A Complete Evaluation of DarkPlaces

[LordHavoc]

also i noticed that certain bloom/corona effects show thru doors (or at least the secret door near the megahealth in e1m1)

The coronas through doors is a bug I observed today too, I'll fix that.

I really do need to implement occlusion query calls to hide coronas properly.

[Baker]

I was referring to testing on windows 98 a few years ago, and I kept acceleration disabled in windows.

Note that DarkPlaces disables acceleration (Windows XP broke the winquake/glquake code for disabling acceleration, but DP has it fixed).

I'll do homework/research on this. I can tell when DINPUT isn't on even in DarkPlaces. If I don't use DINPUT, I feel miserable.

It might just be mouse acceleration or something, but there is something very real that affects only my vertical aim.

In fact, about 2 years ago I switched from being a non-DINPUT user to a DINPUT user because I realized that DINPUT gives superior vertical aim.

It took me 3 weeks to adjust to DINPUT and was a miserable transitition.

It might just be acceleration, but I swear I got the aspect ratio correction from somewhere.

Then again ... maybe a myth is being busted in this thread.

[scar3crow]

LordHavoc - I dont think you broke anything to me, if it did do anything, its negligible, and obviously the difference I experienced was so minor it was just natural on my system. General variance of timedemoing.

[LordHavoc]

Interesting observation:
cl_particles_quake 1 (quake particle effects)
cl_particles_size 3 (3x normal size)
cl_beams_polygons 1 ('straight shaft')

looks like a very competitive gaming config in terms of seeing blood when you hit with weapons.

I think some serious testing to be done on cl_beams_instantaimhack 0/1 to evaluate whether it improves aim to be on, or hurts aim, I think it hurts aim to have it on but am not sure yet, it is off in the Effects: Quake preset, and I may end up turning it off in all presets if a few people test and confirm that it hurts aim.

Actually I think I'll disable cl_beams_instantaimhack in all presets and by default (this WILL affect existing configs that have it on - DP only saves cvars that differ from their default, so configs do adapt automatically to changing defaults in the engine), if a proper evaluation of its good/bad points is done it might be reenabled in the future, but otherwise it's probably better kept off.

[=peg=]

something else i noticed.. lavaballs cast shadows.. that is.. with realtime world lighting [on] lavaballs cast square looking shadows if struck by light from another lightsource.. IMO this looks a bit odd, as the lavaball itself cast a lot of light.. the square shadows dont look like the lavaball shape anyways.. so i guess it would look better if lavaballs dont cast shadows at all..

just thought i'd give some feedback

[LordHavoc]

something else i noticed.. lavaballs cast shadows.. that is.. with realtime world lighting [on] lavaballs cast square looking shadows if struck by light from another lightsource.. IMO this looks a bit odd, as the lavaball itself cast a lot of light.. the square shadows dont look like the lavaball shape anyways.. so i guess it would look better if lavaballs dont cast shadows at all..

just thought i'd give some feedback

DarkPlaces does not mess with content in any way, so to disable the blocky-looking lavaball shadows (the model IS blocky) the qc would have to be altered. The nails have pretty triangular looking shadows too - but so is the model

[Lardarse]

r_glsl 0 was 1-3fps faster in general gameplay on my gf4ti4200
i timedemoed demo1.dem with both, glsl 0 averaged 2 fps higher

If you timedemo twice without restarting, the second one will always be faster.

[LordHavoc]

If you timedemo twice without restarting, the second one will always be faster.

Unfortunately true, this is mainly because of the GLSL shaders being compiled on demand, and there can be some memory allocations during the timedemo as well.

The two solutions to this are to play the demo twice and take only the second result (unfortunately the -benchmark option does not do this), or to change the engine to precompile certain shaders that are likely to be used, and allocate larger memory buffers by default.

I need to do a memory profiling run sometime to figure out when the memory buffers are enlarged so I can find better defaults, and figure out which GLSL shader permutations are needed.

[Lardarse]

Restarting DP is easier and faster

And I don't see the point in doing something that will only improve benchmarks without actually increasing performance...

[LordHavoc]

And I don't see the point in doing something that will only improve benchmarks without actually increasing performance...

It avoids the hitching when the first dynamic light appears (and certain other cases like the first time a dynamic light hits a player model).