Estimate the total number of simd instruction cycles needed

Assignment Help Computer Engineering

Reference no: EM131937116

Problem

Devise a minimum-time algorithm to multiply two 64 x 64 matrices, A = (0,) and 8 = (EV, on an SIMD machine consisting of 64 PEs with local memory. The 64 PEs are interconnected by a 2D 8 x 8 torus with bidirectional links.

(a) Show the initial distribution of the input matrix elements (a4 and (b) on the PE memories.

(b) Specify the SIMD instructions needed to carry out the matrix multiplication. Assume that each PE can perform one multiply, one odd, or one shift (shifting data to one of its four neighbors) operation per cycle. You should first compute all the multiply and odd operations on local data before starting to route data to neighboring PEs. The SIMD shift operations can be either east, west, south. or north with wraparound connections on the COILS. (c) Estimate the total number of SIMD instruction cycles needed to compute the matrix multiplication. The time includes all arithmetic and data-routing operations. The final product elements C = A x 8 = (cf) end up in various PE memories without duplication.

(c) Estimate the total number of SIMD instruction cycles needed to compute the matrix multiplication. The time includes all arithmetic and data-routing operations. The final product elements C = A x 8 = (c1) end up in various PE memories without duplication.

(d) Suppose data duplication is allowed initially by loading the same data element into multiple PE memories. Devise a new algorithm to further reduce the SIMD execution time. The initial data duplication time, using either data broadcast Instructions or data touting (shifting) instructions, must be counted. Again, each result element cl ends up in only one PE memory.

Reference no: EM131937116

Questions Cloud

Model materials to demonstrate dna replication : Present a detailed analysis of DNA replication at one replication fork. Use drawing, descriptions, and/or captions detailing the process.

At the time of the margin call the stocks price must have : In 1 year the investor has interest payable and gets a margin call. At the time of the margin call the stock's price must have been ________.

Differences between prokaryotic and eukaryotic cells : 1. Identify the major similarities and differences between prokaryotic and eukaryotic cells.

Nebulized albuteral treatments : She was in the ED only 6 hours ago with an acute asthma attack, which resolved with nebulized albuteral treatments.

Estimate the total number of simd instruction cycles needed : Estimate the total number of SIMD instruction cycles needed to compute the matrix multiplication. The time includes all arithmetic and data-routing operations.

Calculate the peak performance in g flops with reasoning : Calculate the peak performance in G flops with reasoning in each of the following two vector supercomputers. The Cray Y-MP C-90 with 16 vector processors.

Function and synthesis of these molecular machines : Cells without proteins do not function and synthesis of these molecular machines starts at the DNA level. Transcription of DNA into RNA and translation

Describe an algorithm that compute the given expression : Describe an algorithm using odd. multiply, and data-routing operations to compute the expression s =A1 x8, A2 82 f A32 x 832 with minimum time.

Comparative embryology and the study of evolution : What is the connection between comparative embryology and the study of evolution?

User Account

All Pages