Description

Check impact of sys11 settings where we know that
sorting improves performance

Paul van der Eijk

Small Model of Type : GAMS

Category : GAMS Test library

Main file : syschk1.gms

$title 'Test impact of sys11 setting' (SYSCHK1,SEQ=207)

$onText
Check impact of sys11 settings where we know that
sorting improves performance

Paul van der Eijk
$offText

$set N 200

set i /i1*i%N%/
    j /j1*j%N%/
    k /k1*k5/
    m /m1*m5/;

parameter A(i,j,k,m),BX(m,j),B0(m,j), B1(m,j), B2(m,j);
parameter T0, T(*) 'sys11 timings' /x=0, 0=0, 1=0, 2=0/;

A(i,j,k,m)$(uniform(0,1) < 0.5) = uniform(0,1);

T0 = TimeExec;

BX(m,j) = Sum((k,i), A(i,j,k,m));
T('X')=TimeExec - T0; T0 = TimeExec;

option sys11=0;
B0(m,j) = Sum((k,i), A(i,j,k,m));
T('0')=TimeExec - T0; T0 = TimeExec;

option sys11=1;
B1(m,j) = Sum((k,i), A(i,j,k,m));
T('1')=TimeExec - T0; T0 = TimeExec;

option sys11=2;
B2(m,j) = Sum((k,i), A(i,j,k,m));
T('2')=TimeExec - T0; T0 = TimeExec;

Display T;

Abort$(Abs(T('X') - T('0')) / T('0') > 0.6) 'Sys11 default is not zero';
Abort$(T('1') < 2.0 * T('0'))               'sys11=0 should be the faster than sys11=1';
Abort$(T('1') < 2.0 * T('2'))               'sys11=2 should be the faster than sys11=1';