From: GAMMELJL AT SLU DOT EDU
Date: Sat, 11 Mar 2000 10:22:23 -0600 (CST)
Subject: AMD processors and assembly language
To: djgpp AT delorie DOT com
Message-id: <01JMWIIY2ZT69ASSJ4@SLU.EDU>
Organization: SAINT LOUIS UNIVERSITY  St. Louis, MO
X-VMS-To: IN%"djgpp AT delorie DOT com"
MIME-version: 1.0
Reply-To: djgpp AT delorie DOT com

     Reply to Grendel: Thanks for your message.  I agree with you that
modern compilers are so efficient that there is no need to use assembly
language in most cases.  
     What I do is hydrodynamics.  Because of instabilities, I do the
calculations to more than 500 significant figures.  I have written a
language which I call "genprec" which is like a language such as basic
except that all arithmetic operations are done to whatever precision
the user chooses--provided, of course, that the user has enough ram.
     In high precision arithmetic, the assembly language commands
mul, div, and adc are essential.  In the 486 architecture, these
commands are efficient.  In the alpha architecture, they do not exist.
     AMD's 64-bit processors will follow the 486 architecture.  Intel's
will follow the alpha architecture.  As far as I am concerned, the
AMD processor (code name sledgehammer) will be the processor of the
future.  The Intel processor (code name Itanium ?) is, unfortunately,
not the processor of the future.
     genprec adds two new classes to C++, namely multiple precision
real numbers and multiple precision complex numbers, and utilities  
for printing them.  All arithmetical operations (and the elementary
transcendental functions) are overloaded so that one writes code exactly
as one would write any C++ code).  
     Other such general precision codes exist but I find them harder
to use and less efficient than genprec.  
     I am willing to donate genprec to djgpp (or delorie.com) if anyone
is interested.  I would mail a floppy to an address to someone who
knows how to get it posted as a selectable include option in the djgpp
package.