LLVM 10.0-RC3 Released With The Final Expected Soon For This Big Compiler Update

[phoronix]

Phoronix: LLVM 10.0-RC3 Released With The Final Expected Soon For This Big Compiler Update

LLVM 10.0 was supposed to be released at the end of February but is running slightly behind schedule and now there is a third and unscheduled final release candidate...

http://www.phoronix.com/scan.php?page=news_item&px=LLVM-10.0-RC3-Released

[Raka555]

I am still crossing my fingers that someone will add a PIC8 backend (again).

[CommunityMember]

I am still crossing my fingers that someone will add a PIC8 backend (again).

Unless that someone is you, that seems unlikely, as no one seems to be sufficiently motivated to do so (there was some interest four or five years ago for a new Microchip BE, but as the existing MIPS32 BE worked for the 32bit MCUs, and Microchip itself is focused on the MPLAB IDE, there appears to have never been sufficient motivation in the community).

[Raka555]

Unless that someone is you, that seems unlikely, as no one seems to be sufficiently motivated to do so (there was some interest four or five years ago for a new Microchip BE, but as the existing MIPS32 BE worked for the 32bit MCUs, and Microchip itself is focused on the MPLAB IDE, there appears to have never been sufficient motivation in the community).

I figured that much. Unfortunately all my free time is going into a front-end compiler I am working on.

I am actually surprised at how little interest there is in a PIC8/16 back-end.

I guess most people just use Arduino/AVR, which is a pitty since the PIC chips are really nice.

[Brane215]

Why would you really need that much PIC8 compiler ?
8-bitters are used for specialised things these days, where resources are so tight that the use of assembly is the norm.

And, if you really want a compiler, you can use Microchip's in a pinch.

[Raka555]

Why would you really need that much PIC8 compiler ?
8-bitters are used for specialised things these days, where resources are so tight that the use of assembly is the norm.

And, if you really want a compiler, you can use Microchip's in a pinch.

I do use the Microchip compiler.
I also use Jallib : http://justanotherlanguage.org/

It would have been great to target PIC8 with the compiler I am working on, which uses an llvm backend. Hence my interest.

[Brane215]

I do use the Microchip compiler.
I also use Jallib : http://justanotherlanguage.org/

It would have been great to target PIC8 with the compiler I am working on, which uses an llvm backend. Hence my interest.

OK, but it seems that benefits don't come close to covering the costs, so probably no one will do it.
PIC12/16 cores are VERY quirky. They aren't even remotely similar to AVR etc.
So, while compilers for them do exist, I doubt that quality of generated code is very good.

LLVM is optimized toward being greatest with modern, 32-bit and 64-bit cores. NO one really cares outside of that world.

WRT to JAL - why not Rust ? It offers additional safety checks with essentially zero run overhead, it's nice and simple and is being evolved toward microcontrollers these days.
And, it is based on llvm. It is dfeveloped at prety brisk tempo these days...

[Raka555]

OK, but it seems that benefits don't come close to covering the costs, so probably no one will do it.
PIC12/16 cores are VERY quirky. They aren't even remotely similar to AVR etc.
So, while compilers for them do exist, I doubt that quality of generated code is very good.

LLVM is optimized toward being greatest with modern, 32-bit and 64-bit cores. NO one really cares outside of that world.

WRT to JAL - why not Rust ? It offers additional safety checks with essentially zero run overhead, it's nice and simple and is being evolved toward microcontrollers these days.
And, it is based on llvm. It is dfeveloped at prety brisk tempo these days...

JAL generates PIC code. Can Rust ?

[Brane215]

JAL generates PIC code. Can Rust ?

I think it could generate PIC32 - MIPS32BE code.
It's not PIC8, but for that, there already are good enough solutions.

I'm not great Rust advocate, but they seem to be onto something. And those extra compile-time security checks with zero runtime cost are EXACTLY what doctor ordered for microcontroller work. So, all I am saying, you might want look into that.

[Brane215]

Language advocacy in 2020... The cringe is too strong.

Relax. It was just half question/half a suggestion.

Rust is not just another syntactic tool. It brings quite a few new things to the table. Most of them sorely needed in microcontroller world,
esepecially nowadays that IoT is getting pushed into a toiet paper.