|
|
|
This article
suggests how to get the best performance from an UltraSPARC or
x86/AMD64 (x64) processor running on the latest Solaris systems by
compiling with the best set of compiler options and the latest
compilers. These are suggestions of things you should try, but before
you release the final version of your program, you should understand
exactly what you have asked the compiler to do.
|
|
Large applications
have a particular problem: they have a lot of instructions, and the
processor does not have the capacity to hold the entire application
on-chip at any one time. As a consequence, larger applications spend
some of their run time stalled with the processor waiting to fetch new
instructions from memory. This paper discusses several techniques that
help the processor to hold more useful instructions on-chip,
consequently reducing the time wasted fetching data from memory.
|
|
In the past, most
Sun customers had a Sun workstation running the Solaris Operating
System on their desktop. They could easily develop applications for
deployment on other workstations or on large Sun servers. Today, many
developers have a PC running Microsoft Windows on their desktop
instead. How can these developers use Sun Studio software to develop
their Solaris applications?
|
|
The Fortran 95
compiler, f95, now accepts many of the non-standard features that the
legacy FORTRAN 77 compiler, f77, accepted. This paper outlines those
compatible features and provides useful information for migrating
FORTRAN 77 codes to the Fortran 95 compiler.
|
|
This paper is an
overview of all the new interval features that have been introduced
into Forte Developer Fortran. This is a good paper to start with, and
then go on to other white papers of interest.
|
|
This paper
addresses a number of interesting issues that come up when working with
interval angles or any periodic numbers. The specific issue that
motivated the paper was to rationalize what the containment set (or
cset) is for the ATAN2 Fortran intrinsic function for all possible
interval arguments.
|
|
This paper
introduces readers to the advantages provided by arithmetic systems in
which there are no undefined operators, operands, functions, or
relations. Such systems are referred to as "closed." The obvious
advantages in programming are that exceptional events cannot occur, and
therefore, many special cases need not be explicitly coded.
|
|
Many possible
closed interval systems can be constructed from containment-set (cset)
theory. For a purely software interval implementation, the simple
system described in this paper is a reasonable compromise between
narrow interval results for special edge cases, and speed for cases
that occur more frequently.
|
|
In interval
mathematics, the distinction between constants and variables is even
more important than it is in point mathematics. This paper introduces
language conventions to deal with the problem of transforming a literal
decimal constant into the narrowest possible width interval that
contains the literal constant's value.
|
|
One consequence of
working with a closed interval system is that values outside the domain
of functions such as sqrt and log do not cause an exception. The paper
describes how it possible to numerically find the root of a function
when the root is on the edge, or the boundary of the function's domain
of definition.
|
|
Because intervals
are sets of numbers, it makes sense that there should be an empty
interval. The question is: how should this construct be used? That
question is answered in this paper.
|
|
In Fortran 95 and
2000, literal constants are assumed to have a kind type parameter value
(or precision, as in single, double, or quad). Following this
convention prevents an interval implementation from constructing narrow
intervals when, for example, 0.1 is encountered in a double precision
interval expression. To eliminate this problem and to permit mixed
interval-non-interval expressions, interval expressions are evaluated
using rules that are not the same as standard Fortran. This paper
describes how and why this is done.
|
|
This paper describes in detail why the "no-exception" interval system has been implemented.
|
|
This paper presents
the motivation and theoretical foundation for containment-set (or cset)
theory. The containment set of a given expression is the set of values
that the expression's interval evaluation must contain. Using cset
theory, closed interval systems can be derived.
|
|
This paper is
designed to be read as a document and also used as a reference to
supplement IBLAS man pages. The goal has been to make the material easy
to find and understand.
|
