5 ms·
It's from May 2000 (https://shibaura.pure.elsevier.com/en/publications/lisp-can-be-hard-real-time https://shibaura.pure.elsevier.com/en/publications/lisp-can-..
by auggierose 4y ago
It's from May 2000 (https://shibaura.pure.elsevier.com/en/publications/lisp-can-be-hard-real-time https://shibaura.pure.elsevier.com/en/publications/lisp-can-...)
- mhdhn 4y agoThank you. Side note: who thinks it's a great idea not to put any publication date on a paper? And is there a rule for HN to put at least the year in the title, especially if it's not clearly visible at the top of the linked paper. There oughta be.
- capableweb 4y agoYeah, it's stupid not to include it. Sometimes you can guess it (poorly) by looking at the references, if there is more than a few. If the latest one is at 1998, you can guess the paper is from somewhere around 2-5 years after. In this particular case, latest publish reference is at 1998 and one marked as "to be published" at 2000 so that makes it a bit easier to guess pretty exactly :)
- czbond 4y agoThank you. I was bewildered why one would go to the effort in 2022 to LISP hard, or LISP real time.
- dreamcompiler 4y agoI've built hard realtime systems in Lisp and I did it because it was easier than doing the whole thing in e.g. C. The extra effort for me would have been not using Lisp.
- dunefox 4y agoSounds interesting. Can you elaborate?
- dreamcompiler 4y agoBasic idea is to divide the problem into two parts: The Fast and Dumb part and the Slow and Smart part. You can write the F&D part in assembler or C and implement it as a hard real time process. You can write the S&S part as a Lisp process that only has soft real time constraints. The two processes have to communicate somehow, e.g. message passing etc. As long as the F&D process is simple enough and of bounded length, it can continue to operate even if the Lisp process is delayed by GC. Not every real time problem can be subdivided in this way, but for those that can, Lisp can work well.