Types of Parallelism in Applications Data-level parallelism (DLP) - Instructions from a single stream operate concurrently on several data - Limited by non-regular data manipulation patterns and by memory bandwidth Transaction-level parallelism - Multiple threads/processes from different transactions can be executed concurrently Exploiting Regular (Data) Parallelism Data Parallelism Concurrency arises from performing the same operations on different pieces of data Single instruction multiple data (SIMD) E.g., dot product of two vectors Contrast with data flow Concurrency arises from executing different operations in parallel (in a data driven manner) Contrast with thread ("control") parallelism Instruction Level Parallelism •Instruction-Level Parallelism (ILP): overlap the execution of instructions to improve performance •2 approaches to exploit ILP: 1) Rely on hardware to help discover and exploit the parallelism dynamically (e.g., Pentium 4, AMD Opteron, IBM Power) , and 2) Rely on software technology to find parallelism Example. We can exploit data parallelism in the preceding codes to execute it faster as the arithmetic is loop independent. Parallelization of the matrix multiplication code is achieved by using OpenMP. An OpenMP directive, "omp parallel for" instructs the compiler to execute the code in the for loop in parallel. Instruction Level Parallelism (ILP) ILP: The simultaneous execution of multiple instructions from a program. While pipelining is a form of ILP, the general application of ILP goes much further into more aggressive techniques to achieve parallel execution of the instructions in the instruction stream. 1/30 • Explicit Thread Level Parallelism or Data Level Parallelism • Thread: process with own instructions and data • Thread may be a subpart of a parallel program ("thread"), or it may be an independent program ("process") • Each thread has all the state (instructions, data, PC, register state, and so on) necessary to allow it to • Explicit Thread Level Parallelism or Data Level Parallelism • Thread: process with own instructions and data • Thread may be a subpart of a parallel program ("thread"), or it may be an independent program ("process") • Each thread has all the state (instructions, data, PC, register state, and so on) necessary to allow it to 2007/4/25 2 Outline •Instruction Level Parallelism (2.1) •Compiler techniques for Exposing ILP (2.2) •Reducing Branch Costs with Prediction (2.3) •Overcoming Data Hazards with Dynamic Scheduling (2.4) •Dynamic Scheduling: Examples and the Algorithm (2.5) •Hardware-Based Speculation (2.6) •Exploiting ILP using Multiple Issue and Static different levels of parallelism in scientific applications and give an overview of programming techniques and support for exploiting the parallelism provided. 2.1 Instruction level parallelism A sequential or parallel application offers the potential of instruction level parallelism (ILP), if there are no dependencies between adjacent instructions. Fast Huffman Decoding by Exploiting Data Level Parallelism examples are: task, instruction and data level parallelism, available VLIW and SIMD offers a great deal of parallelism on both an Exploiting Instruction Level Parallelism in Geometry Processing for Three Dimensional Graphics Applications Abstract Three dimensional (3D) graphics applications have