Re: Lisp reader and #1=... abbreviations

Hi

I think I remember some tests for CMUCL from ancient times, where the  
break-even between a-list and hash-table was around 50.
I don't have a CMUCL handy right now, so YMMV.

Cheers

Marco



On Jun 18, 2007, at 6:00 PM, Raymond Toy wrote:

>>>>>> "Jared" == Jared C Davis <[email protected]> writes:
>
>     Jared> Hi,
>
>     Jared> I'm having trouble reading in files which use a lot of
>     Jared> structure-sharing abbreviations.  For example, here is a  
> typical file
>     Jared> (490 KB) which has 18,000+ abbreviations.   
> Unfortunately, reading this
>     Jared> (with "read") takes about 13 minutes in CMUCL 19d on my  
> test machine.
>
> 18000 abbreviations seems like an uncommon situation.
>
>     Jared> We had similar trouble with OpenMCL, and discovered that  
> the reader
>     Jared> was using an association list to manage the  
> abbreviations.  By
>     Jared> replacing this with a hash table, Gary Byers reports  
> speeding up the
>     Jared> process from about 13 minutes to 1.2 seconds.
>
>     Jared> I wonder if it would be possible to speed this up on  
> CMUCL also?
>
> CMUCL uses an alist so it seems possible to replace this with a hash
> table.  I assume the hash table would slow down somewhat the typical
> case, though.  But I haven't tested this.
>
> I'll look into this a bit and see what can be done.
>
> Ray
>

--
Marco Antoniotti