Re^2: Newbie question

I would strongly favor using the abs() function. One of the issues here is that if you have a statement like no warnings 'numeric';, then there should be some comments explaining why that is. abs() requires no explanation.

As far as speed with Perl, I am betting that abs() is really fast - this is not a complex operation for a 2's complement number. Don't know that Perl would be "smart" enough to do this asm code, but this is very fast because no branches to stall the instruction pipe.

; abs(eax), with no branches.
; intel syntax (dest, src)

mov ebx, eax   ; save copy of eax
neg eax
cmovl eax, ebx ; if eax is now negative, restore its saved value
[download]

Comment on Re^2: Newbie question Select or Download Code

Replies are listed 'Best First'.
Re^3: Newbie question by NERDVANA (Priest) on Aug 23, 2022 at 19:57 UTC
Um, I hate to break it to you, but perl is fully interpreted with no just-in-time compilation. Every operation and every argument passed to it in a script is dozens of branches, no matter how innocent. The way to optimize a perl program is to code it with the fewest possible operations, fewest function calls, fewest assignments to a temporary variable, and even fewest curly-brace scopes. Optimizing a perl program is generally the opposite of making it more readable. (but there is a time and a place, etc) For a little mind-bender, try this: `perl -MBenchmark -E ' my $x= 1; timethese(50000000, { mul => q{$x*$x}, sqrt => q{sqrt($x)} })'` [download]	[reply] [d/l]
Re^3: Newbie question by oldB51 (Sexton) on Aug 23, 2022 at 12:58 UTC
Thanks for your help.	[reply]
Re^3: Newbie question by Anonymous Monk on Aug 23, 2022 at 12:46 UTC
Talk of optimization is premature?	[reply]
Re^4: Newbie question by Marshall (Canon) on Aug 23, 2022 at 15:48 UTC
What I was trying to say is that the easiest to "code and understand" way is also likely to be the fastest way. And I was encouraging that. It really doesn't matter how good/bad the detailed implementation of abs() is, the time to execute the worst imaginable implementation is likely faster than making decisions about default behavior in light of an exception (the '-') and also suppressing normal exception behavior. I also suspect that either choice is completely dwarfed by the by the split operation and creation of the list of individual numbers. I don't think any benchmarking or further optimization is warranted here. I vote for the easy to understand way.	[reply]