Using the Cell Processor to Accelerate Kernel Memory Operations
The Cell Processor that has recently been introduced by Sony, Toshiba, and IBM is a heterogeneous multi-core chip. Most attention has focused on the Synergistic Processing Element (SPE) cores and how their incredible computing power can be applied to multimedia and scientific workloads. The DMA capabilities of the SPE are often overlooked. Their asynchronous nature and built in MMU allow for great flexibility in moving memory. In this Seminar we will explore how the Linux Kernel can exploit these features so that common memory operations can be accelerated.
Jimi has been programming computers since he was able to find a Radio Shack store manager that would let him play with the TRS-80 that no one ever wanted to buy. He currently is a member IBM's TJ Watson Research Laboratory where he tinkers with Hypervisors, Processors, K42 Research OS, Linux, Simulators, and Tool-chains. Prior to IBM he worked on other OSes for other companies.
Amos Waterland is a programmer working for IBM.