Search
4:50:47
beach
Good morning everyone!
6:52:05
splittist
Good morning beach!
6:54:45
beach
Long time no see. What's up?
7:34:46
splittist
beach: the usual - overworked, underappreciated (:
8:31:14
shka
did anybody made something like slime inspector, but in clim?
8:31:28
shka
so it is sligthly nicer to use
8:33:26
beach
shka: Yes, there is Clouseau.
8:34:41
shka
beach: heh, now i don't know if you are kidding or you are serious
8:34:57
shka
because Clouseau seems like legit name
8:35:22
beach
moore33 came up with the name, as I recall.
8:40:02
ldb
holly, so i can use a graphic inspector
8:41:02
shka
well, time to check how it works
8:43:57
shka
heh, i forgot that i am connected to a remote machine via ssh tunnel
8:48:29
ldb
emmm, doesn't work that satisfiying
12:00:54
trocado
Hi! Why calling elt on a list (vs nth) gives array-like performance?
12:01:00
trocado
Check this out: https://pastebin.com/TNvsW9L5
12:04:11
beach
trocado: Since you are not using the value of a call to a pure function, SBCL likely removes the call altogether.
12:04:33
beach
trocado: Benchmarks are hard to construct.
12:07:52
shka
heh, don't call on those 148.006 cycles ;-)
12:08:13
shka
sometimes it is useful to even dissasemble
12:08:34
trocado
beach: I guess that's it! I changed "do" to "sum" in the loop and now the results are as expected.
12:16:59
scymtym
beach: this looks like a bug in SBCL to me since ELT is supposed to signal an error for invalid indices
12:17:58
beach
I didn't look very carefully.
12:20:00
scymtym
(and reading a special variable in the inner loop of a microbenchmark can easily ruin the benchmark)
12:31:11
Xof
"should signal an error" mandates an error only in safety 3 code