R
Ravi Kumar Arimilli
Researcher at IBM
Publications - 514
Citations - 8866
Ravi Kumar Arimilli is an academic researcher from IBM. The author has contributed to research in topics: Cache & Cache algorithms. The author has an hindex of 43, co-authored 514 publications receiving 8851 citations.
Papers
More filters
Proceedings ArticleDOI
The PERCS High-Performance Interconnect
Baba Arimilli,Ravi Kumar Arimilli,Vicente Enrique Chung,Scott Douglas Clark,Wolfgang Denzel,Ben Drerup,Torsten Hoefler,Jody B. Joyner,Jerry Don Lewis,Jian Li,Nan Ni,Ramakrishnan Rajamony +11 more
TL;DR: The Blue Waters System, which is being constructed at NCSA, is an exemplar large-scale PERCS installation that is expected to deliver sustained Pet scale performance over a wide range of applications.
Patent
High performance symmetric multiprocessing systems via super-coherent data mechanisms
TL;DR: In this paper, the coherency protocol of a multiprocessor data processing system is described, which includes processing logic that returns to coherent operations with other processing units responsive to an occurrence of a pre-determined condition.
Patent
Multi-level multiprocessor speculation mechanism
TL;DR: In this paper, the authors propose a processor that reduces the issuing of unnecessary barrier operations during instruction processing, which consists of an instruction sequencing unit and a load store unit (LSU) that issues a group of memory access requests that precede a barrier instruction in an instruction sequence.
Patent
Demand-based larx-reserve protocol for SMP system buses
TL;DR: In this paper, the authors propose a method of accessing values in the computer's memory by loading the value from the memory device into all of the SMP caches, and sending a reserve bus operation from a higher-level cache to the next lower level cache only when the value is to be cast out of the higher cache, and thereafter casting out the value in the lower cache after sending the reserved bus operation.
Patent
Multiprocessor system in which a cache serving as a highest point of coherency is indicated by a snoop response
TL;DR: In this paper, the authors propose a method of maintaining cache coherency, by designating one cache that owns a line as a highest point of coherence (HPC) for a particular memory block, and sending a snoop response from the cache indicating that it is currently the HPC for the memory block and can service a request.