Hugo Devillers

Results 29 issues of Hugo Devillers

This PR revives the ~3 year old unfinished SPIR-V backend for Thorin (and proceeds to rewrite most of it anyways). Unlike it's forefather, this version does exactly zero structured control...

The way vector types work in Thorin current matches LLVM: pointers and primitive types can be widened by setting a `length` field to a value other than `1`. I believe...

The following snippet (written using https://github.com/AnyDSL/artic/pull/23): ```rust #[export] fn main (a : i32) -> i32 { test(a) } fn test(a : i32) -> ! { test2(a) } fn test2(a :...

Artic introduced module support a couple years ago, unfortunately it has not seen much adoption. The oft-cited reason by would-be users is that this module support lacks the ability to...

For the purposes of writing tests and such, this would be quite handy. It should possibly be gated behind a compiler option if this is the only use case, but...

``` CMake Error at external/abseil-cpp/CMake/AbseilHelpers.cmake:317 (target_link_libraries): The link interface of target "test_allocator" contains: GTest::gmock but the target was not found. Possible reasons include: * There is a typo in the...

bug

I've been doing this by hand (editing CMakeLists.txt) for years, but I think we need a proper way to do this: disable backends at will, easily. For various reasons (LLVM...

Conceptually reverts https://github.com/AnyDSL/thorin/pull/120 by removing the typedefs (except `f16` for CUDA because it relies on preprocessor magic) Addresses CUDA and OpenCL vector type regressions as pointed out in https://github.com/AnyDSL/thorin/pull/120#issuecomment-2243284502

This PR adds Vulkan Compute support for AnyDSL. It uses shady's JIT capabilities to translate SPIR-V from Thorin's backend into a Vulkan-compliant dialect. Shady also lowers kernel parameters to push...