Mega Code Archive
Utilities for flop (floating-point operation) counting
<![CDATA[
/* Copyright (C) 2006 Univ. of Massachusetts Amherst, Computer Science Dept.
This file is part of "MALLET" (MAchine Learning for LanguagE Toolkit).
http://www.cs.umass.edu/~mccallum/mallet
This software is provided under the terms of the Common Public License,
version 1.0, as published by http://www.opensource.org. For further
information, see the file `LICENSE' included with this distribution. */
//package cc.mallet.grmm.util;
/**
* Utilities for flop (floating-point operation) counting. This is a much better
* way to measure computation than CPU time, because it avoids measuring non-essential
* properties of the implementations.
*
* $Id: Flops.java,v 1.1 2007/10/22 21:37:58 mccallum Exp $
*/
public class Flops {
private static long flops = 0;
// this figures taken from Tom Minka's lightspeed
private static final int EXP_FLOPS = 40;
private static final int LOG_FLOPS = 20;
private static final int DIV_FLOPS = 8;
private static final int SQRT_FLOPS = 8;
public static long getFlops ()
{
return flops;
}
public static void exp ()
{
flops += EXP_FLOPS;
}
public static void log ()
{
flops += LOG_FLOPS;
}
public static void div ()
{
flops += DIV_FLOPS;
}
public static void sqrt ()
{
flops += SQRT_FLOPS;
}
public static void sumLogProb (int n)
{
flops += n * (LOG_FLOPS + EXP_FLOPS);
}
public static void increment (int N)
{
flops += N;
}
public static void log (int N)
{
flops += N * LOG_FLOPS;
}
public static void exp (int N)
{
flops += N * EXP_FLOPS;
}
public static void pow (int N)
{
// Get an upper bound using
// a^b = exp(b*log(a))
Flops.increment (N * (EXP_FLOPS + LOG_FLOPS + 1));
}
public static class Watch {
private long starting;
private long current;
public Watch ()
{
starting = flops;
current = starting;
}
public long tick ()
{
return tick (null);
}
public long tick (String message)
{
long elapsed = flops - current;
current = flops;
if (message != null) System.err.println (message+" flops = "+elapsed);
return elapsed;
}
public long totalFlopsElapsed () { return flops - starting; }
}
}
]]>