Developers Try Again To Upstream Motorola 68000 Series Support In LLVM (phoronix.com) 69
Hobbyist developers are trying once again to get a Motorola 68000 back-end merged into the upstream LLVM compiler. Phoronix reports: The Motorola 68000 series processors have been around since the 80's thanks to the likes of the early Apple Macintosh computers. Fast forward to 2020, the Motorola 68000 is still a popular target for vintage computer enthusiasts and hobbyists. Community developers have worked on improving the Linux kernel support for M68k hardware like early Apple Powerbooks as recently as a few years ago and the compiler support is a continued target. GCC 11 due out next year was looking to drop the M68k target over its unmaintained status. Hobbyists though stepped up there so the M68k support will remain in GCC. Now developers are also looking at adding M68k support to the LLVM compiler.
This isn't the first time that M68k support for LLVM has been brought up albeit never successfully landed to date. Building off the past failures to get the Motorola 68000 series support upstreamed, developers last week sent out new patches proposing this back-end -- this time they are showing more clarity about the developers involved and being committed to supporting the code, the sustainability of the code, and responding quickly to code review comments. This patch series is the latest attempt at upstreaming Motorola 68000 series support in LLVM. Besides all the back-end specific code there is also some common LLVM code changes that fall under greater scrutiny.
This isn't the first time that M68k support for LLVM has been brought up albeit never successfully landed to date. Building off the past failures to get the Motorola 68000 series support upstreamed, developers last week sent out new patches proposing this back-end -- this time they are showing more clarity about the developers involved and being committed to supporting the code, the sustainability of the code, and responding quickly to code review comments. This patch series is the latest attempt at upstreaming Motorola 68000 series support in LLVM. Besides all the back-end specific code there is also some common LLVM code changes that fall under greater scrutiny.
Re: (Score:3)
What, you mean you don't write all your apps with Carbon APIs so that you can run them on Mac OS 8.1?
Re: (Score:3)
Carbon is PPC only.
Re: (Score:2)
Oh, yeah. Good point. I mean, in theory, IIRC, Carbon is a subset of the APIs available on classic Mac OS, so that's mostly a library linkage distinction, but still...
Re: Why? (Score:2)
Not really, it adds a bunch of stuff as well. It gives ways to provide menu definition procedures etc. as code pointers rather than the old MDEF/WDEF/etc. resources. It also provides accessors for structures that become opaque, and the Carbon event loop stuff. None of that's available without CarbonLib, which is PPC-only. You can't make a non-trivial Carbon application build for 68k without a whole lot of #ifdef.
Re: (Score:2)
Bummer. Never mind then. Guess I'll have to write a wrapper library before I rewrite all my Cocoa apps for Mac68k. :-D
Re: (Score:2)
One of the things that cross platform tools improve is hardware abstraction. It makes the tools more portable and capable of being expanded to support new hardware. It forces developers to not tailor the architecture into one niche.
I have NetBSD installed on one of my SE/30s. It isn't something I boot up and use everyday, but it is cool to run the same OS from the same source tree on a 68030 and the latest multicore x86.
Re:Why? (Score:4, Insightful)
Variants of the processor are still around, still used in embedded applications and sold after 40 years. How is it dumb?
Re: (Score:3, Insightful)
You may not need it; but if enough people need it and can/do put in the work you should censor their contributions... you simply need to facilitate a non-mainstream path for various groups to do what they want. open source's strength is in this sort of empowerment; however, you can make things more or less difficult for these fringe users. It is wise to not alienate them while also not to bending over backwards because their contributions may leak into the mainstream project which benefits all.
Furthermor
Re:Why? (Score:5, Insightful)
Because someone still needs this support. The entire reason GCC is popular is because it has support for unpopular CPUs. They don't say "go and get the 1992 version of our compiler and stop bothering us". It's open source, devs can still donate their own time and effort to support this. If you only care about what's the most popular then there are other compiler families to cater to you.
Also what you may think is unpopular may still be widely used. Ie, MIPS, the workstation from the 90s, has its instruction set used in many chip designs. 68000 was used in some relatively popular chips for awhile (Freescale Dragonball in mid to late 90s), and such devices may still be in use today.
Re: (Score:1)
Eh, with all the new hardware CPU vulnerabilities appearing lately, I'm starting to think modern processors are just getting too complex to be secured by a poorly-regulated staff of the lowest-bidding foreign nationals' contractor sweatshop. I think a revival of the these M68K chips is inevitable and overdue. If nothing else, think of the children.
Re: (Score:2)
To revive 68000 to compete with mainstream processors, it would have to be redesigned to have multiple 64 bit cores. In addition to X86, it would have to compete with ARM designs. It would represent a major financial risk for whoever undertook the effort. In all likelihood, there'd be a lot of cruft involved with making a 68040 compatibility mode, and there'd be a lot of griping about it not being a pure 68000 family processor.
I'd love to see it, but I don't think its a good idea.
Re:Why? (Score:4, Interesting)
This seems incredibly dumb. Why add in support for something that is of so little use?
There's still a lot of stuff around that uses 68K derivatives, e.g. Dragonball, Coldfire, 68xx060, etc. Having a modern compiler for them would be a boon.
Re: (Score:2)
There's still a lot of stuff around that uses 68K derivatives, e.g. Dragonball, Coldfire, 68xx060, etc. Having a modern compiler for them would be a boon.
Is GCC10 not recent enough for you?
Coldfire (Score:5, Insightful)
Coldfire, an embedded 68k-based architecture, is still in use in industrial control applications. Those installations don't like to change out hardware. We're talking 20-30 year support timeframes. If developers can get a better compiler for their architecture (I can guarantee LLVM is better than the janky 20-year-old compiler NXT is probably selling you) then it makes their lives that much easier.
Re: (Score:3)
Coldfires are nice little chips, though Motorola (then Freescale, then NXP) never put the work into making them faster, just as they never made the M68xxx parts anywhere near speed-competative with Intel parts. Many people look at the register set and the basic opcodes and assume these are M68K parts, but they are extremely crippled RISC-ified versions of the glorious instruction set of the actual 680x0 family. At a time when the Intel chips had many opcodes that used only certain registers, the 680x0 famil
Re: (Score:2)
... is still in use in industrial control applications.
In CNC machines, the computer control stuff generally goes obsolete well before the machine. You can expect a good thirty years service life out of a machine tool in regular use. Thirty year old computer hardware is practically prehistoric. It was not so long ago that some precision machinists I know were still running Windows 3.1 for driving a CNC machine.
At work, we bought some second-hand pick-and-place machines for electronic assembly. One day, the 5V rail on the PSU went overvoltage, and fried the "br
Re: (Score:2)
In CNC machines, the computer control stuff generally goes obsolete well before the machine.
Big time. 20 years ago I worked for an industrial laser manufacturer/integrator, and plenty of our old machines are still out there and in use even though they cost a fortune to run compared to modern fiber units. The main problem is that the custom control boards were ISA cards, and good luck finding an ISA motherboard that doesn't cost a fortune nowadays. Consequently, a small but healthy market has cropped up
Re: (Score:1)
There still appear to be a lot of PICMG and PCISA boards/backplanes out there.
Its been a while since I dealt with industrial passive backplanes, but they at least used to be a reasonably affordable solution when either old ISA cards, or large numbers of bridged PCI slots were needed. Not cheap, but reasonable, especially when you consider the entire market is built around industrial-grade requirements and the expense of moving to newer controllers.
Went on e-bay to look at used equipment and got a little no
Re: (Score:3)
Mostly because it is not of little use at all. COS 68K systems are still in use, for one. It is a cheap, capable, general purpose CPU which is used for embedded and industrial use.
In addition, there is still development being done for older platforms, like the Amiga, where a more efficient compiler would be extremely useful. And that is growing, not shrinking, since there is newly developed Amiga hardware out there. And with the rise of FPGA machines, that will become even larger.
Something like the 68K is p
Re: (Score:2)
68k has been dying hard ever since ARMs got cheap. You can literally get complete finished devices with ARM cores that blow away literally any 68k processor, for single-digit numbers of dollars for the entire device, not just the processor. 68k was great in its day, but its day is now over, and ARM ended it. It's not like nobody is using it, but nobody is using it for any reasons other than legacy code.
If nvidia ruins ARM we will end up on RISC-V, not 68k. 68k scales, but not cheaply like ARM.
Re: (Score:1)
You can get complete finished 68K devices for that price point as well. If anything is over, it's ARM, and NVidia ended it. ;)
Re: (Score:2)
"You can get complete finished 68K devices for that price point as well."
Yes but they aren't '060s, they are more equivalent to older 68k processors, and the ARMs spank them.
Re: (Score:3, Insightful)
Re: (Score:1)
I used to be fond of having three periods in a row in my posting,
If you are using a Mac you can press option + colon to get an ellipsis as a single character. There are ways to do it on *nix and Windows but they are more cumbersome and I don't know it off the top of my head.
Re: (Score:2)
I used to be fond of having three periods in a row in my posting,
If you are using a Mac you can press option + colon to get an ellipsis as a single character. There are ways to do it on *nix and Windows but they are more cumbersome and I don't know it off the top of my head.
In Linux its the sequence '<compose key> ..' ; I mapped the print screen key to compose years ago to allow easy access to things like this, , å, x and various other things that Slashdot can't display
Re: (Score:1)
When I hit the lameness filter sometimes by hitting submit on accident it was usually due to not having real content (yet), there seems to be a minimum number of words that are not just punctuation. Not a big loss.
What really irritates me is that Slashdot now seems to censor entire posts on occasion. Sometimes I see an interesting post and log in to reply to it. Then I cannot find it again, even with the search function of my browser.
I miss the days when Slashdot did not filter out anything at all. Sure it
Re: (Score:2)
Re: (Score:2)
Sweet...Should be a real boon to the McCarthy-ists of the day in finding those damn commies.
Re: (Score:1)
Re: (Score:2)
It's been abused too much by trolls and political cowards. Seriously it was starting to look like 4chan around here. So they turned it off rather than, I imagine, banning IP address ranges, or other draconic measures.
I would have preferred they allow posting as AC, but that you need to be logged in to your account that is in good standing to do so.
Re: (Score:2)
Re: (Score:2)
Re: (Score:1)
Looks like they finally got their shit together after me ragging on them for not being able to filter the same ascii swastika spammed on every story. In another thread I tried posting some text from Wikipedia and it was flagged as ascii art. The only time I ever posted AC within the past 15 years was to call someone names.
Re: Completely Off Topic, But ... (Score:2)
Come on nobody took that stuff seriously, did they?
Re: (Score:1)
Please stop spamming links to your irrelevant content. This has been going on for more than 4 years now and it's about time it stops!
Along with APK, you are the main reason Slashdot needed to block anonymous posts thus hurting oppressed people all over the world who can't post anymore without being afraid that their oppressing government will get to them.
And now you are trying to profit from it by spamming links about it. Sad, so sad...
Re: (Score:1)
You can still post anonymously as long as you're registered. There is a checkbox in the account preferences. It has to do with the otherwise cost of mitigating a well-funded SPAM operation.
Re: (Score:1)
Well, last I checked anyway it was still there...
Re: (Score:1)
When did Slashdot disable posting as AC? Looks to be around 9/24/2020. It was a great feature to have
It was a great feature for spammers and assholes. It was pretty terrible for everyone else. Accountability is important. It's not like you can't register a slashdot account with a throwaway address. Of course, those are getting harder to find, but they still exist.
Amiga, et al (Score:5, Informative)
The 68000 was also in a number of systems that sold far more units and were more popular (and affordable) than the mac at that time. Like the Commodore Amiga, Atari ST and the Sega Genesis, to name a few.
Re: (Score:3, Informative)
It was also used in 'grown up' computers like the Apollo and Sun workstations, not just in consumer compters sold in department stores.
retro games? (Score:3)
variants of the thing still around and sold. Look up pricing for 68SEC000 in bulk
Re: (Score:2)
Re: (Score:2)
no need for that in 68K case, still made, still used and they're cheap
Re: (Score:2)
Imagine playing space invaders on a 68K fabricated on a 5nm chip. It will probably blaze at 12 Ghz .. that's about 3,000 times faster than 4Mhz, give or take. That's alien ships coming at you 3,000 times faster. Do you have a clue what that would be like?
Re: (Score:2)
There is a more modern 68k compatible cpu on an fpga that's being used for amiga accelerator cards, if you could adapt this to an asic im sure it would be pretty quick.
Re: (Score:1)
Re: (Score:1)
Thanks fun sponge.
Re: (Score:2)
Say you're doing 4K (3840x2160) at 60 frames per second, and each pixel takes six clocks to shift out the six 4-bit nibbles that make up a 24-bit truecolor pixel. 3840*2160*60*6 is darn near close to 3000 MHz.
Re: (Score:1)
Re: (Score:2)
It introduces a temporal problem, though . . . something needs to slow it down so that the user hasn't already lost before letting go of his quarter . . .
hawk
Re: (Score:2)
Finally! (Score:5, Insightful)
Why hasn't DragonBall kept it alive? (Score:3)
Why hasn't the DragonBall processor community kept it alive?
Re: (Score:1)
Maybe Goku found them all?
BLM where is 4004 and 6502 support (Score:1)
Re: (Score:2)
Because none of those were as easy to design boards and write software for as the 68K, and aren't really good for much more than rolling your own Apple II, C64, and other early systems when compared to other 8-bit solutions. The 4004 is even more limited and was discontinued almost 40 years ago. Useful legacy 8-bit processors like the Z80 and 8051 are still made, and the 6809 is available as an FPGA core. For new designs, there are far better/faster 8-bit chips available today.
Re: (Score:2)
Re: (Score:2)
The 68000 has a 32 bit ISA. It's trivial to have GCC support it, I'm surprised they removed support for it in the first place.
They didn't.
Condition codes are used for branching and inter-word propagation of bitwise operations: zero, negative, carry, extend, overflow. GCC was about to remove support for an old condition code framework that only the back end for m68k (the 68000 family) really still used. The Amiga scene managed to crowdfund paying a compiler expert to port m68k to GCC's current condition code framework.
68000 processors are still used (Score:3)
Re: (Score:3)
If there is still a demand for 68000, they may still be manufactured, but at a price. This is sometimes necessary to support military hardware, that has a long service life.
There is a niche market for manufacturers of legacy chips. As far as I know, the legacy chip manufacturer will buy the manufacturing line off the original manufacturer, who no longer uses it. I came across something like this with the continuing manufacture of polycarbonate film capacitors. These were discontinued when the chemical compa
Re: (Score:2)
CMOS version are still sold and used, very cheap in bulk $15-20 each in qty of hundreds to thousands.
Re: (Score:2)
Re: (Score:2)
More recent micros, such as 32 bit PICs, are much better value than the 68000 in hardware terms. I was surprised at the low price, e,g. just a few pounds; not much more than the 16 bit PIC families.
There is maybe a consideration of sticking with a known architecture, which allows for code and hardware design re-use. In the case of my work, it has nearly always been PIC micros, from the smallest 8 bit devices, and approaching the big 32 bit devices. Our biggest usage is 16 bit devices. There is some merit in
it does what nintendont (Score:2)
I bet they're having trouble to make the LLVM properly use the blast processing.