Path: utzoo!utgpu!water!watmath!clyde!cbosgd!ihnp4!alberta!calgary!radford
From: radford@calgary.UUCP (Radford Neal)
Newsgroups: comp.arch
Subject: Re: taken -vs- untaken branches
Summary: FORTRAN Frequency statement vanishment
Message-ID: <1275@vaxb.calgary.UUCP>
Date: 8 Jan 88 21:18:41 GMT
References: <496@cresswell.quintus.UUCP> <638@l.cc.purdue.edu> <18251@bu-cs.BU.EDU>
Organization: U. of Calgary, Calgary, Ab.
Lines: 19

In article <18251@bu-cs.BU.EDU>, hen@bu-cs.BU.EDU (Bill Henneman) writes:
> As I recall, the original FORTRAN (for the IBM 704) had a FREQUENCY
> statement which allowed the programmer to give hints about the
> probability of branches being taken ... I never could understood why that
> statement vanished in later versions...

I read somewhere, once, that the statement was removed after they 
discovered that it was implemented *backwards*, and that no users had
noticed. I haven't been able to confirm this. It's an interesting
comment on optimization and/or marketing policy if true.

I think adding hints such as this is a waste of the programmer's time. The
improvements are likely to be small. Anyway, if you're that concerned about
optimization, you'll get far more and better data by profiling an actual
run of the program and feeding that data into the compiler. I've heard
of someone doing this...

   Radford Neal
   The University of Calgary