OpenEuphoria

Re: Is this a joke?

• Posted by Matthew Lewis <MatthewL at KAPCOUSA.COM> Oct 05, 2000
• 718 views

Robert Craig wrote:

> Some of those sequence operations make intensive use
> of run-time routines, e.g. find(), which aren't made any
> faster by compiling. Also, I get the feeling that WATCOM C
> generates slightly faster code than DJGPP. That would make
> the run-time routines in ec.a a bit slower than those in ec.lib or
> the interpreter. I find that sieve.ex and shell.ex are both
> a bit faster with WATCOM.
>

I guess that makes some sense.  The sequence routines are the real dogs of
the benchmark comparison.  Here is what I've found (for some surrealism, I
ran the benchmarks with David Cuny's eu emulator--eu.ex--both interpreted
and compiled):

(you'll want a fixed width font to view this)
                        .ex             .exe            eu.ex
eu.exe
database                346800  498800  2932            10567
shell                   8099.7  34513.3 16.5            60.3
sequence init   504133  516400  24983           98666
sequence add    163654  124266  25315           65600
sequence append 6327242 4598333 8245            11294
sequence slice  1885382 3669060 5258            7844
sequence look-up        283588  286400  20199           36266
sieve                   5160            14543           11.3            36.3

average performance ratios (left:top)
                .ex     .exe    eu.ex           eu.exe
.ex                     0.81    279.06  140.66
.exe            1.75            605.51  238.77
eu.ex           0.04    0.04                    0.43
eu.exe  0.10    0.11    2.71


Matt Lewis