Send to clients only changed node metadata instad of whole mapblock #3166
Conversation
| @@ -620,6 +622,12 @@ enum ToClientCommand | |||
| u32 id | |||
| */ | |||
|
|
|||
| TOCLIENT_NODEMETA_CHANGED = 0x54, | |||
| /* | |||
| u16 command | |||
There was a problem hiding this comment.
nrz wants to not document the command field, as it exists in any packet (there is no extra information).
There was a problem hiding this comment.
@est31 this is exact, command field is not useful information there
|
You'll have to adjust how the p param of MEET_BLOCK_NODE_METADATA_CHANGED is handled (single position instead of blockpos) everywhere it is used (including rollback_interface.cpp). |
|
Most of the PR looks good, except of things I've already pointed out above, but generally we should be very cautious with applying this. @kwolekr can you review this PR? |
|
|
||
| MapBlock *block = m_env->getMap().getBlockNoCreateNoEx(blockpos); | ||
| if(block) | ||
| block->raiseModified(MOD_STATE_WRITE_NEEDED); |
There was a problem hiding this comment.
same here, we have to take care of storage of the modified block in the map database. We can spare sending it, thats fine, but we shouldn't avoid storing them into the map.
There was a problem hiding this comment.
Done in Server::sendMetadataChanged
|
IMHO I dont have to comment using p parameter as single node coords for this event, it means it for each other case already, MEET_BLOCK_NODE_METADATA_CHANGED was an exception to that rule so it was documented. |
|
Okay, i see you call |
Not all others, all new and modified packets have the command not mentioned. |
Now situation changed: 64a5eec |
RealBadAngel
force-pushed
the
meta
branch
2 times, most recently
from
692173e
to
c3b43dd
Compare
|
Ive removed "command" from packet desc |
|
Provides a small improvement for me - a few fps better than before anyway. Motion seems smoother, less jerky. |
| std::istringstream iss(oss.str(), std::ios::binary); | ||
| meta->deSerialize(iss); | ||
| try { | ||
| m_env.getMap().setNodeMetadata(p, meta); |
There was a problem hiding this comment.
if this return false you MUST delete meta pointer
|
@VanessaE and for info to all others: she really has problem with drivers, with her decent HW setup she can barely hit 40 fps.... |
|
@RealBadAngel Then why do her and I both get way better frame rates with other games, even ones with much fancier graphics? |
|
Just wanting to point out that the bug with this PR that you get all metadata changes on the whole server, not just the ones you are close to still remains unfixed. |
|
@Fixer-007 It required some manual changes to merge it with upstream. |
|
Thank you very much! |
|
continued in #3775 |
thx to sending single meta packet is very small, circa 200-300 bytes for a chest, and it doesnt cause mapblock mesh updates (not only origin block mesh was updated but also neighbours).
Also sending whole block was causing each meta for that block being deleted and restored.
Effect on VE server, known to be laggy as hell and with usually slidshow (~10 fps as on my box):