Re: [dev] New Suckless computer language?

From: lukáš Hozda <luk.hozda_AT_gmail.com>
Date: Sat, 23 Jul 2016 08:59:58 +0200

>Do any one know of a C compiler that can out output snippets of
>machine code, for a wide range of CPU:s ? I'm primarily thinking of
>x86, x64, ARM, ARM-thumb.

As much as the suckless community doesn't like it for it's big size
and a lot unnecessary crap, GCC might be a good example for this.
Taking a look into GCC backends might not be a bad idea,
but due to the size and bloatedness of GCC, you are looking
towards quite a long read.

regards,
magnusi

2016-07-23 5:17 GMT+02:00 Daniel V <daniel.viksporre_AT_gmail.com>:
> Ok!
>
> The latest compiler written in C, and outputs byte code that can run in the VM.
>
> Before I let it out there, I have a few things I would like to fix
> first, like output of x86 binaries. And I like to split the project to
> several projects, so the language, the compiler, and the VM has
> separate names, so it's easier to document and talk about.
>
> Do any one know of a C compiler that can out output snippets of
> machine code, for a wide range of CPU:s ? I'm primarily thinking of
> x86, x64, ARM, ARM-thumb.
>
> The thing that is stopping me to make machine code output, is that I
> have been to lazy to make the effort, to write a ELF head generator
> for the binaries.
>
> The language is a system language. And It's made to make kernel calls directly.
>
> And it compiles fast. It takes somewhere between 10 to 30 instructions
> to compile each token, plus the time to read in the text and tokenize
> it.
>
> The syntax is inspired by C, but it's much clearer what type of
> machine code it will generate.
>
> The benefit is probably that is It's super small codebase (in about
> 300-500 lines of code for a basic compiler functionality), no extra
> library dependencies, easy to understand, easy to extend and compiles
> super fast, and easy to combine with whatever you want. It was
> originally made as a generic language to change the compiler fast, to
> match different instruction set, that you want to use. So it was made
> to be a language, that could be changed on the fly when you make new
> instructions to a VM or a Soft Core for an FPGA.
>
> So it's made to save time in development, and give a higher
> understanding on what exactly is done. Giving more brain power left,
> to solve actual problems.
>
> SCC looks interesting thou.
>
> // Daniel V.
>
>
>
> 2016-07-23 1:15 GMT+02:00, Quentin Carbonneaux <quentin_AT_c9x.me>:
>> On Fri, Jul 22, 2016 at 10:37:54PM +0200, Daniel V wrote:
>>> It's not C. It's a new language. The benefit is that you can get a
>>> basic understanding of what the compiler does fast. It's next to
>>> impossible to hide how it compiles and what type of code it generates.
>>> So it could be used for security reasons, with less need for security
>>> auditing.
>>
>> So, a better compiler. As I said, hold on, we're working on it and
>> made good progress already. SCC's code is really nice and readable.
>> Another good thing about SCC is that the source is public.
>>
>>> And it generates fast code without optimizing. And it compiles
>>> extremely fast. And you can easily change it to fit an other
>>> instruction set, CPU or VM.
>>
>> How fast? Do you have benchmarks of any kind? Also, if you want to
>> sell a language to suckless, you better make sure that you can call
>> Xlib, and also a good bit of POSIX, is that the case?
>>
>>> And the language syntax is readable. It's not a esoteric language.
>>
>> Not esoteric, as in, like C?
>>
>> Also, why haven't we seen any code yet? Are you trolling the list?
>>
>> --
>> Quentin
>>
>>
>
Received on Sat Jul 23 2016 - 08:59:58 CEST