MaX Manycore
Our team has experience programming and performance tuning for architectures like x86 (e.g. Core 2 Duo), Nvidia G80, Cell, Tilera TILEPro, Raw, FPGAs (e.g. Virtex-5), and a handful of cluster-style super computers. We can use cross-platform threading models like MPI, or port applications directly to next generation multi-core processors using sophisticated architecture-specific optimizations. We also use commercially available, open source, and academic research tools to automatically support the identification of computational bottlenecks and opportunities for parallelism.
Advances in VLSI technology, along with a recent wave of novel architecture research, provide a wealth of commercially available processing resources that are massively parallel, yielding greater throughput and energy efficiency. Leveraging these technologies, the next generation of hardware and software systems have the opportunity to alleviate computational bottlenecks at limited power consumption and low cost. However, porting serial applications to new parallel programming models is a daunting task. Even choosing the right processor is challenging, given the diverse set of available technologies. Our customers have a deep understanding of their own applications, but are typically not parallel programmers or multi-core computer architects.
That is where we come in. Our team can optimize the performance and power-consumption of an application or system using cutting edge multi-core technologies:
- Proof-of-concept parallel prototypes
- Production-grade optimized implementations
- Performance tuning
- Accelerated libraries for common algorithms
We see killer applications for multi-core in several areas. Some examples include:
Realizing the promise of multi-core
