Hugo Devillers

Inconsistent shift behavior when shifting by 32

1

While trying to isolate other issues I'm having with Impla (it randomly segfaults when building my bvh builder, will make another issue once I narrow down the problem), I found...

Bare infinite loops may trick scopes analysis

3

This delightful snippet (for Artic) tricks Thorin into thinking a basic block should be the entry of a scope: ```rust #[export] fn zoop() { while true {} } ``` we...

The following Artic snippet (obtained by simplifying an example in rastify/anyosl) ```rust #[import(cc = "C", name = "anydsl_print_string")] fn print_string(_: &[u8]) -> (); struct Vector { x: f32, y: f32,...

Fails to handle multiple physical adapters

1

I found out that plugging more than one ch341 at a time makes the driver very unhappy. The `dmesg` log complains about creating a duplicate `gpio4` file, which to me...

Accessing members of expression involving pointer arithmetic triggers internal error

1

```c struct S { uint32 a; }; S v; void foo(int i) { // fine S* p = (S*) (&v) + i; uint32 a = p->a; // borked uint32 a_...

Incorrect codegen for non-uniform (masked) stores

1

The following example showcases the problematic behavior: ``` varying uint32 stack_ptr = 0; noinline void (f)(bool bruh) { //if (bruh) // uncommenting this 'fixes' the problem stack_ptr = 9; return;...

Release the last year's worth of improvements to the compiler into the wild. For reference, here are some of the relevant major changes: * https://github.com/AnyDSL/thorin/pull/125 Added an explicit mem token...

Reworks the layout of the `Debug` struct to make locations optional, and improves LLVM's debug information output slightly, as well as putting it in a separate file for easier navigation.

This 🚧 draft 🚧 PR adds a `control` construct as found in [Shady](https://github.com:Hugobros3/shady) and performs a significant amount of refactoring on the IR to support closure conversion more robustly. **This...

We need to stop relying on lower2cff

4

I wrote this sample to reduce a problem with my substitution_rework branch, where by accidentally making eta conversion weaker I made the compiler more or less explode. I think it's...