libera/#lisp - IRC Chatlog
Search
15:15:07
pjb
phao: there's https://github.com/mqvist/cl-raytracer and http://vintage-digital.com/hefner/software/bordeaux-fft/manual.html and http://nklein.com/software/cl-fft/
15:15:21
mfiano
phao: That is implementation-specific. You might want to check your implementation's IRC channel if one exists.
15:16:06
phao
Thank you. I'm sorry for asking this (again) in here. I think I've asked this here not so long ago. It's just that I forgot the answers people gave me (sorry).
15:16:39
jackdaniel
phao: this channel is logged, so you may try to find last answers there (see topic)
15:16:43
mfiano
Any efficiency tradeoffs one would make outside of proper algorithms and data structures is usually very compiler-specific.
15:17:49
jackdaniel
magicl provides bindings to fortran libraries (high level interface) - most notably blas and lapack
15:19:29
mfiano
You could check out mfiano/origin on GitHub for a fast (on SBCL at least) linear algebra library suitable for 3D graphics. It is actually as fast or faster (depending on function) as SIMD-optimized Google math-fu library via FFI
15:20:46
mfiano
Yet it is purely Common Lisp. SBCL does a good job of generating efficient machine code.
15:23:30
mfiano
ACTION is not so obsessed with performance these days; it makes extensibility and interactive programming a little more difficult.
15:23:59
mfiano
(I rewrote origin as a different library using CLOS/generic functions; mfiano/gfxmath)
15:24:53
copec
I don't know what, it's enough to make my brain take a double take like it wants to recognize something
15:28:49
wmealing
thank you. I'll ask in case someone knows. I have a form on the repl, and i want to send it back to emacs.. Other than using emacs yank and then insert, is there a function to shunt the current form on the repl back to the lisp buffer ?
15:33:19
edgar-rft
hmm, just tested: C-x C-e in a lisp-mode buffer returns the value from CL in the echo area and the *Messages* buffer but C-j doesn't work in lisp-mode buffers here :-(
15:33:53
mfiano
How is Emacs supposed to know which lisp buffer? There could be many across multiple frames.
15:34:36
wmealing
i anticipated that question mfiano and thought about it.. my hope was that there was a way to insert it back into the most recent active buffer.
15:41:34
edgar-rft
wmealing: C-u C-x C-e does what you want, but it's a bit cumbersome to type three C-<something> combos in a row
15:50:04
wmealing
i'm getting pretty comfortable in my doom.. not sure what that says about my mental state ;)
16:41:47
White_Flame
phao: the biggest is avoiding type tests. SBCL's type inference can go a long way, but when it comes to numbers, adding 2 fixnums might (from the compiler's point of view) overflow into a bignum, so it would need more declarations
16:42:42
White_Flame
if you're doing floating point, then singles are going to be less overhead than doubles on a 64-bit machine, because single floats can be immediate values. Unboxed immediate doubles are possible in certain circumstances, though
18:53:47
jcowan
White_Flame: Unboxed doubles are always possible on x86_64 and ARM64 systems using nanboxing