The following header lines retained to effect attribution:
>Date: Wed, 08 Oct 2003 15:37:05 +0000 (GMT)
>From: Walter Roberson
>Subject: Re: assembly programming on mips (doc pointers req)
>To: info-iris-misc@ARL.ARMY.MIL


>In article ,
>Randolph J. Herber wrote:
>:Since about the 1970's, compilers operating at extreme optimization
>:almost always generate better code than almost all programmers can
>:write because computers can track more details more accurately than
>:humans. IBM did extensive studies on the topic.


>Fortran 77 compilers, sure, but C compilers can still blow optimization
>badly because they have to assume that any two external pointers might
>be aliased.


1) Fortran has the same problems:

REAL A
CALL SUB(A,A)
WRITE(*,*) A
CALL EXIT
SUBROUTINE SUB(X,Y)
REAL X,Y
X=0.0
Y=1.0
RETURN
END

2) Read man ipa

alias[=setting]
Enables or disables alias/mod/ref analysis. setting can be
ON or OFF. The default is OFF.

The MipsPRO compilers are quite capable.

> I don't know if there's destiny,
> but there's a decision! -- Wim Wenders (WoD)


Randolph J. Herber, herber@fnal.gov, +1 630 840 2966, CD/CDFTF PK-149F,
Mail Stop 318, Fermilab, Kirk & Pine Rds., PO Box 500, Batavia, IL 60510-0500,
USA. (Speaking for myself and not for US, US DOE, FNAL nor URA.) (Product,
trade, or service marks herein belong to their respective owners.)