a lot of potentially mispredicted branches and non-linear (all over the place) memory accesses. In other words, lots of stalls, flushes, bubbles, etc. In order to perform well, a processor must have a large L1 and L2 cache along with small penalties (cache miss, branch).

I've been reading reports on instruction-level parallelism -- academic studies which try to determine how efficient superscalar and superpipelined designs will be on typical code. Invariably compilers are one of the hardest applications to execute efficiently. Among the worst is the unix program yacc ("yet another compiler compiler").

This is exactly the kind of code that will give Willy headaches. Take the press releases about "advanced branch prediction" with a grain of salt. Some branches are impossible to predict.

gcc also represents the kind of code that is run on workstations and severs. Unless they can ramp the P4 quickly, Intel will lose this market to AMD.
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext