NAME

Linux::Cpuinfo - Object Oriented Interface to /proc/cpuinfo

SYNOPSIS

  # Old interface ( for single processor devices )

  use Linux::Cpuinfo;

  my $cpu = Linux::Cpuinfo->new();

  die ('Could not find cpu info (does /proc/cpuinfo exists?)')
    unless ref $cpu;

  print $cpu->model_name();


  # New Interface ( copes with SMP ).

  my $cpuinfo = Linux::Cpuinfo->new();

  $cnt  = $cpuinfo->num_cpus();  # > 1 for an SMP system


   foreach my $cpu ( $cpuinfo->cpus() )
   {
     print $cpu->bogomips(),"\n";
   }

DESCRIPTION

On Linux systems various information about the CPU ( or CPUs ) in the computer can be gleaned from "/proc/cpuinfo". This module provides an object oriented interface to that information for relatively simple use in Perl programs.

METHODS

The interface has changed between revisions 1.2 and 1.3 of this module in order to deal with systems with multiple CPUs - now the details of a CPU are acquired by the methods of the Linux::Cpuinfo::Cpu objects returned by the "cpu()" and "cpus()" methods of this class. However in order to retain backward compatibility if the methods described for Linux::Cpuinfo::Cpu are called a Linux::Cpuinfo object then it will work as previously - returning the details of the sole CPU on a single processor system and the last discovered CPU on system with multiple processors ( this was the implicit behaviour on previous versions). Whilst not strictly deprecated this interface is not the recommended one.

cpuinfo: Returns a blessed object suitable for calling the rest of the methods on or a false value if for some reason "/proc/cpuinfo" cant be opened. The first argument can be an alternative file that provides identical information. You may also supply a hashref containing other arguments - the valid keys are

NoFatal: The default behaviour is for the method to croak if an attribute is requested that is not available on this particular CPU. If this argument is supplied with a true value then the method will return undef instead.

num_cpus: Returns the number of CPUs reported for this system.
cpu SCALAR $cpu: Returns an object of type Linux::Cpuinfo::Cpu corresponding to the CPU of index $cpu ( where $cpu >= 0 and $cpu < num_cpus() ) - if $cpu is omitted this will return an object correspnding to the last CPU found.
If $cpu is out of bounds with respect to the number of CPUs then it will be set to the first or last CPU ( depending whether $cpu was < 0 or >num_cpus )
cpus: Returns a list containing objects of type Linux::Cpuinfo::Cpu corresponding to the CPUs discovered in this system. If the method is called in a scalar context it will return a reference to an array of those objects.

PER CPU METHODS OF Linux::Cpuinfo::Cpu

Note that not all of the methods listed here are available on all CPU types. For instance, MIPS CPUs have no cpuid instruction, but might sport a byte order attribute.

There are also some other methods available for some CPUs which aren't listed here.

processor: This is the index of the processor this information is for, it will be zero for a the first CPU (which is the only one on single-proccessor systems), one for the second and so on.
vendor_id: This is a vendor defined string for X86 CPUs such as 'GenuineIntel' or 'AuthenticAMD'. 12 bytes long, since it is returned via three 32 byte long registers.
cpu_family: This should return an integer that will indicate the 'family' of the processor - This is for instance '6' for a Pentium III. Might be undefined for non-X86 CPUs.
model or cpu_model: An integer that is probably vendor dependent that indicates their version of the above cpu_family
model_name: A string such as 'Pentium III (Coppermine)'.
stepping: I'm lead to believe this is a version increment used by intel.
cpu_mhz: I guess this is self explanatory - it might however be different to what it says on the box. The Mhz is measured at boot time by the kernel and represents the true Mhz at that time.
bus_mhz: The MHz of the bus system.
cache_size: The cache size for this processor - it might well have the units appended ( such as 'KB' )
fdiv_bug: True if this bug is present in the processor.
hlt_bug: True if this bug is present in the processor.
sep_bug: True if this bug is present in the processor.
f00f_bug: True if this bug is present in the processor.
coma_bug: True if this bug is present in the processor.
fpu: True if the CPU has a floating point unit.
fpu_exception: True if the floating point unit can throw an exception.
cpuid_level: The "cpuid" assembler instruction is only present on X86 CPUs. This attribute represents the level of the instruction that is supported by the CPU. The first CPUs had only level 1, newer chips have more levels and can thus return more information.
wp: No idea what this is on X86 CPUs.
flags: This is the set of flags that the CPU supports - this is returned as an array reference.
byte_order: The byte order of the CPU, might be little endian or big endian, or undefined for unknown.
bogomips: A system constant calculated when the kernel is booted - it is a (rather poor) measure of the CPU's performance.

EXPORT

None by default.

BUGS

The enormous bug in this is that I didnt realize when I made this that the contents of "/proc/cpuinfo " are different for different processors.

I really would be indebted if Linux users from other than x86 processors would help me document this properly.

The source can be found at

    https://github.com/jonathanstowe/Linux-Cpuinfo

Please feel free to fork, send patches etc.

COPYRIGHT AND LICENSE

See the README file in the Distribution Kit

AUTHOR

Jonathan Stowe, <jns@gellyfish.co.uk>