Search found 5 matches

by bsdfish
Wed Oct 29, 2008 1:42 am
Forum: Common Lisp
Topic: non-linear list conversion to linear
Replies: 14
Views: 32328

Re: non-linear list conversion to linear

Look at your ((listp (car x)) (cons NIL (lin (car x)))) That line means means 'if the first element of your list is a list itself, you call lin on it and attach nil to the front.' What you really want to be doing is appending the linearization of (car x) to the linearization of the remainder of x.
by bsdfish
Tue Aug 26, 2008 12:32 am
Forum: Common Lisp
Topic: Multi-core utilization in SBCL
Replies: 7
Views: 15715

Re: Multi-core utilization in SBCL

Glad it helped. I suspect that the garbage collector is not tuned for parallelism and large memory sizes; I think that increasing the (bytes-consed-between-gcs) is a good idea on any machine with a large amount of memory. In fact, SBCL has some other issues with the address space; last I checked, it...
by bsdfish
Fri Aug 22, 2008 4:39 pm
Forum: Common Lisp
Topic: Multi-core utilization in SBCL
Replies: 7
Views: 15715

Re: Multi-core utilization in SBCL

When I was doing matrix operations in SBCL which I thought should be completely parallelized, I was getting about a 2.5x speedup on a 4 core machine. I suspect that memory issues were the reason -- each core has cache contention issues and also, the memory bandwidth may be limited. If you have a lar...
by bsdfish
Fri Aug 01, 2008 4:26 am
Forum: Common Lisp
Topic: define-compiler-macro
Replies: 11
Views: 30698

Re: define-compiler-macro

Alternatively, lets say you are working on a matrix library. (defun m* (matrix-1 matrix-2 &optional target) (if (null target) (setf target (make-matrix-of-right-size)) ; blas call to multiply matrix-1 by matrix-2 and save to target ) (defun m+ (matrix-1 matrix-2 &optional target) (if (null t...
by bsdfish
Sat Jul 12, 2008 4:36 am
Forum: The Lounge
Topic: New lisp-dialect for probabilistic models
Replies: 1
Views: 5409

New lisp-dialect for probabilistic models

I'm currently at UAI and there is a really interesting paper which uses the code-as-data aspect of lisp to create a language in which you can describe an arbitrary probabilistic model, and do inference on it.

http://uai2008.cs.helsinki.fi/UAI_camer ... oodman.pdf