OpenEuphoria

Re: Referencing Nested Sequences

• Posted by kbochert at copper.net Jun 08, 2003
• 518 views

On 7 Jun 2003 at 23:41, Robert Craig wrote:

> 
> 
> kbochert at copper.net wrote:
> > As an example, Bach has a built-in sort.
> > For a sequence of 6000 chars, Eu managed 27 / sec.
> > while Bach was 125/sec.
> 
> If someone wanted to sort 6000 chars (0-255),
> they could use the bucket sort in euphoria\demo\allsorts.ex,
> and it would "blow away" your sort.  

True enough. On my system the integer sort (allsorts.e) of
5120  elements gives:
EU :    1.25 ms.
Bach:   0.32 ms.
Bucket: 0.12 ms

The figures I quoted are for a sequence of 6000 2-character
sequences, {"a3", "dr", ....} which accounts for the 8ms. (125/sec) 
quoted vs. the .32 ms above (and disqualifies the bucket)

I should also point out that the qsort is slowed a little by a capable
compare routine. The sort may be case insensitive or numerical
(2 before 12) and may start beyond the first element of a
sequence (for 'sorting on fields').

I think these features are very desireable and I would suggest 
adding them to the Eu sort -- except that no Euphorian has ever 
asked for them! 


> When you say "Eu", I assume you mean the Euphoria
> interpreter. Don't forget that the Translator can
> give a very significant boost to the speed of Euphoria code.
> 

I certainly do mean the interpreter 
Bach cannot and never will compete with the translator
for overall speed!  In theory it should match it on a  qsort
benchmark -- after all, a built in qsort is written in C.

Given that Bach cannot translate, bind, trace or profile, it does what 
it can in other areas.

Yours
Karl Bochert