r5368 introduces transparency bug in .hack//G.U. series (NTSC-U) [SLUS-21258][SLUS-21488][SLUS-21489]

[GoogleCodeExporter]

Copied from Forums

"# PCSX2 version: r5368+
# CPU options: "Safest" preset
# Plugins used: GSdx 5365, 5390 (SSE2, hardware and software, dx9, dx11), 
SPU2-X r5317, r5384 2.0.0
# Description: Some textures became transparent, some polygons became plain 
color-filled. Um... Just look at the screenshotsBlush

http://img708.imageshack.us/img708/4173/gsdx20120823083615.th.jpg
http://img687.imageshack.us/img687/7848/gsdx20120823083520.th.jpg
http://img850.imageshack.us/img850/4328/gsdx20120823083507.th.jpg
http://img521.imageshack.us/img521/1479/gsdx20120823083457.th.jpg
http://img191.imageshack.us/img191/192/gsdx20120823083429.th.jpg

* The first screenshot is made with software mode.

It seems like something wrong with emulation, because the bug appears when you 
start pcsx2.exe of revision higher (or equal) than 5368. No matter what mode 
(sw, hw) or plugins you use. Running earlier revisions fix this.

If it'll help I post my pretty ancient configuration:
AMD Athlon II X2 250 3.0GHz, 2.0GB RAM, Sapphire ATI Radeon HD5570 (Win7x64)"

Confirmed on my Geforce / Intel system and logged this issue on the users behalf

Original issue reported on code.google.com by danialho...@gmail.com on 22 Aug 2012 at 10:20

[GoogleCodeExporter]

you really suck at linking to images from the forum, you have to blow them up 
before taking the link ;p

anyway, assigning to me as i killed it.

Original comment by refraction on 23 Aug 2012 at 12:15

• Changed state: Accepted

[GoogleCodeExporter]

people should stop using imagecrap

blowing up images to full screen has been broken for yonks

Original comment by danialho...@gmail.com on 23 Aug 2012 at 8:04

[GoogleCodeExporter]

Original comment by refraction on 23 Aug 2012 at 10:02

• Changed state: New
• Added labels: AMDIssue

[GoogleCodeExporter]

Assigning to Sudo for now, hes good with these AMD only issues.  (Works fine on 
Intel i7 cpu in all revisions up to current)

Original comment by refraction on 23 Aug 2012 at 10:04

[GoogleCodeExporter]

...you're either wrong about the revision it's introduced in or ref's wrong 
about it not happening for him.  And it's certainly not an AMD CPU issue, while 
it MIGHT be an ATI graphics card issue (but probably isn't).

Original comment by sudonim1 on 23 Aug 2012 at 10:49

• Removed labels: AMDIssue

[GoogleCodeExporter]

And regardless of who's wrong about what, make some screenshots which aren't 
postage stamp sized.

Original comment by sudonim1 on 23 Aug 2012 at 10:50

[GoogleCodeExporter]

http://img708.imageshack.us/img708/4173/gsdx20120823083615.jpg
http://img687.imageshack.us/img687/7848/gsdx20120823083520.jpg
http://img850.imageshack.us/img850/4328/gsdx20120823083507.jpg
http://img521.imageshack.us/img521/1479/gsdx20120823083457.jpg
http://img191.imageshack.us/img191/192/gsdx20120823083429.jpg

remove the .th

its that simple

Original comment by danialho...@gmail.com on 23 Aug 2012 at 11:14

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

"And it's certainly not an AMD CPU issue"

I don't put much faith in the word certainly where AMD is concerned

AMD GPU Only bug = 
http://forums.pcsx2.net/Thread-purple-windows-underside-cars-Burnout-revenge-Bur
nout-3-graphic-texture-glitch

GSDX SSE2 affected only on Radeon display drivers.

Original comment by danialho...@gmail.com on 23 Aug 2012 at 11:31

[GoogleCodeExporter]

Are you an idiot? GPUs != CPUs.  CPU bugs are few and far between.

Original comment by sudonim1 on 24 Aug 2012 at 12:36

[GoogleCodeExporter]

Lets not start name calling, it just leads to stupid flamewars where someone 
gets called a biggot and then some foreign country gets invaded for their 
mineral deposits -_-.

In any case, 5367 FINE, 5368 NOT. AMD Affected, Intel Not. (Atleast, Pre 
Bulldozer AMD)

I won't have access to a bulldozer core for atleast the next 4 weeks

Original comment by danialho...@gmail.com on 24 Aug 2012 at 1:04

[GoogleCodeExporter]

pcsx2-r5366.exe is the executable in 5367, this version has no problems with 
the GSDX from the latest svn build.

Replace pcsx2-r5366.exe with pcsx2-r5368.exe from 5368 and then the problem 
occurs.

Original comment by danialho...@gmail.com on 24 Aug 2012 at 1:31

[GoogleCodeExporter]

and Sudonim, its best not to make broad claims like "And it's certainly not an 
AMD CPU issue", it just makes you seem like you've never heard of an Errata 
before.

AMD has quite a few of these on K10 processors involving MOV and stackpoint 
related operations returning invalid results.

Original comment by danialho...@gmail.com on 24 Aug 2012 at 1:42

[GoogleCodeExporter]

I do have a couple of bugs in the recompiler code generation i need to fix to 
do with my changes, but im unsure if it will effect this. 

If you wish to see if im wrong about it not going wrong for me, id posted a 
picture on the forums

http://forums.pcsx2.net/Thread-Bug-Report-hack-G-U-series-NTSC-U-SLUS-21258-SLUS
-21488-SLUS-21489?pid=256672#pid256672

Original comment by refraction on 24 Aug 2012 at 12:50

[GoogleCodeExporter]

As I said in that forum thread, PCSX2 has different code paths for VIF, 
depending on the SSE level supported by the CPU.
This is not a "AMD bug" but a bug that only shows with certain (SSE2 prolly) 
code paths.

Original comment by ramapcsx2 on 24 Aug 2012 at 3:24

[GoogleCodeExporter]

Ah I see yes, ill comment out the sse4 stuff (which Mine is using)

Original comment by refraction on 24 Aug 2012 at 6:34

[GoogleCodeExporter]

Original comment by refraction on 25 Aug 2012 at 5:14

[GoogleCodeExporter]

Original comment by refraction on 25 Aug 2012 at 5:14

• Changed state: Accepted

[GoogleCodeExporter]

can you confirm fixed in r5392 ?

Original comment by refraction on 25 Aug 2012 at 7:10

[GoogleCodeExporter]

confirmed fixed

Original comment by danialho...@gmail.com on 26 Aug 2012 at 12:22

[GoogleCodeExporter]

Good :)

Original comment by refraction on 26 Aug 2012 at 10:06

• Changed state: Fixed