00:09
hodge has joined #pypy
00:11
Hodgestar has quit [*.net *.split]
00:27
muke has quit [Quit: Leaving]
02:01
jcea has quit [Ping timeout: 258 seconds]
02:33
Gustavo6046 has quit [Ping timeout: 272 seconds]
02:39
Gustavo6046 has joined #pypy
03:44
mjacob has quit [Ping timeout: 264 seconds]
04:29
mjacob has joined #pypy
04:30
oberstet has joined #pypy
04:35
mjacob has quit [Ping timeout: 258 seconds]
05:09
xcm has quit [Ping timeout: 256 seconds]
05:18
xcm has joined #pypy
05:19
mjacob has joined #pypy
05:28
<
cfbolz >
muke: rpython/jit/metainterp/resoperation.py
06:36
hodge has left #pypy [#pypy]
06:36
Hodgestar has joined #pypy
07:10
<
LarstiQ >
cfbolz: that's a nice speedup!
07:11
<
cfbolz >
LarstiQ: yep
07:11
<
cfbolz >
It's a bit specialized, most programs probably aren't affected
07:26
<
LarstiQ >
I haven't followed closely what map-improvements does, do you have a writeup yet? Wondering how specialized it is
07:26
<
LarstiQ >
still, 850x is eye-popping even if it's narrowly applicable
07:29
<
cfbolz >
LarstiQ: it's optimizing instance fields that are ints and floats
07:35
<
LarstiQ >
ah, I see
07:37
<
cfbolz >
By storing them without a box
07:37
<
cfbolz >
So reading and writing of such fields is a bit faster
07:42
<
LarstiQ >
right, unboxing in list-strategies seemed like a good idea too
07:50
forgottenone has joined #pypy
08:08
<
cfbolz >
LarstiQ: list strategies are a lot more straightforward 😅
11:24
Dejan has joined #pypy
11:47
danchr_ has joined #pypy
11:49
danchr_ has quit [Read error: Connection reset by peer]
11:49
danchr__ has joined #pypy
11:57
danchr_ has joined #pypy
11:58
danchr_ has quit [Client Quit]
11:58
danchr_ has joined #pypy
12:24
jcea has joined #pypy
12:52
otisolsen70 has joined #pypy
13:00
Rhy0lite has joined #pypy
14:59
Rhy0lite has quit [Quit: Leaving]
15:05
forgottenone has quit [Quit: Konversation terminated!]
15:56
forgottenone has joined #pypy
16:00
forgottenone has quit [Client Quit]
16:07
forgottenone has joined #pypy
18:51
<
cfbolz >
I broke something, but I don't understand how it's possible
18:54
<
LarstiQ >
I had the reverse today, something was changed and I didn't understand why it didn't break things. Turns out there was another source of almost the same data being passed on in exactly 1 case
19:52
otisolsen70_ has joined #pypy
19:53
otisolsen70_ has quit [Remote host closed the connection]
19:55
otisolsen70 has quit [Ping timeout: 258 seconds]
19:57
glyph has quit [Quit: End of line.]
20:00
glyph has joined #pypy
20:05
glyph has quit [Client Quit]
20:06
glyph has joined #pypy
20:28
<
cfbolz >
but in the run two commits earlier there is not
20:29
<
cfbolz >
and the two commits are 1) dead code removed and 2) a comment
20:30
<
mattip >
it didn't reproduce in the subsequent build
20:31
<
cfbolz >
mattip: yes, but that was two commits earlire
20:31
<
cfbolz >
I tried to pinpoint what caused it
20:31
<
cfbolz >
maybe it's not deterministic
20:32
<
mattip >
does it appear on 64 bit?
20:32
<
cfbolz >
32bit only
20:35
<
nulano >
mattip: can you merge default -> py3.7 for the Windows Tcl changes?
20:36
<
nulano >
I'm not sure what is the current py3.6 workflow, but that will also fail to build without it
20:39
<
cfbolz >
I suspect it's a random failure, will wait for tonight's run and then backout stuff
20:39
<
mattip >
I think 3.6 is pretty much abandoned now. Unless someone twists my arm I do not intend on making another 3.6 release
21:12
<
mattip >
I grafted the changes to py3.6 just to prevent bit-rot
21:58
oberstet has quit [Quit: Leaving]
22:44
xcm has quit [Remote host closed the connection]
22:44
xcm has joined #pypy
22:51
Taggnostr has quit [Remote host closed the connection]
22:52
Taggnostr has joined #pypy
22:54
Gustavo6046 has quit [Ping timeout: 246 seconds]
23:03
Gustavo6046 has joined #pypy
23:36
forgottenone has quit [Ping timeout: 265 seconds]