#exult@irc.freenode.net logs for 25 Oct 2012 (GMT)

Archive Today Yesterday Tomorrow
Exult homepage


[00:04:08] <sh4rm4> 1.3GiB [###### ] scale_hq4x.o
[00:04:31] <sh4rm4> we should definitely disable debug info generation for these scalers
[00:04:45] --> Dominus1 has joined #exult
[00:04:50] <sh4rm4> this makes them bloat up into hell
[00:07:52] <-- Dominus has left IRC (Ping timeout: 244 seconds)
[00:09:24] <sh4rm4> --enable-hq2x Enable hq2x scaler support [default no]
[00:09:24] <sh4rm4> --enable-hq3x Enable hq3x scaler support [default no]
[00:09:24] <sh4rm4> --enable-hq4x Enable hq4x scaler support [default no]
[00:09:34] <sh4rm4> hmm i didnt turn on a single one of them
[00:09:47] <sh4rm4> according to this, they should be disabled by default
[00:11:08] <sh4rm4> ah, --enable-all-hq-scalers Enable hq2x, hq3x, hq4x scaler support [default yes]
[00:11:20] <sh4rm4> so the right switch would be --disable-all-hq-scalers to turn em off ?
[00:12:51] <sh4rm4> btw. when i run only ./configure, exult is built with -g (debug infos)
[00:13:07] <sh4rm4> imo -g should only be added when --enable-debug is used
[00:13:53] <sh4rm4> it is even used with --disable-debug
[00:15:02] <sh4rm4> aha, only when i set CXXFLAGS manually, the "-O2 -g" default gets overwritten
[00:15:25] <sh4rm4> CXXFLAGS="-O2" ./configure gets rid of -g
[00:17:28] <sh4rm4> mhm, compiled like this, the exe is only 4MB
[00:17:37] <sh4rm4> and it works with -j9
[00:17:46] <sh4rm4> compiles in ca 50 seconds
[01:19:13] --- ChanServ gives channel operator status to Dominus1
[01:19:14] --- Dominus1 is now known as Dominus
[02:18:52] --- n00bsaibot is now known as Sevalecan
[04:56:41] --> Kirben has joined #exult
[04:56:42] --- ChanServ gives channel operator status to Kirben
[05:38:48] --> Kirben_ has joined #exult
[05:38:49] --- ChanServ gives channel operator status to Kirben_
[05:39:41] <-- Kirben has left IRC (Ping timeout: 255 seconds)
[09:15:01] --> Rottingbeef has joined #exult
[10:27:46] <-- Kirben_ has left IRC ()
[12:03:32] --> TheCycoONE has joined #exult
[12:29:02] --> Marzo has joined #exult
[12:35:24] --> Marzo1 has joined #exult
[12:35:25] <-- Marzo has left IRC (Disconnected by services)
[13:00:42] --> Dominus1 has joined #exult
[13:02:12] <-- Dominus has left IRC (Ping timeout: 248 seconds)
[13:08:46] <-- Marzo1 has left IRC (Ping timeout: 245 seconds)
[13:35:40] <-- Dominus1 has left IRC (Read error: Connection reset by peer)
[14:03:42] <sh4rm4> let me summarize my analysis from yesterday:
[14:04:08] <sh4rm4> no matter if you disable or enable debug info, -g will be added to CXXFLAGS, unless you set it yourself
[14:04:55] <sh4rm4> which will subsequently bloat the hq scaler compilation to use gigabytes of ram
[14:05:07] <sh4rm4> and will lead to a 800 MB big game binary
[14:06:07] <sh4rm4> so this default -g flag has to be removed
[14:06:25] <sh4rm4> also those hq-scalers should be disabled by default
[14:12:26] --> Dominus has joined #exult
[14:12:26] --- ChanServ gives channel operator status to Dominus
[14:20:09] <wjp> I understand your point about hq, but why do you want to remove the -g?
[14:20:33] <wjp> also, how do you "disable debug info"?
[14:21:50] <wjp> (if you mean --enable-debug, that's mainly intended for enabling extra logging output and such)
[14:22:43] <wjp> also note that -g -O2 is autoconf's default set of flags
[14:26:35] <Dominus> I don't think we should disable the hq scalers by default. Normal users want them and testers can quickly disable it in condfigure
[14:28:18] <Dominus> wjp, I think we took down a major bug last night which might very well account for all other crashes not caused by the disappearing objects
[14:28:52] <Dominus> so far only one crash is left in the bug list, I think. sailing ships at the map wrap where all four corners meet
[14:47:04] <wjp> it was definitely one that could hit almost anywhere
[15:20:06] <Eviltar_> i think he's suggesting overriding -g if --debug isn't specified
[15:20:32] <Eviltar_> and asking how to configure all shaders dissabled
[15:20:43] <Eviltar_> if i understood
[15:44:31] <Eviltar_> have you guys looked at the xbr filter?
[15:44:36] <Eviltar_> http://filthypants.blogspot.com/2012/03/xbr-vs-hqx-interpolation-filter.html
[15:44:49] <Eviltar_> it's pretty snazzy
[16:27:25] <Dominus> Eviltar_: really looks nice. Though sounds a bit too good to be true :) Better than HQX and less cpu consuming...
[16:27:47] <Dominus> if you ever feel like doing something big, port this scaler to Exult, Pentagram and Nuvie :)
[16:29:00] <Eviltar_> so far it looks amazing in all the emulator's i've seen it in
[16:32:07] <Dominus> I don't actually doubt it
[16:37:09] <Dominus> hmm, if someone (*hint*hint*hint*eviltar*hint*hint*hint*) would really look into it it seems as if that could be relatively easy to port into Exult's scaler setup
[16:37:31] <Dominus> I'm really a bit spent at the moment and don't know if I can wrap me head around it...
[16:37:59] <Dominus> I'm busy asking wjp to help me further with the autonotes stuff :)
[16:38:24] <Eviltar_> haha if I can figure it out
[16:39:06] <Eviltar_> it woudl look nice
[16:39:29] <Dominus> you'd probably need to look at another scaler that isn't fixed to one size (2x) and compare things etc... That's how I ported the hq4x scaler to exult...
[16:42:07] <Dominus> if you make a patch I'd be happy to add it, too :)
[18:05:00] <sh4rm4> <wjp> I understand your point about hq, but why do you want to remove the -g?
[18:05:16] <sh4rm4> the -g bloats up hq4x to 1.4 GB (!) from 890 KB
[18:05:35] <sh4rm4> because of all the hideous macros and templates
[18:06:57] <sh4rm4> if you compile everything *except* these scalers with -g, the resulting binary will be < 16 MB
[18:07:12] <sh4rm4> but with the scalers, it will become ~ 880 MB
[18:08:53] <sh4rm4> since these scalers dont need to be debugged (they're final), this bloat is completely unnecessary
[18:09:33] <sh4rm4> it justs hurts compile times, memory usage while compiling *and* while using a debug build binary
[18:10:32] <sh4rm4> (note: the scalers impose no problem at all when compiled without -g, they're still slow to compile but need at most 100 MB ram, and compilation is 20x faster)
[18:11:52] <sh4rm4> <wjp> also, how do you "disable debug info"?
[18:12:02] <sh4rm4> i expected --disable-debug would do that
[18:12:33] <sh4rm4> since all --enable flags can be reversed with --disable usually
[18:12:40] <wjp> so it sounds like we should just drop the -g flag for just those hq scalers?
[18:12:48] <sh4rm4> *nod*
[18:12:53] <wjp> yes, but as I said, --enable-debug enables more than just debugging info
[18:13:15] <wjp> hm, now how did that work in automake
[18:13:23] <sh4rm4> even nothing else, by default these scalers should not be debugged with -g
[18:13:34] <sh4rm4> but optimally, nothing would be compiled with -g by default
[18:13:43] <sh4rm4> unless --enable-debug was explicitly used
[18:13:49] <wjp> if anybody has any recent automake experience, please chime in :-)
[18:13:54] <sh4rm4> s/even/if/
[18:14:45] <sh4rm4> s/debugged/compiled
[18:53:27] --> Marzo has joined #exult
[18:54:23] <-- Marzo has left IRC (Remote host closed the connection)
[18:55:09] --> Marzo has joined #exult
[19:20:22] --> Marzo1 has joined #exult
[19:20:23] <-- Marzo has left IRC (Disconnected by services)
[19:25:44] <-- Marzo1 has left IRC (Ping timeout: 255 seconds)
[19:30:40] --> Marzo has joined #exult
[19:45:03] --> Marzo1 has joined #exult
[19:45:03] <-- Marzo has left IRC (Disconnected by services)
[19:59:02] <-- Marzo1 has left IRC (Read error: Connection reset by peer)
[20:00:22] --> Marzo has joined #exult
[20:11:16] <-- Marzo has left IRC (Read error: Connection reset by peer)
[20:11:38] --> Marzo has joined #exult
[20:36:17] <-- Marzo has left IRC (Read error: Connection reset by peer)
[20:36:28] --> Marzo has joined #exult
[20:56:06] <-- Marzo has left IRC (Ping timeout: 265 seconds)
[21:26:41] <-- TheCycoONE has left IRC (Quit: And then there were n-1)
[21:59:46] <-- Rottingbeef has left IRC ()
[23:13:15] --> Kirben has joined #exult
[23:13:15] --- ChanServ gives channel operator status to Kirben