Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!cbosgd!ihnp4!alberta!simon
From: simon@alberta.UUCP
Newsgroups: comp.sys.mac
Subject: Re: LSC Programming Tools: and improved arithmetic.
Message-ID: <226@pembina.UUCP>
Date: Fri, 30-Oct-87 03:49:25 EST
Article-I.D.: pembina.226
Posted: Fri Oct 30 03:49:25 1987
Date-Received: Sat, 31-Oct-87 20:29:03 EST
Organization: U. of Alberta, Edmonton, AB
Lines: 20
Summary: number crunching in LSC? does it have to take so long?


I think this may have been discussed before, but we have a problem now.
Running a program that is primarily arithmetic operations, written in C
and compiled in LSC with floats, is much slower than the equivalent
written in fortran.  This all on a Mac plus, and no data structure
fancier than a 2-d array.  I realise that most, if not all, of the
difference arises from the conversion to double before the arithmetic operation
in SANE, then the conversion back to float.  Memory restrictions prevent us
using double all the way.  (Correct me if any assumptions so far are 
inaccurate.)
So, knowing that the two may be mutually exclusive, what is
  a) the easiest, and b) the best way to regain the competitive edge against
the fortran code runtimes?  Thanks in advance.
  -------------------  
W. Simon Tortike
Dept Min, Met &    UUCP:   {ubc-vision,ihnp4,mnetor}!alberta!simon
  Petroleum Engg   BITNET: stortike@ualtavm
Univ. of Alberta   AGT:    (403) 432-3338
Edmonton, AB
Canada T6G 2G6