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 : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?


To: Scumbria who wrote (127688)2/18/2001 1:01:46 AM
From: Ali Chen  Read Replies (1) | Respond to of 186894
 
"Mainstream apps are not going to be compiled for P4 for a long time"

Even if this may be not completely true, it may not
change much.
The problem with Intel's SPEC compiler is that the compile
is good only for the particular (SPEC) application with
the particular set of data. This is because the Intel's
compiler works in two passes - first it runs the first
cut of code on the data set, learns where weak places
are ("profiles" the data/code), and then uses this
profile to optimize the final code. On the top of this,
a performance test engineer tries variety of hand-picked
compiler switches to improve the run time result.
Hence the compiled code is optimal only for that
particular data set that was frozen by benchmark
designers. If the data size/structure frequently
varies as in real-world applications, the performance
of double-optimized code may not be spectacular.
I am afraid it is what we are seeing in most other
benchmarks that use a "frozen" off-the-shelf code.

Regards,
- Ali