Alien vs Predator Port

[mcobit]

Tested it and it is not any faster.

Also I had top running. It might actually be a mix of cpu and gpu limitation. I get some cpu time of 95% ingame, when there is a lot going on.

But also at the startingpoint of the predator, the game is slow even though it only uses about 60% cpu.

Both at 1000 MHz CC unit.

 

[Silent-Hunter]

OK, apparently this game doesn't work well on the 1GHz Pandora. The graphics are all glitched, I can see through closed doors, and some walls, the radar hud is all messed up. Oddly enough, the same problem occurs with DooM Legacy, and Minetest. Is it possibly because the 1GHz Pandora has a faster graphics chip, maybe it's causing a race condition?

 

[mcobit]

Same goes for DXRebirth apparently.

Might also be a driver issue. Notaz said, that the GHz units cannot get updated drivers or it will break compatibility with the older units.

 

[BAFelton]

I have graphic bugs on mine (1ghz) :

Tried with 1.0.0.1 version and 1.0.0.2

Low or high details, the bugs remain.

Edit : just seen silent-hunter post

 

[mcobit]

Are you playing on a GHz unit?

Edit: Is the speed better on the GHz units at least?

 

[Silent-Hunter]

Same goes for DXRebirth apparently.

Might also be a driver issue. Notaz said, that the GHz units cannot get updated drivers or it will break compatibility with the older units.

Why can't the drivers be updated for just the 1GHz units? Just post the updates, and tell 600Mhz users not to download them?

It's really bad that 1GHz users can't use certain games at all.

BAFelton's picture looks like he's probably having the same issue.

Perhaps tell where to get the updated drivers so we can install them manually?

 

[BAFelton]

Are you playing on a GHz unit?

Edit: Is the speed better on the GHz units at least?

Yes and yes.

 

[mcobit]

Oh, don't ask me...

Can you do a little video just for comaprison?

 

[Rockthesmurf]

Well, the highest profiling function is this:



__int64 aa = (__int64) a;
__int64 bb = (__int64) b;


__int64 cc = aa * bb;


return (int) ((cc >> 16) & 0xffffffff);

The algorithm is to take the mid 32 bits of the 64 bit result of a fixed point multiply. It is called a lot in a frame and accounts for almost 11% of the frame time. For x86 there is some assembly to do this:

int retval;
__asm__(
  "imull	%2			\n\t"
  "shrdl	$16, %%edx, %%eax	\n\t"
  : "=a" (retval)
  : "0" (a), "m" (
  : "%edx", "cc"
);
return retval;

I don't know what 64 bit performance is like on Pandora? On some platforms it is emulated, I haven't checked the generated ASM yet, but I wonder if one of the assembly gods (Exophase? Laurent? one of the many other talented developers) will know a nice optimization here. I would much rather be CPU than GPU limited as there are often a lot more ways to optimize here when there has been a quick port between architectures. FYI CheckForRayIntersectionWithObject is the next hit at around 10%, I *believe* nearly all of this comes down to MagnitudeOfCrossProduct (I am not familiar with the output of gprof yet so I may be wrong here). And MagnitudeOfCrossProduct is short, calling the above multiply function 6 times and then doing a 3 component magnitude (3 muls and a sqrt followed by a float -> int conversion).

 

[Rockthesmurf]

Do we know the actual cause of the graphical glitches on 1 GHz units? It seems plausible we could get to the bottom of it, strip out as much as possible whilst retaining the glitch. Would probably be helpful to get a 1 GHz unit or work with a developer that has one to track this down.

 

[Silent-Hunter]

Do we know the actual cause of the graphical glitches on 1 GHz units? It seems plausible we could get to the bottom of it, strip out as much as possible whilst retaining the glitch. Would probably be helpful to get a 1 GHz unit or work with a developer that has one to track this down.

It's not just your port. DooM Legacy and Minetest have the same problem.

 

[mcobit]

I am pretty sure this is related to older drivers. The GHz units have a newer SGX revision, but the newer drivers don't work on old units so notaz uses the older drivers, that work for all machines.

 

[Silent-Hunter]

I am pretty sure this is related to older drivers. The GHz units have a newer SGX revision, but the newer drivers don't work on old units so notaz uses the older drivers, that work for all machines.

But can't an optional package be provided to update the 1GHz ones? Then 1GHz users can download that to fix their driver problem.

 

[BAFelton]

Oh, don't ask me...

Can you do a little video just for comaprison?

coming here in a moment : 1.2ghz

https://www.youtube.com/embed/OBPvumfQqiE?feature=oembed

 

[mcobit]

From IRC:

<mcobit> That's true, but why not just use the older drivers for the old units and a newer one for the GHz units?
<aTc> and what makes you think that wouldn't happen

 

[notaz]

I am pretty sure this is related to older drivers. The GHz units have a newer SGX revision, but the newer drivers don't work on old units so notaz uses the older drivers, that work for all machines.

But can't an optional package be provided to update the 1GHz ones? Then 1GHz users can download that to fix their driver problem.

The main problem with this is that newer driver does't work at all on any units, but I'm trying to look into this..

 

[Silent-Hunter]

From IRC:

<mcobit> That's true, but why not just use the older drivers for the old units and a newer one for the GHz units?


<aTc> and what makes you think that wouldn't happen

Oh good! So I, and anyone else with a 1GHz unit, are waiting for an update then?

I am pretty sure this is related to older drivers. The GHz units have a newer SGX revision, but the newer drivers don't work on old units so notaz uses the older drivers, that work for all machines.

But can't an optional package be provided to update the 1GHz ones? Then 1GHz users can download that to fix their driver problem.

The main problem with this is that newer driver does't work at all on any units, but I'm trying to look into this..

Oh my! I hope you can fix it!

 

[Silent-Hunter]

Is there a way to detect what type of Pandora a person has? That way the OS Upgrader could only push upgrades appropriate for the Pandora it is being run on.

@Stephen Craft: Sorry for derailing your thread.

 

[Loonie]

This driver crap really needs to be sorted out. I understand not wanting to break drivers in the older units, but the alternative that the new units have to have broken drivers is ridiculous.

Edit: Okay, I had written this message very slowly and sporadically due to constant interruptions at work, and have now seen Notaz's response. My post's tone was rather unpleasant, and is not intended to disparage anyone's efforts. Also, I assisted in thread derailment. My apologies.   ...but I wanna play AvPeeeeeeeee!

 

[Silent-Hunter]

This driver crap really needs to be sorted out. I understand not wanting to break drivers in the older units, but the alternative that the new units have to have broken drivers is ridiculous.

They're prepaing different drivers for the 1GHz units. Right now, those drivers don't work. Notaz is investigating it.