馃憖 Wow, hold your horses! Zig builds are apparently breaking the sound barrier now. 馃殌 After years of intense labor and a bit of #LLVM yeeting, they've achieved the unimaginable: a #compiler that might finally compile something before your coffee goes cold. 鈽曪笍馃敟
https://mitchellh.com/writing/zig-builds-getting-faster #ZigLang #SoundBarrier #FastCompilation #HackerNews #ngated
alcinnz
boosted
New #blog post. Let's write a peephole optimizer for #QBE that operates on #AArch64 assembly code. Three years ago, we did this for #AMD64 assembly code. But now that I have Arm machines, we can replicate the effort for another CPU architecture.
https://briancallahan.net/blog/20250901.html
#compiler #compilers #opensource #freesoftware #unix #bsd #freebsd #openbsd #netbsd #dragonflybsd #linux #illumos #macos #assembler #assembly
New #blog post. Let's write a peephole optimizer for #QBE that operates on #AArch64 assembly code. Three years ago, we did this for #AMD64 assembly code. But now that I have Arm machines, we can replicate the effort for another CPU architecture.
https://briancallahan.net/blog/20250901.html
#compiler #compilers #opensource #freesoftware #unix #bsd #freebsd #openbsd #netbsd #dragonflybsd #linux #illumos #macos #assembler #assembly
Stefano Marinelli
boosted
Two #blog posts in less than one week!
It was considered known that there is no support within GNAT (Ada) for illumos. We cross-build a complete native binutils+gcc toolchain for illumos to test out this claim.
...only to discover that illumos has full support within GNAT. Not a single test fails.
https://briancallahan.net/blog/20250817.html
#Ada #illumos#GNAT#GCC#GNU #freesoftware #unix #bsd #freebsd #openbsd #netbsd #dragonflybsd #linux #compiler #compilers
Two #blog posts in less than one week!
It was considered known that there is no support within GNAT (Ada) for illumos. We cross-build a complete native binutils+gcc toolchain for illumos to test out this claim.
...only to discover that illumos has full support within GNAT. Not a single test fails.
https://briancallahan.net/blog/20250817.html
#Ada #illumos#GNAT#GCC#GNU #freesoftware #unix #bsd #freebsd #openbsd #netbsd #dragonflybsd #linux #compiler #compilers
alcinnz
and 1 other
boosted
The #Rust #compiler #performance #survey 2025 only runs until next Monday, 7th July - so fill it if you haven't yet and help the Rust project make us Rust developers more productive! 馃殌
https://blog.rust-lang.org/2025/06/16/rust-compiler-performance-survey-2025/
alcinnz
boosted
List of #Supercompilation Resources and Papers:
https://github.com/etiams/supercompilation-resources
"Supercompilation is a principiled program transformation technique that symbolically evaluates a given input program into its more efficient version, eliminating as much of computational overhead as possible[...]."
The #Rust #compiler #performance #survey 2025 only runs until next Monday, 7th July - so fill it if you haven't yet and help the Rust project make us Rust developers more productive! 馃殌
https://blog.rust-lang.org/2025/06/16/rust-compiler-performance-survey-2025/
Joel 馃敧 May-Kill 馃敧
boosted
Writing some tests...
Writing some tests...
List of #Supercompilation Resources and Papers:
https://github.com/etiams/supercompilation-resources
"Supercompilation is a principiled program transformation technique that symbolically evaluates a given input program into its more efficient version, eliminating as much of computational overhead as possible[...]."
For my next #compiler project, I want to write the optimization passes myself, but I don't want to deal with generating machine code for multiple platforms. So tell me #programminglanguages #plt #pldev #compilers fedi, what is an IR that I can target that has a non-optimizing compiler to machine code and supports multiple platforms? This rules out most popular IR like LLVM, C, QBE, Cranelift etc.
In short, I want something that does only instruction selection, register allocation and codegen for multiple platforms. I don't need optimization, so I expect this thing to be really small and lightweight, unlike LLVM, GCC etc.
TPDE Compiler Back-End Framework
https://arxiv.org/abs/2505.22610
"TPDE-LLVM: a standalone back-end for LLVM-IR, which compiles 10--20x faster than LLVM -O0 with similar code quality, usable as library (e.g., for JIT), as tool (tpde-llc), and integrated in Clang/Flang (with a patch)."
Holy cow! 馃く
Open Source on GitHub:
https://github.com/tpde2/tpde
As a first post on this account, I've decided to make an explanation on a compiler I'm writing. Please refer to my main account, @Giona_2, if you would like more information on it.
This post is meant to explain each module of my compiler (Optimizer, Tokenizer, Assembler). Each module will have a dedicated paragraph who were their originally their own posts, but I think it's a great way to kick off this first post
# Optimizer
The Optimizer is the most simplistic part of my compiler. It's job is to arrange the raw text file into a more readable format for the Tokenizer.
Essentially, it arranges the source code you wrote into an array where each word (keyword, number, symbol, etc) is it's own element. This allows the Tokenizer to iterate through each word more reliably than if it just read the raw text file verbatim.
# Tokenizer
The Tokenizer is arguably the most important module.
The Tokenizer takes the list generated from the Optimizer and iterates through each word. When it finds a keyword in the list, it goes through the following steps:
1. Finds the end of the declaration the keyword is indicating
2. Sends this full declaration to a function that'll parse it into a token
The final result is arranged as an array of tokens that emulates the steps the final executable must go through.
# Assembler
Finally, the Assembler is what's responsible for turning the token array generated by the Tokenizer and turning it into the final program.
All this does is iterate over each token in the array and translate it into its assembly-instruction equivalent.
This module, by far, is my favorite for one reason and one reason only: It's extremely volatile in the sense that it can turn the token array into pretty much anything you want Unimal to compile to.