Parallel paradigms in optimal structural design

[摘要] ENGLISH ABSTRACT: Modern-day processors are not getting any faster. Due to the power consumption limit of frequencyscaling, parallel processing is increasingly being used to decrease computation time. Inthis thesis, several parallel paradigms are used to improve the performance of commonly serialSAO programs. Four novelties are discussed:First, replacing double precision solvers with single precision solvers. This is attempted in orderto take advantage of the anticipated factor 2 speed increase that single precision computationshave over that of double precision computations. However, single precision routines presentunpredictable performance characteristics and struggle to converge to required accuracies, whichis unfavourable for optimization solvers.Second, QP and dual are statements pitted against one another in a parallel environment. Thisis done because it is not always easy to see which is best a priori. Therefore both are started inparallel and the competing threads are cancelled as soon as one returns a valid point. Parallel QPvs. dual statements prove to be very attractive, converging within the minimum number of outeriterations. The most appropriate solver is selected as the problem properties change during theiteration steps. Thread cancellation poses problems caused by threads having to wait to arrive atappropriate checkpoints, thus su ering from unnecessarily long wait times because of strugglingcompeting routines.Third, multiple global searches are started in parallel on a shared memory system. Problemssee a speed increase of nearly 4x for all problems. Dynamically scheduled threads alleviate theneed for set thread amounts, as in message passing implementations.Lastly, the replacement of existing matrix-vector multiplication routines with optimized BLASroutines, especially BLAS routines targeted at GPGPU technologies (graphics processing units),proves to be superior when solving large matrix-vector products in an iterative environment.These problems scale well within the hardware capabilities and speedups of up to 36x arerecorded.

[发布机构] Stellenbosch University