The performance of a computer is a complicated issue and a function of many interrelated quantities. These quantities include: the application, the language, the implementation, the compiler, the architecture, and the hardware characteristics. The usual method to evaluate the performance is to compose a benchmark of programs. This book presents a useful overview on benchmarking. Over twenty experts contributed papers on five important topics concerning benchmarking advanced scientific computer systems: taxonomy and performance metrics, well-known standard and application benchmarks, compiler benchmarks for parallel computers, and benchmarks for database systems.
What people are saying - Write a review
We haven't found any reviews in the usual places.
algorithm Amdahl's law application programs architecture arithmetic bandwidth benchmark programs benchmark suite bytes cache communication compared compiler computer systems Cray Research CRAY Y-MP cycles database defined degree of parallelism developed Dhrystone distributed efficiency equations EuroBen execution floating point floating-point operations flop Fortran function Gentzsch Hockney hypercube implementation instruction integer Intel Intel iPSC/860 iteration J.J. Dongarra linear Linpack Livermore Fortran Kernels loops machines matrix measure megaflops memory references MFLOPS MIPS module multiprocessor Ncube node number of processors obtained operation count optimization overhead Parallel Computing parallel processing parallel programs parameters PARMACS macros partially vectorized Perfect Benchmarks Performance Evaluation performance metrics problem problem size ratio scalar sequential simulation single processor sizeup solver SPEC benchmarks speed speedup standard subroutine supercomputers Table taxonomy test suite theoretical peak traditional speedup transputer UNICOS values variables vector length vector processors vendors Whetstone workload X-MP