www.digitalmars.com         C & C++   DMDScript  

D - Retire FORTRAN

reply Mark Evans <Mark_member pathlink.com> writes:
Maybe some relevance to the performance nitpicking anti-semantics around
here...in particular read section 4 "Problems and Solutions." -- Mark

http://www.llnl.gov/tid/lof/documents/pdf/217941.pdf

Abstract:

"In the May 1984 issue of Physics Today, Jim McGraw debated David Kuck and
Michael Wolfe on the question of retiring FORTRAN.  They addressed such
questions as:  Is FORTRAN the best tool for decomposing problems for parallel
execution?  Is FORTRAN the programming language we should carry into the 21st
century?  Are there any alternatives?  While McGraw argued forcefully in favor
of retiring FORTRAN, concerns about performance crippled his position [key to
the D debates! -M].  He could not rebut the claim that only FORTRAN could
provide the performance required for scientific computing.  In this report, we
use the current performance of ... SISAL, a functional language for large-scale
scientific computing, to counter that claim.  If McGraw had had our data in
1984, he could have countered what some say is the only defense [left for]
FORTRAN.  The results show that we can move beyond the constraints of imperative
programming.  We can raise the level of abstraction and *retain performance*."
Nov 08 2002
parent reply "Walter" <walter digitalmars.com> writes:
I think that D has the semantic capability of besting FORTRAN for numerical
apps, though the current implementation is not there.

"Mark Evans" <Mark_member pathlink.com> wrote in message
news:aqhmdo$2ovk$1 digitaldaemon.com...

 Maybe some relevance to the performance nitpicking anti-semantics around
 here...in particular read section 4 "Problems and Solutions." -- Mark

 http://www.llnl.gov/tid/lof/documents/pdf/217941.pdf

 Abstract:

 "In the May 1984 issue of Physics Today, Jim McGraw debated David Kuck and
 Michael Wolfe on the question of retiring FORTRAN.  They addressed such
 questions as:  Is FORTRAN the best tool for decomposing problems for


parallel

 execution?  Is FORTRAN the programming language we should carry into the


21st

 century?  Are there any alternatives?  While McGraw argued forcefully in


favor

 of retiring FORTRAN, concerns about performance crippled his position [key


to

 the D debates! -M].  He could not rebut the claim that only FORTRAN could
 provide the performance required for scientific computing.  In this


report, we

 use the current performance of ... SISAL, a functional language for


large-scale

 scientific computing, to counter that claim.  If McGraw had had our data


in

 1984, he could have countered what some say is the only defense [left for]
 FORTRAN.  The results show that we can move beyond the constraints of


imperative

 programming.  We can raise the level of abstraction and *retain


performance*."





 Nov 10 2002



  parent  "Sean L. Palmer" <seanpalmer directvinternet.com>  writes:


The article brings up good points about the need for a class of functions
that are guaranteed to have no side effects, in order that the compiler
won't have to do global analysis just to find out whether a function call
can be hoisted out of a loop.

Sean

"Walter" <walter digitalmars.com> wrote in message
news:aqmj3g$1pop$1 digitaldaemon.com...

 I think that D has the semantic capability of besting FORTRAN for


numerical

 apps, though the current implementation is not there.

 "Mark Evans" <Mark_member pathlink.com> wrote in message
 news:aqhmdo$2ovk$1 digitaldaemon.com...

 Maybe some relevance to the performance nitpicking anti-semantics around
 here...in particular read section 4 "Problems and Solutions." -- Mark

 http://www.llnl.gov/tid/lof/documents/pdf/217941.pdf







 Nov 10 2002