Starting from:

$25

QUANTUM-Exercise 1 Number Precision and Test Performance Solved

Exercise 1: Setup

(a)    Create a working directory.

(b)    Open emacs and write your first program in FORTRAN(c) Submit a test job.

(d) Connect to the cluster spiro.fisica.unpd.it via ssh and repeat the execution

Exercise 2: Number precision

Integer and real numbers have a finite precision. Explore the limits of INTEGER and REAL in Fortran.

(a)    Sum the numbers 2.000.000 and 1 with INTEGER*2 and INTEGER*4

√ 

(b)    Sum the numbers π · 1032 and 2 · 1021 in single and double precision.

Exercise 3: Test performance

Matrix matrix multiplication is many times the bottleneck of linear algebra computations.

(a)    Write explicitely the matrix-matrix multiplication loop in two different orders.

(b)    Use the Fortran intrinsic function.

(c)     Increase the matrix size and use the Fortran Function CPUTIME to monitor the code performance.

(d)    Use the compiler different optimization flags and monitor the performances

More products