[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
LispMachine File Reading Time.
Did you use :ELEMENT-TYPE 'STRING-CHAR streams? If not,
you were in some sense comparing apples and oranges, and the
real Symbolics times will be noticably faster.
It is very important when reporting benchmarks to report
all relevant issues like this.
Date: Thu, 18 Jan 1990 8:37:02 PST
From: Keith Price <email@example.com>
With all the talk of file reading times, here are the results of an internal
benchmark for Lisp reading character data and storing in a structure.
Further testing of the Symbolics version reveals that 90%+ of the time goes
to the (read).
Meaning? I would guess that by this you mean that 90% of the time was spent
inside READ exclusive of time spent inside READ-CHAR or :TYI. Is this what
The comparison is with programs written in Lisp (i.e. the
same program, no tuning for the system, no changes for each system). The
comparison is various versions of Suns (3s and 4s) in different
configurations, Symbolics 3645 and XL400, TI Explorer II, and Lisp on a VAX
After several conversations with Symbolics people both here in LA and
in MA, I am expecting great improvements in this program with Genera 8.0 (I
was "promised" such operations were addressed), so these times reflect old
data. All the times are for local file accesses.
I assume local means local LMFS for the LISPM case?
I should also state that
the general benchmark results were strongly in favor of the Symbolics, even
compared to the Sun 4.
Input function runtimes. Primarily Read the file, and put into the data
structure. The times depend on the disk, GC, and conversion times.
Sun3/160 Sun3/260 Sun IV(S) SUN IV(U) Symbolics TI-II 8600
Real/Run Real/Run Real/Run Real/Run Real(Paging)
Small 20.8/20.4 17.0 /16.1 9.58/9.10 8.2/8.0 17.1(1.1) 6.85 3.78
Small 31.7/29.0 14.1 /13.4 10.01/9.61 12.0/11.9 16.4(0.3) 6.93 3.74
Large 539.7/179.7 134.6/131.8 78.2/75.0 47.8/47.4 155.8(20.7) 57.89
Large 350.6/117.5 85.6/ 77.8 39.1/37.6 35.9/35.6 75.7(10.5) 26.48
(Times are in seconds, XL400 didnot fit on the list, but was not that much
different from 3645 for this case. All system overhead is included on
Symbolics and not included for the Unix boxes. You want real speed get a VAX
8600, not a Sun4.)
If you're going to report times including Paging time on a Symbolics machine,
you had better compare memory sizes and disk organization. Is the Symbolics
machine doing paging and file I/O from the same disk drive? Is the LMFS
scattered around in several FEP files on one drive? Is it fragmented? What
kind of disk(s)?
Similar questions apply to the other configurations being tested.
Why no 8600 times for the large case? What configuration? What kind of disk(s)?
Did you include file opening & closing time, or just file processing time?
I don't mean to imply that your benchmark isn't useful, but it would be a lot
more useful if you were more careful about reporting your test conditions.
Too often, people do good benchmarking and then spoil it by reporting piecemeal