On Saturday, March 12, 2016 at 7:50:43 AM UTC+5:30, Chris Angelico wrote: > On Sat, Mar 12, 2016 at 12:16 PM, BartC wrote: > > You'd be surprised how much any kind of program relies on adhoc integer > > operations. It doesn't need to work with large int arrays for them to be > > important. (Look at the benchmark below: the inner loop is nearly all > > integer ops.) > > Sure, but in real-world code, there are other considerations than just > integer performance. If someone waves a magic wand and improves > machine-word integer performance, great, but there are other calls on > core dev time.
I guess that BartC (or is it Bartc?) is describing something that is a commonplace in compiler world but not so well known elsewhere, viz. a simple operation like an array/attribute-access etc, which from a HLL pov may have no 'operations' at all may emit a slew of integer operations when compiled. Which is not so surprising if you consider that apart from control-flow there is nothing going on in a CPU beside arithmetic; there is no datatype beside integers of various sizes and (un)signed combos. -- https://mail.python.org/mailman/listinfo/python-list
