68060 based QL system?

[Peter]

I have agitated for porting to ARM for many years. It wouldn't be trivial, but it would be amazing.

Few things are more unrealistic than porting SMSQ/E to the ARM architecture. So I never even start thinking of it. And personally I like the beauty of the 68K architecture and assembler language.

Nasta's solution: The big step up in performance came with the '040, so have multiple core '040 systems.

I wouldn't call it a solution, but wishful thinking. At least wishful thinking that is more realistic than a port to ARM, and multicore is well suited for FPGA.

The '060 didn't add much except higher clocks at disproportionately higher prices.

I disagree here. The 68060 is significantly faster per clock, consumes much less power, runs at a useful logic voltage.

[Dave]

I have agitated for porting to ARM for many years. It wouldn't be trivial, but it would be amazing.

Few things are more unrealistic than porting SMSQ/E to the ARM architecture. So I never even start thinking of it. And personally I like the beauty of the 68K architecture and assembler language.

Well, thank you for putting so much thought into it before sh!tt!ng all over it. I wonder how far Sinclair would have gone if every time he had an idea, the first thing he heard was "that's unrealistic"... It's a big job. Maybe bigger than the community could ever do. It is *technically* feasible. There is just a manpower problem. But hey, thanks for putting so much thought into it.

Nasta's solution: The big step up in performance came with the '040, so have multiple core '040 systems.

I wouldn't call it a solution, but wishful thinking. At least wishful thinking that is more realistic than a port to ARM, and multicore is well suited for FPGA.

What do you think the whole point of Goldfire was? That Coldfire core was a 68060 with multiprocessing logic.

Or what if you have a $2 68SEC000 running the QL and IO, and a 68060 running SMSQ/E with its own video system, and a small dual port IO area shared between them? The '000 would be a helper processor doing IO for the '060. Feasible? Yes! Reasonable? Yes! Cheaper than an FPGA? In manufacturing cost about the same, in design cost far cheaper. Have the '000 run background tasks that don't require screen space? SURE! Have the '060 do the heavy lifting? GREAT! It the CPUs were symmetrical? Even better!

The '060 didn't add much except higher clocks at disproportionately higher prices.

I disagree here. The 68060 is significantly faster per clock, consumes much less power, runs at a useful logic voltage.

Did you fall out of the wrong side of the bed, land on lego, break your coffee maker then decide it would be good to post on QL Forum?

*grins* It's ok, I forgive you. We all have bad days sometimes.

[Peter]

What do you think the whole point of Goldfire was? That Coldfire core was a 68060 with multiprocessing logic.

No, it was a single CPU accelerator card based on the MCF5102, essentially a cutdown 68EC040, that never materialized.

[Peter]

The '060 didn't add much except higher clocks at disproportionately higher prices.

I disagree here. The 68060 is significantly faster per clock, consumes much less power, runs at a useful logic voltage.

Did you fall out of the wrong side of the bed, land on lego, break your coffee maker then decide it would be good to post on QL Forum?

No, but I designed mainboards with both 68040 and 68060 that actually shipped.

Oversimplified, major improvements of the 68060 are double sized caches, dual integer execution units and branch prediction/cache. At same clock rate, 1.7 times the 68040 performance was typical. And that has to be multiplied with the clock ratio. Power consumption: If you don't want to read the datasheet, just place a finger on a 40 MHz 68040.

[XorA]

Few things are more unrealistic than porting SMSQ/E to the ARM architecture. So I never even start thinking of it.

Good thing AmigaOS guys are more positive (although they do have advantage of bigger usebase)

[dilwyn]

Did you fall out of the wrong side of the bed, land on lego, break your coffee maker then decide it would be good to post on QL Forum?
*grins* It's ok, I forgive you. We all have bad days sometimes.

He he, Dave swallowed his Sarcasm pills again - he obviously speaks fluent sarcasm

(Dilwyn runs for cover before he cops it too...)

[Dave]

Hey, at least I'm *taking* my meds

[mk79]

On one hand, I see that without a second, non-680x0 CPU, QL style hardware may never get support for some modern peripherals. On the other hand, dealing with that second CPU is relatively boring, not an interesting challenge for me.

Well, then go get a max3421 chip and write the SMSQ/E driver, that should be relatively "un-boring" task while actually still being feasible. Thought about it myself, actually have the chip here but didn't find the time to play with it.

Marcel

[mk79]

I have agitated for porting to ARM for many years. It wouldn't be trivial, but it would be amazing.

Few things are more unrealistic than porting SMSQ/E to the ARM architecture. So I never even start thinking of it. And personally I like the beauty of the 68K architecture and assembler language.

Well, thank you for putting so much thought into it before sh!tt!ng all over it. I wonder how far Sinclair would have gone if every time he had an idea, the first thing he heard was "that's unrealistic"... It's a big job. Maybe bigger than the community could ever do. It is *technically* feasible. There is just a manpower problem. But hey, thanks for putting so much thought into it.

You're right, but even after thinking about it for a few seconds it's still about as realistic as the QL community launching its own Mars probe. Still, it's *technically* feasible! The Mars probe at least.

Marcel

[Dave]

You're right, but even after thinking about it for a few seconds it's still about as realistic as the QL community launching its own Mars probe. Still, it's *technically* feasible! The Mars probe at least.

As a thought exercise: if you omit the not insignificant manpower problem - what is the show-stopper missing feature in ARM cores that makes this less practicable than rocket science?

OS X is two or three orders of magnitude more complex than SMSQ/E, and it has been ported from 68K and PowerPC to x86 to x86-64 to ARM.