GSdx

[rama]

- AA1 is a GS feature. It's not forced on scenes but gets applied when the game specifies it. Hence it should be always enabled in software mode anyway (but it causes a bit of a speed drop when used).
- If AA1 changes the font then it means it is enabled and the font will only look perfect with a perfect AA1 algorithm.
- It looks like the current AA1 is not pixel perfect, which we'll look into some day

[Ilya-v]

Great thanks.

[Cleretic]

Okay, I think this is the right place to post this, if it's not please point me in the right direction.

Simply put: I can see through geometry in Shin Megami Tensei Nocturne. Apparently this issue cropped up a couple years ago and was more or less fixed, but somehow I'm having it. No amount of fiddling with settings will solve this (except for going into software rendering, but I don't really want to do that since it looks terrible), I assure you I have tried flipping every switch within reach.

I'm using the latest version of PCSX2, which apparently comes with GSdx 0.1.16, having upgraded from PCSX2 0.9.8 specifically in attempts to solve this issue. All my specs that I can figure would be relevant here:

OS: Windows 7 Home Premium, 64-bit
Processor: Intel Core i7-2720QM
Graphics card: NVidia GeForce GT 540M, 1GB
RAM: 8 GB

[dwapar]

hello freinds,

I have log file which i attached ,i got the error:

Path: H:\PCSX2 1.0.0\Plugins\gsdx32-avx-r5350.dll
File is not a valid dynamic library.
Some kinda plugin failure: H:\PCSX2 1.0.0\Plugins\gsdx32-avx-r5350.dll

How to rectify this error

[Bositman]

Delete gsdx32-avx-r5350.dll. Your processor does not support AVX instructions so you can't use that plugin.

[KrossX]

So, I finished Xenosaga Ep1 and some texture edge bug was annoying me, but meh. I started playing Episode 2 and finally made a GS snap to mess around with. Found a thread about it (oddly enough with a shot at the same place I took it) HERE. Gs snap HERE.

[Image: gsdx_20120425015236.jpg]
(using that thread's image while at it, cuz I'm lazy )

I've traced it to tfx.fx:244, but I'm too sleepy to do anything about it. The d3d9 part is fine as it doesn't show the issue. The line doesn't seem to have been touched in the... googlecode history, the other wrapps have seen a fix or two though.

Code:

uv = (float4)(((int4)(uv * WH.xyxy) & MskFix.xyxy) | MskFix.zwzw) / WH.xyxy; // <<<<<<<====== EVIL!!

Code:

float4 wrapuv(float4 uv)
{
    if(PS_WMS == PS_WMT)
    {
/*
        if(PS_WMS == 0)
        {
            uv = frac(uv);
        }
        else if(PS_WMS == 1)
        {
            uv = saturate(uv);
        }
        else
*/
        if(PS_WMS == 2)
        {
            uv = clamp(uv, MinMax.xyxy, MinMax.zwzw);
        }
        else if(PS_WMS == 3)
        {
            #if SHADER_MODEL >= 0x400
            uv = (float4)(((int4)(uv * WH.xyxy) & MskFix.xyxy) | MskFix.zwzw) / WH.xyxy; // <<<<<<<====== EVIL!!
            #elif SHADER_MODEL <= 0x300
            uv.x = tex1D(UMSKFIX, uv.x);
            uv.y = tex1D(VMSKFIX, uv.y);
            uv.z = tex1D(UMSKFIX, uv.z);
            uv.w = tex1D(VMSKFIX, uv.w);
            #endif
        }
    }
    else

#EDIT: Perhaps it's a Radeon thing? If someone with an nVIDIA could check the gs_snap, that would be nice.

#EDIT2: Hackzoring... Screenshot

Code:

uv = (float4)(((int4)(uv * WH.xyxy) & MskFix.xyxy) | MskFix.zwzw) / WH.xyxy;
uv.w = uv.y;

There, evil edge lines are gone. Side effect, there's no filtering in that direction now.

#EDIT3: Hackzoring 2... Screenshot

Code:

uv.yw += -1.0f / WH.y;
uv = (float4)(((int4)(uv * WH.xyxy) & MskFix.xyxy) | MskFix.zwzw) / WH.xyxy;

This one works nicely.

[dralor]

Yup but it looks like their is still something weird going on with the filtering on his left shoulder just not as noticable.

[Gvaz]

It looks a lot better now though.

[dralor]

I don't dispute that in fact the filtering messiness might not have anything to do with his code just something I noticed in the screenshot.

[KrossX]

It's just a bad texture that gets a close up.