Re: An idea, now that we have dynamic loading

[Tom Tromey <tom@tromey.com>]

>>>>> "John" == John Wiegley <johnw@gnu.org> writes:

>>>>> "TT" == Tom Tromey <tom@tromey.com> writes:
TT> Yes, you can see a random selection of functions converted this way here:
TT> https://github.com/tromey/emacs/commits/compiler-rewrites-2

John> Neat; any benchmark numbers for us?

Only what I wrote on github; I haven't done any real benchmarking, and I
haven't done any recently -- nor have I compared it against the JIT.

What I wrote there is:

    Using the bubble sort benchmark from
    http://www.emacswiki.org/emacs/EmacsLispBenchmark (with the list bumped
    to 1000 elements), with 100 runs, I got some timings:

    Approach 	Seconds
    interpreted 	54.874574673000005
    byte-compiled 	13.390510359999999
    el-compilador 	4.312016277000001

What I did for this test was compile the bubble sort code from Lisp to
C, then build Emacs with it linked in.

Tom