#mmz projects

1 messages · Page 1 of 1 (latest)

dawn moon
#

in case someone wants to give some advice

devout magnet
#

are you making your own vphysics module because you want to behavior to be 100% identical to vanilla vphysics?

#

rather than implementing a new physics engine like joshua's module does with jolt

dawn moon
#

correct

#

i want to be able to replace only the serverside dll instead of both. When you use a different physics engine than IVP, you need to do both server and client, or else, prediction will be screwed.

#

@devout magnet

#

With every step toward multithreading, it makes so that we dont need to rely on high frequency CPUs to power servers, but with high thread count, that means the capacity of everything increases, that is prop count, player count, entity count, etc...

#

And then for the client we just need to omit physics of far away props

devout magnet
#

off topic but does vjolt mess up prediction?

#

thats actually a really good point abt different engine fucking up prediction

dawn moon
#

i dont really know, but it cant possibly be the same, of course if you use both client and server the same physics then its fine, but its impractical to require all clients of a big server to replace the dll every time the game updates.

#

The only way it would be the same is if vjolt physics engine used the same equations for collision detection, constraints, gravity simulation, etc... Which is quite unlikely

#

@devout magnet

supple cipher
#

for testing on linux 64bits i need to put:
gmod_better_vphysics\projects\linux\gmake2\x86_64\Debug\vphysics.so

in:
/serverfiles/bin/linux64

or am i missing something ?

supple cipher
#

@dawn moon

dawn moon
#

Yes, that's correct, I'll check my Linux testserver later to check, but that should do it, btw there is a possibility of it launching in 32 bits, it needs to run 64 bits to catch the 64 bit library

supple cipher
# dawn moon Yes, that's correct, I'll check my Linux testserver later to check, but that sho... 
A wide merge error has occurred.  Qhull has produced a wide facet due to facet merges and vertex merges.
This usually occurs when the input is nearly degenerate and substantial merging has occurred.
See http://www.qhull.org/html/qh-impre.htm#limit
QH6380 Qhull precision error (qh_check_maxout): a facet or vertex merge produced a wide facet: v1 below f54 distance  0 (2797.3x). Trace with option 'TWn' to identify the merge.  Allow with 'Q12' (allow-wide)
ERRONEOUS FACET:
f54
flags: bottom simplicial
normal:    0.8714  -0.4906 -6.666e-05
offset:   74.79082
vertices: p256(v15) p272(v12) p170(v2)
neighboring facets: f172 f105 f110

While executing:  | qhull Qs QJ0.00144255 C-0 Pp W1e-14 E1.0e-18
Options selected for Qhull 2020.2 2021/09/27:
  run-id 825595115  Qsearch-initial-simplex  QJoggle 0.0014
  Centrum-premerge-  0  Pprecision-ignore  W-outside 1e-14
  Distance-roundoff 1e-18  _zero-centrum  _run 1  _joggle-seed 16807
  _max-width 46  _one-merge 7e-18  _near-inside 3.5e-17  Visible-distance 2e-18
  U-max-coplanar 2e-18  _wide-facet 1e-14  _narrow-hull 4.4e-09
  _maxoutside 2.2e-14
Last point added to hull was p34.

At error exit:

Convex hull of 289 points in 3-d:

  Number of vertices: 43
  Number of facets: 82

Statistics for:  | qhull Qs QJ0.00144255 C-0 Pp W1e-14 E1.0e-18

  Number of points processed: 51
  Number of hyperplanes created: 223
  Number of distance tests for qhull: 4020
  Input joggled by: 0.0014
  Maximum distance of point above facet: 0.0025
  Maximum distance of vertex below facet: -0.0025
#

I get spam with this error and then the server restart (Linux 64bits on sandbox with no addons)

dawn moon
supple cipher
dawn moon
#

Sure

dawn moon
#

I found some uncommited changes in the repo, i commited and pushed them, im compiling the linux version right and im going to test it right away

dawn moon
#

of course it errors it bruh undefined reference to _V_memmove

dawn moon
#

just use the commit b80501d2aefa471183d941a119a4d3c4487a4e12 for now, its before the SSE2 optimization, so it is slower than original, but it does prove that it works at all

#

Tested: win32, win64, linux32 and linux64

supple cipher
charred bobcat
#

@dawn moon, hey, is there any precompiled versions of VPhysics for linux64? I kinda can’t compile myself…

dawn moon
#

i can pass to you in the dms