On Tue, Jan 06, 2004 at 06:40:06PM +0100, Lars Gullik Bjønnes wrote: > Angus Leeming <[EMAIL PROTECTED]> writes: > > | $ size trial_case? > | text data bss dec hex filename > | 7850 816 8 8674 21e2 trial_case1 > | 9510 840 16 10366 287e trial_case2 > | 9530 872 8 10410 28aa trial_case3 > > With gcc 3.4: > > size trial_case? trial_case?.o > text data bss dec hex filename > 5435 356 164 5955 1743 trial_case1 > 6907 356 172 7435 1d0b trial_case2 > 7120 356 168 7644 1ddc trial_case3 > 2926 8 1 2935 b77 trial_case1.o > 4367 8 10 4385 1121 trial_case2.o > 4583 8 6 4597 11f5 trial_case3.o > > I did some performance test as well, running the find_if 10000000 > times for each of the cases. > > runtime ratio loops/sec > case 1: 5.070s 1.0 1972386 > case 2: 7.965s 1.57 1255492 > case 3: 11.341s 2.236 881756 > > I am not afraid of using any of these techniques in real code.
case 1 seems to be uniformly better. So why not use it? Andre'
