[prev in list] [next in list] [prev in thread] [next in thread] 

List:       haskell
Subject:    Re: Benchmarks
From:       Stephen Bailey <steph%candide.uchicago.edu () yalevm ! ycc ! yale ! edu>
Date:       1991-03-20 9:06:09
Message-ID: 9103181730.AA27259 () candide
[Download RAW message or body]

Original-Via: uk.ac.nsf; Mon, 18 Mar 91 21:42:39 GMT
To: haskell@edu.yale.cs
Subject: Re: Benchmarks
Date: Mon, 18 Mar 91 11:30:15 CST
From: Stephen Bailey <steph%candide.uchicago.edu@yalevm.ycc.yale.edu>
Sender: haskell-request@cs.glasgow.ac.uk

>> The figures I am interested in should look something like:
>>
>>    Test results for evaluation of fib (26) = 392835 on Sun 3/50
>>    (Average unix user "time" in seconds)
>>              Average   nfib/sec
>>    Lazy FAM  15.0      26189
>>    C         2.9       135460   (FAM/C ratio 0.19)
>>
>>    Test results for evaluation of fib (23) = 92735 on Vax 750
>>    (Unix user "time" in seconds)

Note that user time is often not a good benchmark figure.

In particular, languages with very dynamic storage behavior and poor
locality of reference may spend a great deal of time paging.  Some of
this overhead appears as system time charged to the process, but most
of it isn't charged to any process at all!

Therefore, still use my trusty stop watch for benchmarks.

I've seen systems which are much slower than user time would indicate.

Steph







[prev in list] [next in list] [prev in thread] [next in thread]