Valentina Salapura
Valentina Salapura | |
|---|---|
| Alma mater | Technische Universität Wien; University of Zagreb |
| Occupation(s) | Researcher in high-performance computing, supercomputers, computer architecture |
| Known for | Contributions to Blue Gene supercomputer series (BlueGene/L, BlueGene/P, BlueGene/Q), Frontier supercomputer; HPC software libraries; processor architecture |
| Notable work | Development of Blue Gene series, Frontier system architecture, HPC software libraries |
Valentina Salapura is a distinguished researcher specializing in high-performance computing (HPC), supercomputers, and computer architecture. Her expertise encompasses the design of cutting-edge computing platforms, with a particular emphasis on scalable systems, parallel computation, and energy-saving technologies. Her contributions bridge both theoretical research and practical industry implementations.
Education and early work
Salapura obtained her doctorate in computer science from the Technische Universität Wien located in Vienna, Austria. Additionally, she holds master's degrees in computer science and electrical engineering from the University of Zagreb in Croatia. Her initial scholarly pursuits concentrated on enhancing computer architecture designs and formulating novel computational techniques.
Career
Throughout her professional journey, Salapura has held positions at prominent organizations including IBM, AMD, and Google. During her tenure at IBM, she played a pivotal role in the creation of the Blue Gene supercomputer series and contributed significantly to the architecture of the Power8 processor.
At AMD Research, her focus was on advancing distributed computing and supercomputing innovations. She led efforts in developing HPC software libraries and shaped software architecture for the Frontier system—recognized as the world’s inaugural publicly accessible exascale machine.
Contributions to high-performance computing
Salapura’s work in HPC involves integrating heterogeneous computing elements and accelerators within hyperscale data center environments. Her research prioritizes creating systems that achieve an optimal balance between superior performance and energy conservation.
She was instrumental in leading the development of the BlueGene family, contributing to multiple generations including BlueGene/L, BlueGene/P, BlueGene/Q, as well as the Frontier supercomputer.
BlueGene/L
The Blue Gene/L system employed numerous low-frequency, low-power embedded PowerPC cores equipped with floating-point accelerators. This design favored energy efficiency over raw processor speed, making it well-suited for massively parallel workloads. By utilizing many energy-efficient cores operating concurrently, it achieved reduced power consumption.
BlueGene/P
Building upon its predecessor, the Blue Gene/P increased core density substantially. Each rack contained 1,024 nodes totaling 4,096 processor cores. The system maintained a strong focus on power efficiency, earning top positions on the Green500 list during 2007–2008 for its environmentally conscious supercomputing capabilities.
BlueGene/Q
The BlueGene/Q platform, exemplified by the Sequoia installation at Lawrence Livermore National Laboratory, reached a peak performance of 16.32 petaflops by employing 1,572,864 cores—marking it as the first supercomputer to surpass one million cores. The configuration included 96 racks, 98,304 compute nodes, and 1.6 petabytes of memory. Primarily cooled by water, Sequoia demonstrated notable enhancements in power efficiency compared to earlier models.
Frontier
The Frontier system—developed collaboratively by Hewlett Packard Enterprise and AMD, and housed at Oak Ridge National Laboratory—achieved the milestone of being the first publicly available exascale supercomputer in May 2022. Delivering 1.194 exaflops on the high-performance Linpack benchmark, it incorporates 8,699,904 CPU and GPU cores alongside HPE's Slingshot 11 network infrastructure for data communication. Frontier utilizes an advanced water cooling solution capable of circulating 60,000 gallons per minute.
Contributions to computer science specialties
Valentina Salapura has made significant advancements in high-performance computing (HPC), supercomputing, and distributed systems, notably enhancing simulations in quantum chromodynamics. Her influence extends to cloud computing, where she has concentrated on virtualization techniques and system resilience. Moreover, her pioneering research in processor architecture and microarchitecture remains a cornerstone for ongoing innovations within the technology sector.
Selected publications
- Salapura, Valentina; Hamann, Volker: Design of Fuzzy Controllers Using Statecharts and Embedded VHDL (1994)
- Salapura, Valentina: Efficient Neural Network Implementation via Bit Stream Arithmetic (IEEE International Symposium on Electrical Insulation, 1994)
- Gschwind, Michael; Salapura, Valentina; Maischberger, Oliver: Implementing Space-Efficient Neural Networks (FPGA '94)
- Gschwind, Michael; Salapura, Valentina: FPGA Design Methodology Based on VHDL (FPL '95)
- Salapura, Valentina; Hamann, Volker: VHDL and Statecharts for Fuzzy Control System Implementation (EURO-DAC '96)
- Georgiou, Christos J.; Salapura, Valentina; Denneau, Monty: Scalable and Programmable Platforms for Next-Generation Networking (Network Processor Design, 2004)
- Salapura, Valentina; Gschwind, Michael; Maischberger, Oliver: Rapid FPGA-Based General-Purpose Neuron Implementation (FPL '94)
- Salapura, Valentina: Instruction-Set Architecture and Vector Extensions (Encyclopedia of Parallel Computing, 2011)
- Blumrich, Matthias; Salapura, Valentina; Gara, Alan: Architectural Exploration of Stream Register-Based Snoop Filters (Transactions on High-Performance Embedded Architectures And Compilers III, 2011)
- Salapura, Valentina; Karkhanis, Tejas; Nagpurkar, Priya; Moreira, Jose: Enhancing Business Analytics Performance (IEEE International Symposium on High-Performance Comp Architecture, 2012)
- Hsu, Ching-Hsien; Shi, Xuanhua; Salapura, Valentina (eds.): Proceedings of the 11th IFIP WG 10.3 International Conference on Network and Parallel Computing NPC 2014
- Li, Min et al.; Salapura, Valentina: SparkBench: A Comprehensive Benchmark Suite for Spark In-Memory Data Analytics Platform (2015)
Honors and distinctions
In recognition of her contributions to Blue Gene/L development, Salapura received the prestigious ACM Gordon Bell Prize in 2006. She holds co-inventor status on more than 500 patents and was honored as an IEEE Fellow in 2012 for her exceptional work in multiprocessor system design.
ISC 2023
Salapura presented insights into the shifting dynamics between hyperscale cloud providers and the HPC community.
ICCD 2012
She discussed the adoption trends of cloud computing alongside its effects on virtualization strategies and system robustness.
Grace Hopper Celebration of Women in Computing 2007
Salapura emphasized the paradigm shift from increasing single-processor speeds to embracing multiprocessor architectures.