Time  Nick         Message
07:48 RealBadAngel https://github.com/minetest/minetest/pull/1487
07:48 RealBadAngel ^^some changes to mapblock mesh
08:18 RealBadAngel celeron55, here?
09:17 celeron55    RealBadAngel: so whate's the benchmark?
09:17 celeron55    where's*
09:18 celeron55    (no, i don't know how i could write that so wrong)
09:19 celeron55    RealBadAngel: so where's the benchmark? <- i'll just repeat this to not sound so stupid
09:21 RealBadAngel we are tryin the code now
09:22 RealBadAngel definitely we can play with much larger view distances
09:22 VanessaE     celeron55: I'm testing that pull now.  I'd have to guess a 20% fps improvement with no visible regressions, at least on my R9 280X, with one exception - HavenNG's water is dark, almost black at close range (parallax mapping issue?), but default textures and HDX look fine.
09:23 RealBadAngel ahh, the water, i have to fix it
09:23 VanessaE     what is the cause of the black water?
09:23 RealBadAngel forget that, its the issue with my normalmap
09:23 VanessaE     ok
09:23 VanessaE     so not your code.
09:26 VanessaE     one thing I have noticed is that no matter how much improvement is done, there is an awful lot of work being done by the CPU to render the display, enough to peg one core at 100%, far more than I would expect should be needed even with a fairly simple scene and default textures with this badass GPU
09:26 VanessaE     which in turn is limiting the fps/view range
09:27 VanessaE     by "default textures" I should clarify:  default textures, no shaders, no filters, no effects.
09:31 VanessaE     the standard "acid test", looking north from the spawn on my Creative server:  32-37 fps (variable), 41m view range.  default textures.
09:33 VanessaE     it re-tuned itself slightly, 30-33 fps, 44.5m range.
09:33 VanessaE     only 39 entities in the area, btw.
09:40 RealBadAngel celeron55, so whats your opinion on the code?
09:41 celeron55    well at least you totally bastardized TileSpec which was intended to be a very simple data structure that is fast to copy in all cases
09:42 celeron55    that is why i really want to see performance as objective numbers
09:43 RealBadAngel tilespec is the right place to hold all the data needed to display it
09:43 RealBadAngel and we dont really need to copy it
10:09 VanessaE     RealBadAngel: got a glitch here...
10:10 VanessaE     normalmaps for fenceposts are rotated 90
10:11 VanessaE     90 degrees*
10:12 VanessaE     seems to affect anything using the fence drawtype.
10:12 VanessaE     (if it has a normalmap in the texture pack)
10:17 RealBadAngel can you make a screenshot before and after?
10:27 VanessaE     sure, gimme a few mins.
10:29 RealBadAngel that shouldnt happen at all, but maybe fences are using rotation modifiers, instead of rotating the vertices
10:32 VanessaE     before (i.e. git HEAD):  http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2007152014%20-%2006%3a32%3a11%20AM.png
10:32 VanessaE     after (with your patch): http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2007152014%20-%2006%3a29%3a41%20AM.png
10:34 RealBadAngel yeah, the drawtype needs to be fixed
10:38 VanessaE     celeron55: on a second look, I'd give RBA's code an estimate of 5% improvement, not 20%, at least when using HDX.  this CPU core being zonked to 100% is definitely limiting fps at this point.  gotta be.
10:39 VanessaE     now that I'm back on mainline, the same scene gives me maybe a slightly lower fps with a similar view range as with his code
10:39 RealBadAngel propably you will notice bigger difference with real huge amount of different and animated nodes around
10:40 VanessaE     in this scene, the only animation I can see is the result of shaders waving stuff around, and I think there's some flowing water off to one side, just off-camera.
10:41 VanessaE     (it's the same scene you saw in those screenshots above, in fact)
10:41 RealBadAngel i mean animated textures, not geometry shaders
10:41 VanessaE     (except I usually pitch up a bit)
10:41 VanessaE     yes, I know.
10:41 VanessaE     hence the mention of the flowing water.
12:14 sapier       RealBadAngel: still our main code has to be optimized to regular case, optimizing a worst case scenario to run bad instead of not at all at cost of making 99% scenario run slow instead of good isn't an option
12:18 sapier       Please keep this in mind on optimizing ;-)
14:25 VanessaE     HAH!
14:25 VanessaE     I FOUND an interesting glitch
14:26 VanessaE     while on my Survival server, if I grab and drag the Minetest window, the window content flickers between the current scene and what looks like some area from near that server's spawn
14:26 VanessaE     it also flickers between those two frames when the window gains/loses focus
14:27 VanessaE     it's not reproducable though
17:52 sapier       can someone explain to me why lua_isnil check doesn't work with luajit?
17:52 sapier       maybe too with lua, haven't tried
17:54 kaeza        sapier, you mean to check if no value was specified?
17:54 sapier       yes
17:55 sapier       I want optional parameters so I tried to check
17:55 sapier       if(!lua_isnil(L, 3))
17:55 sapier       value = lua_tonumber(L, 3);
17:55 kaeza        if (!lua_isnoneornil(L, i)) { /* value is nil */ } else if (lua_isnil(L, i)) { /* value not specified */ }
17:56 kaeza        err no
17:56 kaeza        in your case, probably just lua_isnumber()
17:56 sapier       but I don't wanna set value to 0 ;-)
17:56 sapier       it's preinitialized
17:57 sapier       isnoneornil works ... but there are various locations within minetest where my variant is used ... wonder how much of those cause bugss
17:57 sapier       -s
18:07 ShadowNinja  sapier: In Lua unspecified arguments are nil, but in C there's a difference between none (not passed) and nil (passed as nil).
18:08 ShadowNinja  Use isnumber or innone or innoneornil or type() == TNUMBER.
18:35 sapier       https://github.com/minetest/minetest/pull/1489
18:35 sapier       please check as I don't have a usecase for it
18:38 Calinou      very nice
18:38 Calinou      but, about mob movement… a mob may move several times in the same or a similar direction
18:38 Calinou      so won't the mob stop every now and then?
18:52 sapier       that's up to you if you set it prior time exceeds the mob wont stop
18:52 sapier       if you don't set the additional parameters behaviour is as before
18:52 kaeza        still, much better than entities drifting endlessly in one direction ;)
18:52 kaeza        thanks sapier
18:53 sapier       finetuning like adding a lua function "walk_to_pos" is for someone else ;-)
18:57 Krock        ShadowNinja, It's not possible to convert 'std::vector<_Ty>' to 'const Json:n_value' #1480
19:31 ShadowNinja  Krock: Hmmm, keep the loop then, but don't puch each item individually.
19:47 Krock        meh. Can't figure out how to do that automatically. foreach loops do not exist in C++
19:49 * Krock      capitulates
19:49 Krock        ShadowNinja ^
19:50 LemonLake    Krock: loop through 0 to the length of the array, item = array[i]
19:50 LemonLake    (psuedo)
19:50 Krock        and how to get the Key of that KeyValuePair-like variable?
19:51 Krock        nvm
19:51 ShadowNinja  Krock: for (const &item : container).  But that's C++11.  Anyway, there's already a loop doing that.
19:55 Krock        ok. found something at c_content.cpp , let's see if that loop works
19:57 * Krock      expects a test-compiling time of 10 minutes
20:05 VanessaE     hmmmm: did you make any progress on your idea of doing one(three) irrlicht scene node per mapblock?
20:50 hmmmm        honestly i haven't gotten to work on it
21:09 VanessaE     ok, just wondered.