Front page | perl.perl5.porters |
Postings from July 2010
From: Jim Cromie <jim.cromie@gmail.com> Date: 31.07.2010 06:37 > On Fri, Jul 30, 2010 at 3:26 PM, <w.c.humann@arcor.de> wrote: > >> From: Marvin Humphrey <marvin@rectangular.com> >> Date: 30.07.2010 17:34 >> >>>> * Why the "+10"? Why have a extra minimum at all? Why "10" instead >>>> of another number? Given your examples, for a large string the 10 is >>>> a trivial fraction of the total and for a small string, the efficiency >>>> either doesn't matter or the string quickly grows into a large string >>>> anyway. >>>> >>>> >>> That's similar to my analysis. I think it's more useful to round up to a >>> multiple of the pointer size for small values. >>> >>> >> If I understand correctly (which I'm not certain about), this is done anyway >> right afterwards in >> newlen = PERL_STRLEN_ROUNDUP(newlen); >> >> Wolfram >> >> >> > > what happens if you presize the variable 1st ? > then assign an empty string, > then start appending the additions. > > this will avoid all the sv_grow calls, > all the reallocs, and just use the one-time preallocated memory. > > this avoids all the mods to the heuristics youre proposing > (modest tho they might be), and has no unintended effects > on other benchmarkable use cases. > There are many cases when you don't know the the final size of your string. In that case you would need to invent a strategy similar to what I proposed. Take my original script: $ps_text = qx( pdftops some_pdf_file.pdf - ); The PDF file is about 2MB. I know it will grow during PDF -> Postscript conversion. But it's hard to estimate how much it will grow. If I play safe, I will likely waste much more memory than some well designed perl-internal strategy. Also, any perl compiled to use the internal malloc/realloc will "round up to a certain geometric progression anyway" (Ilya Zakharevich). If you look at Marvin Humphrey's post, you'll see that such a geometric growth strategy (although with different details concerning acceptable memory overhead etc.) is very common. WolframThread Previous | Thread Next