DumpSave for W95 Needs Major Reengineering
- To: mathgroup at smc.vnet.net
- Subject: [mg5151] DumpSave for W95 Needs Major Reengineering
- From: Mark Evans <evans at gte.net>
- Date: Wed, 6 Nov 1996 01:33:29 -0500
- Organization: None
- Sender: owner-wri-mathgroup at wolfram.com
I don't want to start what is called a "flame war," but I want to
vent some frustration and solicit advice.
I don't think WRI put enough effort into some fundamental aspects of
technical computation for version 3.0.
The typesetting is all very wonderful, but I can't believe how long
it is taking me to do very simple I/O tasks with binary data files
that are not all that large. The freebie utilities I get off the
Internet don't have any trouble loading, modifying, and saving them,
but Mathematica acts like it is having a baby.
This deficiency is one of long standing. Mathematica has never been
able to read and write binary files in any efficient manner. I
don't understand why WRI did not incorporate the FastBinaryFiles
concept into version 3.0.
The topic of this post is DumpSave[]. I was one who lobbied WRI to
restore this kind of capability in version 3.0. I'm glad to have
it, but very disappointed in my first attempt to use it. Maybe
someone can tell me what I'm doing wrong.
I used Miscellaneous`Audio` to read in a mono, 8-bit, 44.1 kHz,
6-second sound file. That in itself was a 5-minute affair on a P166
machine. I took the precaution of turning these "exact" integers
into machine numbers with N[]. The data file contained about
250,000 numbers; in megabytes, this amount of data is about 2 MB.
I processed the data a little bit before trying to DumpSave[] the
result. The bandpass filter came from MathSource and took 31.5
minutes to process the whole file. By contrast, the Internet
freeware took 1.2 minutes to do the same filtering.
At least I got a result. I can't say the same for DumpSave[],
unfortunately. The command was simply
DumpSave["filename",audiodata];
I figured this command should only take a minute at most. Well, I
waited ten; then twenty; then I went to the grocery store, thinking
that it would surely be done before I came back. Not so! It was
still cranking away forty minutes later. I decided that this amount
of time was ridiculous and force-quit the kernel.
The problem may lie in Windows 95 virtual memory. My virtual memory
is on the same partition as Mathematica. I have only one hard
drive, albeit a large 2 GB model.
I have another theory as well. I get the strong impression that
Mathematica is composed of myriads of neat, well-desigend,
computer-science bullet-proof, politically correct modules, which in
spite of all that behave like molasses.
I've been trying to use Mathematica to help a client process his
data. It's getting harder and harder to convince him (and myself)
that Mathematica is up to the job. What a bummer!
Mark Evans
evans at gte.net