Re: JIT threshold

[Michael D. Godfrey]

On 07/09/2013 08:50 PM, Rik wrote:
> I liked a suggestion you made about making the threshold for JIT
> acceleration customizable so I added a new function called jit_startcnt to
> Octave
> (http://hg.savannah.gnu.org/hgweb/octave/rev/e6ddaa65a777?revcount=120).
> Eventually if we collect enough parameters we can have a jit_options
> function.  For the time being, it was much easier to just have a single
> variable with a getter/setter routine.  The startcnt is the minimum number
> of loop iterations required before JIT acceleration takes place.  I left
> the default at the current value of 1000.
>
> Cheers,
> Rik
Thanks!  This will be useful for debugging and for performance analysis.
As soon as JIT learns to deal with llvm 3.3 I will use it.  (Max said he 
will work
on this soon.)

Michael