|Languages that give "hints" to the compiler jhummel@cy4.ICS.UCI.EDU (Joe Hummel) (1994-07-18)|
|Re: Languages that give "hints" to the compiler firstname.lastname@example.org (1994-07-29)|
|Re: Languages that give "hints" to the compiler email@example.com (1994-07-19)|
|Re: Languages that give "hints" to the compiler firstname.lastname@example.org (1994-07-21)|
|Re: Languages that give "hints" to the compiler email@example.com (1994-07-31)|
|Languages that give "hints" to the compiler firstname.lastname@example.org (1994-08-01)|
|Re: Languages that give "hints" to the compiler email@example.com (1994-08-02)|
|Re: Languages that give "hints" to the compiler firstname.lastname@example.org (1994-08-03)|
|From:||email@example.com (Josh Fisher)|
|Keywords:||Fortran, history, optimize|
|Organization:||Hewlett-Packard Laboratories, Palo Alto, CA|
|Date:||Thu, 21 Jul 1994 15:26:35 GMT|
Our esteemed moderator wrote:
>[The original FORTRAN compiler had a FREQUENCY statement that let you give
>a hint to the compiler how many times a DO loop would run or which was the
>most likely branch of an IF. It disappeared in Fortran II. Legend says
>that the final nail in its coffin was when someone found that it was
>implemented backwards and nobody had noticed. -John]
I believe it was Lois Haibt at Yorktown who 10 years ago told me the story
in response to my mentioning the branch frequency directive in the Yale
Bulldog Compiler. As I recall, she said she'd released the 7090? code
generator--it used the FREQ statement for some different treatment of the
index register--and then went on vacation. While she was out, someone
upstream from her code reversed the sense. She found it herself a year
and a half later.
This was in the days of suspicion of compilers and hacking of object code.
No one ever caught it; she never got a bug report.
-- Josh Fisher // HP Labs Cambridge Research
Return to the
Search the comp.compilers archives again.