gcl-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gcl-devel] advice


From: Camm Maguire
Subject: [Gcl-devel] advice
Date: 05 Sep 2008 13:30:53 -0400
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2

Greetings!  When tuning lisp performance, 

1) What is considered a good balance between gc time and non-gc time?
   I.e. when does one stop worrying about generating garbage?

2) The rate limiting step in the compiler now (which is getting rather
   peppy these days), is type-and.  GCL computes unique types in
   standard lisp (i.e. list) form, and eq hashes the normalized
   result.  type-and is guaranteed to receive two such normalized types.
   Right now, we check each in an eq hash-table which returns a
   hash-table, in which we check the other for a possibly precomputed
   result.  We only cache a hardcoded list of 'useful' types for this
   purpose.  Failing that, we cons on the stack `(and ,t1 ,t2),
   uniqueify this, which requires a slow 'equal hash, normalize and
   simplify and return the result.  

   This seems a lot like your hons mechanism, i.e. quickly returning a
   unique object as a function of two other unique objects.  What is
   the best cache-management/hashing algorithm here?

Take care,
-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah




reply via email to

[Prev in Thread] Current Thread [Next in Thread]