NEC PC Engine / TGFX16

[fossil pumice]

That 1 pixel line till reach an even resolution is filled with garbage

If another core generates an odd resolution, I bet the line will be also there

@dreamy tiger were u able to reproduce the garbage line?

[dreamy tiger]

I’m not at home. I will look when I am

[fossil pumice]

Aps, sorry

[grand moss]

Well, I don't know what your video pipeline looks like, but I use a 720p HDMI monitor on MiSTer, and I don't see that line with those games. I'm pretty sure I'm using a much more vanilla configuration though.

[fossil pumice]

But do you have the vscale_mode=5?

Understand me. There's no problem at all. Just that something is off and I think it's interesting to try to replicate the problem and see if it's something more general or just an isolated case

Vanilla config sets vscale_mode=0 or, in other words, the default vscale_mode value if no config line is present equals to 0

[grand moss]

I definitely don't have that (vscale_mode=5)

and it doesn't even make sense if the output is to a 720p monitor, does it ?

as far as my mister.ini says, it's:

vscale_mode=1          ; 0 - scale to fit the screen height.
                       ; 1 - use integer scale only.
                       ; 2 - use 0.5 steps of scale.
                       ; 3 - use 0.25 steps of scale.

I'm not aware of any possible value beyond 3

So anyway, I don't think it's part of the core - though the scaler is a possiblity (and that's part of the framework, meaning it could show on some other cores)

[fossil pumice]

I think your base ini is old. Mine has up to 5 commented modes (I think it is at least 1 year old) and in the mister docs there is info about 4 and 5 modes. I also agree with you that this should also happen in other cores outputting an odd vertical resolution

I also think it is something related with "an obligation" of sending an even vertical resolution.

[dreamy tiger]

Looks fine here

[fossil pumice]

Overscan = hidden?

[dreamy tiger]

That resolution tho

Your monitors just don’t like the weird resolutions

[fossil pumice]

Tomorrow I'll try with the tv

Does your monitor have osd info on the resolution shown?

Tomorrow I'll also compare your ini with mine. Maybe another video option is changed and I forgot about it

[grand moss]

Well, since the original console doesn't send an even number of scanlines in that mode, the core is just reflecting the original device. (So there is no "obligation" to do so)

[dreamy tiger]

The core is fine. This is just a matter of displays disagreeing with the (sometimes) weird resolutions that vscale_mode 4 and 5 land on

[coarse shadow]

An external scaler is always the better way. Yes, some are expensive but it's worth.

[dreamy tiger]

300 bucks gets you into a morph 4k. I would argue that, unless you have brain worms like Jack and I, it's the best scaler for the MiSTer

[coarse shadow]

Yeah, just for MiSTer HDMI -> Direct Video the morph 4K is just fine.

It's too annoying to mess with the MiSTer scaling/resolutions functions.

[fossil pumice]

I'll be seriously considering buying an scaler the day they were able to deform the image (curved corners) like and old CRT. I searched a bit and was unable to find one that do it at the moment.

[fossil pumice]

I ment via hdmi. I have to do more tests with more displays. I don't know how to check if the mister is sending an odd resolution and the monitor rounding to an even one or if it is the mister rounding it on its side before send it out

[fossil pumice]

What makes me rise an eyebrow is that the consequence of being an odd resolution is the same in two completely different screens (an initial green garbage line at the bottom)

I have two mister pi. Maybe is related with it. I'll check with the second one

[fossil pumice]

With the TV (4k), the vertical res is 4 times the original (under 1080p), so the mister outputs an even res. The res is set to 1684x924

[fossil pumice]

@dreamy tiger , the picture u sent showing your res, in which AR and res monitor was it taken?

[dreamy tiger]

video_mode=1920,48,32,80,1200,3,6,26,154000,+hsync,+vsync

it's a 16:10 monitor

[fossil pumice]

Ok. I'll try to find more monitors and test there

[stiff lotus]

Not sure if you’re able to do that with zero latency. That sounds like an added framebuffer cost.

Might as well do software emulation at that point?

[summer rivet]

I can’t believe you said that

[stiff lotus]

Wasn’t me

[fossil pumice]

In theory it should be possible in one movement: at the moment you have the pixel you apply the transformation at the moment of write it to the output buffer (and you write it at the new coordinate). Little extra work for blending the pixels because more than one will land in the same place

I mean, if the scaler is already writing to a buffer, that probably it is already doing

In fact, I think mister could have a core that is an upscaler... With a second mister u could have a nice scaler for a relative low price

[stiff lotus]

Hey good points! Didn’t think of it that way.

[regal valley]

there is (or was) a version of OSSC based on the DE-10. Not MiSTer, but same FPGA board

[fossil pumice]

Dexx. Interesting

[grand moss]

Yeah, it was the "pro" - at least "pro-totype".

Also, DE10-Nano is used in that disk flux reader, "Pauline"

[fossil pumice]

Today I tested the normal 1080p monitor and was not able to reproduce the green line at the bottom like before 😤🤦🏻🤷

[dreamy tiger]

And the choir sang

🎶 The core is fine. This is just a matter of displays disagreeing with the (sometimes) weird resolutions that vscale_mode 4 and 5 land on 🎶

[fossil pumice]

That the core is fine I was already convinced 😅

Just that this core happens to have an odd resolution 🙂

[grand moss]

So it was able to coerce your monitor into showing its problem...