Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

FPGA implementation of a configurable cache/scratchpad memory with virtualized user-level RDMA capability

Pnevmatikatos Dionysios, Vassilis Papaefstathiou, George Nikiforos, George Kalokerinos, Xiaojun Yang

Full record


URI: http://purl.tuc.gr/dl/dias/73F41371-2CA9-4267-A400-CC209CC96951
Year 2009
Type of Item Conference Full Paper
License
Details
Bibliographic Citation G.Kalokerinos, V. Papaefstathiou, G. Nikiforos, S. Kavadias, M.Katevenis, Pnevmatikatos , X. Yang, "FPGA implementation of a configurable cache/scratchpad memory with virtualized user-level RDMA capability,"in 2009 Intern. Conf. on Embedded Comp. Systems: Architectures, Modellin and Simulation,pp.149 - 156.doi:10.1109/ICSAMOS.2009.5289226 https://doi.org/10.1109/ICSAMOS.2009.5289226
Appears in Collections

Summary

We report on the hardware implementation of a local memory system for individual processors inside future chip multiprocessors (CMP). It intends to support both implicit communication, via caches, and explicit communication, via directly accessible local (ldquoscratchpadrdquo) memories and remote DMA (RDMA). We provide run-time configurability of the SRAM blocks near each processor, so that part of them operates as 2nd level (local) cache, while the rest operates as scratchpad. We also strive to merge the communication subsystems required by the cache and scratchpad into one integrated Network Interface (NI) and Cache Controller (CC), in order to economize on circuits. The processor communicates with the NI in user-level, through virtualized command areas in scratchpad; through a similar mechanism, the NI also provides efficient support for synchronization, using two hardware primitives: counters, and queues. We describe the block diagram, the hardware cost, and the latencies of our FPGA-based prototype implementation, which integrates four MicroBlaze processors, each with 64 KBytes of local SRAM, a crossbar NoC, and a DRAM controller on a Xilinx-5 FPGA. One-way, end-to-end, user-level communication completes within about 30 clock cycles for short transfer sizes.

Services

Statistics