From: James Rogers (jamesr@best.com)
Date: Fri Dec 13 1996 - 13:37:25 MST
At 10:36 AM 12/13/96 +0100, you wrote:
>James Rogers wrote:
>
>> >novel:
>> >http://194.182.38.101/m&k/gallery/maxmcorp/GNov/main.htm
>>
>> I checked out the novel. Pretty righteous! Just out of curiousity, what
>> did you do to create it?
>
>3dstudio V4, Photoshop 2.5, a camera, a scanner some friends and 3
>months.
>btw it was part of a sci-fi project witch consisted of a sci-fi book,
>the graphic novel + a music CD on a mixed mode CD. I wanted to be the
>first in the world to make a major "multimedia project" single-handed. I
>think i succeded. I took me two years in all and was released in '92 as
>far as i remember.
Pretty good work, especially at the time. It had a strong "cyber" flavor;
quite Gibson-esque. Takes me back to the thriving hacker scene of the mid
to late 80's and very early 90's before the flood of lamers kind of killed
that scene.
Sigh...The good old days...
>> I think that soundcards will take over the low end, and may creep into the
>> mid-range, but I doubt they will impact many people in the mid to high-end
>> range.
>
>We will soon only need the soundcard for previewing and analog input.
>Today i burn straight to CD. My DAT is sitting idle. And an eight
>processor 200 Mhz Pentium Pro in a SMP system will be able to do nicely
>as specialised DSP's. Not even more expensive than a good synth(ok a
>pretty good one.). An when you don't make music on it you can make a
>little music video.
>
Yes and no. What operating system are you going to use on the SMP system?
I have a room full of powerful Unix and NT boxes, but for things like MIDI,
I still use Macs and Windows 3.1. One negative feature of pre-emptive
multitasking "advanced" OSs like NT is that the pre-emptive architecture
sucks for any task that requires real-time precision timing.
Running a MIDI sequencer at 960ppqn or even 240ppqn is impossible in this
type of environment because your sequencer may or may not be preempted just
before it is supposed to send out a MIDI signal. The practical MIDI
resolution is very low, and the timing is off at any resolution. In my case
the problem is worsened because I usually use 8-16 MIDI ports in a fairly
intensive manner, increasing the requirements on the system. I used to run
8 ports on a 386dx40 without any problems using Windows 3.1, but running 1
port under NT on an SMP system usually gives poorer performance, especially
at intermediate resolutions.
However, there *is* a positive side to SMP systems such as NT. I used to
write small digital audio processing software under NT. Most DSP algorithms
for digital audio multi-thread really well. This means that audio DSP
performance scales almost linearly across multiple processors. Non-realtime
audio works great under these systems. I do believe that eventually
software on computers will replace all the dedicated hardware, but the synth
software currently available (AFAIK) just doesn't work as well as dedicated
hardware. But I think most of the limitations are in getting real-time
performance out of non-real-time operating systems, rather than the actual
processing capabilities of the software, which is very impressive in some cases.
NEW THOUGHT ON THE FLY: I think the reason digital audio timing for
playback and recording is unaffected by preemptive multi-tasking is that the
audio cards pre-buffer the data. If someone designed a MIDI interface card
with a large buffer and real-time microcontroller to run the buffer, 1) the
timing would be unaffected by preemption, and 2) you could run *a lot* of
MIDI channels through that card. Writing the device driver might be a
little more complicated than usual, but workable. Most MIDI interfaces I've
worked with do not have significant buffering, and definitely nothing like a
pre-transmission queue. Hmmm...
-James Rogers
jamesr@best.com
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:35:53 MST