You could do 16 bit transfers rather than 8 bit (there wasn't enough address lines to do 16 bit writes on the ROM port).
You could also add IRQ, I think I was going to use the one on the printer port. I wasn't much good at software back so knew the hardware worked and my test code showed the IRQ working but didn't take it any further.
You would need to re-write the ASM parts that read/write the data in the driver - thats it really, probably just a few lines. Then move the base address I guess to somewhere and decode it in address space. We used the ROM3 and ROM4 to control the reads and write lines (which are bascially just address decoded by the GLUE). Don't know if the source is available for the drivers, I have the Ozk version of driver as source code, was told not to release that so stuck by my word.
Sounds like even with a cleaner interface to the RTL8019AS you aren't going to improve speed. Perhaps the limitation is the lack of IRQ or just the TCP/IP stack. I seem to remember getting better performance from Mintnet than Sting.
LAN project for someone using 68K expansion slot on the H5
-
Stimpy
- Posts: 36
- Joined: 03 Apr 2018 21:19
-
exxos
- Site Admin

- Posts: 28357
- Joined: 16 Aug 2017 23:19
- Location: UK
Re: LAN project for someone using 68K expansion slot on the H5
I'm not really sure how fast the GLUE deals with the ROM port. I'm not really looking at improving speed at this point. But ultimately with a faster CPU with alt-ram things could potentially speed up a fair bit anyway.Stimpy wrote: 29 Nov 2022 18:54 Sounds like even with a cleaner interface to the RTL8019AS you aren't going to improve speed. Perhaps the limitation is the lack of IRQ or just the TCP/IP stack. I seem to remember getting better performance from Mintnet than Sting.
-
mikro
- Posts: 820
- Joined: 28 Aug 2017 23:22
- Location: Kosice, Slovakia
Re: LAN project for someone using 68K expansion slot on the H5
... and this is exactly, from the first to the last line, what the Falcon LAN adapter did. ;-) It uses 16-bit transfer, it doesn't use IRQ, it changed a few asm lines in the original driver, sting was indeed way slower than mintnet aaaand.... the speed in general wasn't that much different from ROM port, despite Steve's expectations.Stimpy wrote: 29 Nov 2022 18:54 You could do 16 bit transfers rather than 8 bit (there wasn't enough address lines to do 16 bit writes on the ROM port).
You could also add IRQ, I think I was going to use the one on the printer port. I wasn't much good at software back so knew the hardware worked and my test code showed the IRQ working but didn't take it any further.
You would need to re-write the ASM parts that read/write the data in the driver - thats it really, probably just a few lines. Then move the base address I guess to somewhere and decode it in address space. We used the ROM3 and ROM4 to control the reads and write lines (which are bascially just address decoded by the GLUE). Don't know if the source is available for the drivers, I have the Ozk version of driver as source code, was told not to release that so stuck by my word.
Sounds like even with a cleaner interface to the RTL8019AS you aren't going to improve speed. Perhaps the limitation is the lack of IRQ or just the TCP/IP stack. I seem to remember getting better performance from Mintnet than Sting.
-
Stimpy
- Posts: 36
- Joined: 03 Apr 2018 21:19
Re: LAN project for someone using 68K expansion slot on the H5
Sure, but it might be useful to try IRQ. And it looks like that Falcon adaptor was based on the older driver from Thomas Redelberger.mikro wrote: 30 Nov 2022 07:52... and this is exactly, from the first to the last line, what the Falcon LAN adapter did. ;-) It uses 16-bit transfer, it doesn't use IRQ, it changed a few asm lines in the original driver, sting was indeed way slower than mintnet aaaand.... the speed in general wasn't that much different from ROM port, despite Steve's expectations.Stimpy wrote: 29 Nov 2022 18:54 You could do 16 bit transfers rather than 8 bit (there wasn't enough address lines to do 16 bit writes on the ROM port).
You could also add IRQ, I think I was going to use the one on the printer port. I wasn't much good at software back so knew the hardware worked and my test code showed the IRQ working but didn't take it any further.
You would need to re-write the ASM parts that read/write the data in the driver - thats it really, probably just a few lines. Then move the base address I guess to somewhere and decode it in address space. We used the ROM3 and ROM4 to control the reads and write lines (which are bascially just address decoded by the GLUE). Don't know if the source is available for the drivers, I have the Ozk version of driver as source code, was told not to release that so stuck by my word.
Sounds like even with a cleaner interface to the RTL8019AS you aren't going to improve speed. Perhaps the limitation is the lack of IRQ or just the TCP/IP stack. I seem to remember getting better performance from Mintnet than Sting.
Anyway, if it's simplicity that's required I'd just stick with the "hacky" original design - it works and there doesn't seem to be any advantage of changing it. Then just change the base address away from the cartridge port in busenex.i include file.
Who is online
Users browsing this forum: ClaudeBot and 5 guests