msx vaporware - true or false ?

Page 18/24
11 | 12 | 13 | 14 | 15 | 16 | 17 | | 19 | 20 | 21 | 22 | 23

By ARTRAG

Enlighted (6323)

ARTRAG's picture

20-07-2012, 00:29

This one is much lighter (27.9990 tiles updated per frame in average)
https://sites.google.com/site/testmsx/Home/double-buffer-in-...

Tomorrow with more time I'll try to explain my code a bit more
About shadows it is doable, but needs a lot of hand work to set proper colors on right faces, 3d was just an experiment, anyway the result could be nice, we will see.
It is more urgent to implement an optimal strategy of tile updating based on the knowledge of witch tile will be reused in next frames and witch tile is not used anymore and can be replaced with the tiles needed for the current frame

By hit9918

Prophet (2882)

hit9918's picture

20-07-2012, 01:25

I did not mean something with shadows,
I meant just picking different colors for the planes that are already in the demo.

Like e.g. one object is red, its polygons got 3 colors from bright red to dark red.
The walls you look at directly in bright color,
the side walls medium color,
the tilted one dark color.
I expect some nice illumination feel, giving more 3D feel.

From new demo I conclude you like it more with black lines? I like it much more without black lines, because then it is with practically no colorbleed noise.

And, e.g. right at the beginning of the demo are two cyan faces.
Only with black line there is a clear edge.
But with said lighting style, the two polygons would have dark blue for the tilted polyon and cyan for the other polygon, and then need no black lines to get a clear edge.

The 80s workstations on TV too did such stuff too without black lines, but lighting style Smile

By hit9918

Prophet (2882)

hit9918's picture

20-07-2012, 02:06

About speed, I found a code "tileupdate:" that does vram setup for every tile.
With multiplying HL * 8. That info could maybe be precalculated in the stream?
And the demo goes with violating VDP speed limit Tongue

But I feel like looking at old code. Can't see instructions that would set up a mapper in that loop.

Taking it to the extreme, the encoder make multi-char chunks that go without vram setup and without page switch.
I still got no idea what is going on in the nametable, is it speed critical?

Another brainstorm, pgtbase and pctbase get added 16bit before vram setup,
the whole thing probably goes just as well with adding the hi byte with akku, leaving free HL and DE.

By ARTRAG

Enlighted (6323)

ARTRAG's picture

20-07-2012, 07:16

run makex2ani.bat and see if it gives a valid rom

PS
I updated the 3d.rar just now with the latest data

By NYYRIKKI

Enlighted (5425)

NYYRIKKI's picture

20-07-2012, 18:31

Very impressive again. Too bad that the tile conversion engine still fails quite miserably. ie When there is dark red wall, gray floor and black line. I can see colors like light red, yellow, light green, light blue and white as well. I think you should try to select 2 most used colors. That would make black lines pretty much dissappear, but I think the result would be better anyway.

By ARTRAG

Enlighted (6323)

ARTRAG's picture

20-07-2012, 23:07

I've loaded a wrong rom: try now at the same link, the file is without black lines and almost no bleeding

By Manuel

Ascended (15969)

Manuel's picture

21-07-2012, 09:50

What's the URL?

By ARTRAG

Enlighted (6323)

ARTRAG's picture

21-07-2012, 10:25

By NYYRIKKI

Enlighted (5425)

NYYRIKKI's picture

24-07-2012, 12:12

Without black lines the problem is much more invisible, but I can see the wrong color problem is still there. I understand that there is 2 color limit on 8-pixel area, but that does not explain why the colors are totally incorrect.

By ARTRAG

Enlighted (6323)

ARTRAG's picture

24-07-2012, 19:58

It is the dithering that spreads in space errors due to the 2 colors per 8 pixels limit

Page 18/24
11 | 12 | 13 | 14 | 15 | 16 | 17 | | 19 | 20 | 21 | 22 | 23