MiSTer FPGA Forum

I noticed this was on the unofficial core updated script that is in Update All (but not selected to run as default).

https://github.com/Kyp069/zx48-MiSTer

Seems it's a new ZX 48 core that was added 18 days ago. Hadn't seen any mention of it.

Do we need another ZX Spectrum core?

antonie wrote: ↑Thu Apr 22, 2021 12:51 pm Do we need another ZX Spectrum core?

Why not? Just because two cores simulate the same system it doesn't mean they are the same.

The core is a little bit light on documentation for my taste, from what I see the core supports real tape loading, I wonder how it is realised.

This is a core for perfectionists, it has very accurate timings for contended memory, unattainable in a regular core, it can be seen only in special tests. The same kernel exists for zx128 as well

To get 1:1 pixel aspect ratio I had to set a custom aspect ratio of 352:305. Screenshots are not working, by the way.

So could this core wind up replacing a current one?

I get a flickery screen with this and don't have a clue how to use it. The other Speccy cores are definitely superior so far on a usability basis.

Moondandy wrote: ↑Thu Apr 22, 2021 10:09 pm So could this core wind up replacing a current one?

This core aims to simulate a ZX Spectrum 48K only (it seems the core was originally developed for the ZX Uno board and that there is a ZX Spectrum 128K also available. This core seems to be developed for quite a while now and I guess it is pretty stable. But I think it is not intended to replace the current core. It is more an addition to what is already here. From what I have seen the core currently supports only VHD loading from the SD card (via ESXDOS).

caffeinekid wrote: ↑Thu Apr 22, 2021 11:07 pm I get a flickery screen with this and don't have a clue how to use it. The other Speccy cores are definitely superior so far on a usability basis.

you have to set vsync enabled for this core to prevent that.

azesmbog wrote: ↑Thu Apr 22, 2021 5:36 pm This is a core for perfectionists, it has very accurate timings for contended memory, unattainable in a regular core, it can be seen only in special tests. The same kernel exists for zx128 as well

Why? (to the bold part)

Can you point me to these tests?

pgimeno wrote: ↑Fri Apr 23, 2021 5:40 pm

azesmbog wrote: ↑Thu Apr 22, 2021 5:36 pm This is a core for perfectionists, it has very accurate timings for contended memory, unattainable in a regular core, it can be seen only in special tests. The same kernel exists for zx128 as well

Why? (to the bold part)

Can you point me to these tests?

I think he is talking about these tests:

http://www.zxspectrum4.net/op_timing.php

Thanks. Well, considering that the official Spectrum core passes the 35 normal tests, it doesn't seem unattainable to me. It only fails the 2 "hidden" tests (36 and 37). Fuse and JSpeccy pass all tests 1-37, so there's room for improvement. ZXNext is failing the pre-test, but that seems like an upstream issue.

Curiously, JSpeccy, Fuse and the Spectrum core all fail some of the 128K tests: Fuse fails 4, 17, 18, 26 and 33, and both JSpeccy and the Spectrum core fail the same as Fuse plus the hidden 36 and 37. That makes me wonder who's right. When I documented how contention works, I made some mistakes which have been corrected since, but maybe not all of them were corrected when Fuse, JSpeccy and the Spectrum core were created.

ZX Spectrum core is aimed to be fully accurate in timings. If there are fixes for it, then anyone is welcome to improve. I don't see a reason to start another core instead improving existing one.
ZX Spectrum supports basically everything this core supports and more on top.

more cowbell, more zx spectrum cores!

Failed 3 tests in that app are related to IN/OUT instructions to specific ports. Probably has something to do with bits read from those ports.
A lot of work has been done on T80 CPU improvements by different people including me and which is used in zx48 core. Without this work, zx48 wouldn't be correct. So author of zx48 instead making just another ZX core to fix these 3 tests could improve existing ZX core.

btw, zx48 core hangs after tests 34-37.

Sorgelig wrote: ↑Sat Apr 24, 2021 11:31 am btw, zx48 core hangs after tests 34-37.

Алексей!
Твое ядро лучшее, хотя у меня и не работает
Ну портировал человек свое ядро на МиСТер, ну хорошо же
Зависает после 34-37 - это из-за портов DivMMC - если их отключАть- все проходит до конца.
чтобы на IN/OUT не фейлилось - можно Кемпстон отключить) Это на нем чаще всего.
на этом ядре снег работает. У тебя - не знаю. по причинам выше.
Может все же ядро АТМ3 соберешь? там есть кое что интересное)
Сорри за мой французский..

azesmbog wrote: ↑Sat Apr 24, 2021 12:35 pm Сорри за мой французский..

я тебе в скайпе пишу, а тебя там нет.

New release passes all tests

Great, always nice to see that you pay attention to the details. I know it is academic if the core passes all tests, but as the tests are passed on the real hardware it is nice that they are passing on the MiSTer now too.

In order to access DIVMMC I had to use (from http://www.esxdos.org/index.html) the version ESXDOS 0.8.8 Public BETA

created a small boot.vhd which I placed in a directory named "zx48" in "fat/games".

Mount boot.vhd in windows (this works also from network share) and unpack the esxdos zip in the boot.vhd then eject/unmount the share drive.
That's it.

F5 just launches the browser as expected to load games...
Dot commands work as well.

_______________
Note: boot.vhd is created in windows "Disk Management" or "Computer Management" system console
Note2: I did not succeed with newer esxdos beta.