Terriblefire Accelerators

Help & news on accelerators from TF, Amiga, Atari, CD32 etc

Moderator: terriblefire

Forgottenmyname
Trusted Guru
Trusted Guru
Posts: 553
Joined: Fri Mar 23, 2018 8:22 pm

Re: Terriblefire Accelerators

Post by Forgottenmyname » Thu Nov 08, 2018 5:22 pm

That's good news as I'm very slowly building one to pop into a Mega 4.

terriblefire
Moderator Team
Moderator Team
Posts: 71
Joined: Mon Aug 28, 2017 10:56 pm
Location: Glasgow, UK
Contact:

Re: Terriblefire Accelerators

Post by terriblefire » Thu Nov 08, 2018 6:28 pm

Forgottenmyname wrote:
Thu Nov 08, 2018 5:22 pm
That's good news as I'm very slowly building one to pop into a Mega 4.
Current status with the ST firmware..

The 030 does not like to format a floppy with identical firmware to the 020. I need to investigate but time is limited.

User avatar
rpineau
Site Admin
Site Admin
Posts: 386
Joined: Thu Aug 17, 2017 6:08 pm
Location: USA
Contact:

Re: Terriblefire Accelerators

Post by rpineau » Thu Nov 08, 2018 7:06 pm

IT's probably a cache issue. You should only allow caching for the RAM, TOS and eventually fastRam (AltRam).
This is the VHDL from the test we did on the 68020 board when we put a 68030 on it (via a small adapter) :

Code: Select all

	-- 68030 experimentation
	CIIN <= '1' when  (Address_bus >= x"00E00000" and Address_bus < x"00F00000") -- tos
				or (Address_bus >= x"FFE00000" and Address_bus < x"FFF00000") -- tos shadow
				or (Address_bus >= x"00000000" and Address_bus < x"00400000") -- ram
				or (Address_bus >= x"FF000000" and Address_bus < x"FF400000") -- ram shadow
				or (Address_bus >= x"01000000" and Address_bus < x"09000000") -- fast ram 128MB
			else '0';
Working ones : MegaSTE (68020) / TT030 / Falcon with AB040 & Eclipse / 1040STF
Need testing : Falcon with CT2

terriblefire
Moderator Team
Moderator Team
Posts: 71
Joined: Mon Aug 28, 2017 10:56 pm
Location: Glasgow, UK
Contact:

Re: Terriblefire Accelerators

Post by terriblefire » Thu Nov 08, 2018 7:41 pm

rpineau wrote:
Thu Nov 08, 2018 7:06 pm
IT's probably a cache issue. You should only allow caching for the RAM, TOS and eventually fastRam (AltRam).
This is the VHDL from the test we did on the 68020 board when we put a 68030 on it (via a small adapter) :

Code: Select all

	-- 68030 experimentation
	CIIN <= '1' when  (Address_bus >= x"00E00000" and Address_bus < x"00F00000") -- tos
				or (Address_bus >= x"FFE00000" and Address_bus < x"FFF00000") -- tos shadow
				or (Address_bus >= x"00000000" and Address_bus < x"00400000") -- ram
				or (Address_bus >= x"FF000000" and Address_bus < x"FF400000") -- ram shadow
				or (Address_bus >= x"01000000" and Address_bus < x"09000000") -- fast ram 128MB
			else '0';
Possibly. I have the CDIS jumper set though. So i dont think thats the issue here in this case. Caches were my first thought too.

In my code I also assert CIIN any time i access anything but the fast ram. My fast ram is from 0x40000000 to 0x403FFFFF

User avatar
rpineau
Site Admin
Site Admin
Posts: 386
Joined: Thu Aug 17, 2017 6:08 pm
Location: USA
Contact:

Re: Terriblefire Accelerators

Post by rpineau » Thu Nov 08, 2018 9:15 pm

Yep, with CDIS jumpered if it still fails it obviously not a cache issue.
The floppy uses the DMA so may be a BR/BG/BGACK issue with retaking the bus to quickly with DMA data lingering on the bus.
On our 68020 board we resync these (I assume you do the same) to the 8MHz clock.
If I think of something I'll post here.
Working ones : MegaSTE (68020) / TT030 / Falcon with AB040 & Eclipse / 1040STF
Need testing : Falcon with CT2

terriblefire
Moderator Team
Moderator Team
Posts: 71
Joined: Mon Aug 28, 2017 10:56 pm
Location: Glasgow, UK
Contact:

Re: Terriblefire Accelerators

Post by terriblefire » Thu Nov 08, 2018 9:17 pm

I only allow BGACK to be sampled to the CPU when the CPU_AS and AMIGA_AS are the same value. which has the same effect.

If you havent seen my LA setup already checkout the YT channel. I can actually capture the DMA cycles at 500Mhz.

EDIT: The odd bit is this working with the same HDL code on the 020. Wonder if there is some difference in behaviour.

8 Bit Dreams
Posts: 27
Joined: Fri Nov 09, 2018 7:12 am

Re: Terriblefire Accelerators

Post by 8 Bit Dreams » Fri Nov 09, 2018 7:19 am

Hello,
Happy to see the story continues,
Still hope to design a CPU relocator with ribbon cable for my ST

8 Bit Dreams
Posts: 27
Joined: Fri Nov 09, 2018 7:12 am

Re: Terriblefire Accelerators

Post by 8 Bit Dreams » Fri Nov 09, 2018 11:58 am

Have updated info on my For Sale thread on Amibay regarding TF5XX cards

terriblefire
Moderator Team
Moderator Team
Posts: 71
Joined: Mon Aug 28, 2017 10:56 pm
Location: Glasgow, UK
Contact:

Re: Terriblefire Accelerators

Post by terriblefire » Fri Nov 09, 2018 12:28 pm

Excellent.

One thing worth mentioning with all the TF 030 cards on the ST.. they *read*from floppy fine. But write is broken. It doesn’t write a single sector correct.

I’ve yet to debug properly but at first I thought it was the 7416245 chip that cuts off the day bus from the 030 section. But the problem is identifical on the Tf530 which doesn’t have it.

All three TF53x versions have the same issue with cache disabled and identical code. The Tf520 works with the same code. It’s v odd.

terriblefire
Moderator Team
Moderator Team
Posts: 71
Joined: Mon Aug 28, 2017 10:56 pm
Location: Glasgow, UK
Contact:

Re: Terriblefire Accelerators

Post by terriblefire » Sat Nov 10, 2018 11:10 pm

Ok.. new board announcement. TF330 - New CD32 Board. Its an 030 with a buttload of RAM.

Specs to be discussed tomorrow.. but please speculate here :)

Image
Attachments
Capture.JPG
Capture.JPG (277.29 KiB) Viewed 152 times

Post Reply