Search
Thursday, 12th of December 2019, 15:39:00 UTC
17:14:21
stassats
have a rough solution to #1398785
17:14:24
specbot
https://bugs.launchpad.net/bugs/1398785
17:14:36
stassats
that "Hard." is not so hard five years later
17:14:51
flip214
XachX: die eierlegende Wollmilchsau
17:28:54
karlosz
i thought i had done something wrong, but (find ... (sset-vector ...)) returns T while sset-member returns NIL
17:29:35
stassats
did you change the sset-element-number at any time?
17:30:53
stassats
as find-dfo changes block numbers
17:30:59
karlosz
well, this code is just checking for back edges by looking at dominators
17:31:20
karlosz
https://paste.gnome.org/pdk49c449
17:31:56
stassats
do you run find-dfo in between set operations?
17:32:31
stassats
which is to say, do your set operations span several ir1-optimize iterations
17:33:07
karlosz
no, i run this optimization after find-dfo
17:33:29
karlosz
but this is self contained after delete-if-no-entries in main
17:34:29
karlosz
should i expect find-dfo or dfo-as-needed to invalidate the dominator ssets?
17:35:10
stassats
yes, that's the known place where it's broken
17:35:34
stassats
but no point in fixing that without throwing sset away
17:35:49
stassats
something reasonably fast
17:50:14
stassats
need to count how many times 1398785 is encountered in real code
17:54:28
stassats
can't find any in some common code
17:59:27
karlosz
you mean replacing the implementation of sset rather than the concept itself?
18:00:29
stassats
sset has some degenerate states beside not keeping the blocks where they should be
18:01:03
stassats
reflecting on performance, not correctness
18:04:52
karlosz
well, the implementation has been replaced at least once already, couldn't hurt to do it again
21:02:44
stassats
the unbox-box is actually encountered in sbcl itself, so not a total waste of time
21:17:21
Xach
How can I find where SB-UNIX is created?
21:17:32
Xach
I'd like to look at its definition and exports
21:18:04
Xach
M-. says Error: DEFINITION-SOURCE of package SB-UNIX did not contain meaningful information.
21:21:14
stassats
in package-data-list.lisp-expr
21:23:29
stassats
and sb-unix isn't something you ought to be using
21:23:39
stassats
but you can look, sure
21:24:33
Xach
stassats: I wanted to look because i was confused by #.path-max failing when using C-c C-c in run-program.lisp
21:24:52
Xach
But recompiling the whole thing worked fine so that's good
21:25:57
stassats
are you on or openbsd freebsd dragonfly?
21:26:23
stassats
it might fail C-c C-c but not C-C C-k because of that conditional
21:26:46
Xach
stassats: I am in fact adding darwin to the list because run-program :pty t does not work properly right now on darwin
21:26:53
Xach
(trying to construct good tests to demonstrate)
21:29:25
Xach
stassats: C-c C-k gives me many package lock errors for that file
21:30:34
stassats
i don't know what will happen if you don't have #+sb-devel
21:30:41
stassats
probably why the path-max error
21:38:59
stassats
added a new HACKING section
21:39:00
stassats
https://github.com/sbcl/sbcl/blob/master/HACKING
22:36:06
karlosz
is it frowned upon to annotate nodes in hash tables? is there a better way to mark nodes during an optimization?
22:36:53
karlosz
ir1-optimize seems too big and scary, i'd like to optimize separately
22:38:06
karlosz
theres also block-flag too
22:38:20
stassats
well, that's for blocks
22:39:18
karlosz
speaking of which, IR1-TO-DOT confused the hell out of me until i realized it was doing DFOs on predecessors rather than successors
22:40:09
karlosz
i've got working loop invariant code motion now, at least for the simple example of (defun f (x) (dotimes (i 10) (print (+ x x))))
22:40:38
karlosz
now to just make it smart with SETs and maybe send a patch to the list for some feedback
22:41:52
stassats
ir1-to-dot is backwards because back-propagate-one-dx-lvar goes backwards
22:45:59
stassats
but what is nth-vop even
22:46:45
stassats
git log -S nth-vop => "Mon Sep 18 01:26:16 2000 +0000"
22:47:05
Xach
I can't seem to simply reproduce the issues I had last year regarding pty use on darwin. So I will just start using them and see what happens.
22:49:59
stassats
and then "Tue Feb 6 11:33:52 1990 +0000"
22:50:40
stassats
i don't really see how it's useful for debugging, but ok
22:51:33
stassats
Xach: maybe apple fixed it
Friday, 13th of December 2019, 3:39:00 UTC