MSX HDMI mutlimedia card

صفحة 11/57
4 | 5 | 6 | 7 | 8 | 9 | 10 | | 12 | 13 | 14 | 15 | 16

بواسطة maxis

Champion (512)

صورة maxis

07-04-2014, 12:36

PingPong wrote:

my first wish is to have something like F18A
Another interesting thing i would add is a screen 4 tiled mode with no colour clash tiles.

Thank you for the feedback. I'm working currently on the trial run core of 9958 w/o command engine. I saw your link about the analyzed VRAM access timing (thanks for pointing that out), this at least will save me quite a bit of time.
Looks like the VRAM access engine of 9938/9958 was originally designed for something else Smile
Also read the original notes from TI designer (Hans tech library) and partially implemented the original memory access timing in the case of TMS9918 in the current design. Why "partially", because several clock domains due to the HDMI and SDRAM not running from the same time base. Once I'll get the video subsystem working I'll redesign the VDP again to optimize the architecture before going any further.
Could you, please, suggest me what games/tests are readily available, which can be used in the following order for compatibility check?
1. TMS9918 features only, SCREEN1&2 no sprites.
2. TMS9918 features only, no sprite collision detection.
3. FULL TMS9918 regression test.
4. 9938 no sprites, no commands
5. 9938 sprites (no collision), no horizontal line interrupt
6. 9938 heavy test for the precise timing compatibility, but no commands
I'll stop here for the moment ...

بواسطة Pencioner

Scribe (1609)

صورة Pencioner

07-04-2014, 13:57

I cannot say about particular games (which features were used in game etc) but i'm pretty sure that the demos are the best suitable for full and regression tests Smile you know those are hacky and very often do use some exact timings and other stuff not even used at regular basis by game titles. Also, you can find more info on features used by demos than by games

بواسطة maxis

Champion (512)

صورة maxis

07-04-2014, 14:51

Pencioner wrote:

the demos are the best suitable for full and regression tests

Entirely agree. If I could find a short demo for MSX1, it would be grateful! Cool So, then I'll start from 512byte demos. I'm wondering, which one doesn't use the sprites?

بواسطة Pencioner

Scribe (1609)

صورة Pencioner

07-04-2014, 16:38

Hm, 512 bytes could be just explored for not using that sprites residence VRAM part. If i get a bit of free time at weekend i'll try to disassemble some.

بواسطة maxis

Champion (512)

صورة maxis

07-04-2014, 17:11

Pencioner wrote:

Hm, 512 bytes could be just explored for not using that sprites residence VRAM part. If i get a bit of free time at weekend i'll try to disassemble some.

Thank you, this would be a great help. Also it doesn't have to be necessarily 512 bytes. Also, IMHO, demo coders are quite accessible, so we can always ask the question regarding the techniques they used.
Simply a small demo is easier to track with the real-time Z80 disassembler for the Agilent logic analyzer. To detect a failing sequence would be quite possible by a naked eye without writing a complex script, I hope.

بواسطة PingPong

Enlighted (4155)

صورة PingPong

07-04-2014, 20:58

Quote:

1. TMS9918 features only, SCREEN1&2 no sprites.
2. TMS9918 features only, no sprite collision detection.
3. FULL TMS9918 regression test.
4. 9938 no sprites, no commands
5. 9938 sprites (no collision), no horizontal line interrupt
6. 9938 heavy test for the precise timing compatibility, but no commands
I'll stop here for the moment ...

Let's see. It's not an easy task, because you have a partially working prototype in terms of features.
1. As you can imagine, there aren't pratically games that avoid sprites except zx speccy ports. But the kind of test they do is always the same. Dump RAM to VRAM, VRAM to RAM in standard tile mode (screen 2)
2. Pratically all msx native games, because they do not rely on collision detection (because of scanline limit you also get collision not detected so games prefer the brute force scan via CPU)
3. ...... easy (1)+(2) :-)
4. there is pratically no sw that uses v9938 in this way. Maybe some demo? not sure.
5. most of msx2 games you can choose any but avoid those that have on top or bottom a wide screen scorebar. It's used with scanline ints.
6. not so easy. most games uses vdp engine and vdp engine takes vram bandwith, so if you want to test vram I/O you also need commands running. Some sw do CPU data port write WHILE a command is executed. on v9958 this work with no trouble, the vdp pause current VRAM I/O to give priority to CPU request.

بواسطة maxis

Champion (512)

صورة maxis

08-04-2014, 14:46

PingPong wrote:

Let's see. It's not an easy task, because you have a partially working prototype in terms of features.

Thank you.
What would be the speccy port titles?

بواسطة Pencioner

Scribe (1609)

صورة Pencioner

08-04-2014, 15:10

as i remember, Chase HQ was the one of (there ia a lot of them but i didn't play too much Speccy games though)

بواسطة maxis

Champion (512)

صورة maxis

08-04-2014, 15:59

Pencioner wrote:

as i remember, Chase HQ was the one of (there ia a lot of them but i didn't play too much Speccy games though)

Excellent, thanks! Just finished the smart trigger script for LA, so then any access to the sprite attribute/pattern table will trigger it.

بواسطة Grauw

Ascended (10819)

صورة Grauw

08-04-2014, 16:47

Screen 4 games probably don’t use the VDP commands. E.g. Space Manbow / Manbow 2.

صفحة 11/57
4 | 5 | 6 | 7 | 8 | 9 | 10 | | 12 | 13 | 14 | 15 | 16