Pi1MHz Software

[dp11]

New 6502code.bin file which should add the system via wite.

Code: Select all

    LDA #&7F
    STA &FE4E
    LDA #200
    LDX #3
    JSR OSBYTE
    JMP (&FFFC) ; Reset

If you are short of space, I suspect the subseqent OSEBYTE (*FX 200,3) is now superfluous, as a power up BREAK always triggers a memory clear.

Dave

Always glad to have space Dave , thanks.

[Mince]

Have I missed something subtly changing between 1.13 and this latest version published yesterday (20240428), at least as far as the Electron is concerned?

I'm using the ESWMMFS version of MMFS1, with the Jim RAM build, loaded into a SWRAM bank.

If I format the card with the SD Card Association formatter, then copy everything from the Pi1MHz_20240428_1442_root directory on, then BEEB.MMB, I get "disc not formatted" for disc 0 and a corrupted *DCAT same as Ken got a while ago here: https://www.stardot.org.uk/forums/viewt ... 33#p419533

If I re-format the card, then put on the 1.13 build and copy the BEEB.MMB back on, I get a perfectly working MMFS with *DCAT, etc.

[KenLowe]

Have I missed something subtly changing between 1.13 and this latest version published yesterday (20240428), at least as far as the Electron is concerned?

I'm using the ESWMMFS version of MMFS1, with the Jim RAM build, loaded into a SWRAM bank.

If I format the card with the SD Card Association formatter, then copy everything from the Pi1MHz_20240428_1442_root directory on, then BEEB.MMB, I get "disc not formatted" for disc 0 and a corrupted *DCAT same as Ken got a while ago here: https://www.stardot.org.uk/forums/viewt ... 33#p419533

If I re-format the card, then put on the 1.13 build and copy the BEEB.MMB back on, I get a perfectly working MMFS with *DCAT, etc.

I think you probably need a later version of ESWMMFS for Pi1MHz. You'd need to build it from the sources here:

https://github.com/dp111/MMFS

I could do that for you, if it would help.

Edit: Give this version a try. It's not the very latest. It's a version I built yesterday (shortly after commit 1f6e065), but it should still work with this latest version of Pi1MHz. I'm not sure you'll be able to use the special helper function, that allows you to load a ROM image from Pi1MHz to a SWR bank, with the Electron though. It will depend on how RAM banks are selected on the Electron.

[Mince]

Thanks — I'd just assembled that up myself but I'm just getting hanging when I try *. or *DCAT, now. That's the same whether I use your version or the one I assembled.

(Just to confirm, I'm using the one in build/MMFS/1/ESWMMFS.rom — I presume that's the correct one?)

The ADFS HDD in the BeebSCSI folder works fine, and P. $&FD00 gives Pi1MHz v1.16. Have I missed another change out here, like BEEB.MMB now going in a folder?

[KenLowe]

Thanks — I'd just assembled that up myself but I'm just getting hanging when I try *. or *DCAT, now. That's the same whether I use your version or the one I assembled.

(Just to confirm, I'm using the one in build/MMFS/1/ESWMMFS.rom — I presume that's the correct one?)

The ADFS HDD in the BeebSCSI folder works fine, and P. $&FD00 gives Pi1MHz v1.16. Have I missed another change out here, like BEEB.MMB now going in a folder?

I think that's the correct ROM. That's the one I attached to my last post.

Have you tried the ESWMMFS.rom in the MMFS2 directory (the one that doesn't need beeb.mmb) to see if it behaves any differently? Also, what about the EMMFS version? Does that work? I presume that's a version that lifts PAGE, in the same way as the beeb version does, but I'm not certain because I'm not that familiar with the Electron.

Edit: Here's the details of the various versions. I'd certainly try the build/MMFS/1/EMMFS.rom version:

Code: Select all

EMMFS.rom    - Electron MMFS (PAGE=&1900)
ESWMMFS.rom  - Electron MMFS using SWRAM (PAGE=&E00)
ZEMMFS.rom   - Electron MMFS using SWRAM (PAGE=&E00) wrapped with bootloader
EMMFSDB.rom  - Electron MMFS with debug enabled (PAGE=&1900)

[Mince]

EMMFS gives the same hang.

Loading EMMFSDB and typing *DCAT, it prints the following and then hangs, if this helps anyone:

FF400000000095:01
FF4100000000FF:00
FF7700000000FF:00
FF6940000000FF:00
FF7A00000000FF:00
FF5000000200FF:00
FF4A00000000FF:00

[dp11]

There have been quite a lot of changes and improvements. But it appears I've broken the electron version.
From power up what does X%=0:CALL FC88 give?

[Mince]

That just prints lots of blank spaces and scrolls off the screen.

Looking in EXMON and disassembling, I can see there is nonsense at &FD00 but, when I write 0 to &FC88, I can see the code at &FD00 copied in and it looks sensible but it's just printing the blank spaces when run. I haven't looked any further than that yet, if there's something specific you want me to check for.

[dp11]

I think that implies the pi is seeing a load of glitches and the auto increment gets confused. Just to check can you single step the code at FD00 and if silly data is returned?

[hoglet]

There have been quite a lot of changes and improvements. But it appears I've broken the electron version.

This optimization looks dubious to me:

The BMI on line 279 is now branching on sign of X rather than A.

[Mince]

Single stepping the code at &FD00 in EXMON II seems to loop around &FD12. LDA &FCD9 seems to yield &FF each time.

Would this affect MMFS too?

[dp11]

There have been quite a lot of changes and improvements. But it appears I've broken the electron version.

This optimization looks dubious to me:
bug1.PNG
The BMI on line 279 is now branching on sign of X rather than A.

Yes you are right the INX needs to be before the JSR. I remember working that out, but clearly mucked up the commit.

[dp11]

Single stepping the code at &FD00 in EXMON II seems to loop around &FD12. LDA &FCD9 seems to yield &FF each time.

Would this affect MMFS too?

Can you read back FDD6 onwards and see if they match what was written? I particular each read of FDD9 should increment FDD6

[hoglet]

Can you read back FDD6 onwards and see if they match what was written? I particular each read of FDD9 should increment FDD6

The issue with the ElectronAP5 is that addresses &FCDx are not mapped to the 1MHz bus, because they are used by several Elk cartridges as the lock /unlock registers. In fact, &FCD7 is the AP5 version register.

Currently the following blocks are mapped: &FC0x, &FC1x, &FC2x (write only), &FC3x, &FC4x, &FC8x, &FCAx, &FCFx

Could the Electron use &FCAx?

Dave

[dp11]

Can you read back FDD6 onwards and see if they match what was written? I particular each read of FDD9 should increment FDD6

The issue with the ElectronAP5 is that addresses &FCDx are not mapped to the 1MHz bus, because they are used by several Elk cartridges as the lock /unlock registers. In fact, &FCD7 is the AP5 version register.

Currently the following blocks are mapped: &FC0x, &FC1x, &FC2x (write only), &FC3x, &FC4x, &FC8x, &FCAx, &FCFx

Could the Electron use &FCAx?

Dave

Well spotted . I can move it to FCAx .

I'll recompile everything tonight.

[KenLowe]

Does the Electron also use &FE34 to page in RAM / ROM banks?

[dp11]

Latest version attached. I have also attached the MMFS builds as well. Thanks for all the test and comments so far.

[hoglet]

Does the Electron also use &FE34 to page in RAM / ROM banks?

No it doesn't - it uses FE05 so a different ROM loader will be needed.

And it's a bit more complicated than just a singe write:

Code: Select all

        MACRO  page_rom_x
        pha
        lda    #&0F
        sta    &f4
        sta    &fe05
        pla
        stx    &f4
        stx    &fe05
        ENDMACRO

Dave

[dp11]

So If my SWR loader could detect if it was running on an electron can I use the same SWMMFS.rom as the beeb and master ? No idea how to create the code space yet.

[KenLowe]

You would need to use the ESWMMFS rom, I think:

https://www.stardot.org.uk/forums/viewt ... 13#p424013

[hoglet]

So If my SWR loader could detect if it was running on an electron can I use the same SWMMFS.rom as the beeb and master ?

There is one small difference: flashing the the LEDs to indicate activity:

Code: Select all

.SetLEDS
IF _ELECTRON_
   LDA &282
   EOR #&80
   STA &282
   STA &FE07
ELSE
   LDX #&6
   STX &FE40
   INX
   STX &FE40
ENDIF

The Beeb code would I think mess with the interrupt enable register in the Elk ULA.

Shame there isn't a clean way to do this - it would be nice to get rid of Electron specific builds.

Dave

[dp11]

So If my SWR loader could detect if it was running on an electron can I use the same SWMMFS.rom as the beeb and master ?

There is one small difference: flashing the the LEDs to indicate activity:

Code: Select all

.SetLEDS
IF _ELECTRON_
   LDA &282
   EOR #&80
   STA &282
   STA &FE07
ELSE
   LDX #&6
   STX &FE40
   INX
   STX &FE40
ENDIF

The Beeb code would I think mess with the interrupt enable register in the Elk ULA.

Shame there isn't a clean way to do this - it would be nice to get rid of Electron specific builds.

Dave

We have space todo a runtime test if we could find a bit to hold the machine state unless the OS already has one.

[hoglet]

We have space todo a runtime test if we could find a bit to hold the machine state unless the OS already has one.

Actually, there is more than just this difference.

The most significant other difference is the tube data register is at a different address (FCE0 vs FEE0)

There is some cosmetic stuff as well (like the ROM title), and the Elk build excludes *ROMs as this is present in the AP1/6 ROM.

Dave

[Mince]

Latest version attached. I have also attached the MMFS builds as well. Thanks for all the test and comments so far.

Thanks for these — I've loaded that in and now CALL &FC88 gives me the help screen for 1.17.

MMFS (from the SSD you posted) is still not working, although it's now giving me "Card format?" rather than hanging (the previous version still hangs). Is that related to the further conversation, after the updates?

[dp11]

Latest version attached. I have also attached the MMFS builds as well. Thanks for all the test and comments so far.

Thanks for these — I've loaded that in and now CALL &FC88 gives me the help screen for 1.17.

MMFS (from the SSD you posted) is still not working, although it's now giving me "Card format?" rather than hanging (the previous version still hangs). Is that related to the further conversation, after the updates?

Shouldn't be. Can you try the electron mmfsv2 and DCAT? Also using the debug version of electron mmfs what do you get ?

Sorry this might take a little time while I work out what I've done to break it.

[Mince]

No problems — thanks for your help and the work so far!

Starting with EMMFSDB, a *DCAT gives (hoping I've got the right number of zeroes here!):

FF400000000095:01
FF4100000000FF:00
FF7700000000FF:00
FF6940000000FF:00
FF7A00000000FF:00
FF5000000200FF:00
FF4A00000000FF:00
FF5100000000FF:00
FF5100010E00FF:00

For MMFS2, after loading the ROM, I get a startup banner:

Code: Select all

Acorn Electron @ 64K

Electron MMFS2 SWRAM Pi1MHz


Card format?
*DCAT
*.

Drive empty
*

I've not used MMFS2 on the P1MHz before, though, so I might be doing something wrong. Do I just put the SSDs in the root directory of the card, or have to put them in a special directory?

[KenLowe]

For MMFS2, I think your card needs to be formatted FAT32. Then it's just a case of adding .ssd / .dsd files to the root of your SD card. You can probably add then to a specific directory, but I can't remember what the beeb command is to change directories. *DCAT should then list all the files and then you just *DIN or *DBOOT the specific disc image to mount the image.

[hoglet]

Just a quick note to say it all seems to work correctly for me on the Elk/AP5:

I'm using MMFSv2 ESWMMFS build just now from Dominic's MMFS fork, manually loaded into slot 4 on my AP6.

Also using the latest Pi 1MHz firmware which declares itself as 1.17.

And with the EMMFSDB build I get:

I suspect a genuine card format issue.

Any chance you could image the whole card?

Dave

[dp11]

No problems — thanks for your help and the work so far!

Starting with EMMFSDB, a *DCAT gives (hoping I've got the right number of zeroes here!):

FF400000000095:01
FF4100000000FF:00
FF7700000000FF:00
FF6940000000FF:00
FF7A00000000FF:00
FF5000000200FF:00
FF4A00000000FF:00
FF5100000000FF:00
FF5100010E00FF:00

For MMFS2, after loading the ROM, I get a startup banner:

Code: Select all

Acorn Electron @ 64K

Electron MMFS2 SWRAM Pi1MHz


Card format?
*DCAT
*.

Drive empty
*

I've not used MMFS2 on the P1MHz before, though, so I might be doing something wrong. Do I just put the SSDs in the root directory of the card, or have to put them in a special directory?

Do you know how the card is formatted ? Ideally it should be FAT32. Fat16 works for MMFS also

[Mince]

Thanks — I'll check tonight. I format the card with the SD Card Association Formatter but it's 2GB, so I think it should be FAT16. That probably explains MMFS2 but I'm not sure about MMFS, but I'll try that: I have a larger card and can try that instead to see how it behaves.