[blowfish]

Neoterran, not septuple posting plz.  Just edit your posts to add new information

And I don't think the FPS weirdness is related to the issue at hand.  I've seen it too (it existed before the GLSL fix).

[neoterran]

Well it doesn't exist at all when no_glsl is turned on. And before, on macs, you if you had no_glsl off, then you had all white models in full screen and totally ****ty fps in window mode.

Sorry for the multiple posts. I made them over several hours of testing on my time. I was hoping it might be appreciated a little better. 

[blowfish]

Well it doesn't exist at all when no_glsl is turned on. And before, on macs, you if you had no_glsl off, then you had all white models in full screen and totally ****ty fps in window mode.

Sorry for the multiple posts. I made them over several hours of testing on my time. I was hoping it might be appreciated a little better. 

It doesn't exist when no_glsl is turned on because it has to do with shaders.

[neoterran]

ummm.... "And I don't think the FPS weirdness is related to the issue at hand."

isn't the issue at hand fixing the shader / normal mapping via shader support  ?

If it was just to fix the lighting and env mapping when no_glsl is turned on, then you already succeeded. That is definately better now.

However, when people are telling me "hey, that FPS drop you see - that's just because you turned the shaders on" - and those people are SCP team members - I think I should say, no, hey, there's another problem.

[blowfish]

Well, the issue at hand was fixing a specific bug causing graphics to not work properly in full-screen mode.  While there were shaders involved in the issue, it wasn't specific to shaded rendering.

[neoterran]

Okay, well then maybe we should open another Mantis entry that says "Shaders on OS X result in a completely unacceptable performance loss".

whatever. I'm through wasting any more of my time giving feedback on this.

[blowfish]

To be fair, shaded rendering does take up a lot more processing power than fixed rendering does...

[neoterran]

No. You don't lose close to 90 percent of your framerate enabling shaders on windows, and the fact you're still bringing that up makes me think that you're not even reading the thread. I posted all this info so that the team would have an empirical example of what a typical macbook user sees using the game. The fact that it's still being blown off (for the second time, I might add) as "you enabled the shaders, that's why there is a FPS loss" is just really making me mad, because it means I've totally wasted hours trying to document and expose the issue. 

I mean look at this : 110 fps


vs this : 2 fps (only difference : no_glsl off)


Anyway, as I said, my final post on the matter.

[chief1983]

I think he means the point of this particular issue is to make sure that nothing got worse with this new fix.  Your detailed reporting concerns issues with OS X that existed before this fix.  Don't worry, the feedback is appreciated but the problem isn't new to this patch so it shouldn't block it from being committed.  And, as long as -no_glsl works, OS X shouldn't be in too much worse shape than they were before the GLSL commit from July, which means this particular issue shouldn't be holding up 3.6.10.

[Tinman]

[...]you enabled the shaders, that's why there is a FPS loss" is just really making me mad, because it means I've totally wasted hours trying to document and expose the issue. 

Anyway, as I said, my final post on the matter.

neoterran, please calm down.

I understand your point completely, but believe me, all people working out problems and document them are working a lot of their spare time on this, like you. And it will take its time before most of the problems are solved - and there is no "wasted hours trying" to do something to help.

With this large project things have to be changed step by step.  We should be lucky that some developers have time to work on the MacOS X codebase, because as I see from the d/ls there are not so many Mac users around - this may change in the next years, i think. 
So please don't take the posts here to seriously...   

edit:
just played around on the F3 screen

I do not know if it is helpful, but i have 4 fps with the Ulysses and when I turn on "fixed render pipeline" in the render options the fps jump to 60 (which is the limit by vsync) and the cpu load is considerable lower than with "fixed render pipeline" disabled.

I have not the least inkling what "fixed render pipeline" is, though.

[taylor]

I do not know if it is helpful, but i have 4 fps with the Ulysses and when I turn on "fixed render pipeline" in the render options the fps jump to 60 (which is the limit by vsync) and the cpu load is considerable lower than with "fixed render pipeline" disabled.

I have not the least inkling what "fixed render pipeline" is, though.

"Fixed render pipeline" just means that you aren't using shaders (ie, -no_glsl).

The FPS problem, along with the CPU load difference simply indicates that at least some part of the shaders is running in software mode.  If it were all hardware then the speed would be about the same as -no_glsl.  This basically just indicates a driver problem, where the shaders are trying to use something that only has a software fallback.  I'm not sure what shader set you are using, my 1119 set or what is in the MediaVPs, but I would just suggest trying whichever set you aren't using now and seeing if it makes a difference.

[Tinman]

I'm not sure what shader set you are using, my 1119 set or what is in the MediaVPs, but I would just suggest trying whichever set you aren't using now and seeing if it makes a difference.

I use the mv_complete.vp of the 3610mediavps, usually.

All comments are for the Ulysses LOD 0 windowed (wim) mode and fullscreen (fsc) mode:

the shaders in mv_complete.vp show 4fps in F3 screen with MacOS 10.5.6 all the time with no change.

the 1119 shaders  show 7-8fps in F3 screen with MacOS 10.5.6 for about 40-60s , CPU load both cores 100%
               then the fps rises from 7 to 30fps(wim)/43fps(fsc), the cpu load is then decreasing to 80% for both CPUs
               then after about 20s the fps rises to 60, cpu load is then decreasing to 40-50% for both CPUs
               (this is showing again after leaving and reentering F3-screen)

the mv_complete.vp shaders  show 60fps in F3 screen with MacOS 10.4.11 immediately
the 1119 shaders  show 60fps in F3 screen with MacOS 10.4.11 immediately

[Androgeos Exeunt]

* Androgeos Exeunt tests r5020 rosebay on -mod mediavps without -window.

Hey! It works! I entered the techroom and it worked! Everything looks smooth!

* Androgeos Exeunt tests out his TSM-69.

All right, I've managed to get fourteen screenshots. Here they are:








Is that how it's supposed to look?


If you look closely at the Colossus, you see somebody's computer.


















If you look closely at the flak turret to the left, you can see a portion of somebody' computer.

Apart from someone's computer being reflected, everything else looks all right to me while I was flying about in that mission.

[chief1983]

That sounds like the environment mapping bug that blowfish was talking about.  Sounds like the next biggest problem with OS X graphics.  Only seems to  happen with shaders enabled.

[blowfish]

That sounds like the environment mapping bug that blowfish was talking about.  Sounds like the next biggest problem with OS X graphics.  Only seems to  happen with shaders enabled.

I've seen it both with and without shaders.

[chief1983]

Ok my bad then.

[chief1983]

This is committed now.  Will be in the next nightly r5028 or higher.

[squidsquasher]

Fantastic to seem normal maps working on Mac!!! 

But - I'm having the exact same symptoms of frame rate degredation as Neoterran - with the exact same computer (macbook pro 2.4GHz, early 2008). I'm running nightly build FS2_Open-20090106_r5034.

the shaders in mv_complete.vp show 4fps in F3 screen with MacOS 10.5.6 all the time with no change.

the 1119 shaders  show 7-8fps in F3 screen with MacOS 10.5.6 for about 40-60s , CPU load both cores 100%
               then the fps rises from 7 to 30fps(wim)/43fps(fsc), the cpu load is then decreasing to 80% for both CPUs
               then after about 20s the fps rises to 60, cpu load is then decreasing to 40-50% for both CPUs
               (this is showing again after leaving and reentering F3-screen)

the mv_complete.vp shaders  show 60fps in F3 screen with MacOS 10.4.11 immediately
the 1119 shaders  show 60fps in F3 screen with MacOS 10.4.11 immediately

Does that mean it is an OS X 10.5 driver issue maybe? In my opinion apple has been getting a little more sloppy lately, so it wouldn't surprise me. I could test on my old mac that runs 10.4.11, but it is a lot slower (1.25GHz single G4).

[Tinman]

looks like a driver issue on MacOS X 10.5.x

[Echelon9]

I'll try to take a look at the performance impact on Mac over the next few weeks. Just to be clear, yes I do feel there is an issue when enabling shaders causes a drop in FPS to low 10s while enabling shaders in Windows on the same hardware only leads to a marginal drop in FPS.

This issue is very unlikely to have been caused by the changes in revision 5028 that stopped white models appearing on Mac fullscreen, merely we are noticing it now because running in windowed mode with shaders (which was the only option prior to r5028) was always going to see a performance hit with the desktop chrome still being rendered and OpenGL not receiving dedicated access to the hardware.

In fact, just thinking about it then, these performance problems and the odd issue with the reflectivity on ships actually reflecting your desktop (Finder window, browser, iTunes) rather than space around Capella could be related...

In any case, I'll be using a new thread and a new release should I make headway with this.