HPC Performance Engineer

NVIDIA•US, OR, Remote, MA

1d•$152,000 - $241,500

About The Position

NVIDIA is seeking an HPC Performance Engineer for its NVHPC compilers & tools group. This role involves analyzing High Performance Computing (HPC) applications to improve their performance through compiler optimizations and library enhancements. The engineer will work with a variety of applications, from small benchmarks to large-scale multi-node, multi-GPU systems. Responsibilities include identifying optimization opportunities and collaborating with compiler development and application engineering teams to implement solutions.

Requirements

BS/MS or equivalent experience in Computer Science or a related engineering field.
5+ years of programming experience.
Solid understanding of Fortran, C, C++, and programming techniques, especially for parallel architectures, preferably for compilers.
Experience with OpenACC, OpenMP, MPI, CUDA and Standard Language Parallelism.
Strong skills in performance analysis and tuning, as well as a broad understanding of parallel applications development tools and runtime environments.
Strong mathematical fundamentals, including linear algebra and numerical methods.
Understand performance considerations, tradeoffs, and impact.
Expert interpersonal skills, a logical approach to problem-solving, good time management, and task prioritization skills.
Excellent written and verbal communication skills, along with the ability to work in a dynamic development team.

Nice To Haves

Deep understanding of machine architectures and micro-architectures.
Experience with debugging and porting as well as assembly language programming is a significant advantage.

Responsibilities

Analyze the effect of different compilers and compiler optimizations on HPC application performance.
Evaluate whether and how the compiler and runtime can improve the performance of key HPC applications on CPU and CPU/GPU platforms.
Communicate limits of compiler capabilities with application engineering to guide the best path forward towards improving performance.
Port HPC applications to different platforms and programming models.
Evaluate the impact of NVHPC package components, device driver, and system configuration on application performance.