Chat log from the meeting on 2020-09-29

From OpenSimulator

Jump to: navigation, search

[11:04] Andrew Hellershanks: Hello, everyone.
[11:04] Bill Blight: Howdy Andrew
[11:04] Kayaker Magic: I just noticed the yeng command in console help, how long has that been there?
[11:04] Ada Radius: Hi Andrew
[11:04] Gavin.Hird Hi Andrew
[11:04] Bill Blight: since Yengine has been there
[11:05] Bill Blight: the Yeng command is handy for finding runaway scripts
[11:05] Kayaker Magic: Yes, does look useful!
[11:05] Gavin.Hird Evening Ubit
[11:05] Ada Radius: Hi Ubit
[11:05] Ubit Umarov: (hi all )
[11:06] Kayaker Magic: Ubit!
[11:06] Andrew Hellershanks: Hello, Ubit.
[11:07] Bill Blight: yeng ls -max=20 -topcpu is my fav, use it a lot
[11:07] Ada Radius: Hi Leighton
[11:07] Gavin.Hird Hi Leighton
[11:07] Leighton.Marjoram Hello everyone good to see you all :)
[11:07] Andrew Hellershanks: The yeng command could be useful if you can't use top scripts in the viewer.
[11:08] Kayaker Magic: But still have access to the console?? Seems backwards.
[11:08] Andrew Hellershanks: Hm... I just remembered I haven't used one of the known bad scripts to determine what is considered a bad runtime in YEngine.
[11:09] Andrew Hellershanks: Hello, Leighton.
[11:09] Bill Blight: Hard to say, would have to be a known bad in Yengine, since sometimes a bad Xengine script works fine in Yengine
[11:09] Kayaker Magic: I'm using it to watch copies of one script that I hope will demonstrate heap issues. One copy has been running for weeks, racking up the miliseconds!
[11:10] Andrew Hellershanks: Bill, I have some known bad XEngine scripts and their fixed counterparts. I would compare the numbers of the bad and fixed versions in YEngine.
[11:11] Andrew Hellershanks: IIRC, the bad scripts I have are currently stored away in an IAR file that I haven't loaded in to my standalone.
[11:11] Ubit Umarov: last time i checked yegine did pass all sl tests on a test script they have
[11:12] Bill Blight: I've never subscribed to the "How long can you keep a region running" contest ... Even SL does rolling restarts ... I restart all my regions at least once a week.
[11:12] Ubit Umarov: hmm except one about number of chars on float.tostring or something
[11:12] Ubit Umarov: xengine running that.. well nm ;)
[11:12] Andrew Hellershanks: Ubit, too many or too few?
[11:12] Ubit Umarov: a few more i thing
[11:13] Ubit Umarov: think
[11:13] Kayaker Magic: I discovered the yeng command by typing help to the console, on my system it also lists vivox in the help list (I do have it running) but that help command does not work.
[11:13] Ubit Umarov: damm folish use of double
[11:13] Ubit Umarov: i should change all that fo float :p
[11:13] Bill Blight: " help vivox

For more information, type 'help all' to get a list of all commands,
or type help <item>' where <item> is one of the following:
vivox debug <on>|<off> - Set vivox debugging
[11:14] Ubit Umarov: guess only some bad math scripts will notice it
[11:14] Ubit Umarov: but well its a big change.. both on Yeng and xengine.. bit pain
[11:14] Ubit Umarov: big
[11:15] Ubit Umarov: but double is a total waste..
[11:15] Kayaker Magic: I thought you already changed all the doubles to floats?
[11:15] Ubit Umarov: that was on some mysql
[11:15] Andrew Hellershanks: If it is just a .tostring() call it isn't likely to be much of an issue.
[11:16] Ubit Umarov: that was a fun thing also.. things that are really floats where stored on DBs as double
[11:16] Ubit Umarov: no idea what gets into ppl minds sometimes..
[11:17] Kayaker Magic: that's because MySQL internally stores floats as double even if you declare them as small floats.
[11:17] Andrew Hellershanks: Changing variable types can be a big pita as it is often opening a big can of worms. I have a program that should use long instead of int. To change that requires changing the variable type in a ton of places.
[11:18] Ubit Umarov: mysql now has type float
[11:18] Ubit Umarov: does not store it as double
[11:18] Gavin.Hird the viewer code is brilliant for variable types...
[11:19] Ubit Umarov: UUID is a DBs silly thing also
[11:19] Ubit Umarov: guess only postg does it right
[11:20] Kayaker Magic: Yes, if you define something as a float in MySQL, send it a string with too many digits, verify with a search that the result is truncated, change the data type to double, do a search again, the extra digits come back.
[11:20] Ubit Umarov: ie as a native 16byte type
[11:20] Ubit Umarov: we store them in many places as a large char and UTF8 Yeackk
[11:21] Ubit Umarov: but as andrew said if i try to fix that, some grids will be down while mysql updates all that crap
[11:23] Gavin.Hird indra/llcommon/stdtypes.h
[11:23] Ubit Umarov: so what new do you have about opensim?
[11:23] Ubit Umarov: news..
[11:24] Kayaker Magic: we keep hoping you will tell us!
[11:24] Gavin.Hird you did not push any changes over the last few
[11:24] Andrew Hellershanks is trying to type with a cat standing on his lap
[11:24] Andrew Hellershanks: There have been changes related to YEngine and Profile this past week.
[11:24] Ubit Umarov: well that thing you added to dayt about parcels group should work now
[11:25] Gavin.Hird yes, you said, but no changes pushed :-)
[11:25] Ubit Umarov: totally useless flag since viewer had all info there. but well
[11:25] Ubit Umarov: ofc i pushed all there
[11:26] Gavin.Hird not to bitbucket
[11:26] Ubit Umarov: thinki did to that also
[11:26] Ubit Umarov: but look again :p
[11:26] Gavin.Hird haha
[11:26] Ubit Umarov: ( push to bitbucket is by hand.. sometimes i forget :p )
[11:26] Gavin.Hird better late than never, hehe
[11:27] Ubit Umarov: well my region on yr box has it and i update it using git
[11:27] Ubit Umarov: just our box git
[11:27] Gavin.Hird I know, but RL is rolling by
[11:28] Ubit Umarov: made some push during this last week
[11:28] Ubit Umarov: many about HG uris
[11:29] Ubit Umarov: namely gatekeeper and home ( that is a agents service, not avatar home )
[11:30] Ubit Umarov: added proper validation of the strings.. old code just did read them and kept trying to use everywhere
[11:30] Ubit Umarov: in past i added checks but local to each module
[11:31] Ubit Umarov: now i centralized that, scene knows its grid gatekeeper and "home"
[11:31] Ubit Umarov: so most modules get them from it
[11:32] Ubit Umarov: if you look to old ini you see gatekeeper could be set on a ton of modules..
[11:32] Ubit Umarov: only seem by each, and each could have own value...
[11:32] Gavin.Hird it could
[11:32] Ubit Umarov: that makes no sense at all on a region
[11:32] Ubit Umarov: so i removed all that.
[11:33] Ubit Umarov: gatekeeper and that "home" ( bad name) belong to section [hypergrid] only
[11:33] Ubit Umarov: ( startup is also checked )
[11:33] Ubit Umarov: [hypergrid] is used on on both
[11:34] Ubit Umarov: sure the fancy name "microservices" did popup.. but thats nonsense on a region
[11:35] Ubit Umarov: those things are global to a grid and are also used as identifiers, like on avatar names
[11:37] Ubit Umarov: did fix a bug on remoteparcels, used by landmarks about
[11:37] Gavin.Hird maybe someone envisioned regions running across server instances
[11:37] Bill Blight: I have helped fix a few grids where they decided to set the gatekeeper in multiple locations and had a typo .. Always thought that was a really bad idea
[11:37] Ubit Umarov: yeha there where some dreams about that
[11:37] Ubit Umarov: forgetting the communications kill all that
[11:37] Ubit Umarov: like intel DSG
[11:38] Ubit Umarov: they even considered distributed physics engine
[11:38] Ubit Umarov: well fancy ideas.. total fail
[11:38] Ubit Umarov: because communications COST a LOT
[11:39] Ubit Umarov: well even inside cpus those NUMA crossings.. err nm
[11:39] Gavin.Hird NUMA
[11:39] Andrew Hellershanks: Bill, I have wondered why the program has the option of setting certain HG settings in multiple places. If someone reads the comments they do tell you which setting is the master.
[11:39] Gavin.Hird that was a very hot buzzword in IBM back when
[11:40] Ubit Umarov: see in this things context, loading textures to GPU; devs consider the CPU to be 2 slow, and new GPUs can read directly from disk
[11:40] Ubit Umarov: thats very local communications..
[11:40] Gavin.Hird Apple's new GPU will share memory with the CPU
[11:41] Gavin.Hird on the same die
[11:41] Gavin.Hird dye
[11:41] Ubit Umarov: even so.. they want gpu to control ( fun with diferent file systems.. )
[11:41] Gavin.Hird it might die
[11:41] Ubit Umarov: :)
[11:42] Ubit Umarov: what i said is already in use on some consoles
[11:42] Andrew Hellershanks: Gavin, I hope they dual port the RAM.
[11:42] Ubit Umarov: gpu gets the data from disk directly, decodes etc
[11:42] Ubit Umarov: dream andrew
[11:42] Gavin.Hird they do
[11:42] Ubit Umarov: pseudo dual port at most
[11:43] Ubit Umarov: and with pci.ex in the middle
[11:43] Andrew Hellershanks: Sharing memory between GPU and CPU without dual porting it is something that was done in the past and caused performance issues.
[11:43] Ubit Umarov: well thats on current memory share
[11:44] Gavin.Hird but we have not seen it yet
[11:44] Ubit Umarov: wlel the intel embedded gpus guess do direct share
[11:44] Ubit Umarov: so actually memory share is a old thing
[11:45] Ubit Umarov: so thats how communications are considered on simulations
[11:45] Ubit Umarov: thinking using slow crap http internel to bind time critical parts.. well
[11:45] Ubit Umarov: nm :p
[11:46] Ubit Umarov: fun ideas on paper. sure
[11:46] Ubit Umarov: when we can't do crap region crossings.. duhhh
[11:46] Andrew Hellershanks: The Amiga computer had a way of interleaving access to RAM for use by the CPU and video circuitry without requiring dual port memory. It was a clever design.
[11:46] Bill Blight: Loved my Amiga
[11:46] Bill Blight: And my Video Toaster
[11:46] Ubit Umarov: only on a shared memory bank
[11:47] Ubit Umarov: not all ram
[11:47] Ubit Umarov: that was a problem on more recent amigas
[11:47] Ubit Umarov: only chip got more access but still not all ram
[11:47] Andrew Hellershanks whispers: Bill: Nice. I still have my A2000HD with the 68030 accelerator card. I need to dig it out one day and see how badly the leaky battery has affected the motherboard.
[11:48] Andrew Hellershanks whispers: Bill: Nice. I still have my A2000HD with the 68030 accelerator card. I need to dig it out one day and see how badly the leaky battery has affected the motherboard.
[11:48] Ubit Umarov: and that in part was made possible because how 68k worked with ram
[11:48] Andrew Hellershanks whispers: bah. Why did it whisper that (again)? :P
[11:48] Ubit Umarov: guess not that easy on the 68030
[11:48] Ubit Umarov: 6809 was very easy
[11:49] Andrew Hellershanks whispers: The '030 board also had additional memory.
[11:49] Ubit Umarov: slow ofc and only did use ram on half the clock cycle
[11:49] Ubit Umarov: the other half was free for any dma
[11:49] Andrew Hellershanks whispers: The 6809 is my favourite 8-bit CPU.
[11:49] Bill Blight: looks like you have clicked the button to the right and set all your chat to whisper Andrew
[11:50] Andrew Hellershanks: No, it still says "Say".
[11:50] Ubit Umarov: fs sometimes changes that, no idea why
[11:50] Ubit Umarov: i mena does whisper
[11:50] Bill Blight: sticky key
[11:50] Andrew Hellershanks: I've switched to the main chat bar now instead of the "nearby chat" window.
[11:51] Ubit Umarov: let me see commits history on last week :)
[11:52] Ubit Umarov: yeah mostly about those HG URIs
[11:52] Ubit Umarov: have another push on that waiting :p
[11:53] Ubit Umarov: fun .net core team worries a lot more abotu perfomance now
[11:54] Ubit Umarov: adding native UTF8 support instead of crap utf16
[11:55] Ubit Umarov: and they also had the doubles thing.. system.math is only doubles
[11:55] Ubit Umarov: now they are adding mathf
[11:55] Ubit Umarov: and more use of SIMD
[11:55] Gavin.Hird they probably broke their back on utf16 when implementing WLS 2
[11:55] Ubit Umarov: utf16 was a bad choice
[11:56] Ubit Umarov: the worse of both utf32 and utf8
[11:56] Gavin.Hird I'd be happy to see that out of the Win viewer
[11:56] Ubit Umarov: but well we can't use core
[11:57] Ubit Umarov: trying to do that is still a wast eof time
[11:57] Ubit Umarov: they keep changin it
[11:57] Ubit Umarov: guess we should wait for .net5 release
[11:57] Ubit Umarov: then at least 6 months after that
[11:58] Andrew Hellershanks: Wow. This hour has flown by.
[11:58] Ubit Umarov: several opensim code whould just die on core 2.0
[11:58] Ubit Umarov: they removed a ton of things
[11:58] Andrew Hellershanks: Does anyone have a question for todays meeting before we start to wrap things up?
[11:58] Ubit Umarov: then seen how bad that was and did add some back on 3.x
[11:59] Ubit Umarov: milenium coders telling everyone should remake all applications :p
[11:59] Ubit Umarov: ( yes they said that on blogs.. old .net code would need to be converted )
[12:00] Ubit Umarov: that is .net4.8 code
[12:00] Gavin.Hird ha, try tell old COBOL guys that
[12:00] Ubit Umarov: or fortran :)
[12:00] Gavin.Hird yep
[12:00] Andrew Hellershanks: Gavin, :)
[12:01] Ubit Umarov: fact is that if oyu want fast math, you need to use fortran, even today
[12:01] Gavin.Hird that is correct Ubit
[12:01] Ubit Umarov: fast and acurate..
[12:01] Ubit Umarov: c++ and c# math libs are not that good
[12:02] Gavin.Hird the oil and gas industry would die without fortran - but hey, they want them to die anyway...
[12:02] Ubit Umarov: stock standard ones i mean
[12:02] Bill Blight: I need to bail, will leave the logger running
[12:02] Bill Blight: laters
[12:02] Ubit Umarov: well .net system numerics for vectors etc was just BAD
[12:02] Ubit Umarov: tons of math issues
[12:02] Andrew Hellershanks: ok, bill. tc.
[12:03] Ubit Umarov: well and we have the cpus
[12:03] Ubit Umarov: that intel fpu bug LOL
[12:03] Ubit Umarov: even using the windows calculator it gave worng results :)
[12:03] Gavin.Hird we all remember that
[12:03] Ubit Umarov: yeah hard to forget :)
[12:04] Ubit Umarov: but bugs are all around on modern cpus also
[12:04] Andrew Hellershanks: I rarely if ever used the Windows calculator.
[12:04] Gavin.Hird I think Excel had some spaced out calcs too for a while
[12:04] Andrew Hellershanks: I always had my HP-41C nearby.
[12:05] Ubit Umarov: hand calculator is 50cm way from mouse, so i use the win one :p
[12:05] Gavin.Hird I just aks Siri
[12:05] Kayaker Magic: I use the HP-42S emulator.
[12:05] Gavin.Hird ask*
[12:05] Andrew Hellershanks: I have an HP41 emulator on my computer and on my PDA. Both are within easy reach.
[12:06] Ubit Umarov: but that issue was not the calculator software.. was the cpu
[12:06] Ada Radius: Excel did have spaced out calcs. I was a CPA then, it was a big deal
[12:06] Ubit Umarov: the fpu was just broken
[12:07] Andrew Hellershanks: Any last minute OS related issues for today?
[12:07] Gavin.Hird LL released 2 viewers since the last meeting
[12:07] Andrew Hellershanks: 2?
[12:07] Gavin.Hird some good fixes in there that will find their way into my viewer
[12:07] Gavin.Hird yes 2
[12:07] Ubit Umarov: released the eep fixes?
[12:07] Gavin.Hird first a maint release, and the a love me render release
[12:08] Andrew Hellershanks: hehe
[12:08] Gavin.Hird yes, ysterday
[12:08] Ubit Umarov: wonder if our eep still works with those changes :)
[12:08] Gavin.Hird haha
[12:08] Gavin.Hird moving target
[12:08] Ubit Umarov: yeah
[12:09] Ubit Umarov: Second Life Release (64bit) is what i have
[12:10] Gavin.Hird right now they run a link for a new merch store on their MOTD, so might be strapped for cash now they can't sell new regions
[12:10] Gavin.Hird
[12:10] Gavin.Hird is the latest (love me render)
[12:10] Ubit Umarov: lets see if mine auto updates
[12:11] Leighton.Marjoram well I need to head off good meeting see you again next week
[12:11] Leighton.Marjoram ttfn
[12:11] Gavin.Hird tc Leighton
[12:11] Ubit Umarov: it did
[12:12] Gavin.Hird have fun Ubit
[12:12] Gavin.Hird wow
[12:12] Andrew Hellershanks: Bye, Leighton.
[12:12] Andrew Hellershanks: Well, this seems like the time to wrap up todays gathering.
[12:13] Andrew Hellershanks: Thank you all for coming. See you again next week.

Personal tools
About This Wiki