| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8
Comments:
<0> Are there any other lisp OSes/machines that can be run in bochs or qemu? And nyef, do you intend to publish instructions on the net once you do it successfully?
<1> Hopefully soon I'll be running !cold-init.
<1> Mitja: For bochs/qemu, there's Movitz. And there are emulators for the CADR and Explorer I LispMs. And possibly the interlisp D machines.
<2> nyef: do you know why you get exception 28 there_
<2> ?
<1> frodef: Yes. It's IRQ8, the RTC.
<1> It's spectacularly boring when I handle it correctly, you know?
<0> nyef, thanks
<2> so how do you plan to deal with interrupts_
<2> ?
<1> Any interrupt or exception causes a priv transition to ring 0. The ring 0 code does a context-switch to the manager thread, p***ing the faulting thread in EDX.
<3> The CADR and Explorer emulators and Medley (D-machine) run under an OS, though, don't they all?
<2> so.. the ring 0 code is some sort of micro-kernel? in sbcl?
<1> Yeah...
<1> Sortof.
<1> Look at the patch at http://www.lisphacker.com/temp/lispos/ for src/os/temp.lisp and src/os/interrupt-stubs.lisp
<1> rydis: Yeah, they do.
<1> s/temp.lisp/test.lisp/
<1> The ring0 lisp code for the interrupt path is common-interrupt-handler.
<1> And boot-kernel is also ring0 code. Anything above boot-kernel is expected to run in both ring 0 and ring 3.
<2> sounds good
<1> I also have an ATA hard disk driver that is known to be good for reads, so I'll be moving the cold-core to an HD image instead of a floppy image and demand-paging dynamic space.
<4> ...no writes?
<5> so, what do folks use when they need a poor man's cluster to run CPU-intensive SBCL jobs on many CPUs?
<6> my multi-CPU pcs :)
<7> slyrus: Kyoto CL / GCL has an MPI implementation. Maybe it's portable. (Har har.)
<8> nfroyd: Say, do you have a recent glibc on an x86 machine? Mine is quite old.
<7> slyrus: More realistically speaking, spend the two or four thousand dollars on some Solaris hardware that's bigger than a PC :P
<4> suns don't scale very big
<4> (for the moment)
<5> I'm not looking for a single box that scales big, but rather something for a farm of many CPUs
<4> slyrus: perhaps mpi through cffi, if mpi really solves your needs... mpi really doesn't DO all that much though and would be fairly trivial to re-implement
<4> cmucl had some sort of 'wire' thing
<6> rtoy_: 2.3.6 good enough?
<5> I can always make an executable core and ship that out to the nodes
<5> I'm not so worried about the interprocess communication; this stuff is pretty obviously parallelizable. I just need to run things on a bunch of CPUs. well, want, really.
<3> slyrus: I /think/ there are some MPI/PVM thingies on CLiki. ParGCL is supposedly quite good.
<6> rtoy_: still uses fsin for sin, if that's what you want to know. there's some kind of cleanup loop after the fsin, but I don't know 387 asm
<5> ok, thanks rydis
<6> doesn't call anything to do arg reduction etc.
<8> nfroyd: I don't know. Is that recent? I just wanted to see what the dis***embly for sin was. Getting the source for glibc won't really tell me that because I won't know what really gets built.
<9> how do you redirect sbcl debug-io to a file ?
<8> nfroyd: Oh, yeah, that's it. Does it use fprem?
<3> lhz: (setf *debug-io* (open ...))?
<5> or I can just wait for these damn squirrels
<3> (But that's probably not a good idea, since there's an "i" in io.)
<6> rtoy_: I think that's pretty recent (last couple months or so)
<9> rydis: it is already printing on terminal, doesn't it have a fd number to redirect ?
<6> rtoy_: yeah, fprem1
<9> oh, I meant how do you redirect it from unix
<8> nfroyd: Ok. So x86 glibc still prefers speed over correctness. :-) Although I think some people will debate what (sin 1d100) should really be, pretending that 1d100 is some kind of "approximate" FP value.
<6> rtoy_: yeah, kind of unfortunate
<10> what steps does sbcl take to disable the crash reporter from popping up?
<10> on mac os x
<8> nfroyd: But if I take fdlibm's arg reduction code, x86 will get accurate sin/cos/tan, then. At the cost of slower evaluation.
<11> slava: none, afaik
<10> jsnell: ok. i'm looking for a simpler solution than what i'm currently using (mach exception handlers) since i don't understand the code and it doesn't work on x86
<6> rtoy_: and then we can use the code in nyef's sos :)
<12> is there any example code of how to define a binary structure and then output it to a stream?
<6> it's too bad that some of the fp libs since fdlibm are GPLd
<11> slava: I was just about to suggest you try using mach exceptions ;-)
<8> nfroyd: I wasn't going to translate it to Lisp. :-)
<3> dinosaur: Structures are printable/readable by default.
<8> The cephes library is pretty good. And not GPLed.
<12> rydis: yes, but in binary format?
<12> rydis: for instance, in C I could make a structure
<12> brb
<1> Dinosaur: I think that-dead-***y-book has an example of parsing and/or generating binary structures from a file.
<12> nyef: I'm a lisp beginner, with no college education. what is that-dead-***y-book? sicp?
<1> minion: Tell Dinosaur about that-dead-***y-book
<13> Dinosaur: please look at that-dead-***y-book: pcl-book: "Practical Common Lisp", an introduction to Common Lisp by Peter Seibel, available at http://www.gigamonkeys.com/book/ and in dead-tree form from Apress (as of 11 April 2005).
<12> oh. that one. I'm on chapter 3.
<6> rtoy_: huh, interesting. is it faster or slower than fdlibm?
<12> its pretty good.
<8> nfroyd: Don't know either. But I do know it doesn't have 1500 bits of pi stored away for arg reduction. :-) Claimed accuracy is pretty good, though.
<6> man, where's libm on osx?
<14> in libSystem.dylib i believe
<15> What would you guys suggest as a good socket library besides the one clisp comes with? (I haven't found anything wrong with it, I'm just curious and would like to explore other libraries too)
<16> minion: tell Yawgmoth7 about usocket
<13> Yawgmoth7: have a look at usocket: usocket is a sockets networking library providing a portability layer encapsulating implementation specific sockets programming details. http://www.cliki.net/usocket
<17> usocket, if you don't need server sockets
<3> Depends on what you want to do. trivial-sockets work well if they do what you want to do.
<4> and if they don't you're kinda up a creek
<3> Oooh. usocket seems new and shiny.
<17> it is
<17> eh put a lot of thought into it
<17> night.j
<15> I'll try both.
<18> evening, #lisp. any ecl users around?
<15> I won't need server sockets anyways
<3> Is usocket drop-in compatible with trivial-sockets? (For common operations.)
<16> I don't think so, but I could be wrong.
<3> Doesn't matter much, really, since it can't differ /that/ much in the trivial stuff. :)
<5> I suppose I should learn to make my packages asdf-installable one of these days
<15> asdf is so handy :)
<3> I don't really know ASDF, but there seems to be no simple way to hook in block-compilation of multi-file compilation-units and context-declarations/optimize-interface.
<15> I was really happy once I started using asdf
<12> a year ago I said lisp needed an equivalent of cpan. it looks like peter van eynde and others have come close to it. is that in afct the case?
<15> It would be pretty cool if lisp had a cpan
<19> rydis: not now, but working on it
<3> Dinosaur: CCLAN was there three or so years ago. It didn't really catch on. ASDF-INSTALL is somewhat similar, but not. cl-user.net is also somewhat similar, but not.
<12> the key to cpan was how easy it was to download missing modules.
<12> does any of those have that?
<3> eh: Cool. I hacked together something trivial for just block-compiling entire systems, but it was fairly ugly.
<12> ie, some "cpan installer" that will scream "missing module, where should I download it from?"
<5> so is there a lispier tar replacement yet?
<3> Dinosaur: That'd be ASDF-INSTALL.
<12> rydis: ok. and is there a wrapper that will check for missing dependencies and install missing modules?
<12> rydis: or is it easy to generate such a wrapper?
<3> No, ASDF-INSTALL will do that, itself.
<12> sounds like a cpan winner.
<3> Except for the "C" part, perhaps.
<12> lpan?
<12> ctan, cpan, lpan?
<3> Oh, wait. It's comprehensive, not centralized. Never mind.
<3> It's the "an" part that's not there.
<3> Dinosaur: ASDF-INSTALL indirects through CLiki pages with a certain stanza in them.
<12> rydis: nifty. cpan PLUS wiki....
<3> Which means that if the site that's put in the CLiki stanza is down, you lose.
<3> It also means that there's no "official" stamp of approval on asdf-install:able packages.
<3> So it's rather different from CTAN, CPAN et al.
<3> But if you're just after the aspect "in normal operation, get this package and make sure you get all the packages it depends on", ASDF-INSTALL is what you want.
<5> I don't suppose there's an easy way to make a symbolic link from SBCL?
<20> slyrus: why don't you suppose that?
<21> well, having loaded sb-posix before APROPOSing helps
<5> ah yes, thanks.
<5> that's what I was looking for
<18> why not have apropos also do a google for you, while you're at it? :)
<20> slyrus: you are full of lisp surprises!
<5> lisp is full of slyrus surprises, you mean
<3> Hmm. There's popped up a NewLisp evangelist on cll now?
<20> yes
<16> I think he's the author of newLisp.
<3> Riastradh: I see. That'd explain it, at least. ;)
<12> what is newlisp?
<22> A new Lisp dialect. :)
<5> FSVO new
<12> trivially new, or significantly new?
<3> "New" in the meaning "recently implemented". The language it implements is a bit old-fashioned.
<22> It has dynamic scope by default (or as the only choice?), for example.
<22> It is allegedly designed for scripting tasks.
<3> Interpreted only, too, IIRC.
<12> nice. is it small?
<12> be nice to have a small scripting language
<12> most guys that try that go for a scheme dialect rather than lisp
<3> But it's said to have a big library, and is single-implementation, so people who like Python/Perl/Ruby would probably like it. Most people here don't.
<3> Dinosaur: Scheme is about as much Lisp as NewLisp, I believe.
<22> I don't see why scsh would not be suitable for scripting tasks if you insist that CL is not suitable.
Return to
#lisp
or
Go to some related
logs:
shell-init cwd error retrieving current directory
#lisp
how to boot to gnome
#ldap
#perl
what_is_it_sm.jpg
#css
#suse
home_maildir
#web
|
|