Host introspection

Host introspection
You are encouraged to solve this task according to the task description, using any language you may know.

Print the word size and endianness of the host machine.

See also: Variable size/Get

Ada

with Ada.Text_IO;  use Ada.Text_IO;
with System;       use System;
 
procedure Host_Introspection is
begin
   Put_Line ("Word size" & Integer'Image (Word_Size));
   Put_Line ("Endianness " & Bit_Order'Image (Default_Bit_Order));
end Host_Introspection;

Sample output on a Pentium machine:

Word size 32
Endianness LOW_ORDER_FIRST

ALGOL 68

Works with: ALGOL 68 version Revision 1 - no extensions to language used

Works with: ALGOL 68G version Any - tested with release 1.18.0-9h.tiny

INT max abs bit = ABS(BIN 1 SHL 1)-1;
INT bits per char = ENTIER (ln(max abs char+1)/ln(max abs bit+1));
INT bits per int = ENTIER (1+ln(max int+1.0)/ln(max abs bit+1));
 
printf(($"states per bit: "dl$,max abs bit+1));
printf(($"bits per char: "z-dl$,bits per char));
printf(($"bits per int:  "z-dl$,bits per int));
printf(($"chars per int: "z-dl$,bits per int OVER bits per char));
 
printf(($"bits width: "z-dl$, bits width));
 
STRING abcds = "ABCD";
FILE abcdf;
INT abcdi;
 
INT errno := open(abcdf, "abcd.dat",stand back channel);
put(abcdf,abcds); # output alphabetically #
reset(abcdf);
get bin(abcdf,abcdi); # input in word byte order #
STRING int byte order := "";
FOR shift FROM 0 BY bits per char TO bits per int - bits per char DO
  int byte order +:= REPR(abcdi OVER (max abs bit+1) ** shift MOD (max abs char+1))
OD;
printf(($"int byte order: "g,", Hex:",16r8dl$,int byte order, BIN abcdi))

Output (Intel i686):

states per bit:  2
bits per char:   8
bits per int:   32
chars per int:   4
bits width:  32
int byte order: ABCD, Hex:44434241

On older CPUs the results would vary:

ALGOL 68R	ALGOL 68RS
~ bits per char: 6 bits per int: 24 chars per int: 4	ICL 2900 bits per char: 8 bits per int: 32 chars per int: 4	Multics bits per char: 6 bits per int: 36 chars per int: 6

C

#include <stdio.h>
#include <stddef.h> /* for size_t */
#include <limits.h> /* for CHAR_BIT */
 
int main() {
    int one = 1;
 
    /*
     * Best bet: size_t typically is exactly one word.
     */
    printf("word size = %d bits\n", (int)(CHAR_BIT * sizeof(size_t)));
 
    /*
     * Check if the least significant bit is located
     * in the lowest-address byte.
     */
    if (*(char *)&one)
        printf("little endian\n");
    else
        printf("big endian\n");
    return 0;
}

On POSIX-compatible systems, the following also tests the endianness (this makes use of the fact that network order is big endian):

#include <stdio.h>
#include <arpa/inet.h>
 
int main()
{
  if (htonl(1) == 1)
    printf("big endian\n");
  else
    printf("little endian\n");
}

C#

static void Main()
{
  Console.WriteLine("Word size = {0} bytes,",sizeof(int));
 
  if (BitConverter.IsLittleEndian)
    Console.WriteLine("Little-endian.");
  else
    Console.WriteLine("Big-endian.");
}

Clojure

(println "word size: " (System/getProperty "sun.arch.data.model"))
(println "endianness: " (System/getProperty "sun.cpu.endian"))

Common Lisp

Common Lisp doesn't provide a native way to reliably determine this (though some unlike other languages, you rarely, if ever, need this information).

The Environment has some implementation-specific functions that might provide a good hint, e.g.,

(machine-type) ;; => "X86-64" on SBCL here

The *features* list also provides useful information, e.g., some compilers declare :LITTLE-ENDIAN there.

The cl-trivial-features library standardizes this, so you will always get either :LITTLE-ENDIAN or :BIG-ENDIAN. It also adds the CPU (:X86, :X86-64, :PPC, :PPC64, etc.), from which you can probably derive the word size, but it's not (yet) available as a separate flag.

D

import std.stdio, std.system;
 
void main() {
  writefln("word size = ", size_t.sizeof * 8);
  writefln(endian == Endian.LittleEndian ? "little" : "big", " endian");
}

Delphi

program HostIntrospection ;
 
{$APPTYPE CONSOLE}
 
uses SysUtils;
 
begin
  Writeln('word size: ', SizeOf(Integer));
  Writeln('endianness: little endian'); // Windows is always little endian
end.

Erlang

To find the word size:

1> erlang:system_info(wordsize).
4

In the case of endianness, Erlang's bit syntax by default has a 'native' option which lets you use what is supported natively. As such, there is no function to find endianness. However, one could write one by using bit syntax, setting endianness and then comparing to the native format:

1> <<1:4/native-unit:8>>.
<<1,0,0,0>>
2> <<1:4/big-unit:8>>
<<0,0,0,1>>
3> <<1:4/little-unit:8>>.
<<1,0,0,0>>

And so the following function would output endiannes:

endianness() when <<1:4/native-unit:8>> =:= <<1:4/big-unit:8>> -> big;
endianness() -> little.

Factor

USING: alien.c-types io layouts ;
"Word size: " write cell 8 * .
"Endianness: " write little-endian? "little" "big" ? print

Forth

: endian
  cr 1 cells . ." address units per cell"
  s" ADDRESS-UNIT-BITS" environment? if cr . ." bits per address unit" then
  cr 1 here ! here c@ if ." little" else ." big" then ."  endian" ;

This relies on c@ being a byte fetch (4 chars = 1 cells). Although it is on most architectures, ANS Forth only guarantees that 1 chars <= 1 cells. Some Forths like OpenFirmware have explicitly sized fetches, like b@.

Fortran

Works with: Fortran version 90 and later

INTEGER, PARAMETER :: i8 = SELECTED_INT_KIND(2)
INTEGER, PARAMETER :: i16 = SELECTED_INT_KIND(4)
INTEGER(i8) :: a(2)
INTEGER(i16) :: b
 
WRITE(*,*) bit_size(1)     ! number of bits in the default integer type
                           ! which may (or may not!) equal the word size
 
b = Z'1234'                ! Hexadecimal assignment
a = (TRANSFER(b, a))       ! Split a 16 bit number into two 8 bit numbers
 
IF (a(1) == Z'12') THEN    ! where did the most significant 8 bits end up
  WRITE(*,*) "Big Endian"
ELSE
  WRITE(*,*) "Little Endian"
END IF

F#

A lot of research before I finally came up with an answer to this that isn't dependent on the machine it was compiled on. Works on Win32 machines only (obviously, due to the interop). I think that strictly speaking, I should be double checking the OS version before making the call to wow64Process, but I'm not worrying about it.

open System
open System.Runtime.InteropServices
open System.Diagnostics
 
[<DllImport("kernel32.dll", SetLastError = true, CallingConvention = CallingConvention.Winapi)>]
extern bool IsWow64Process(nativeint hProcess, bool &wow64Process);
 
let answerHostInfo =
    let Is64Bit() =
        let mutable f64Bit = false;
        IsWow64Process(Process.GetCurrentProcess().Handle, &f64Bit) |> ignore
        f64Bit
    let IsLittleEndian() = BitConverter.IsLittleEndian
    (IsLittleEndian(), Is64Bit())

Go

package main
 
import (
    "fmt"
    "io/ioutil"
    "strings"
    "unsafe"
)
 
func main() {
    // inspect an int variable to determine endianness
    x := 1
    if *(*byte)(unsafe.Pointer(&x)) == 1 {
        fmt.Println("little endian")
    } else {
        fmt.Println("big endian")
    }
    // inspect cpuinfo to determine word size (unix-like os only)
    c, err := ioutil.ReadFile("/proc/cpuinfo")
    if err != nil {
        fmt.Println(err)
        return
    }
    ls := strings.Split(string(c), "\n")
    for _, l := range ls {
        if strings.HasPrefix(l, "flags") {
            for _, f := range strings.Fields(l) {
                if f == "lm" { // "long mode"
                    fmt.Println("64 bit word size")
                    return
                }
            }
            fmt.Println("32 bit word size")
            return
        }
    }
    fmt.Println("cpuinfo flags not found")
}

Output:

little endian
64 bit word size

Alternative technique:

package main
 
import (
    "debug/elf"
    "fmt"
    "os"
)
 
func main() {
    f, err := elf.Open(os.Args[0])
    if err != nil {
        fmt.Println("  ", err)
        return
    }
    fmt.Println(f.FileHeader.ByteOrder)
    f.Close()
}

Output:

LittleEndian

Groovy

Solution follows Java:

println "word size:  ${System.getProperty('sun.arch.data.model')}"
println "endianness: ${System.getProperty('sun.cpu.endian')}"

Output:

word size:  64
endianness: little

Haskell

import Data.Bits
import ADNS.Endian -- http://hackage.haskell.org/package/hsdns
 
main = do
  putStrLn $ "Word size: " ++ bitsize
  putStrLn $ "Endianness: " ++ show endian
      where
        bitsize = show $ bitSize (undefined :: Int)

J

   IF64 {32 64
64

This returns 32 in 32 bit J.

Note that this mechanism is testing the interpreter, and not the OS or Hardware. (Though, of course, you cannot run a 64 bit interpreter on a machine that does not support it.)

That said, this does not deal with endianness. For the most part, J programs do not need to know their own endianness. When converting to and from binary format you can specify "native", "little endian" and "big endian", and it's rare that you have an interface with conflicting needs. That said:

   ":&> (|: 32 64  ;"0 big`little) {"_1~ 2 2 #: 16b_e0 + a. i. 0 { 3!:1  '' 
64    
little

Java

Java conceals the byte order of its integers, but reports the native byte order through java.nio.ByteOrder.nativeOrder().

Works with: Java version 1.4

import java.nio.ByteOrder;
 
public class ShowByteOrder {
    public static void main(String[] args) {
        // Print "BIG_ENDIAN" or "LITTLE_ENDIAN".
        System.out.println(ByteOrder.nativeOrder());
    }
}

Some JVMs also have system properties for the word size and byte order.

System.out.println("word size: "+System.getProperty("sun.arch.data.model"));
System.out.println("endianness: "+System.getProperty("sun.cpu.endian"));

Mathematica

If[$ByteOrdering > 0, Print["Big endian"], Print["Little endian" ]] 
$SystemWordLength "bits"

x86 Output:

Little endian
32 bits

MATLAB / Octave

The concept of "word size" is not meaningful in Matlab and Octave, uint64 is also available on 32bit-platforms, and there are no pointers. Endianity can be tested with the function below:

  function [endian]=endian()
    fid=tmpfile();
    fwrite(fid,1:8,'uint8');
 
    fseek(fid,0,'bof');
    t=fread(fid,8,'int8');
    i8=sprintf('%02X',t);
 
    fseek(fid,0,'bof');
    t=fread(fid,4,'int16');
    i16=sprintf('%04X',t);
 
    fclose(fid);
 
    if strcmp(i8,i16) endian='big';
    else endian='little';
    end;
 

Output:

  octave:128> computer 
  x86_64-unknown-linux-gnu
  octave:129> endian
  endian = little

Modula-3

MODULE Host EXPORTS Main;
 
IMPORT IO, Fmt, Word, Swap;
 
BEGIN
  IO.Put("Word Size: " & Fmt.Int(Word.Size) & "\n");
  IF Swap.endian = Swap.Endian.Big THEN
    IO.Put("Endianness: Big\n");
  ELSE
    IO.Put("Endianness: Little\n");
  END;
END Host.

Output (on an x86):

Word Size: 32
Endianness: Little

Bold text

NetRexx

Translation of: Java

NetRexx can access this information from the Java virtual machine in the same way as the Java sample above.

/* NetRexx */
options replace format comments java crossref savelog symbols nobinary
 
wordSize = System.getProperty('sun.arch.data.model')
endian   = System.getProperty('sun.cpu.endian')
 
say ' word size:' wordSize
say 'endianness:' endian
 

Objective-C

Endianness:

switch (NSHostByteOrder()) {
  case NS_BigEndian:
    NSLog(@"%@", @"Big Endian");
    break;
  case NS_LittleEndian:
    NSLog(@"%@", @"Little Endian");
    break;
  case NS_UnknownByteOrder:
    NSLog(@"%@", @"endianness unknown");
    break;
} 

Architecture: Template:Works on

switch ([NSRunningApplication currentApplication].executableArchitecture) {
  case NSBundleExecutableArchitectureI386:
    NSLog(@"%@", @"i386 32-bit");
    break;
 
  case NSBundleExecutableArchitectureX86_64:
    NSLog(@"%@", @"x86_64 64-bit");
    break;
 
  case NSBundleExecutableArchitecturePPC:
    NSLog(@"%@", @"PPC 32-bit");
    break;
 
  case NSBundleExecutableArchitecturePPC64:
    NSLog(@"%@", @"PPC64 64-bit");
    break;
 
  default:
    NSLog(@"%@", @"Unknown");
    break;
}

OCaml

Printf.printf "%d\n" Sys.word_size; (* Print word size *)
Printf.printf "%s\n" Sys.os_type;   (* Print operating system *)

Endianness is hidden in ocaml, but there are tricks. For example in Linux or Unix variants, one may use the uname shell command :

let uname arg =
  let arg = if arg = "" then "-" else arg in
  let ic = Unix.open_process_in ("uname -" ^ arg) in
  (input_line ic)
;;
 
# uname "sm";;
- : string = "Linux i686"

In most cases, endianness can be infered from informations given by uname.

One may also read files in the /proc directory in order to get informations about the host, only under linux :

(* Reading all the lines from a file.
If the loop is implemented by a recursive auxiliary function, the try...with breaks
tail recursion if not written carefully *)
let lines name =
  let f = open_in name
  and r = ref []
  in
  (try
     while true do
       r := (input_line f)::!r
     done
   with End_of_file -> close_in f);
  (List.rev !r)
;;
 
# lines "/proc/meminfo";;
- : string list =
["MemTotal:      2075240 kB"; "MemFree:        469964 kB";
 "Buffers:         34512 kB"; "Cached:        1296380 kB";
 "SwapCached:         96 kB"; "Active:         317484 kB";
 "Inactive:      1233500 kB"; "HighTotal:     1178432 kB";
 "HighFree:        45508 kB"; "LowTotal:       896808 kB";
 "LowFree:        424456 kB"; "SwapTotal:     2650684 kB";
 "SwapFree:      2650588 kB"; "Dirty:             228 kB";
 "Writeback:           0 kB"; "AnonPages:      220036 kB";
 "Mapped:          67160 kB"; "Slab:            41540 kB";
 "SReclaimable:    34872 kB"; "SUnreclaim:       6668 kB";
 "PageTables:       1880 kB"; "NFS_Unstable:        0 kB";
 "Bounce:              0 kB"; "WritebackTmp:        0 kB";
 "CommitLimit:   3688304 kB"; "Committed_AS:   549912 kB";
 "VmallocTotal:   114680 kB"; "VmallocUsed:      5172 kB";
 "VmallocChunk:   109320 kB"; "HugePages_Total:     0";
 "HugePages_Free:      0"; "HugePages_Rsvd:      0";
 "HugePages_Surp:      0"; "Hugepagesize:     4096 kB"]

Same methods can be used to get the results of commands lshw, dmidecode...

Pascal

program HostIntrospection(output);
begin
  writeln('Pointer size: ', SizeOf(Pointer), ' byte, i.e. ', SizeOf(Pointer)*8, ' bit.');
{ NtoBE converts from native endianess to big endianess }
  if 23453 = NtoBE(23453) then
    writeln('This host is big endian.')
  else
    writeln('This host is little endian.');
end.

Output:

>: ./HostIntrospection
Pointer size: 4 byte, i.e. 32 bit.
This host is little endian.

Perl

Most basic example:

use Config;
print "UV size: $Config{uvsize}, byte order: $Config{byteorder}\n";

Example output:

UV size: 4, byte order: 1234

More verbose example:

use 5.010;
use Config;
my ($size, $order, $end) = @Config{qw(uvsize byteorder)};
given ($order) {
    when (join '', sort split '') { $end = 'little' }
    when (join '', reverse sort split '') { $end = 'big' }
    default { $end = 'mixed' }
}
say "UV size: $size, byte order: $order ($end-endian)";

Example outputs:

UV size: 4, byte order: 1234 (little-endian)

UV size: 4, byte order: 3412 (mixed-endian)

UV size: 8, byte order: 87654321 (big-endian)

PicoLisp

We inspect the ELF header of the executable file (the 'cmd' function returns the path to the command that invoked the interpreter). Note that this (like most other contributions to this task) only tells how the binary was compiled/assembled/linked, not necessarily the nature of the underlying system.

(in (cmd)                              # Inspect ELF header
   (rd 4)                              # Skip "7F" and 'E', 'L' and 'F'
   (prinl
      (case (rd 1)                     # Get EI_CLASS byte
         (1 "32 bits")
         (2 "64 bits")
         (T "Bad EI_CLASS") ) )
   (prinl
      (case (rd 1)                     # Get EI_DATA byte
         (1 "Little endian")
         (2 "Big endian")
         (T "Bad EI_DATA") ) ) )

Output:

64 bits
Little endian

PowerShell

Write-Host Word Size: ((Get-WMIObject Win32_Processor).DataWidth)
Write-Host -NoNewLine "Endianness: "
if ([BitConverter]::IsLittleEndian) {
    Write-Host Little-Endian
} else {
    Write-Host Big-Endian
}

Note that endianness is essentially a moot point with PowerShell, as there is only a Windows implementation currently and current Windows versions don't run on big-endian systems. But in theory this check should work.

PureBasic

Enumeration 
  #LittleEndian
  #BigEndian
EndEnumeration
 
ProcedureDLL EndianTest()
  Protected Endian = #LittleEndian
  Protected dummy.l= 'ABCD'
  If "A"=Chr(PeekA(@dummy))
    Endian=#BigEndian
  EndIf
  ProcedureReturn Endian  
EndProcedure
 
;- *** Start of test code
If OpenConsole()  
  PrintN("Your word size is "+Str(SizeOf(Integer)) +" bytes,")
  Select EndianTest()
    Case #LittleEndian
      PrintN("and you use Little Endian.")
    Default
      PrintN("and you use Big Endian.")
  EndSelect
EndIf

Python

>>> import sys, math
>>> int(round(math.log(sys.maxint,2)+1)) # this only works in Python 2.x
32
>>> import struct
>>> struct.calcsize('i') * 8
32
>>> sys.byteorder
little
>>> import socket
>>> socket.gethostname()
'PADDY3118-RESTING'
>>>

R

Word size

8 * .Machine$sizeof.long # e.g. 32
# or
object.size(0L)          # e.g. 32 bytes

Endianness

.Platform$endian         # e.g. "little"

Retro

These introspections are possible through the standard variations library.

Word Size

needs variations'
^variations'size

Returns the number of bits per cell. This is normally 32, though may be smaller or larger on embedded systems and under special cases.

Endianness

needs variations'
^variations'endian

Returns 0 for little endian, and 1 for big endian.

REXX

Since all variables in the REXX language are stored as characters, the wordsize is immaterial (REXX supports variable precision for numbers). This also applies to the "endiness" of words or how they are stored.
The REXX language was designed for scripting and interfacing with the operating system.
However, there is a STORAGE built-in function that allows a programmer to look at (local) storage, and if there is an indicator stored anywhere in the virtual address space, that it can be examined.

Ruby

# We assume that a Fixnum occupies one machine word.
# Fixnum#size returns bytes (1 byte = 8 bits).
word_size = 42.size * 8
puts "Word size: #{word_size} bits"
 
# Array#pack knows the native byte order. We pack 1 as a 16-bit integer,
# then unpack bytes: [0, 1] is big endian, [1, 0] is little endian.
bytes = [1].pack('S').unpack('C*')
byte_order = (bytes[0] == 0 ? 'big' : 'little') + ' endian'
puts "Byte order: #{byte_order}"

With MRI, ri Fixnum states, "A Fixnum holds Integer values that can be represented in a native machine word (minus 1 bit)." This bases our claim that a Fixnum occupies one machine word.

Some other implementations of Ruby are different. With JRuby, a Fixnum is always 64 bits, because it is a Java long (1). JRuby uses the correct native byte order by calling java.nio.ByteOrder.nativeOrder() (2).

Scheme

Works with: Chicken Scheme

(define host-info
  (begin
    (display "Endianness: ")
    (display (machine-byte-order))
    (newline)
    (display "Word Size: ")
    (display (if (fixnum? (expt 2 33)) 64 32))
    (newline)))

Output:

Endianness: little-endian
Word Size: 32

Slate

inform: 'Endianness: ' ; Platform current endianness.
inform: 'Word Size: ' ; (Platform current bytesPerWord * 8) printString.

Output:

Endianness: LittleEndian
Word Size: 32

Tcl

This is very straightforward in Tcl. The global array tcl_platform contains these values. In an interactive tclsh:

% parray tcl_platform
tcl_platform(byteOrder)   = littleEndian
tcl_platform(machine)     = intel
tcl_platform(os)          = Windows NT
tcl_platform(osVersion)   = 5.1
tcl_platform(platform)    = windows
tcl_platform(pointerSize) = 4
tcl_platform(threaded)    = 1
tcl_platform(user)        = glennj
tcl_platform(wordSize)    = 4

TI-89 BASIC

Disp "32-bit big-endian"