NVIDIA CUDA Tile support has been successfully ported to Java through TornadoVM, enabling tile kernels to be chained with JIT SIMT kernels in a single task graph. The implementation includes a new tornado-api tile package with working JVM fallbacks, C++ tile kernels compiled to CUDA 13 cubins using tensor cores, and comprehensive compiler integration for tile operations with support for compute capability 8.0 and toolkit 13.3.
Metacarp is a new self-hosting compiler for the Carp programming language, written in Carp itself. It successfully compiles the reference test suite, itself, and generates byte-identical C code, featuring modular compiler phases as independent libraries and experimental LLVM and JIT backends. While not yet a complete drop-in replacement for the Haskell reference implementation, it demonstrates a working second compiler for the language.
AgentJIT is a compiler that traces dynamic LLM agent workflows and converts them into optimized Python code, reducing execution time from seconds to milliseconds and eliminating API token costs for recurring patterns. It works by capturing tool calls on the first run, analyzing data dependencies, and generating type-safe Python pipelines with automatic fallback to the LLM agent if inputs diverge.
JEP 544 introduces ahead-of-time (AOT) code compilation for Java's HotSpot JVM to improve startup and warmup times by pre-compiling application code to native code during a training run and caching it for production use. The system dynamically regenerates native code if workloads change, combining AOT and JIT compilation benefits without requiring application code changes or new workflows.
DeepJIT is a header-only C++20 JIT runtime that provides a unified interface for compiling and caching GPU kernels on NVIDIA CUDA and Huawei Ascend devices. It enables kernel libraries to share runtime infrastructure including caching, lazy initialization, and distributed filesystem support across both backends.