Thanks Komat. I appreciate your effort developing this wrapper.

I tested one more level: Indonesian Free TV(aka TV station).

Unfortunately this level is a tough one. The only thing that appears to work is the shadows.

Right from the start , just after you drop from the ledge to the ground there is a light poll in front of you. If you want to aim at it(center aim at the light - red dot active) you will see that it dissapears. Move the dot out of an invisble area the dot appears. It's like 'Cloak field' effect .😀
See here the poll in question :http://imgur.com/Krimyrb

Staying right there also if you move the view slowly upwards you'll see that the at one point it doesnt rain any more. No more rain drops on the character. At one point you'll start to see rain drops but they fall only till the middle of the screen. This is similar to the reflection problem from the first level with water(looking at it from a certain angle ).
See here 'middle rain': http://imgur.com/WlEHB1L and here no rain at all : http://imgur.com/qwbixJ6

It seems also that the enviroment is not lit in certain areas. For example the garden or just at the end of the level. Also no beams from the main screen inside the TV station.
In the sewer area reflections in the water is missing. Water is opaque. Enviroment is not lit.

See here:
http://imgur.com/DkKWa3x
http://imgur.com/M7vVzI7
http://imgur.com/wwrQ3Ex

Well i think it's enough for now. I' am sure that the same problems will apear in the next levels also.
When a new version comes out i will try to do a regresion test to see if everthing is ok.

Here is a link with the first level where artefact near the ladder happens,at the begging of the level, doesnt exist here:
https://www.youtube.com/watch?v=5HDnUxSfFsI
At leat i cant see it on this monitor.

PS: the author is also part of this forum 😀

Edit1: I've said to myself a few minutes ago: from where that light might come from? So, i open up the game and had a quick testing session.

Indeed while you walk towards the problem you can spot on the right a light source. If you shoot that light source, guess what , the problem dissapears. You can move through there and everything seems ok.
So, i believe that, there is indeed a bug near the ladder. if you get close to the ladder and move your view upwards you can see that the ceiling is lit.
Another argument here is that you can see this bug on earlier versions like PS3 HD. And this means that older drivers fail to render this.(double fail - one nvidia one ubisoft)

It appears that this game has some rendering fails(on modern hardware) if you look upwards 😀. See water reflection, raining bug and others.

I am still fighting with some fixes.

The missing beams and some lights were just case of another matrix which it did not detect properly.

Edit: The color problem seems to be related to precision of the color buffer. The beams are created by many additive layers so even 1/256 delta in one color component will after many additions accumulate to visible difference. Older cards did support HW dithering although I am not sure that it was applicable to 32 bit mode.

The game also has some bug and it is sometimes selecting incorrect lights when generating shadowmaps. I noticed the problem when debugging the colored lighs. That light is spot light so the shadowmap should be always generated from the same location however depending on angle of the view, when generating the shadowmap for rendering of the beams, the game sometimes uses projection from entirely different place. When it tries to use it, it will provide proper light as well as when rendering shadowmap for the floor. The strange thing is that it does not happen with the period-correct HW so maybe it is some uninitialized variable or something else depending on the capabilities features reported by the GPU. Similar problem likely happens on the wall to which the walkway with blue beams leads. Entire wall becomes lit or shadowed depending on where you look.

Here are some videos of the first 4 mission played on Windows XP and a GeForce 7 card:

https://www.youtube.com/playlist?list=PL5T8bm … XdG4_AERAQuJujx

Might be helpful. I'm afraid I don't have footage of any later levels. I did play the jungle level though but other projects pulled me away from playing more Splinter Cell 😒

Komat wrote:

I am still fighting with some fixes. […]
Show full quote

I am still fighting with some fixes.

The missing beams and some lights were just case of another matrix which it did not detect properly.

Edit: The color problem seems to be related to precision of the color buffer. The beams are created by many additive layers so even 1/256 delta in one color component will after many additions accumulate to visible difference. Older cards did support HW dithering although I am not sure that it was applicable to 32 bit mode.

The game also has some bug and it is sometimes selecting incorrect lights when generating shadowmaps. I noticed the problem when debugging the colored lighs. That light is spot light so the shadowmap should be always generated from the same location however depending on angle of the view, when generating the shadowmap for rendering of the beams, the game sometimes uses projection from entirely different place. When it tries to use it, it will provide proper light as well as when rendering shadowmap for the floor. The strange thing is that it does not happen with the period-correct HW so maybe it is some uninitialized variable or something else depending on the capabilities features reported by the GPU. Similar problem likely happens on the wall to which the walkway with blue beams leads. Entire wall becomes lit or shadowed depending on where you look.

Ubisoft coding at it's 'best' this game is 😀
Komat, take your time with these bugs. When you think you have an update for us i will gladly test and report back.

I dont expect to be easy for you and i think that's why Ubisoft doesnt bother with this game. From a programmer point of view this game could be a mess. Time and effort for them could mean a lot of money that would not generate profit
I am a programmer myself but in another field so i kind of understand the complexity and time required for debugging and applying a solution without breaking something else.

PS: really love your explanations. Tells a lot about a person and his way of thinking. 😉

Edit:

I've done a small test. I installed my 7950gx 2(in place of my 680 SLI) on Win 7 + your wrapper and your guess is rigth about something related to the video card.The artifacts dont happen.
So what the diffrence may be?. Can it be something specific to DX 8, like the shadow problem?!

daniel_u wrote:

I've done a small test. I installed my 7950gx 2(in place of my 680 SLI) on Win 7 + your wrapper and your guess is rigth about something related to the video card.The artifacts dont happen.
So what the diffrence may be?. Can it be something specific to DX 8, like the shadow problem?!

It could be some feature which is reported differently by those cards however I do not know what it could be. Using Apitrace it is possible to record traces of both cards and examine what device capabilities they report. You might need a modified version of Apitrace wrapper dll (attached) because SC is using a API feature which the default wrapper can not handle (namely use of CreateVertexShader to create a fixed function vertex shader by specifying NULL) and will crash.

Another possibility is that the driver changes some setting which affects calculations done by the game. E.g. FPU precision, rounding mode while executing path for some GPU. I will try to check for this this weekend.

The worst case scenario is that memory layout of something changes and the game is accessing some uninitialized variable somewhere. There is indication that something like that might be happening. When I use Dr.Memory tool, it will report uninitialized access when my wrapper is examining some matrix. This might indicate that some part of the calculation used to calculate that matrix might use something uninitialized. Now the problem is that by default the tool only checks the uninitialized state of variables on comparisons and other similar operations to avoid false positives when objects with uninitialized padding are copied so the report does not pinpoint the location where the problem started. While the tool has option to enable reporting of any such access, trying to use it will crash the tool so I will have to find another way to trace that (or fix the crash). I tried to force clears of newly allocated memory or portion of stack before each ret instruction however it did not change anything.

I am currently planing to try to record all memory accesses done to render single frame and hopefully I will find the problematic one that way. It is still possible that this is not related to the problem we are looking for.

I love these reverse-badass-engineer posts, you're hacking the very core of the game 😁

Speaking of..... https://www.eff.org/deeplinks/2015/04/videoga … ms-and-archives

Keep on "hacking"!

DosFreak wrote:

Speaking of..... https://www.eff.org/deeplinks/2015/04/videoga … ms-and-archives

Keep on "hacking"!

"for those who modify games to keep them working after the servers they need are shut down".
No game files are modfied here. Not even one !

Also American laws most of the time are plain stupid and made for greeady corporations.

@Komat I will send you the files later this day.

DosFreak wrote:

Speaking of..... https://www.eff.org/deeplinks/2015/04/videoga … ms-and-archives

Keep on "hacking"!

Oh man. Are they ever gonna wake up? 🙄

F2bnp wrote:

DosFreak wrote:

Speaking of..... https://www.eff.org/deeplinks/2015/04/videoga … ms-and-archives

Keep on "hacking"!

Oh man. Are they ever gonna wake up? 🙄

Disturbing, really? 😁

Here you go :http://www.megafileupload.com/2AS3/SplinterCell2.rar

Also while checking the wrapper.log i saw there multiple entries like this:

Call failed 0x88760868.

The archived attached it an rar file. I used max compression with dictionary 1024MB , and rar format RAR5.

The trace files are rather big, i made the mistake completing the first level. The file has whooping 9 GB. 😀 If it's case i'll upload that.😀

Best of luck.

daniel_u wrote:

Call failed 0x88760868.

This error is normal when the application is loosing focus.

I examined the trace. It is strange. While it reports slightly different capabilities, namely some limits are lower, the difference is not too big and emulating the same reply did not have any effect. Another interesting thing is that according to the caps it enables SW processing of the vertex data (unlike my trace) however simultaneously it does not use another flags it is supposed to provide to some operations when operating in the SW mode. Also when I try to replay the trace, it fails at one point.

I tried to enable the SW mode however some shaders fail to be created in that case. I do not know why yet.

The capture contains some overlay information. What SW is generating it? It is possible that it is affecting the rendering in some way.

The capture contains some overlay information. What SW is generating it? It is possible that it is affecting the rendering in some way.

I've been using MSI Afterburner. Hmm, i was not aware it will affect the results. Sorry. 🙁
I'll try again this night. Also, do you want a longer trace?

Here's another trace without MSI Afterburner:

http://www.megafileupload.com/V7t/SplinterCell2.rar

Hope this helps.

Edit:
I forgot to mention. The api tracer wrapper does not quite work . I mean it's like the wrapper it is not there. The game runs but i dont have the 'features' developed in previous wrapper releases.

daniel_u wrote:

I forgot to mention. The api tracer wrapper does not quite work . I mean it's like the wrapper it is not there. The game runs but i dont have the 'features' developed in previous wrapper releases.

That is expected, that tracer dll is from a third party application aimed at recording API calls and does not contain my tweaks. The version 0.05 I just added on my web can chain to another version of d3d8.dll if that other version is named d3d8_trace.dll. Now with the afterburner disabled the trace uses hw vertex processing as expected and I can play it back.

When you were playing it with the tracer wrapper, did you see the shimmering light artifact near the ladder or was it gone?

EDIT: I added version 0.05 with fixes to the light beams (I think that the beams are still more intensive that they should be) and more shadows visible in the Jerusalem level. I am still fighting with some shadow artifacts so I did not try to fix the other issues yet.

When you were playing it with the tracer wrapper, did you see the shimmering light artifact near the ladder or was it gone?

I did not see the light artifact. The enviroment was not lit.

(I think that the beams are still more intensive that they should be) and more shadows visible in the Jerusalem level.

Yes, your observation is corect.
The beams are somewhat more intensive but i also saw other differences(this bug i was holding it off for later posting) , like the pattern and form. Hard to put it into english. I will try to add some pictures for comparisons.
Also i saw some more shadows in the first level.
In Jerusulem level, I noticed with the 7950gx2 that some lights are colored. Some are all yellow , others have beems with different colors like blue, green, white.

Take your time Komat, as i said, i dont think it you'll have a easy time with this game. For now i will stop adding new bugs until the lighting/shadow artifacts are solved, hopefully.

Now i'am recording a walktrough for every level of the game with 7950gx2 so i can compare against an original source.

I've been comparing the wrapper with my recorded levels and i have to say that it looks quite similar. For the untrained eye could be perfect.
Let me give you my feedback.(i will not post specific bugs now)

Every level played with your wrapper has some(mostly minor) problems. They can be split into 2 types of problems:
(i do not count the issue you are working on)
-Extra shadows and lights on various levels.(some very hard to spot/recognize)
-Beams not displaying right. Missing or extra colors, spreading pattern changes with mouse movement and form do not match. Some are barely visible others are more intesive.

Some levels that have beems are perfect. No differences.
But the beams are there and this is important.

Funny that some problems happen on the old card also. 😀)

Also after the bugs have been ironed out i have a few propositions to make as improvements like: (sky, fire, some extra shadows and beams,infrared mode enhancement).

Cheers.

Edit: I've been comparing my first recorded level with PS3 HD edition. Ohh man , you should see the difference. You can consider the PC version + wrapper top notch. The PS3 adds even more extra shadows in the wrong place and some are missing.

What puzlles me is that some rendering errors are the same as generated by the wrapper.

daniel_u wrote:

(i do not count the issue you are working on)
-Extra shadows and lights on various levels.(some very hard to spot/recognize)
-Beams not displaying right. Missing or extra colors, spreading pattern changes with mouse movement and form do not match. Some are barely visible others are more intesive.

Those things likely have the same cause. For some reason the game is sometimes using a different lights that it should. The situations when the beams change patterns or the light/shadow disappears happen when there are several shadowed lights on the scene and the game selects a wrong one for that piece of geometry.

daniel_u wrote:

What puzlles me is that some rendering errors are the same as generated by the wrapper.

Because they are not generated by the wrapper, they are generated by the game.

I found several errors with uninitialized memory however none from them was apparently causing it.