activate renderer/network isolation with host_maxfps >= 73

[j4reporting]

No description provided.

[vsonnier]

Hi @j4reporting thanks for the change. On my machine I don't see much of a difference compared to master in the Immortal Lock, but I don't see a loss of performance either, so we not taking it ? For the Immortal Lock though, host_maxfps 0 stay unplayable no matter the version.

[j4reporting]

then the cpu is stil able to handle the load fast enough (host_maxftp >=73 )
check with host_speeds 1 , it becomes a problem if time spend in server code exceeds 30. or 40. ? Then there is a significant drop in FPS.

There is no safeguard how to deal with this situations. The previous implementation took this in to account with the accumtime

this change makes my slow nuc playable with host_maxftp 73

diff --git a/Quake/host.c b/Quake/host.c
index 8ee60f59..7ccb0bd5 100644
--- a/Quake/host.c
+++ b/Quake/host.c
@@ -922,7 +922,8 @@ void _Host_Frame (double time)
                double realframetime = host_frametime;
                if (host_netinterval && isDedicated == 0)
                {
-                       host_frametime = sv.active ? (listening ? q_min (accumtime, 0.017) : host_netinterval) : accumtime;
+                       //host_frametime = sv.active ? (listening ? q_min (accumtime, 0.017) : host_netinterval) : accumtime;
+                       host_frametime = q_max(accumtime, host_netinterval);
                        accumtime -= host_frametime;
                        if (host_timescale.value > 0)
                                host_frametime *= host_timescale.value;

that's the old topic.

[vsonnier]

@j4reporting Indeed I see a big difference using host_frametime = q_max(accumtime, host_netinterval) : with this host_maxfps 0 is playable with the Immortal Lock save, and the fps seems now more or less constant whatever host_maxfps is 0, 58 , 72, 73 or whatever.

So, I've commited the change in 60957d1

[j4reporting]

this needs more tests though, Not sure if something else needs to be reverted too. test branch?
I prefer the strict timing, but sometimes it would be nice to have the option to switch to the q_max(accumtime, host_netinterval) version. This helps playing maps with extrremely high number of edicts and/or high wpoly count (r_speeds) on slower systems.