Wow, I had a look at his slides, and although the rest of the course seems pretty good he clearly misses the boat with Perl. He calls it out specifically as being difficult to work with for codes longer than about 100 lines, which is manifestly not the case. Worse, as you (Demian) pointed out, he underestimates the speed of Perl by a lot. By a factor of about 3,000, in fact -- I benchmarked the codes in his slides, and the PDLPP snippet performs only 1.8x slower than C for his example, compared to Python at about 75x slower than C and his pure-Perl "solution" at about 5,000x slower.
On Nov 17, 2013, at 5:19 PM, Craig DeForest <[email protected]> wrote: > Here's a mandel.pdl suitable for autoloading. Try it like this (or > substitute your favorite plotting commands): > > $locs = (ndcoords(1000,1000)/500 - 1)/3000 + pdl(0.748973,0.0570852); > $foo = mandel($locs,2000); > $w=gpwin(x11,size=>[9,9]); > $w->image($locs->using(0,1),$foo->sqrt,{j=>1,title=>"Cool!"); > > <mandel.pdl> > > On Nov 17, 2013, at 4:33 PM, Derek Lamb <[email protected]> wrote: > >> Xavier Calbet did this about 6 years ago (but did not use a Python >> comparison) >> >> http://www.freesoftwaremagazine.com/articles/cool_fractals_with_perl_pdl_a_benchmark >> >> Yes, if that python example uses scipy, then by not using Perl's equivalent >> (PDL), there should be no surprise that Perl is 70x slower! >> >> cheers, >> Derek >> >> On Nov 17, 2013, at 1:10 PM, Demian Riccardi wrote: >> >>> Hello everyone! >>> >>> While cruising the web for who knows what, I found this: >>> >>> http://www.physics.rutgers.edu/grad/509/Intro0.pdf >>> >>> which are slides from the first class of a computational physics course at >>> Rutgers. There's a little language comparison and a calculation of the >>> Mandelbrot Set using Fortran, C++, Perl (with Math::Complex), and Python >>> (with numpy/scipy, which is the main language for the course it seems). >>> Perhaps it would be nice if someone sent some awesome PDL code for the Prof >>> to replace in the introduction so at least the comparison is fair (slide >>> 35)? I'd do it, but I'm swamped, and still not using PDL enough for it to >>> be the most beautiful ever (which it should be for this). >>> >>> Demian >>> >>> >>> _______________________________________________ >>> Perldl mailing list >>> [email protected] >>> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl >> >> _______________________________________________ >> Perldl mailing list >> [email protected] >> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl >
_______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
