Re: Elisp native profiler

[Stefan Monnier]

> The first lines in what I got (still only the memory profiler, the
> itimer code is in the works and not ready yet) look suspicious to me
> in the last column:

Yes, the profiler.el code that handles the percents is victim of
overflows, especially with the memory profiler where the numbers are
typically much higher.  There's a FIXME for it in profiler.el.

> Here's the same with the + signs expanded:

>   - call-interactively                                       46,352,691   2%
>     - scroll-up-command                                      43,373,729   0%
>       - scroll-up                                            43,373,729   0%
>       - jit-lock-function                                  41,648,995   0%
>         - jit-lock-fontify-now                             41,648,995   0%
>           - byte-code                                      41,619,059   0%
>             - run-hook-with-args                           41,619,059   0%
>               - font-lock-fontify-region                   41,619,059   0%
>                 - c-font-lock-fontify-region               41,619,059   0%
>                   - font-lock-default-fontify-region       40,842,125  -1%
>                     - font-lock-fontify-keywords-region    39,963,822  -2%
>                       - c-font-lock-declarations           34,602,780   1%
>                         - c-find-decl-spots                34,602,780   1%
>                           + #<compiled 0x40567bf>          28,867,024  -2%
>                             looking-at                      2,817,588   1%
>                             re-search-forward               1,791,738   1%
>                           - c-forward-sws                     413,348   0%
>                               looking-at                      355,304   0%
>                               put-text-property                26,844   0%
>                           - c-beginning-of-current-toke       150,676   0%
>                               looking-at                      150,676   0%
>                           - c-beginning-of-macro               66,408   0%
>                               looking-at                       66,408   0%
>                           - c-backward-sws                     39,468   0%
>                               c-beginning-of-macro             12,480   0%

That looks fine (modulo the known bug).


        Stefan