Zandronum Chat on our Discord Server Get the latest version: 3.2
Source Code
My View | View Issues | Change Log | Roadmap | Zandronum Issue Support Ranking | Rules | My Account
  

View Issue Details [ Jump to Notes ] [ Issue History ] [ Print ]
IDProjectCategoryView StatusDate SubmittedLast Update
0001956Zandronum[All Projects] Bugpublic2014-10-11 20:202018-09-30 23:36
ReporterIjon Tichy 
Assigned ToDusk 
PrioritynormalSeverityminorReproducibilityalways
StatusclosedResolutionfixed 
PlatformOSOS Version
Product Version1.3 
Target Version1.4Fixed in Version1.4 
Summary0001956: Clientside ACS calls are bulkier than one would expect bandwidth-wise
DescriptionFor whatever reason, two clientside ACS calls per tic sends 10KB/s worth of bandwidth. For tracer scripts, which must have the tracer drawing clientsided if they don't want to disconnect everyone online, this is not desirable for any large amount of pellets (like, say, a minigun with realistic fire rates).

If there's any way to minimize the size of these calls, it should be pursued so that weaker connections (people still use dial-up!) can handle online play more easily.
Steps To Reproduce1) Start local testing server with scriptlag.pk3, and cheats on.
2) Open up some network monitoring tool. I used iftop.
3) Join local testing server.
4) give "Pistol Beam"
5) Hold down +attack with the Pistol Beam out.
6) Note the sudden 8KB/s jump.
Additional InformationI adapted Metroid: Dreadnought's Chroma Storm for this. Hence the FABULOUSNESS.
Attached Files? file icon scriptlag.pk3 [^] (247,126 bytes) 2014-10-11 20:20

- Relationships

-  Notes
User avatar (0010510)
Dusk (developer)
2014-10-11 20:31
edited on: 2014-10-12 01:28

 Hmm. The arguments are sent as three longs and the map name is sent as a string. We could probably cut down on those.
User avatar (0010511)
Watermelon (developer)
2014-10-11 21:14
edited on: 2014-10-12 01:49

 Interesting its sending the map name, that should be at most a single byte I'm guessing.
User avatar (0010519)
Dusk (developer)
2014-10-12 01:36
edited on: 2014-10-12 01:37

 To cut down on the args, we could send a byte header for them. For each of the three args, let (header >> (2 * arg)) & 8) be 0 if the arg is not sent, 1 if sent as byte, 2 if sent as short and 3 if sent as long.

In the best case this chops off 11 bytes (a common case where args are all 0), at worst it adds one byte if all 3 args are sent and are at least 2^16 (AFAIK a very rare corner case).
User avatar (0010521)
Watermelon (developer)
2014-10-12 01:55
edited on: 2014-10-12 02:01

 I was thinking about what you said there, and I have an idea:

void SERVERCOMMANDS_ACSScriptExecute( ULONG ulScript, AActor *pActivator, LONG lLineIdx, char *pszMap, bool bBackSide, int iArg0, int iArg1, int iArg2, bool bAlways, ULONG ulPlayerExtra, ULONG ulFlags )


Using what you said, arg1/arg2/arg3 can be cut down to 2 bits. Since it's weird to send a byte/short to represent 24 bits, here's a possible idea:

0x00 = no args (don't read a byte for this argument)
0x01 = arg is a byte
0x02 = arg is a short
0x03 = arg is a long

This would require 6 bits total (2 bits per 3 args).

We can also bundle the two booleans into 1 bit each. This is 2 more bits.

Therefore we can put all the header information in 1 byte.



The map can be a byte and reference the map index number, so that saves probably on average 4-5 bytes.


Net inflate: 2 bytes
Net deflate: approx 8-17 bytes (probably 17 a lot when sending 0/0/0 args).

Sounds like a good gain! This should lead to most commands being 8-9 bytes in size, rather than 26-27 that they are now.
User avatar (0010527)
Dusk (developer)
2014-10-12 12:59
Quote

0x00 = no args (don't read a byte for this argument)
0x01 = arg is a byte
0x02 = arg is a short
0x03 = arg is a long

This would require 6 bits total (2 bits per 3 args).


This is exactly what I said in my post. Here's what I have in mind:'http://pastebin.com/kKBspCh8 [^]'

In the best case this would be 9 bytes total should the map be sent as a byte index.
User avatar (0010561)
Dusk (developer)
2014-10-12 23:41

 'https://bitbucket.org/Torr_Samaho/zandronum-stable/pull-request/78 [^]'

Note that this will cause a merge conflict when merging to 2.0. Just overwrite SERVERCOMMANDS_ACSScriptExecute with this new one.
User avatar (0010572)
Torr Samaho (administrator)
2014-10-13 19:10

 The "Beam Pistol" example from'https://zandronum.com/tracker/view.php?id=1956 [^]' doesn't work properly anymore if I apply your patch. For instance, if I use the Beam Pistol on Doom2 MAP01 from the player 1 start the beam doesn't originate from the player body. Did you test this example?
User avatar (0010631)
Dusk (developer)
2014-10-19 09:44

 'https://bitbucket.org/Torr_Samaho/zandronum-stable/pull-request/86 [^]'

Yeah I forgot to test the beam pistol with this. Problem was that the code didn't take negative numbers into account.
User avatar (0010634)
Torr Samaho (administrator)
2014-10-19 16:07
edited on: 2014-10-19 16:14

 The updated patch seems to work with the example wad now, but it doesn't solve the problem: The net traffic is still very high when firing the pistol. I'll have a closer look at what's going on.

EDIT: The influence of SVC_ACSSCRIPTEXECUTE is negligible for the example wad. SVC_GIVEINVENTORY is eating all the bandwidth.
User avatar (0010636)
cobalt (updater)
2014-10-19 19:39

 Issue addressed by commit ce30242c5baf: - significantly optimized the SVC_ACSSCRIPTEXECUTE packet, making server-to-client ACS script calls not use up as much bandwidth (fixes 1956)
Committed by Teemu Piippo [Dusk] on Sunday 19 October 2014 12:41:49

Changes in files:
 docs/zandronum-history.txt | 1 +
 src/cl_main.cpp | 74 +++++++++++++++++++++++++-----------
 src/p_acs.cpp | 2 +-
 src/p_lnspec.cpp | 50 ++++++++----------------
 src/sv_commands.cpp | 92 +++++++++++++++++++++++++++++----------------
 src/sv_commands.h | 2 +-
 6 files changed, 130 insertions(+), 91 deletions(-)
User avatar (0011835)
Dusk (developer)
2015-03-16 01:28

 Since this modified a rather integral CLIENTSIDE feature, if there were any problems, surely we would've heard of them by now..

Issue Community Support
This issue is already marked as resolved.
If you feel that is not the case, please reopen it and explain why.
Supporters: No one explicitly supports this issue yet.
Opponents: No one explicitly opposes this issue yet.

- Issue History
Date Modified Username Field Change
2014-10-11 20:20 Ijon Tichy New Issue
2014-10-11 20:20 Ijon Tichy File Added: scriptlag.pk3
2014-10-11 20:31 Dusk Note Added: 0010510
2014-10-11 21:14 Watermelon Note Added: 0010511
2014-10-12 01:28 Dusk Note Edited: 0010510 View Revisions
2014-10-12 01:36 Dusk Note Added: 0010519
2014-10-12 01:37 Dusk Note Edited: 0010519 View Revisions
2014-10-12 01:37 Dusk Note Edited: 0010519 View Revisions
2014-10-12 01:47 Watermelon Note Edited: 0010511 View Revisions
2014-10-12 01:48 Watermelon Note Edited: 0010511 View Revisions
2014-10-12 01:49 Watermelon Note Edited: 0010511 View Revisions
2014-10-12 01:49 Watermelon Note Edited: 0010511 View Revisions
2014-10-12 01:55 Watermelon Note Added: 0010521
2014-10-12 01:56 Watermelon Note Edited: 0010521 View Revisions
2014-10-12 02:00 Watermelon Note Edited: 0010521 View Revisions
2014-10-12 02:00 Watermelon Note Edited: 0010521 View Revisions
2014-10-12 02:01 Watermelon Note Edited: 0010521 View Revisions
2014-10-12 12:59 Dusk Note Added: 0010527
2014-10-12 19:54 Dusk Assigned To => Dusk
2014-10-12 19:54 Dusk Status new => assigned
2014-10-12 23:41 Dusk Note Added: 0010561
2014-10-12 23:41 Dusk Status assigned => needs review
2014-10-13 19:10 Torr Samaho Note Added: 0010572
2014-10-13 19:14 Torr Samaho Status needs review => feedback
2014-10-19 08:42 Dusk Status feedback => assigned
2014-10-19 09:44 Dusk Note Added: 0010631
2014-10-19 09:44 Dusk Status assigned => needs review
2014-10-19 16:07 Torr Samaho Note Added: 0010634
2014-10-19 16:14 Torr Samaho Note Edited: 0010634 View Revisions
2014-10-19 19:39 cobalt Status needs review => needs testing
2014-10-19 19:39 cobalt Target Version => 1.4
2014-10-19 19:39 cobalt Description Updated View Revisions
2014-10-19 19:39 cobalt Steps to Reproduce Updated View Revisions
2014-10-19 19:39 cobalt Note Added: 0010636
2015-03-16 01:28 Dusk Note Added: 0011835
2015-03-16 01:28 Dusk Status needs testing => resolved
2015-03-16 01:28 Dusk Fixed in Version => 1.4
2015-03-16 01:28 Dusk Resolution open => fixed
2018-09-30 23:36 Blzut3 Status resolved => closed




Questions or other issues? Contact Us.

Links

Copyright © 2000 - 2025 MantisBT Team
Powered by Mantis Bugtracker