La investigación en computadoras de alto rendimiento incluye varios campos que van desde la arquitectura de la computadora por si misma hasta los algoritmos y aplicaciones. [1.2.3]. En el campo del diseño de computadoras, en esta década se ha dado más énfasis al desarrollo de arquitecturas de computadoras en paralelo, debido a que éstas tienen el enorme potencial de incrementar sustancialmente la capacidad de cómputo. Esto se debe al hecho de que un número considerable de procesadores trabajando en paralelo reducen en gran medida el tiempo de cómputo, siempre y cuando la computadora tenga la arquitectura adecuada y que los “algoritmos” de los programas sean susceptibles a una división en procesos paralelos. Nombre Significado SISD Single Instruction Stream Single Data Stream. Un Solo flujo de Instrucciones Un Solo Flujo de Datos. SIMD Single Instruction Stream Multiple Data Stream. Un Solo flujo de Instrucciones Múltiple Flujo de Datos. MISD Multiple Instruction Stream Single Data Stream. Múltiple flujo de Instrucciones Un Solo Flujo de Datos. MIMD Multiple Instruction Stream Multiple Data Stream. * En la arquitectura SISD una sola instrucción es procesada o actúa sobre cada dato a la vez; en esta arquitectura están basadas todas las máquinas monoprocesador y computadoras personales que existen en la actualidad (Modelo de Von Newman).

Introducción a la computación paralela

Developments in parallel architectures are an important branch in computer science. The success of such architectures derives from their inherent ability to improve the program performances. However, their ability to improve the performance on programs depends on the parallelism extraction strategies, which are always limited by the logic of each sequential program. Speculation is the only known alternative to overcome these constraints and increase the parallelism. In this paper, we study the explicit speculative parallelism using a library of thread-level speculation. We present the design of this library and study different speculative models: speculation of decision structures, speculation of loops, and speculation of critical sections. Our study evaluates different cases taken from SPEC CPU 2000, allowing acceleration of about 1.8x in multicore architectures (four core) with coarse-grained

/pdf/an-explicit-parallelism-study-based-on-thread-level-4di5hfv3qg.pdf

An explicit parallelism study based on thread-level speculation

The introduction of the speculative parallelism into any models can improve the performance and provides significant benefits and increases the Instruction level parallelism (ILP) and Thread level Parallelism (TLP). A General Purpose Graphics Processing Unit (GPGPU) is the future computing technology working with both Graphics Processing Unit (CPU) and GPU to solve many real-world problems not only the graphics problems but also the general purpose applications. As GPU uses data parallelism tasks, the dynamic memory creation and the splay trees which are self adjusting allows for the increase in throughput and load balancing. The frequently used nodes near to the root are an advantage for finding locality of threads as well as for caching and garbage collection. A technique used to render and to study complex scenes into images and to render color, intensity of pixels, distance between pixels is referred as Ray tracing. Multithreading is a promising technique which increases the performance of the computer systems by increasing the instruction level parallelism and thread level speculation. In this paper a new technique is proposed for workload balancing on the Graphics processors and CPU that can be implemented on the graphics processors along with the CPU which provides the optimal result with the speculation techniques and Lorentz Transformation, which is used to determine color and brightness of the ray which are refracted or reflected and also the relative distance between the thread spawning which results in time dilation and contraction. A Graphics Processing Unit Ocelot (GPUOCELOT) is a compilation framework, a simulator used for the execution of the programs which has resulted in the increase in the performance of the instructions which uses the amortized cost.

Splay thread cooperation on ray tracing as a load balancing technique in speculative parallelism and GPGPU.

In this paper we describe the design and implementation of a library (based on pthreads and C + +) to use explicit parallelism thread-level speculation. The library consists of several models proposed for speculation, guaranteeing sequential consistency: decision structures speculation (where the branches are executed in parallel to the previous section), loops speculation (where all iterations are run in parallel), critical sections speculation, and speculation by cooperating threads. The library was evaluated through case studies taken from SPEC CPU 2000, allowing acceleration of about 1.5 x for a multicore architecture (four cores) with coarse-grained multithreaded.

Kahlil Campero

Papers

An explicit parallelism study based on thread-level speculation

A library for parallel thread-level speculation