#!/usr/local/bin/perl
#
# Script to read a file specifying the packages to which every function belongs
# in VIS and then process the file output of purify and create a table of
# memory usage per package.
#
# Abelardo Pardo <abel@vlsi.colorado.edu>
#
# Revision: [$Id: memoryaccount,v 1.8 1997/01/23 03:50:43 hsv Exp $]

require 5.001;
use Getopt::Long;

$version = "1.8";

# Define option and default variables
#
$opt_f = 0;
$opt_h = 0;
$opt_p = 0;
$opt_v = 0;
$opt_u = 0;

# Read the options
# 
$optionResult = GetOptions("f=s","h","m=s","p","u=s","v");

# Read the result of the options
#
if ($opt_f) {
    $file = $opt_f;
}
else {
    $file = "purify.log";
}

# Print the help message if required
#
if ($opt_h || !$optionResult) {
    goto usage;
}

if ($opt_u) {
    if ($opt_u eq "k") {
	$unit = "Kbytes";
	$factor = 1000.0;
	$precission = 2;
    }
    if ($opt_u eq "m") {
	$unit = "Mbytes";
	$factor = 1000000.0;
        $precission = 2;
    }
    if ($opt_u eq "g") {
	$unit = "Gbytes";
	$factor = 1000000000.0;
	$precission = 2;
    }
    if ($opt_u eq "b") {
	$unit = "bytes";
	$factor = 1.0;
	$precission = 0;
    }
}
else {
    $unit = "bytes";
    $factor = 1.0;
    $precission = 0;
}

# Print the version if required
#
if ($opt_v) {
    print <<ENDOFMESSAGE;
$0 -- Version: $version -- by Abelardo Pardo <abel\@vlsi.colorado.edu>
ENDOFMESSAGE
    exit;
}

# Initial value of certain variables 
#
$detectedmiu = 0;
$totalchunks = 0;
$totalmemory = 0;

# Read in the function map files
#

if ($#ARGV == -1) {
    push(@ARGV, "./.fmap");
}

foreach $filename (@ARGV) {
    if (open(INPUT, $filename)) {
        while (<INPUT>) {
            @fields = split(/\s/,$_);
            $func2pkg{$fields[2]} = $fields[0];
	    $total{$fields[0]} = 0;
	}
	close(INPUT);
    }
}

# Open the input file
#
open(INPUT, $file) || die "Unable to open file $file\n";
while (<INPUT>) {
    chop;

    # Detect the version of purify being executed
    if (/\s\s\*\sPurify\s([0-9\.]+)\s/) {
	$purifyVersion = $1;
	print "Memory Map of $pname at $pdate obtained with Purify $1\n";
	if ($purifyVersion eq "3.2") {
	    $totalMemoryExp = "^Memory\\sin\\-use:\\s([0-9]+)\\sbytes";
	}
	if ($purifyVersion eq "4.0.1") {
	    $totalMemoryExp = "^New\\smemory\\sin\\-use:\\s([0-9]+)\\sbytes";
	}
    }

    # Detect the first line telling when this was executed
    if (/^\*\*\*\*\s\sPurify\sinstrumented\s(.+)\s\((.+)\sat\s(.+)\)/) {
	$pname = $2;
	$pdate = $3;
    }

    # Detecting when the data structure has to be flushed
    if (/^Purify:\sSearching\sfor\sall\smemory\sin\-use\.\.\.$/) {
	foreach $name (keys %total) {
	    $total{$name} = 0;
	}
	$parsing = 1;
    }

    if ($parsing) {
	# Detect the line saying how much memory in use is visible
	if (/^Memory\sin\-use:\s([0-9]+)\sbytes\s+\(([0-9\.]+)%\s.+\)/) {
	    $memInUse = $2;
	}
	
	# If we are inside a MIU statement 
	if ($detectedmiu == 1) {
	    
	    # Detect end of MIU statement
	    if ($_ eq "") {
		$detectedmiu = 0;
		if ($notallocatedyet == 1) {
		    $total{"unclaimed"} += $portioninuse;
		}
	    }
	    else {
		
		# Obtain the function name of the allocation
		if ($notallocatedyet == 1) {
		    if (/\s+([a-zA-Z0-9_]+)\s+\[(.+)\]/) {
			$function = $1;
			if (defined($func2pkg{$function})) {
			    $pkg = $func2pkg{$function};
			    $total{$pkg} += $portioninuse;
			    $notallocatedyet = 0;
			}
		    }
		}
	    }
	}
	
	if (/^MIU:\s([0-9]+)\sbytes/) {
	    $detectedmiu = 1;
	    $portioninuse = $1;
	    $notallocatedyet = 1;
	}
	
	if (/^\s\s\s\s\sTotal\sAllocated\s+([0-9]+)\s+([0-9]+)/) {
	    $totalchunks = $1;
	    $totalmemory = $2;
	    $parsing = 0;
	}
    }
}

close(INPUT);

if ($memInUse ne "100") {
    print "Warning: The program is using $memInUse% of its allocated memory\n";
}
print "Profile of the memory currently in use:\n";
foreach $name (sort keys %total) {
    if ($opt_p || $total{$name} != 0) {
	$tmemoryunits = $total{$name}/$factor;
	$percentage = 100.0*$total{$name}/$totalmemory;
	printf "%9s :  %12.${precission}f %s %5.2f %%\n", 
	       $name, $tmemoryunits, $unit, $percentage;
    }
}
print "-------------------------------------------------------------------------------\n";

if ($totalchunks == 0) {
    print "No information found in file $file\n";
    exit;
}

$memoryunits = $totalmemory/$factor;
$memoryperchunk = $totalmemory/$totalchunks;
printf "Total Memory Allocated = %12.${precission}f %s\n", $memoryunits, $unit;
print "Memory allocated in $totalchunks portions. ";
printf "Average %12.2f bytes per portion\n", $memoryperchunk;

exit;

usage:
    print <<ENDOFMESSAGE;

$0 - Program to analyze and collect statistics about the memory usage of VIS.
    This is how the filter works. It reads a file mapping function names to
    package names. The filename to read may be provided in the command
    line. This map may be created with the script "createfunctionmap". The
    application first reads in that function map. If a function is mapped in
    more than one file, the last one read in the one that prevails (for that
    reason it is recommended to read in first the global map and then the map
    in the user area). After that, it opens the file "purify.log" (this value
    may be overwritten with the "-f" option). It parses this file to detect the
    beginning and end of the purify dump of the memory in use. Among other
    things, the parsing analyzes every MIU (Memory in use) statement. The MIU
    statement gives the call stack whenever a memory allocation of a portion
    that is still in use is performed. The script parses this call stack
    searching for function names. When a function name that is present in the
    map previously read is found, that memory is added to a counter for the
    package to which that function belongs. Since purify prints a finite number
    of functions in this call stack, it might happen that none of the function
    names in that stack is in the map. In that case, the memory allocated is
    accounted to the "unclaimed" category.  If while using this script, the
    unclaimed category represents a significant portion of the memory in use,
    there are two things that can be done: Either the length of the call stack
    is increased by using the "-chain-length" option in purify whose default is
    6, or more function names are included in the map file, such that the
    "unclaimed memory" situation does not occur that often.  When the script is
    done parsing the purify file, it prints a summary of the memory allocated
    to each and every package present in its map. Purify makes a distinction
    between the memory in use and the allocated memory. The former is the
    memory that is still accessible to the program. The latter is the total
    memory allocated to the program. If the memory being used by the program is
    not the total memory allocated by the program, the script issues a
    warning. This situation means that there is a portion of the memory that
    cannot be accessed and therefore leaked. Please refer to the purify user's
    guide Chapter 14, "Purify API" for more information about the type of
    information provided by purify.

Usage:
  $0 [-f <filename>] [-h] [-p] [-v] [-u <units>] filenames

Options:
  -f <filename>     File to read the dump from. The default is "purify.log"
  -h                Print this message
  -p                Print also the packages that do not use any memory.
  -v                Print the version
  -u <units>        Units to print the memory usage in. It may be "b" for bytes
                    "k" for kilobytes, "m" for megabytes and "g" for gigabytes.
		    The default is bytes.
  filenames         Files containing the function map to be read. Default is
                    ./.fmap 

Bugs:
  This script has been tested for purify versions 3.2 and 4.0.1. Since the
  parsing is sensitive to the messages written in the dump file, a new version
  of purify might have different messages that are not matched in the
  script. If this situation arises, chances are the results are completely
  bogus. 

Author: Abelardo Pardo <abel\@vlsi.colorado.edu>


ENDOFMESSAGE
    exit;