Coding challenge: Maximize-contiguous-chipmem bootblock

[ross]

Originally Posted by meynaf

Under V39 on my A1200 it's really in chip if i don't move it.
I suppose it's the same for other kick versions. If Exec is in chip, you've got big chances supervisor stack also is. And it is by default.

Yes, for A1200 there is more cases.
The problem is that accelerators memory is not directly exec recognized, so is added by the small initialization board ROM.
But is to late: exec.library and sstack are already set..

(A500 have usually fixed memory zones: ranger at 0xc00000 and fast at 0x200000, that exec maps immediately)

Regards,
ross

[Photon]

Stacks are at their default small sizes just before opening startup-sequence, and I think you're going off in too many directions at once.

Typically this,

Originally Posted by Photon

I just mean "click to disable DF1"+"Add21k", but do it automatically in the bootblock before running the startup-sequence.

is the only thing preventing some people from running "a big" demo or game.

[hooverphonique]

So here we go.. a couple of pieces of code. I don't remember how they work in detail, but that can be worked out by looking at them..

TheBoot can be written directly to floppy using the asmone WS/CC commands, then reboot and experiment (I don't have any drives other than df0: atm). A1 contains the trackdisk.device io request when entering the bb code. I think it enables/disables drives by writing/clearing a longword representing the pulse train emitted on the RDY pin by extra drives (more on that can be found here on the board in other threads).

The other piece of code let's you detect the boot drive, so you only disable 'extra' drives (on V36+, you can boot off of any floppy drive).

Hope that helps..

[ross]

Hi, addchip.bootblock done.
I thought it was simpler.. it was not trivial to figure out how to make it working (and in every config)

Attached the result.
Test done on A500 KS1.2 512KB CHIP DF0+DF1.
(consider it a beta, need to be tested better )

I 've to refine a bit the source then I publish a binary bootblock and the code.

Cheers,
ross

[ross]

And.. here we go.

Greetings
ross

[hooverphonique]

Originally Posted by ross

And.. here we go.

Greetings
ross

"Evergreen Amiga Board" ???

[ross]

Originally Posted by hooverphonique

"Evergreen Amiga Board" ???

Yes this need an explanation: in italian evergreen (sempreverde) not only refers to plants but is a synonymous of timeless.
(in other language also?)

So fashionable, distinguished and renowned, which despite the passing of time preserves intact its style.

Evergreen members who are forever young!

Regards,
ross

[NorthWay]

Originally Posted by NorthWay

You could relocate the header for chipmem to (fake) fastmem and then free it. That is some 16-32 bytes IIRC.

As I have tested and found out: You can't FreeMem() the header. At least not just like that.
It might be as simple as subtracting MH_SIZE from MH_LOWER before trying it.

UPDATE: Don't try that if you have Mungwall running... better make sure it is done before you run Mungwall.

[ovale]

Well done Ross!

[NorthWay]

"Evergreens" are known as songs that are timeless and sung by artist both old and young.

[ross]

Originally Posted by ovale

Well done Ross!

I think that 48000 bytes is a very good result for a 512KB machine
(and memory is as contiguous as possible).

Usability from 1.2 to 3.9 is nice too.
The best result on A500+ (KS2.0) where you can gain even ~64KB (the Height parameter works on v37..).

[EDIT: values related to DF1: present]

Ciao!
ross