Use MMU for creating fake slow-fast memory

[Lisko]

Hi like the title says, would it be possible to remap 0x00c00000 to fast memory using the mmu? The new aca1234 by icomp has this memory as selling point and I was wondering if with an mmu we can achieve the same thing on a blizzard 1230. If possible would be amazing to integrate it in a custom 1.3 kickstart, maybe with a blizkick module, that when booting it will automatically make this memory available.

[jotd]

Sure it should be possible but I don't see the point. Which app needs 0xC00000 in fastmem?

only games can do this, and in that case, use whdload which remaps this slow mem to real fast when possible (sometimes it can fail to use real 32 bit mem because of wrong programming, but that is really a rare case)

[Lisko]

Originally Posted by jotd

Sure it should be possible but I don't see the point. Which app needs 0xC00000 in fastmem?

only games can do this, and in that case, use whdload which remaps this slow mem to real fast when possible (sometimes it can fail to use real 32 bit mem because of wrong programming, but that is really a rare case)

I know but someone could prefer to use floppies for nostalgic or trainers purposes.

[Thomas Richter]

Originally Posted by Lisko

Hi like the title says, would it be possible to remap 0x00c00000 to fast memory using the mmu?

Yes. See MuManual.lha on Aminet. It also contains sources of some of the MuTools which show how to do that.

[ross]

This is my code from ~25 years ago written for myself (probably not even BlizKick existed yet).
I haven't checked the code and it's probably rough and doesn't have any kind of refinement, but it sure worked because I remember running disks that only went on KS1.3/A500_512chip+512fake@$c00000 on my Blizzard 1230.
Take it as it is, I don't know if I want to make a more functional or complete version

EDIT:
I hadn't seen Thomas's message yet.
His is certainly a better and generalized solution, therefore absolutely recommended.

[Lisko]

Originally Posted by ross

This is my code from ~25 years ago written for myself (probably not even BlizKick existed yet).
I haven't checked the code and it's probably rough and doesn't have any kind of refinement, but it sure worked because I remember running disks that only went on KS1.3/A500_512chip+512fake@$c00000 on my Blizzard 1230.
Take it as it is, I don't know if I want to make a more functional or complete version

EDIT:
I hadn't seen Thomas's message yet.
His is certainly a better and generalized solution, therefore absolutely recommended.

Hi thank you very much. Do you think could work also with relokick's modded kickstart which lets you disable the caches? Also I don't have any assembler on my miggy yet, which one to install? Do I need also the ndk?
Thanks to you both

[ross]

Originally Posted by Lisko

Hi thank you very much. Do you think could work also with relokick's modded kickstart which lets you disable the caches? Also I don't have any assembler on my miggy yet, which one to install? Do I need also the ndk?
Thanks to you both

No need for a modded kickstart, in my code there is a line that disable the cache (I should look at what it actually changes, but it sure worked ).
Yes, you need the NDK, but the source can be easily modified so as not to require it (a few EQUs are enough).

However it would be interesting to make a slightly more complete version of it to also support KS1.2, a toggle for enabling icache, support for NTSC and 1MB of chip-ram..
I don't really know how useful an application of this kind actually would be, I think about it and see if I feel like it

[Lisko]

Originally Posted by ross

No need for a modded kickstart, in my code there is a line that disable the cache (I should look at what it actually changes, but it sure worked ).
Yes, you need the NDK, but the source can be easily modified so as not to require it (a few EQUs are enough).

However it would be interesting to make a slightly more complete version of it to also support KS1.2, a toggle for enabling icache, support for NTSC and 1MB of chip-ram..
I don't really know how useful an application of this kind actually would be, I think about it and see if I feel like it

I tried with asm one 1.48 but I get a comma expected error on line 43 (there is a lea instruction there). Also what I don't understand is if your code actually performs the softkicking itself or if it only modifies a kickstart which needs to be softkicked by other means. I would like if it's the latter so we can use the maprom function of the accelerators.

[ross]

Originally Posted by Lisko

I tried with asm one 1.48 but I get a comma expected error on line 43 (there is a lea instruction there). Also what I don't understand is if your code actually performs the softkicking itself or if it only modifies a kickstart which needs to be softkicked by other means. I would like if it's the latter so we can use the maprom function of the accelerators.

Well, the source codes are provided for just that: to tailor them to your needs .
My code softkick by itself, but of course it could be modified to use the blizzard maprom function.
The MMU table is very optimized for the 030 (er.., it only works with it..) and is created to handle 24bit memory in 512KB chunk as you like.
The ATC is so tiny that fully fit in the MMU cache and the performance impact is negligible.

The asm-one error at line 43 is probably due to the

crp

label that the assembler simply confuses with the register, change it to something different at both line 43 and line 61 (crp->

crpm

)

[Lisko]

Originally Posted by ross

Well, the source codes are provided for just that: to tailor them to your needs .
My code softkick by itself, but of course it could be modified to use the blizzard maprom function.
The MMU table is very optimized for the 030 (er.., it only works with it..) and is created to handle 24bit memory in 512KB chunk as you like.
The ATC is so tiny that fully fit in the MMU cache and the performance impact is negligible.

The asm-one error at line 43 is probably due to the

crp

label that the assembler simply confuses with the register, change it to something different at both line 43 and line 61 (crp->

crpm

)

Thanks very useful. After changing crp to crpm I get an illegal operator at line 44 pmove.d (a0), crpm

[ross]

Originally Posted by Lisko

Thanks very useful. After changing crp to crpm I get an illegal operator at line 44 pmove.d (a0), crpm

I never said to change line 44

[a/b]

crp/drp/srp is fine, you have to change pmove.d to pmove.q if you are using asm-one&family.

[ross]

Originally Posted by a/b

crp/drp/srp is fine, you have to change pmove.d to pmove.q if you are using asm-one&family.

I just checked with

vasm

and it also accepts

.q

But it also works without specifying any size, like

pmove

solo or whatever you want (.b, .w, .l, .d, .q ..)
It's probably the right choice because

crp

is 64-bit anyway