|Re: failure due to compiler? firstname.lastname@example.org (1996-07-04)|
|failure due to compiler? email@example.com (1996-07-09)|
|Re: failure due to compiler? firstname.lastname@example.org (1996-07-10)|
|Re: failure due to compiler? email@example.com (1996-07-13)|
|Re: failure due to compiler? firstname.lastname@example.org (1996-07-15)|
|Re: Using PL/I efficiently email@example.com (1996-08-09)|
|From:||firstname.lastname@example.org (++ robin)|
|Date:||9 Aug 1996 14:11:58 -0400|
|Organization:||Comp Sci, RMIT, Melbourne, Australia|
|Expires:||1 November 1996 00:00:00 GMT|
|References:||96-07-041 96-07-056 96-07-064 96-07-079 96-07-100 <96-07-123|
>(Declaring a bit string ALIGNED meant operations on it were done inline
>instead of by subroutine call -- about a 15-1 performance improvement as I
>Not particularly interesting, until we started considering that during IBM's
>compiler development -- all of alpha testing -- all of beta testing -- and
>*a year of release* -- no one had considered it worth while to declare any
>bit strings ALIGNED.
---Not entirely surprising. Another common one is to
specify full optimization OPTIMIZE(TIME), but to omit
to specify REORDER for the program.
---Without REORDER specified, the compiler is not usually able
to perform best optimization, typically of the type where
code can be shifted outside loops!
Return to the
Search the comp.compilers archives again.