Diana Picus
MTS Hacking On LLVM For AMDGPUs
Stockholm, Sweden
Summary
🤩
Rockstar🎓
Top SchoolDiana Picus is a compiler engineer with a decade of experience specializing in LLVM backends, currently driving AMDGPU support at AMD from Stockholm. She has a strong track record at Linaro and Freescale working on ARM/AArch64 toolchains, GlobalISel, Flang, LLDB for CHERI hardware, and upstream CI/buildbot maintenance. Her open-source contributions include implementing AMDGPU-specific instructions and dynamic VGPR stack allocation in the widely used llvm-project, as well as targeted Clang and ARM backend fixes in mirrored repos. Comfortable in both research and production contexts, she pairs deep low-level expertise with hands-on build and release responsibilities. Colleagues rely on her for subtle, architecture-specific fixes that improve correctness and portability across diverse targets.
10 years of coding experience
10 years of employment as a software developer
Colegiul National "Petru Rares" Piatra NeamtBachelor of Science (B.Sc.), Computer Engineering, Bachelor of Science (B.Sc.), Computer Engineering at Universitatea „Politehnica” din București
Github Skills (20)
assembly10
c-language10
llvm10
c1110
c1710
compiler-compiler10
cplus10
x86-6410
assembler10
instruction-set10
arm10
compiler10
g10
cpp10
cprogramming-language10
Programming languages (8)
Github contributions (5)
llvm/llvm-project
Aug 2022 - Aug 2022
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
Role in this project:
Back-end Developer (focus on compiler infrastructure)
Back-end Developer (focus on compiler infrastructure)Contributions:99 reviews, 2 commits, 84 PRs in 6 days
Contributions summary:Diana contributed to the LLVM compiler project, primarily focusing on the AMDGPU backend. Their work involved implementing and refining specific instructions, such as `llvm.amdgcn.init.whole.wave` and `S_ALLOC_VGPR`. The user also made modifications to the frame lowering process, including the allocation of stack space for dynamic VGPRs in compute shaders with CWSR (compute queue). Furthermore, the user made improvements related to the documentation for the IRTranslator.
compilerllvmtoolchain
apple/swift-llvm
Jul 2017 - Nov 2017
Role in this project:Back-end Developer
Back-end DeveloperContributions:30 commits in 4 months
Contributions summary:Diana primarily contributed to the ARM backend of the LLVM project. Their work involved fixing issues related to instruction selection within the GlobalISel framework, specifically for the ARM architecture. They added and refined support for various ARM-specific instructions, including G_GLOBAL_VALUE, G_FDIV, and G_FMUL, as well as shift operations. They also focused on enabling PIC mode support for the ARM target, ensuring compatibility across different relocation models.
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.