[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Alpha recommendations

I'll take a look at libffm to see what's included.  Those routines are 
optimized and faster than glibc, but are accurate for only a restricted 
domain, or so I've been told...

I've heard that rth is working to integrate some of the new math routines 
into glibc2.1.

99% of my work is integer related, which Alpha excels at too... so 
unfortunately I don't have extensive experience optimizing fp routines.  
The irony is that Alpha has such a reputation for fp work, that people 
sometimes don't realize its potential for data-processing, etc.


On Thu, 15 Oct 1998, Scott Lewis wrote:
> >     CPU/OS                   Integer      FP
> >     -----------------------  -------  ------
> >     Pentium 233MHz, WinNT4:     2.63    2.22
> >     Sun UltraSPARC, 350MHz,
> >     Solaris 2.6:                3.54    3.11
> >     Alpha 164SX, 533MHz,
> >     Debian/Linux:               8.17    2.92
> > 
> > (Tests were performed with GCC 2.7.2 on Pentium and UltraSPARC, and 
> > egcs-1.1 on Alpha.)  Clearly my 164SX does well on integer.  The fp 
> > results are disappointing, but not suprising since they depend on sqrt(), 
> > sin() and cos() which are poorly optimized in glibc.
> I heard that some folks are working on sqrt, but what about other fp ops?
> I would love to see what my Alpha can really do, especially since I intend
> to start a research project with some pretty heavy math in it.  (I intend
> to _try_ to manipulate current ray-tracing algorithms to use acoustic
> parameters rather than optical in an attempt to model room acoustic
> performance, if anybody cares.)  Well, for that matter, what is involved
> (generally) in optimizing said operations?  Is it a matter of
> reformulating the algorithm, or do you actually need to do alpha-specific
> stuff?
> Scott Lewis   
> Computer Support
> Department of Civil and Environmental Engineering
> Georgia Institute of Technology
> scott.lewis@ce.gatech.edu   (404) 894-2210
> "Make it idiot proof and someone will make a better idiot."

Reply to: