Search
Wednesday, 13th of June 2018, 11:13:15 UTC
11:40:26
Xof
I think it's probably sensible for me to merge it nevertheless
11:40:51
stassats
does it break things? no - merge it
11:44:13
stassats
trying to base SSET on cl:hash-table, and it doesn't work somewhere, of course the place it breaks is not near sset usage
11:47:40
stassats
rereading the code for the fifth time, everything looks correct
12:28:06
Xof
loke: method tracing is now available in mainline sbcl
12:28:40
Xof
well, when this git push finishes it will be
13:36:47
corci
Project sbcl-master build #3476: FIXED in 1 hr 3 min: http://ci.cor-lab.de/job/sbcl-master/3476/
18:39:54
_death
what's the idea making it an sbcl contrib instead of a cl-capstone library?
21:05:18
stassats
oh damn, the elements that go into ssets might actually have duplicate numbers
21:07:21
stassats
and that's for blocks, which do change their numbers
21:10:30
stassats
how did it ever work? the current SSET doesn't store sset-element-number directly, but it hashes it, while sset-adjoin would work ok, sset-member will not work correctly
21:11:12
stassats
or does it depend on copy-sset rehashing the number or something?
21:32:36
stassats
using sset-element as a hash-key, not sset-element-number, and it finally works, doh
21:33:35
stassats
good thing i had started with cl:hash-table and not rolling a better hashing algorithm and found this problem first
21:37:15
stassats
well, sset-on-cl:hash-table: a fourfold increase in regalloc-bound test performance
21:37:36
stassats
unless i'm testing it wrong
21:44:59
stassats
no, that's not right, but i'm not planning on using cl:hash-table anyway
Wednesday, 13th of June 2018, 23:13:15 UTC