3 ms·
they did work ok in an unix environment. they had ethernets, you could run them with NFS, you could telnet into them sorta. but they were expensive. like reall
by convolvatron 8y ago
they did work ok in an unix environment. they had ethernets, you could run them with NFS, you could telnet into them sorta.
but they were expensive. like really expensive. not just a few 10s of k to buy, but you basically had to have a really fat support contract to keep them running. and they were really single user, so to have a lab supporting 8 users was a well over a million dollar investment. a deskside sun with terminals or pizza box sun 3s as stations was alot cheaper. and they were fragile. i'm pretty sure one of the ones we had was wire wrapped.
and they were slow. I may be misremembering, but the flashing status line at the bottom of the monitor would sometimes say the equivalent of 'i am performing a floating point operation now'. we would often reboot them overnight (it took a couple hours) so we could run for a while in the morning before the GC started bogging everything down. the little one they came out with near the end, I think it was the 3620, was largely unusable.
the difference was so marked in the end that if you wanted a lisp and didn't want the ferrari model, you could get a mac or a sun and run lucid or allegro and such for basically 1/10th the cost. and it would be faster.
just an additional historical note, the other vertical they had besides DoD AI was graphics. for an additional 20k (I think) you could get a 24 bit frame buffer with a nice trinitron monitor and they had some really pleasant software to work with.
- gnulinux 8y ago> they were really single user How is this a limitation in the hardware? Operating system could switch between tasks using context switching right? Or are you suggesting these machines didn't have any mechanism to make sense of time/ticks or didn't have interrupts?
- rjsw 8y agoThey had a single address space, tasks were more like threads in a modern system. The MIT/LMI/TI ones didn't take interrupts, the microcode needed to poll the interrupt register at suitable points.
- mrguyorama 8y agoIt could have been due to their lack of speed and performance. Can't run multiple users' software if it can barely run one user's software
- convolvatron 8y agomore importantly the whole point of the symbolics was the environment. one hi definition mono display and one space keyboard meant one 3600
- sedachv 8y agoThere was no concept of name/address space or device virtualization in the Lisp Machine operating systems. I don't know the instruction set, but I have never seen references to privileged instructions or protection rings. Not that that would be an impediment - you would really want to do virtualization at the software level in a Lisp OS anyway.
- lispm 8y ago> we would often reboot them overnight (it took a couple hours) A mid 80s Symbolics Lisp Machine boots in three minutes into a customized Lisp world. > the little one they came out with near the end, I think it was the 3620 The 3620 was hardly near the end in 1986. In 1988 a completely new generation with a in-house designed microprocessor was released and the last models of those were released in 1992. Some of the rest is also not really what you say.