Profiling

Donald Knuth wrote:

"We should forget about small efficiencies, say about 97% of the time: premature optimization is the root of all evil"

A little example:

The old IBM OS PL/I compiler on z/OS has an option to count how many times each statement is executed. I still have a version of one of my programs, and the total statement count comes to 1,257,761,317, with one single procedure accounting for 1,238,974,804 of those executed statements (98.50%)

The current version of the program, which has rather a lot more functionality, has a rather nicer executed number of statements count of 14,778,811 (a reduction of 98.82%), and the nasty procedure, using a completely different algorithm executes just 110,204 statements (a reduction of more than 99.99%)

Obviously, this being a compiler and not an assembler, some statements result in calling library routines, but it makes patently clear that selecting the right algorithm has far more influence than anything else!

As for the

Code:

        cmp     [wmsg],WM_CREATE
        je      wmcreate
        cmp     [wmsg],WM_COPYDATA
        je      wmcopydata
        cmp     [wmsg],WM_GETMINMAXINFO
        je      wmgetminmaxinfo
        cmp     [wmsg],WM_SIZE
        je      wmsize
        cmp     [wmsg],WM_SETFOCUS
        je      wmsetfocus
        cmp     [wmsg],FM_NEW
        je      fmnew
        cmp     [wmsg],FM_OPEN
        je      fmopen
        cmp     [wmsg],FM_SAVE
        je      fmsave
        cmp     [wmsg],FM_COMPILE
        je      fmcompile
        cmp     [wmsg],FM_SELECT
        je      fmselect
        cmp     [wmsg],FM_ASSIGN
        je      fmassign
        cmp     [wmsg],FM_GETSELECTED
        je      fmgetselected
        cmp     [wmsg],FM_GETASSIGNED
        je      fmgetassigned
        cmp     [wmsg],FM_GETHANDLE
        je      fmgethandle
        cmp     [wmsg],WM_INITMENU
        je      wminitmenu
        cmp     [wmsg],WM_COMMAND
        je      wmcommand
        cmp     [wmsg],WM_NOTIFY
        je      wmnotify
        cmp     [wmsg],WM_DROPFILES
        je      wmdropfiles
        cmp     [wmsg],WM_CLOSE
        je      wmclose
        cmp     [wmsg],WM_DESTROY
        je      wmdestroy

earlier in this thread, WM_CLOSE and WM_DESTROY were apparently put at the bottom, but did anyone actually bother to temporarily add counts to all other compares and order them in an descending order? For example, I would expect that an IDE would use FM_COMPILE rather a lot more than FM_NEW...