Re: Specifying time limits in source code ?

hat@se-46.wpa.wtb.tue.nl wrote:
> Therefore, it would be better if a hard guarantee could be given that
> the processor will never take more than x (micro) seconds for a
> certain piece of code. A compiler can help here.


If you know how long each instruction takes (and it sounds like you
are running on a more restricted class of machines where the problem
fits in memory and stuff so issues of "where is the data coming from
aren't problematic") can you just generate the assembly and then
simply annotate this with the amount of time each instruction takes
(you stated you had this information as well). IBMs xlc/xlf compilers
currently have this capability. From this you can generate a whole
slew of statistics including runtime of any given fragment given that
you know things like the number of iterations a loop will execute
(which you seemed willing to provide).


If you want to be conservative, as it sounds you do, you can ignore
issues of pipelining and out-of-order issue to get a conservative
upper bound.
--
Peace,


KSG
Droppin' Science in REAL AUDIO on KSDT Sundays 8pm-10pm (PST)
ARCHIVED SHOWS: http://scw.ucsd.edu/droppinscience
Personal: http://www.cs.ucsd.edu/~kgatlin

Return to the comp.compilers page.
Search the comp.compilers archives again.