Guys,
this patch makes some small updates to the MD5 files.
* Remove some code which was retained in case changes for 64-bit processors didn't work
* Convert some macro temps to be .locals in calling function
* Omit 'library' path in load_bytecode calls
* General cleanup
Leo's previously reported that the wrong checksum is reported on big endian systems, but I don't have access to one to investigate. Offers kindly accepted.
[I know that MD5 is best implemented in C, but this library has been useful in shaking out a few issues in the past, and now useful for checking for regressions, especially in the GC and JIT. Breakage is very easy to spot! It would also be interesting for benchmarking, but I haven't got around to trying]
Regards,
Nick
Index: runtime/parrot/library/Digest/MD5.imc
===================================================================
--- runtime/parrot/library/Digest/MD5.imc (revision 8032)
+++ runtime/parrot/library/Digest/MD5.imc (working copy)
@@ -1,5 +1,5 @@
-# Parrot md5sum; Nick Glencross <[EMAIL PROTECTED]>
-# Improvements from Leo and Jens Rieks
+# Parrot MD5 library; Nick Glencross <[EMAIL PROTECTED]>
+# Improvements from Leo and Jens Rieks
#
# Based on md5.c, from md5sum
# written by Ulrich Drepper <[EMAIL PROTECTED]>, 1995.
@@ -10,30 +10,30 @@
=head1 SYNOPSIS
- load_bytecode "library/Digest/MD5.imc"
+ load_bytecode "Digest/MD5.imc"
$P0 = _md5sum("foo")
_md5_print($P0)
or
- load_bytecode "library/Digest/MD5.imc"
+ load_bytecode "Digest/MD5.imc"
$P0 = _md5sum("bar")
$S0 = _md5_hex($P0)
=head1 DESCRIPTION
-This is a pure Parrot MD5 hash routine. Therefore you should run it
-with the JIT core if possible.
+This is a pure Parrot MD5 hash routine. You should run it with the JIT
+core if possible.
=head1 SUBROUTINES
=head2 _md5sum
-Pass in a string, returns an Integer array with the result
+Pass in a string, returns an Integer array with the result.
=head2 _md5_hex
-Pass it the Integer array to get the checksum as string.
+Pass it the Integer array from _md5sum to get the checksum as string.
=head2 _md5_print
@@ -41,14 +41,20 @@
=head1 BUGS
-Only tested so far on i386.
+This section provides a list of bugs.
=over 4
-=item * Might work on 64 bit platforms
+=item *
-=item * Might not work on big endian systems (confirmed)
+By all accounts, doesn't seem to work on big endian systems, but I
+don't have access to investigate.
+=item *
+
+There is some scope to optimise things a little further, perhaps removing
+some of the 'and's (for 64-bit) which may be superflouous.
+
=back
=cut
@@ -63,14 +69,6 @@
$P0 = _config()
-# $I0 = $P0["intvalsize"]
-# if $I0 == 4 goto is_4byte_word
-#
-# printerr "This doesn't seem to be a 32 bit processor: "
-# printerr "Please verify the MD5 checksum\n"
-#
-#is_4byte_word:
-
$I0 = $P0["bigendian"]
unless $I0 goto is_little_endian
endian = 1
@@ -93,7 +91,7 @@
.include "library/config.imc"
###########################################################################
-# Main entry point
+# Main backend entry point
.namespace ["Digest"]
@@ -123,41 +121,41 @@
# A parrot rol instruction might be good (as it can often be JIT'd)
.macro rol (x,n, out)
- .out = .x << .n
- $I1000 = 32 - .n
- $I1000 = .x >>> $I1000
- .out |= $I1000
+ .out = .x << .n
+ tmp2 = 32 - .n
+ tmp2 = .x >>> tmp2
+ .out |= tmp2
.endm
.macro FF (b,c,d)
- tmp = .c ~ .d
- tmp = .b & tmp
- tmp = .d ~ tmp
+ tmp1 = .c ~ .d
+ tmp1 = .b & tmp1
+ tmp1 = .d ~ tmp1
.endm
.macro FH (b,c,d)
- tmp = .b ~ .c
- tmp = tmp ~ .d
+ tmp1 = .b ~ .c
+ tmp1 = tmp1 ~ .d
.endm
.macro FI (b,c,d)
- tmp = ~.d
- tmp = .b | tmp
- tmp = .c ~ tmp
+ tmp1 = ~.d
+ tmp1 = .b | tmp1
+ tmp1 = .c ~ tmp1
.endm
###########################################################################
# Higher level MD5 operations
.macro common (a, b, k, s, T)
- .a += tmp
+ .a += tmp1
.a += .T
- $I99 = .k + idx
- tmp = buffer[$I99]
- .a += tmp
+ tmp2 = .k + idx
+ tmp1 = buffer[tmp2]
+ .a += tmp1
.a &= 0xffffffff
- .rol (.a, .s, tmp)
- .a = .b + tmp
+ .rol (.a, .s, tmp1)
+ .a = .b + tmp1
.a &= 0xffffffff
.endm
@@ -221,12 +219,12 @@
.end
###########################################################################
-# Create a buffer from the requested buffer
+# Create an internal scratchpad buffer
.sub _md5_create_buffer
.param string str
- .param pmc buffer
- .param int endian
+ .param pmc buffer
+ .param int endian
.local int counter
.local int subcounter
@@ -240,7 +238,7 @@
# Work out how many words to allocate
.local int words
- words = len + 8
+ words = len + 8
words = words | 63
words = words + 1
words = words / 4
@@ -271,11 +269,11 @@
inc subcounter
if subcounter != 4 goto md5_create_buffer_loop
- if endian goto endian_ok
+
+ if endian goto endian_ok1
.swap (word)
-endian_ok:
+endian_ok1:
-
buffer[slow_counter] = word
word = 0
@@ -294,6 +292,7 @@
.local int shift
shift = 8*subcounter
word = word << shift
+
if endian goto endian_ok2
.swap (word)
endian_ok2:
@@ -317,13 +316,13 @@
###########################################################################
.sub _md5_process_buffer
- .param pmc context
+ .param pmc context
.param pmc buffer
.local int A, B, C, D
.local int A_save, B_save, C_save, D_save
- .local int tmp, idx, len
+ .local int tmp1, tmp2, idx, len
idx = 0
len = elements buffer
@@ -440,6 +439,7 @@
$P0 = new FixedIntegerArray
$P0 = 4
+
$P0[0] = A
$P0[1] = B
$P0[2] = C
@@ -451,7 +451,7 @@
###########################################################################
-# Retrieve the final checksum
+# Retrieve the final checksum as a string
.sub _md5_hex
.param pmc context
@@ -477,7 +477,7 @@
.return($S0)
.end
-# Convenience subroutine
+# Convenience subroutine to print the MD5 string
.sub _md5_print
.param pmc context
Index: t/library/md5.t
===================================================================
--- t/library/md5.t (revision 8032)
+++ t/library/md5.t (working copy)
@@ -14,7 +14,6 @@
use Parrot::Config;
-my $intsize = $PConfig{intsize};
my $bigendian = $PConfig{bigendian};
SKIP: {
Index: examples/assembly/md5sum.imc
===================================================================
--- examples/assembly/md5sum.imc (revision 8032)
+++ examples/assembly/md5sum.imc (working copy)
@@ -1,7 +1,7 @@
# Parrot md5sum; Nick Glencross <[EMAIL PROTECTED]>
# Improvements from Leo
#
-# Harness for the library/Digest/MD5.imc library
+# Harness for the Digest/MD5.imc library
=head1 NAME
@@ -28,7 +28,7 @@
.param pmc args
.local int size
- load_bytecode "library/Digest/MD5.imc"
+ load_bytecode "Digest/MD5.imc"
# Argument count
$I0 = args
$I0 = $I0 - 1
