Neat coding style and good habits

[phx]

Originally Posted by Thorham

I want REAL 68k with REAL Amiga chipset, not an FPGA computer with Amiga stickers on it.

Have to agree here. A Minimig with an FPGA chipset and a real 68SEC000 is already questionable, and certainly the limit for me. But when replacing the CPU by FPGA then the system loses its heart.

[daxb]

The heart all the same. The soul is important.

[Pyromania]

@Mrs Beanbag

Thanx for starting a great thread!

@phx

FPGA's are people too and they have a heart.

[matthey]

Originally Posted by meynaf

Not a big deal, as varargs aren't very common. Nearly only printf-like stuff use it.

Most of the standard C vararg functions are I/O related.

fprintf(), printf(), sprintf(), snprintf(), vprintf(), vfprintf(), vsprintf(), vsnprintf(), fscanf(), scanf(), sscanf(), vscanf(), vfscanf(), vsscnanf()

There can be user defined varargs though too. They are security risks as well as less efficient. Too bad standard C alternative functions were not offered with C99 or C11 instead of increasing varargs support. It is not like the embedded or server C guys are properly represented in the C standards committees though.

Originally Posted by meynaf

Whose the fault ? Poor 68k isn't responsible of compilers doing stupid things. In fact the greatest handicap of 68k compared to other cpus (in the ancient times where it was still mainstream), was poor compiler generated code.

The 68k ISA is not the easiest for compilers. It lacks many modern and mainstream functions. Ease and efficiency of compilers for the CPU are probably *more* important than peak performance which most CPU designers are targeting usually while ignoring compiler design and implications. The huge blunders I mentioned could only take place with an almost complete disconnect between CPU designers (and their managers) and compiler designers. There was and is a problem which keeps repeating.

Originally Posted by meynaf

I wouldn't regard MOV3Q as useful, even for that case. What's the win ? 2 bytes every time you push a small value. Big deal.

Yes. MOV3Q replaces PEA to put small immediates on the stack.

Originally Posted by meynaf

True. Some additions are neat, though. MVZ, MVS, BITREV, BYTEREV are quite missing in classical 68k.

Some of the new functionality is good. MVZ and MVS are powerful and useful instructions with good encodings which help compilers but where did those names comes from? Not the 68k where they might be SXTB.L, SXTW.L, ZXTB.L, ZXTW.L, BREV.L and REVB.L. There may be other alternatives which could be voted on but some projects are dictatorships with the dictators surrounded by their yes men.

Originally Posted by meynaf

It seems not many people care at all.
But good programming starts with a good architecture.

You are one of the few who cares. I wish you would respect the importance of compilers more though. A CPU goes nowhere today without good compiler support.

Originally Posted by meynaf

Perhaps the 68k needs some kind of worthy successor, which unfortunately isn't gonna happen either.

I agree. It wouldn't be difficult to make the most powerful CISC ISA in the world (more powerful and much prettier than x86_64) but nobody would build it.

Originally Posted by Thorham

That's certainly not what I want. I want REAL 68k with REAL Amiga chipset, not an FPGA computer with Amiga stickers on it.

Originally Posted by phx

Have to agree here. A Minimig with an FPGA chipset and a real 68SEC000 is already questionable, and certainly the limit for me. But when replacing the CPU by FPGA then the system loses its heart.

I think we would all like to have a hard (ASIC) CPU instead of FPGA CPU. A FPGA CPU is usable and affordable today and that is how they are designed and debugged. An FPGA chipset doesn't have as many disadvantages and has several advantages.

[modrobert]

Originally Posted by matthey

I think we would all like to have a hard (ASIC) CPU instead of FPGA CPU. A FPGA CPU is usable and affordable today and that is how they are designed and debugged. An FPGA chipset doesn't have as many disadvantages and has several advantages.

I agree.

Logic gates in FPGA or ASIC makes no difference to the core functionality, it is just implemented differently in hardware. There seems to be a lot of confusion on the subject where "emulation" is associated incorrectly.

The problem is more about reverse engineering the original CPU to make the FPGA design fully compatible, and actually putting the main effort into compatibility before adding any new instructions.

Personally I would rather have a 68060 in FPGA thoroughly tested "cycle exact" at 50MHz which can be overclocked reliably to 100MHz, than a beefed up 68040 at 200MHz with added 64 bit instructions and compatibility issues.

[meynaf]

Originally Posted by matthey

The 68k ISA is not the easiest for compilers. It lacks many modern and mainstream functions.

I do not see anything "modern and mainstream" that the 68k lacks.
Sure in todays cpu there can be a lot of stuff that's not in the 68k. But i don't think we need all that bloat.
Neat asm coding style and good habits lead nowhere if the cpu is a complete mess.

Originally Posted by matthey

You are one of the few who cares. I wish you would respect the importance of compilers more though. A CPU goes nowhere today without good compiler support.

The situation is quite balanced. You know what's needed for compilers, i know what's needed for handwritten asm.
The only problem is that we both got nearly fired out of the design team.

[Thorham]

Originally Posted by matthey

I think we would all like to have a hard (ASIC) CPU instead of FPGA CPU.

I already have what I want: A 68030 CPU.

[matthey]

Originally Posted by meynaf

I do not see anything "modern and mainstream" that the 68k lacks.
Sure in todays cpu there can be a lot of stuff that's not in the 68k. But i don't think we need all that bloat.
Neat asm coding style and good habits lead nowhere if the cpu is a complete mess.

The 68k has certainly aged better than the original x86 and ARM as it has fewer limitations. The lack of modern and mainstream support just makes code slower and compiler support more difficult as the functionality is still possible without it. Code has changed a lot since the early days of the 68k when assembler code was not unusual, integers were more commonly 8 and 16 bits, floating point code and an FPU were rare and C and OOP were leading edge technologies embraced by the Amiga.

Originally Posted by meynaf

The situation is quite balanced. You know what's needed for compilers, i know what's needed for handwritten asm.
The only problem is that we both got nearly fired out of the design team.

Originally Posted by Thorham

I already have what I want: A 68030 CPU.

I'm glad that you like your 68030 but you might be one of the few Amiga users who wouldn't use a free 68060 accelerator. I guess there are some gamers who would prefer the 68030 but you claim that you aren't a big gamer either. Then you don't think the AmigaOS was very good. Your classification might be unique .

[meynaf]

Originally Posted by matthey

The 68k has certainly aged better than the original x86 and ARM as it has fewer limitations. The lack of modern and mainstream support just makes code slower and compiler support more difficult as the functionality is still possible without it. Code has changed a lot since the early days of the 68k when assembler code was not unusual, integers were more commonly 8 and 16 bits, floating point code and an FPU were rare and C and OOP were leading edge technologies embraced by the Amiga.

But what exactly do you call "modern and mainstream support" ?

[Thorham]

Originally Posted by matthey

The lack of modern and mainstream support

Like meynaf, I'd like to know what that means.

Originally Posted by matthey

I'm glad that you like your 68030 but you might be one of the few Amiga users who wouldn't use a free 68060 accelerator.

Not as my main system if I don't have to, no. Too fast. 68000 is too slow, 68060 is too fast. This is programming related.

Originally Posted by matthey

I guess there are some gamers who would prefer the 68030 but you claim that you aren't a big gamer either.

Not on the Amiga, no. Practically the only thing I play on my Amiga is meynaf's Dungeon Master port.

Originally Posted by matthey

Then you don't think the AmigaOS was very good.

I don't think it's very good considering it could be better and faster on the same hardware.

Originally Posted by matthey

Your classification might be unique .

Perhaps.

[matthey]

Originally Posted by meynaf

But what exactly do you call "modern and mainstream support" ?

Most modern processors include instructions like the ColdFire MVS, MVZ and BYTEREV for example. These operations were less common when the 68k was new but are common and practically expected today. Later 68k processors removed the integer 64 bit result MULS.L/MULU.L, several fp instructions, etc. which even later processors like ARM and x86_64 added. The 68k needs too many branches because of lack of a few instructions like shift with immediates >8 although fast bit field instructions, predication, etc. can help this. Like I said, the type of code has changed dramatically from when the 68k was created to now. A more modern CPU design makes a bigger difference for modern code but ISA/ABI changes can be helpful too. Compiler designers aren't going to spend much time with a small market CPU like the 68k so we want to impress them with how easy to use, consistent, standard and modern an enhanced 68k can be. An attractive face lift for the 68k might attract some outside attention or bring 68k/CF fans to take a 2nd look. A few changes may not seem like much but look what ARM Holdings has done for 15 billion processors to be sold in 2015 alone, employ more than 3000 and deserve an ~40% buyout premium recently when Brexit gave an opportunity. An enhanced 68k can give better single core performance, a smaller foot print and better ease of use than ARM's Thumb 2. There really isn't that much difference but ARM rules the world and the 68k is thrown in the trash bin. ARM has an advantage in power efficiency with simple processors but I don't think that is true for higher performance processors where OoO and many weak cores are needed for performance. Modern support alone is not enough to get the 68k to the success of ARM but the 68k status quo is going nowhere.

[Thorham]

Originally Posted by matthey

Modern support alone is not enough to get the 68k to the success of ARM but the 68k status quo is going nowhere.

Unless you become extremely rich, 68k is never going to go anywhere.

As for FPGA computers: If you can do anything you want, why not try something completely new? If you want to move ahead, then you have to let Amiga go, it's just that simple.

[alpine9000]

Originally Posted by Thorham

Unless you become extremely rich, 68k is never going to go anywhere.

As for FPGA computers: If you can do anything you want, why not try something completely new? If you want to move ahead, then you have to let Amiga go, it's just that simple.

For people interested in an FPGA core with a fully supported modern tool chain check out j-core

The Super-H instruction set is quite nice compared to other "modern" processors.

[meynaf]

Originally Posted by matthey

Like I said, the type of code has changed dramatically from when the 68k was created to now.

Code mostly changed quantitatively. The building blocks remain the same. 68k only lacks few things, and a lot less than any other architecture.

Originally Posted by matthey

Compiler designers aren't going to spend much time with a small market CPU like the 68k so we want to impress them with how easy to use, consistent, standard and modern an enhanced 68k can be. An attractive face lift for the 68k might attract some outside attention or bring 68k/CF fans to take a 2nd look.

Compiler designers will care only when we have a significant market share. Else they won't, regardless of how 'nice' the cpu is.
And we won't get there without a good Ghz implementation, which isn't gonna happen.

Originally Posted by matthey

An enhanced 68k can give better single core performance, a smaller foot print and better ease of use than ARM's Thumb 2.

No isa changes are needed for that. We're already better than ARM's thumb 2. The Mhz range is our only handicap.

Originally Posted by matthey

There really isn't that much difference but ARM rules the world and the 68k is thrown in the trash bin. ARM has an advantage in power efficiency with simple processors but I don't think that is true for higher performance processors where OoO and many weak cores are needed for performance. Modern support alone is not enough to get the 68k to the success of ARM but the 68k status quo is going nowhere.

ARM rules the world because there is no serious concurrence in its market segment. It's maybe fine for handheld devices but it will never beat the others in high performance and so it'll remain where it is. Same with x86, which will never kick the ARM out because of its stupidly high (and useless) complexity.

[chaos]

Originally Posted by Thorham

Unless you become extremely rich, 68k is never going to go anywhere.

As for FPGA computers: If you can do anything you want, why not try something completely new? If you want to move ahead, then you have to let Amiga go, it's just that simple.

Exactly

[demolition]

Originally Posted by meynaf

ARM rules the world because there is no serious concurrence in its market segment. It's maybe fine for handheld devices but it will never beat the others in high performance and so it'll remain where it is.

ARM's new 64-bit server CPUs are gaining serious traction since they offer very high performance with significantly lower power requirements than any of the competitor's offerings. So this 'it will never beat the others in high performance' is definitely not correct..

[meynaf]

Originally Posted by demolition

ARM's new 64-bit server CPUs are gaining serious traction since they offer very high performance with significantly lower power requirements than any of the competitor's offerings. So this 'it will never beat the others in high performance' is definitely not correct..

High performance has never meant better performance.
Sure you have lower power requirements. But the performance, though maybe good enough, is lower as well.

[demolition]

'Better performance' is a relative term. Why does one have to compare the performance of one CPU of one model with one CPU of another model? One could just as well compare the performance you could get from a 1U rack with Xeon CPUs and another 1U rack with ARM CPUs. Since the cooling might be the limiting factor in how much CPU die area you could squeeze in there, the ARM rack could fit a higher CPU die area and end up with overall better performance with the same power cost.

For servers, one of the most important factors are also the performance/power since any dissipated power needs equal amounts of active cooling.

For gaming desktop PCs this may not so relevant, but they make up a very little part of the computer market. Laptops, tablets and phones are very prominent and they all benefit heavily from lower power requirements.

[meynaf]

Originally Posted by demolition

'Better performance' is a relative term. Why does one have to compare the performance of one CPU of one model with one CPU of another model? One could just as well compare the performance you could get from a 1U rack with Xeon CPUs and another 1U rack with ARM CPUs. Since the cooling might be the limiting factor in how much CPU die area you could squeeze in there, the ARM rack could fit a higher CPU die area and end up with overall better performance with the same power cost.

That may be true, however performance doesn't scale linearly with the number of cpus. I'd like to see real numbers.

Originally Posted by demolition

For servers, one of the most important factors are also the performance/power since any dissipated power needs equal amounts of active cooling.

Right, but I still don't see arm kicking the others out of that market...

Originally Posted by demolition

For gaming desktop PCs this may not so relevant, but they make up a very little part of the computer market. Laptops, tablets and phones are very prominent and they all benefit heavily from lower power requirements.

Yet if you benchmark the average x86 vs the average arm, the arm is everything but impressive. I don't want that in my laptop.

[demolition]

Originally Posted by meynaf

Yet if you benchmark the average x86 vs the average arm, the arm is everything but impressive. I don't want that in my laptop.

The new ARM 64-bit platform that is coming out is nothing like the ARM performance we are used to. It is very promising and performance-wise much better than the predecessors so it can't be compared. I have heard from many places that unless Intel comes up with something significant very soon, they will lose a major part of the server market very rapidly.

The problem here with the consumer PC market is that people want to be able to run x86 code and because of that it is hard to move them to any other platform. Even if ARM's new 64 bit platform was 50% better than x86, it would still be uphill due to the lacking compatibility.