Abkürzungen

Partitioned Global Address Space

Computational Science and Engineering

(classic application area of parallel computing)

Cycles Per Instruction

Instruction Level Parallelism

(hardware discovers parallelism in instruction stream, VLIW, EPIC)

Very Long Instruction Word

Explicit Parallel Instruction Computing in Intel terminology

Single Instruction Multiple Data

Simultaneous Multi-Threading

(combine ILP and TLP)

Thread-level Parallelism

(independent threads of control, to program multiple-cores)

Data-Level Parallelism

(same operation, different data run in parallel, vector processing and SIMD)

Uniform Memory Access

(all memory accessible with same speed)

Non-Uniform Memory Access

(some parts of memory accessible faster)

Hardware View

Unified Parallel C

(extension to C)

Compute Unified Device Architecture

(parallel computing platform and API developed by NVIDIA, allows developers to use GPU for general purpose taskjs)

Graphics Processing Unit

(specialized processor to handle parallel computations)

Performance Application Programming Interface

(high-level API)

Open specification for Multi-Processing

Architecture Review Board

(standard Open MP)

Message Passing Interface

(standardized communication protocol for parallel computing in distributed memory systems)

Symmetric MulitProcessing

(2 or more processors share single memory and operate under single operating instance)

Basic Linear Algebra Subprograms

(highly optimized library)