Performance Validation Engineer
Location : Austin, TX- Hybrid
Duration : 1 yr with possible extension
- Hybrid onsite (3 days a week minimum) is required.
Summary
Performance engineering role focused on benchmark development, execution, and performance validation of next generation Client server architecture and AI accelerators.Education
Bachelors degree is required for candidates.Day to Day Responsibilities
Analyzing benchmarks and performance characteristics of next generation AMD servers and AI Accelerators.Develop and maintain benchmark workloads, configuration files, build and run scripts.Research trade-offs for our server architecture, especially as related to the memory hierarchy, storage systems, scheduling, and I / O devices.Collect data from benchmark results, data from CPU performance counters, and other related performance information to identify bottlenecks and other performance limitations of the benchmark itself, SoC and system.Evaluate system performance by executing industry standard benchmarks (SPEC CPU, STREAM, HPL, DGEMM, SPECjbb, etc.) with reproducible methodologies across platforms.Collaborate with compiler and OS teams to perform code level analysis to improve benchmark performance and stability.Required Experience
Experience using Linux performance analysis tools and methods.Experience executing industry standard benchmarks or other performance test workloads.Test automation experience.Highly motivated independent contributor with a strong desire to learn and develop a deep understanding of the technology.Understanding of computer system architecture.Programming / scripting skills (e.g. C / C++. BASH, Python, etc).Experience working with different compilers (GCC / LLVM / AOCC), optimizations techniques (LTO, OpenMP etc.); NUMA pinning / affinity strategies.Experience with benchmarks such as STREAM, Multichase, SPECCPU2017, HPL, DGEMM, GUPS, SPECJbb and others.Preferred Experience
Experience benchmarking memory technologies such as DDR5, CXL, HBM and others.Experience with vectorization / SIMD tuning; memory allocator selection.Experience with analyzing and reproducing public benchmark results and understanding of run rules.