Alexandru Voicu
PMTS Software Development Engineer at AMD
Royal Leamington Spa, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolAlexandru Voicu is a PMTS Software Development Engineer at AMD with nine years of experience specializing in compiler back-end development, performance engineering, and heterogeneous compute. He has contributed significant features to high-profile open-source projects like ROCm/hip and the LLVM project, improving module loading, kernel dispatch, and AMDGPU/SPIR-V code generation. His work spans low-level codegen, address-space and sret handling, and performance-focused refactors that bridge HIP, OpenCL and AMDGCN specifics. Based in Royal Leamington Spa, he combines industry roles at AMD and prior experience at Microsoft and Imagination Technologies with editorial insight from running Beyond3D. Unusually for a systems engineer, he also holds doctoral-level training in economics, bringing a quantitative, analytical lens to technical problem-solving.
9 years of coding experience
7 years of employment as a software developer
Academia de Studii Economice BucurestiDoctor of Philosophy (Ph.D.), Economics, Doctor of Philosophy (Ph.D.), Economics at Bucharest Academy of Economic Studies
english, french
Github Skills (17)
spirv10
c-language10
llvm10
kernel10
hip10
compiler-design10
code-generation10
performance-optimization10
kernel-mode10
cuda10
amdgpu10
clangd10
cprogramming-language10
clang10
back-end-development9
Programming languages (7)
Github contributions (5)
ROCm/hip
Oct 2017 - Mar 2020
HIP: C++ Heterogeneous-Compute Interface for Portability
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:2 releases, 179 commits, 59 PRs in 2 years 5 months
Contributions summary:Alexandru primarily contributed to the HIP (Heterogeneous-Compute Interface for Portability) repository by implementing features related to module loading and kernel dispatch. Their work included adding support for global variables within the HIP module loading API and refactoring the dispatch mechanism to leverage the existing module API. The changes involved modifying the compilation process and incorporating specific calls to improve performance and compatibility. These changes also included addressing issues with CUDA vector type functionality and the efficient handling of memory operations for optimized code generation.
cudaheterogeneousgpuportabilityhip-runtime
llvm/llvm-project
Jun 2023 - Apr 2025
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
Role in this project:Back-end Developer & Compiler Engineer
Back-end Developer & Compiler EngineerContributions:249 reviews, 80 PRs, 53 pushes in 1 year 10 months
Contributions summary:Alexandru primarily contributed to the LLVM project by modifying and extending Clang's code generation capabilities, specifically for the AMDGPU and SPIR-V targets. Their work includes implementing support for AMDGCN-flavored SPIR-V, which involves adding features like AMDGCN inline assembly and target-specific builtins. They also focused on improving the handling of address spaces, dynamic casts, and the integration of HIP and OpenCL code, ensuring correct code generation for these specialized targets. The user addressed issues related to sret arguments and compiler optimizations, indicating a focus on ensuring that generated code is correct, efficient, and aligns with the target architecture's requirements.
compilerstechnologiesclangsubmittoolchain
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.