Revelator's TDM Branch

[Tels]

Is that a performance increase from the changes or from the change of location in the screen shot? Sorry to be a nag, but would it be possible to get a side by side of the same location, just for easier comparison? It's hard to get a feel from the different locations.

 

At least it proves it compiles and runs - let's ship it :D

[New Horizon]

At least it proves it compiles and runs - let's ship it :D

 

I had no doubts about it compiling.

[nbohr1more]

Yeah, it's a bit premature to be looking for major improvements yet. This has yet to reach parity with his revelation branch AFAIK.

Even there, we are still probably (mostly) bottle-necked by CPU skinning so my major outlook on this will be for framework improvements

(GLEW, boost removal, 64-bit executable, cruft removal, etc) It would help if these also worked on Linux though

Edited November 20, 2014 by nbohr1more

[Bikerdude]

blackbog hollow on darkmod with GLEW backend and GC garbadge collector running, all effects on.

original engine shot.

Its hard to see whats changed where because the second shot is from a different angle/location. And having the before & after shots in the same post would make it easier to see.

[Bikerdude]

let's ship it

I assume your just jesting, meaning we need to get it beta tested etc.

[Tels]

I had no doubts about it compiling.

 

Only on Windows

[revelator]

Allmost same location just a different angle was trying to avoid a guard while i took it hehe.

 

_aligned_malloc / _aligned_free do not exist on linux i found out :S but could make a replacement from my GC wrapper, lets see what i can come up with.

[revelator]

Ok created replacement functions for linux and other unix based OS that do not have these, changes have been comitted.

 

Cant exactly tell if theres a real FPS increase since the old version of the FPS counter puts out some wildly varying values compared to the new one, but as far as i could see the old fps counter variated between 47 to 63 fps so there might be some small increase measurable.

[SteveL]

The first s/shot has less in view than the other which makes a big difference to our FPS.

 

@all: Even if fps is the same so far, that's not a fail by the way as there hasn't been much optimization gone in yet. Improved fps would be a nice surprise, but still a surprise. I've only been half following revelator's progress because I'm scrambling to get more stuff into 2.03, but the parts I've seen so far are about preparing the way for us to make improvements later: bug fixes, cleanup of redundant code and an easier openGL interface. The pay-off will come later.

[revelator]

Ok as close as i could get but non conclusive as i waries a lot dependent on location,

 

Old TDM

 

New TDM

[revelator]

Toatally agree Steve this is just the groundwork

[revelator]

Just to blow it totally out the window heres a shot with 16x aniso 16x antialiasing adaptive vsync and all the gfx shebang i could enable.

 

 

Me card still likey but atleast now it shows some fps dropout.

[motorsep]

Lighting seems to be different. Normal maps look different.

[revelator]

Noticed that as well the old one seems a bit more grainy when looking at the wall textures, though im not sure if its caused by me not having the 100% same location when taking the shot.

[Tels]

Lighting seems to be different. Normal maps look different.

 

The two shots need to be taken from the exact same camera angle, and at the exact same timeframe (otherwise particles will look different).

[Tels]

Ok created replacement functions for linux and other unix based OS that do not have these, changes have been comitted.

 

Thanx. It progresses a lot further now:

 

 

 

scons: building `build/release/core/glimp/sys/gllog/gl_extensions.o' because it doesn't exist
g++ -o build/release/core/glimp/sys/gllog/gl_extensions.o -c -pipe -Wall -Wno-unknown-pragmas -fmessage-length=0 -fpermissive -fvisibility=hidden -m32 -O3 -march=pentium3 -ffast-math -fno-unsafe-math-optimizations -fomit-frame-pointer -DXTHREADS -fno-strict-aliasing -DGLIMP -Ibuild/release/core/glimp/sys/scons -Isys/scons -Iglimp/sys/scons -Iinclude -Iinclude/zlib -Iinclude/minizip -Iinclude/libjpeg -Iinclude/devil -I. -I. build/release/core/glimp/sys/gllog/gl_extensions.cpp
build/release/core/glimp/sys/gllog/gl_extensions.cpp:15: warning: missing braces around initializer for ‘glExtName_t’
build/release/core/glimp/sys/gllog/gl_extensions.cpp:19: error: ‘GLExtension_t’ does not name a type
build/release/core/glimp/sys/gllog/gl_extensions.cpp:17: warning: ‘void StubFunction()’ defined but not used
include/boost/system/error_code.hpp:221: warning: ‘boost::system::posix_category’ defined but not used
include/boost/system/error_code.hpp:222: warning: ‘boost::system::errno_ecat’ defined but not used
include/boost/system/error_code.hpp:223: warning: ‘boost::system::native_ecat’ defined but not used
scons: *** [build/release/core/glimp/sys/gllog/gl_extensions.o] Error 1
scons: building terminated because of errors.

 

 

Edited November 20, 2014 by Tels

[New Horizon]

Didn't mean to imply that I was expecting any performance boosts, I was simply curious if the shots were suggesting a minor increase.

[AluminumHaste]

Okay some issues, between SVN and experimental branch I get 30 fps in Lords and Legacy wheras in the SVN build I get 60 fps. Specifically at this spot:

 

 

There's also an issue with the sky, it pulses and doesn't look right.

 

[Bikerdude]

Ok as close as i could get but non conclusive as i waries a lot dependent on location,

As Ive not feeling this thread that closely, what am I looking at? whats better in the new version compared to the old...?

[AluminumHaste]

As for just regular scenes they look almost the same:

 

2.02 SVN:

 

 

Experimental branch

 

[nbohr1more]

Okay some issues, between SVN and experimental branch I get 30 fps in Lords and Legacy wheras in the SVN build I get 60 fps. Specifically at this spot:

 

 

There's also an issue with the sky, it pulses and doesn't look right.

 

http://www.youtube.com/watch?v=2wGfqBw2TqE

 

Did you test with:

 

r_useArbMapBufferRange

 

toggled 1 vs 0 ?

[revelator]

If you have an older AMD gfx card you should turn it off since it does not work to well with those 5x series and older atleast as far as i know.

Hmm have not touched anything in the skybox code curious ? not sure what causes that effect.

Experimental branch is mostly a cleanup project, making a solid code base from where the devs can start adding in the funny stuff

a few optimizations might also happen but mostly its to get rid of outdated cruft, correcting some mistakes in the code, and generally shining this thing up with a bit of spit and polish.

[revelator]

Ok removed some leftover code in gl_stub.cpp that is not needed anymore with GLEW, changes have been comitted, i hope it runs to the end now on linux

[AluminumHaste]

Using the video card in my sig, Nvidia 780ti latest drivers. I'll try the toggle.

[revelator]

Should be fine with that card as its sufficiently new but by all means try it out.