Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.lang.forth > #17811 > unrolled thread

SHA-512

Started bymhx@iae.nl (Marcel Hendrix)
First post2012-12-02 23:48 +0200
Last post2012-12-08 22:08 +0200
Articles 11 on this page of 31 — 5 participants

Back to article view | Back to comp.lang.forth


Contents

  SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-02 23:48 +0200
    Re: SHA-512 jzakiya@gmail.com - 2012-12-07 14:47 -0800
      Re: SHA-512 Coos Haak <chforth@hccnet.nl> - 2012-12-08 00:49 +0100
        Re: SHA-512 jzakiya@gmail.com - 2012-12-07 16:54 -0800
          Re: SHA-512 Coos Haak <chforth@hccnet.nl> - 2012-12-08 18:21 +0100
        Re: SHA-512 jzakiya@gmail.com - 2012-12-08 18:36 -0800
          Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-09 09:14 +0200
            Re: SHA-512 jzakiya@gmail.com - 2012-12-09 17:23 -0800
              Re: SHA-512 jzakiya@gmail.com - 2012-12-09 17:29 -0800
                Re: SHA-512 jzakiya@gmail.com - 2012-12-09 20:46 -0800
                Re: SHA-512 m.a.m.hendrix@tue.nl - 2012-12-10 03:53 -0800
                  Re: SHA-512 jzakiya@gmail.com - 2012-12-10 13:05 -0800
                    Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-10 23:17 +0200
                      Re: SHA-512 jzakiya@gmail.com - 2012-12-13 06:30 -0800
                        Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-13 21:33 +0200
                          Re: SHA-512 jzakiya@gmail.com - 2012-12-13 19:38 -0800
                            Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-14 22:52 +0200
                              Re: SHA-512 jzakiya@gmail.com - 2012-12-14 16:39 -0800
                                Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-15 19:11 +0200
                                  Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-16 16:00 +0200
                                    Re: SHA-512 jzakiya@gmail.com - 2012-12-16 17:33 -0800
                                      Re: SHA-512 jzakiya@gmail.com - 2012-12-16 22:35 -0800
                                        Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-17 10:14 +0200
                                      Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-17 10:51 +0200
                                        Re: SHA-512 jzakiya@gmail.com - 2012-12-17 09:13 -0800
                                          Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-17 22:24 +0200
                                            Re: SHA-512 jzakiya@gmail.com - 2013-01-15 20:38 -0800
                                            Re: SHA-512 jzakiya@gmail.com - 2013-01-15 20:54 -0800
                        Re: SHA-512 Howerd <howerdo@yahoo.co.uk> - 2012-12-13 13:06 -0800
                      Re: SHA-512 Howerd <howerdo@yahoo.co.uk> - 2012-12-13 13:03 -0800
      Re: SHA-512 mhx@iae.nl (Marcel Hendrix) - 2012-12-08 22:08 +0200

Page 2 of 2 — ← Prev page 1 [2]


#18040

Fromjzakiya@gmail.com
Date2012-12-16 17:33 -0800
Message-ID<cd9d5192-14e2-46d9-8117-84df2b72d9cd@googlegroups.com>
In reply to#18039
On Sunday, December 16, 2012 9:00:07 AM UTC-5, Marcel Hendrix wrote:
> mhx@iae.nl (Marcel Hendrix) writes Re: SHA-512
> 
> 
> 
> > jzakiya@gmail.com writes Re: SHA-512
> 
> [..]
> 
> > I have looked at my own (PICK-less) idea of SHA-512 again, and it couldn't work. 
> 
> > However, I found a way to improve your algorithm (literalize H[x]). ATM  test3  
> 
> > runs in 11.048 seconds on my 2.66 GHz i7 system. SHAfile needs 2.741 seconds 
> 
> > to process 385,241,618 bytes (64.9.. 50 MB/s/GHz). Hopefully your fix of 15 PICK 
> 
> > will create additional room for speedup.
> 
> [..]
> 
> 
> 
> The PICK fix works now too, but the speedup is only slight.
> 
> Performance is 65.95 MB/s/GHz for strings, 52.16 MB/s/GHz for file. 
> 
> This should be faster than the NIST reference (99MB/s on a 1.8 GHz Core2).
> 
> 
> 
> The following code is not portable.
> 
> 
> 
> -marcel
> 
> -- 
> 
> (*
> 
>  * LANGUAGE    : ANS Forth with extensions
> 
>  * PROJECT     : Forth Environments
> 
>  * DESCRIPTION : SHA-512 64-bit Hash algorithm
> 
>  * CATEGORY    : Utility
> 
>  * AUTHOR      : Copyright (c) 2012 Jabari Zakiya -- jzakiya@mail.com  12/07/2012 
> 
>  * LAST CHANGE : December 16, 2012, Marcel Hendrix 
> 
>  *)
> 
> 
> 
> 
> 
> 
> 
> 	NEEDS -miscutil
> 
> 
> 
> 	REVISION -sha-512 "--- SHA-512             Version 1.01 ---"
> 
> 
> 
> 	PRIVATES
> 
> 
> 
> DOC
> 
> (*
> 
>    NIST spec at:  http://csrc.nist.gov/encryption/tkhash.html
> 
> 
> 
> 	FORTH> SHAtest
> 
> 	SHA-512 test suite:
> 
> 
> 
> 	""
> 
> 	cf83e1357eefb8bd f1542850d66d8007 d620e4050b5715dc 83f4a921d36ce9ce 47d0d13c5d85f2b0 ff8318d2877eec2f 63b931bd47417a81 a538327af927da3e
> 
> 	cf83e1357eefb8bd f1542850d66d8007 d620e4050b5715dc 83f4a921d36ce9ce 47d0d13c5d85f2b0 ff8318d2877eec2f 63b931bd47417a81 a538327af927da3e
> 
> 
> 
> 	"abc"
> 
> 	ddaf35a193617aba cc417349ae204131 12e6fa4e89a97ea2 0a9eeee64b55d39a 2192992a274fc1a8 36ba3c23a3feebbd 454d4423643ce80e 2a9ac94fa54ca49f
> 
> 	ddaf35a193617aba cc417349ae204131 12e6fa4e89a97ea2 0a9eeee64b55d39a 2192992a274fc1a8 36ba3c23a3feebbd 454d4423643ce80e 2a9ac94fa54ca49f
> 
> 
> 
> 	"abcdefghbcdefghicdefghijdefghijkefghijklfghijklmghijklmnhijklmnoijklmnopjklmnopqklmnopqrlmnopqrsmnopqrstnopqrstu"
> 
> 	8e959b75dae313da 8cf4f72814fc143f 8f7779c6eb9f7fa1 7299aeadb6889018 501d289e4900f7e4 331b99dec4b5433a c7d329eeb6dd2654 5e96e55b874be909
> 
> 	8e959b75dae313da 8cf4f72814fc143f 8f7779c6eb9f7fa1 7299aeadb6889018 501d289e4900f7e4 331b99dec4b5433a c7d329eeb6dd2654 5e96e55b874be909
> 
> 
> 
> 	"The quick brown fox jumps over the lazy dog"
> 
> 	07e547d9586f6a73 f73fbac0435ed769 51218fb7d0c8d788 a309d785436bbb64 2e93a252a954f239 12547d1e8a3b5ed6 e1bfd7097821233f a0538f3db854fee6
> 
> 	07e547d9586f6a73 f73fbac0435ed769 51218fb7d0c8d788 a309d785436bbb64 2e93a252a954f239 12547d1e8a3b5ed6 e1bfd7097821233f a0538f3db854fee6
> 
> 
> 
> 	2 million copies of ASCII 'a' (61h)
> 
> 	9bc68759247e3332 bec1c79d128d28a8 931d0c9f96c8aa97 5731b563475fdddd df7f873c25086908 effe270e23c5a01e 5dfb3289bf5d091d 8fb454b1bcf98dda
> 
> 
> 
> 	400,000 copies of ASCII BL (20h)
> 
> 	19ef4876e03c9347 6e00b486c62108d4 d7136f6d4782817b 449bacf44d0963e0 3f5ab474cb6822c2 963a979e8e15298c d60f22841c51f616 1620c0e031c65f4a
> 
> 	 ok
> 
> 
> 
> 	FORTH> SHAbm
> 
> 	SHA-512 test for EX1, 1000 loops: 0.001 seconds elapsed.
> 
> 	SHA-512 test for EX2, 1000 loops: 0.003 seconds elapsed.
> 
> 	SHA-512 test for EX3, 1000 loops: 10.872 seconds elapsed. ok
> 
> 
> 
> 	2e6 (bytes) 1000e f* ( loops) 10.872e ( s) f/  2.66e ( GHz) f/  20e f2^x ( 1MB) f/  f. ( 65.95 MB/s/GHz )
> 
> 	( NIST ref.: 99 MB/s on 1.8 GHz Core2 )
> 
> 
> 
> 	FORTH> TIMER-RESET S" C:\idfwforth\df_snapshot_4.0_2010_08_19.tar.gz" @SHAfile .ELAPSED
> 
> 	Bytesize: 385241618
> 
> 	SHA-512 : 45800c7d8fe7c13f 6670176b93b82c34 6f6f072ff1f50aee 586ccd6e7b02ef7c e27eb8f5a3cb814a 74e1d18b3107ea9b 8fdf171bf573a420 2a666c67703437bf
> 
> 	2.648 seconds elapsed. ok
> 
> 	385241618e ( bytes) 2.741e ( s) f/  2.66e ( GHz) f/  20e f2^x ( 1MB) f/  f.  ( 52.16 MB/s/GHz )
> 
> *)
> 
> ENDDOC
> 
> 
> 
> #64 =: CELLSIZE	 PRIVATE -- CPU bitsize
> 
> 
> 
> 0. DVALUE SHAlen PRIVATE		  -- Holds byte length of string < 2^128 bits|2^125 bytes
> 
> CREATE SHAval    PRIVATE   8 CELLS ALLOT  -- Holds hash after each block
> 
> CREATE SHAsh     PRIVATE #88 CELLS ALLOT  -- Fully extended hash array
> 
> CREATE W         PRIVATE #16 CELLS ALLOT  -- Holds message block
> 
> ALIGN1024				  -- this helps, why??	
> 
> CREATE X         PRIVATE #80 CELLS ALLOT  -- Holds temp processed data
> 
> 
> 
> 0 VALUE =X       PRIVATE -- Pointer to addr of temp data
> 
> 0 VALUE =H       PRIVATE -- Pointer to addr of hash value H for each round
> 
> CREATE cbuffer   PRIVATE  $8000 CHARS ALLOT
> 
> 
> 
> : >cb       cbuffer CELLPLACE+  BL cbuffer CELLCHAR+ ; PRIVATE ( c-addr u -- ) 
> 
> : InitMake  SHAsh TO =H  X TO =X  cbuffer 0!  S" : SHA512 ( -- )" >cb ; PRIVATE
> 
> : EndMake   S" ;" >cb  cbuffer @+ EVALUATE ; PRIVATE 
> 
> : :=: 	    CREATE PRIVATE ,   DOES> @ (H.) >cb ; PRIVATE
> 
> 
> 
> $428a2f98d728ae22 :=: K0   $7137449123ef65cd :=: K1   $b5c0fbcfec4d3b2f :=: K2   $e9b5dba58189dbbc :=: K3
> 
> $3956c25bf348b538 :=: K4   $59f111f1b605d019 :=: K5   $923f82a4af194f9b :=: K6   $ab1c5ed5da6d8118 :=: K7
> 
> $d807aa98a3030242 :=: K8   $12835b0145706fbe :=: K9   $243185be4ee4b28c :=: K10  $550c7dc3d5ffb4e2 :=: K11
> 
> $72be5d74f27b896f :=: K12  $80deb1fe3b1696b1 :=: K13  $9bdc06a725c71235 :=: K14  $c19bf174cf692694 :=: K15
> 
> $e49b69c19ef14ad2 :=: K16  $efbe4786384f25e3 :=: K17  $0fc19dc68b8cd5b5 :=: K18  $240ca1cc77ac9c65 :=: K19
> 
> $2de92c6f592b0275 :=: K20  $4a7484aa6ea6e483 :=: K21  $5cb0a9dcbd41fbd4 :=: K22  $76f988da831153b5 :=: K23
> 
> $983e5152ee66dfab :=: K24  $a831c66d2db43210 :=: K25  $b00327c898fb213f :=: K26  $bf597fc7beef0ee4 :=: K27
> 
> $c6e00bf33da88fc2 :=: K28  $d5a79147930aa725 :=: K29  $06ca6351e003826f :=: K30  $142929670a0e6e70 :=: K31
> 
> $27b70a8546d22ffc :=: K32  $2e1b21385c26c926 :=: K33  $4d2c6dfc5ac42aed :=: K34  $53380d139d95b3df :=: K35
> 
> $650a73548baf63de :=: K36  $766a0abb3c77b2a8 :=: K37  $81c2c92e47edaee6 :=: K38  $92722c851482353b :=: K39
> 
> $a2bfe8a14cf10364 :=: K40  $a81a664bbc423001 :=: K41  $c24b8b70d0f89791 :=: K42  $c76c51a30654be30 :=: K43
> 
> $d192e819d6ef5218 :=: K44  $d69906245565a910 :=: K45  $f40e35855771202a :=: K46  $106aa07032bbd1b8 :=: K47
> 
> $19a4c116b8d2d0c8 :=: K48  $1e376c085141ab53 :=: K49  $2748774cdf8eeb99 :=: K50  $34b0bcb5e19b48a8 :=: K51
> 
> $391c0cb3c5c95a63 :=: K52  $4ed8aa4ae3418acb :=: K53  $5b9cca4f7763e373 :=: K54  $682e6ff3d6b2b8a3 :=: K55
> 
> $748f82ee5defb2fc :=: K56  $78a5636f43172f60 :=: K57  $84c87814a1f0ab72 :=: K58  $8cc702081a6439ec :=: K59
> 
> $90befffa23631e28 :=: K60  $a4506cebde82bde9 :=: K61  $bef9a3f7b2c67915 :=: K62  $c67178f2e372532b :=: K63
> 
> $ca273eceea26619c :=: K64  $d186b8c721c0c207 :=: K65  $eada7dd6cde0eb1e :=: K66  $f57d4f7fee6ed178 :=: K67
> 
> $06f067aa72176fba :=: K68  $0a637dc5a2c898a6 :=: K69  $113f9804bef90dae :=: K70  $1b710b35131c471b :=: K71
> 
> $28db77f523047d84 :=: K72  $32caab7b40c72493 :=: K73  $3c9ebe0a15c9bebc :=: K74  $431d67c49c100d4c :=: K75
> 
> $4cc5d4becb3e42b6 :=: K76  $597f299cfc657e2a :=: K77  $5fcb6fab3ad6faec :=: K78  $6c44198c4a475817 :=: K79
> 
> 
> 
> : H[H]  =H             (H.) >cb ; PRIVATE -- Return H addr
> 
> : H[G]  =H  1 CELLS  + (H.) >cb ; PRIVATE -- Return G addr
> 
> : H[F]  =H  2 CELLS  + (H.) >cb ; PRIVATE -- Return F addr
> 
> : H[E]  =H  3 CELLS  + (H.) >cb ; PRIVATE -- Return E addr
> 
> : H[D]  =H  4 CELLS  + (H.) >cb ; PRIVATE -- Return D addr
> 
> : H[C]  =H  5 CELLS  + (H.) >cb ; PRIVATE -- Return C addr
> 
> : H[B]  =H  6 CELLS  + (H.) >cb ; PRIVATE -- Return B addr
> 
> : H[A]  =H  7 CELLS  + (H.) >cb ; PRIVATE -- Return A addr
> 
> 
> 
> : SHAinit ( -- )  \ Load initial hash values H0 - H7
> 
> 	$6a09e667f3bcc908 ( H0)  $bb67ae8584caa73b ( H1)
> 
> 	$3c6ef372fe94f82b ( H2)  $a54ff53a5f1d36f1 ( H3)
> 
> 	$510e527fade682d1 ( H4)  $9b05688c2b3e6c1f ( H5)
> 
> 	$1f83d9abfb41bd6b ( H6)  $5be0cd19137e2179 ( H7)
> 
> 	SHAsh  7 0 DO  TUCK  !  CELL+  LOOP  !		\ Put initial hash in SHAsh array
> 
> 	SHAsh  SHAval  8 CELLS  CMOVE ; PRIVATE        	\ Put copy in SHAval array
> 
> 
> 
> : UpDateHash ( a b c -- )  \ Update hash values and load arrays with new values
> 
> 	8 0 DO  DUP >R @  SWAP  DUP >R @  +  DUP        \ Compute updated hash subvalue
> 
> 	  	R@ !  OVER !  CELL+  R> CELL+  R> CELL+ \ Store updated hash subvalue
> 
> 	  LOOP  3DROP ; PRIVATE				\ Clear stack when done
> 
> 
> 
> : >offs	  ( u -- addr ) CELLS =X + (H.) >cb ; PRIVATE
> 
> 
> 
> : sig0    DUP  DUP   1 ROR  SWAP   8 ROR  XOR  SWAP  7 RSHIFT  XOR ; PRIVATE 	( x -- n ) 
> 
> : sig1    DUP  DUP #19 ROR  SWAP #61 ROR  XOR  SWAP  6 RSHIFT  XOR ; PRIVATE 	( x -- n ) 
> 
> : Wi	  #-16 >offs S" D@ sig0 +" >cb  -7 >offs S" @ +" >cb  -2 >offs S" @ sig1 +  DUP" >cb  0 >offs S" !" >cb  CELL +TO =X ; PRIVATE ( ..Wi --..Wi' Wi') 
> 
> : Wi@     S" DUP @ DUP" >cb  0 >offs S" !" >cb   CELL +TO =X  ; PRIVATE    \ dup @ tuck 
> 
> : Ch      H[F] S" 2@  OVER  AND  SWAP  INVERT" >cb  H[G]  S" @  AND  XOR" >cb ; PRIVATE ( -- n ) 
> 
> : Maj     H[C] S" DUP >R  CELL+  2@  OVER  AND  SWAP  R@ @  AND XOR  R> 2@ AND  XOR" >cb ; PRIVATE ( -- n ) 
> 
> : T1x     Ch  H[E] S" @  DUP >R  #14 ror  R@  #18 ror  XOR  R>  #41 ror  XOR  +" >cb  H[H] S" @ +" >cb ; PRIVATE ( -- n ) 
> 
> : T2      Maj H[A] S" @  DUP >R  #28 ror  R@  #34 ror  XOR  R>  #39 ror  XOR  +" >cb ; PRIVATE ( -- n ) 
> 
> : subrnd  S" DUP"  >cb  H[D] S" +!" >cb  T2 S" +" >cb  H[A] S" CELL+ !" >cb  CELL +TO =H ; PRIVATE 
> 
> : rndi[   Wi@  T1x  S" +"  >cb ; PRIVATE 
> 
> : rndn[   Wi   T1x  S" +"  >cb ; PRIVATE   
> 
> : ]rndi   S" +" >cb subrnd S" CELL+" >cb ; PRIVATE 
> 
> : ]rndn   S" +" >cb subrnd ; PRIVATE 
> 
> : ~rndi	  S" DROP" >cb ; PRIVATE 	
> 
> : ~rndn	  S" SHAsh SHAval" >cb  H[H]  S" UpDateHash" >cb ; PRIVATE 
> 
> 
> 
> InitMake
> 
>   rndi[ K0  ]rndi  rndi[ K1  ]rndi  rndi[ K2  ]rndi  rndi[ K3  ]rndi
> 
>   rndi[ K4  ]rndi  rndi[ K5  ]rndi  rndi[ K6  ]rndi  rndi[ K7  ]rndi
> 
>   rndi[ K8  ]rndi  rndi[ K9  ]rndi  rndi[ K10 ]rndi  rndi[ K11 ]rndi
> 
>   rndi[ K12 ]rndi  rndi[ K13 ]rndi  rndi[ K14 ]rndi  rndi[ K15 ]rndi ~rndi
> 
>   rndn[ K16 ]rndn  rndn[ K17 ]rndn  rndn[ K18 ]rndn  rndn[ K19 ]rndn 
> 
>   rndn[ K20 ]rndn  rndn[ K21 ]rndn  rndn[ K22 ]rndn  rndn[ K23 ]rndn  
> 
>   rndn[ K24 ]rndn  rndn[ K25 ]rndn  rndn[ K26 ]rndn  rndn[ K27 ]rndn  
> 
>   rndn[ K28 ]rndn  rndn[ K29 ]rndn  rndn[ K30 ]rndn  rndn[ K31 ]rndn  
> 
>   rndn[ K32 ]rndn  rndn[ K33 ]rndn  rndn[ K34 ]rndn  rndn[ K35 ]rndn  
> 
>   rndn[ K36 ]rndn  rndn[ K37 ]rndn  rndn[ K38 ]rndn  rndn[ K39 ]rndn  
> 
>   rndn[ K40 ]rndn  rndn[ K41 ]rndn  rndn[ K42 ]rndn  rndn[ K43 ]rndn  
> 
>   rndn[ K44 ]rndn  rndn[ K45 ]rndn  rndn[ K46 ]rndn  rndn[ K47 ]rndn  
> 
>   rndn[ K48 ]rndn  rndn[ K49 ]rndn  rndn[ K50 ]rndn  rndn[ K51 ]rndn  
> 
>   rndn[ K52 ]rndn  rndn[ K53 ]rndn  rndn[ K54 ]rndn  rndn[ K55 ]rndn  
> 
>   rndn[ K56 ]rndn  rndn[ K57 ]rndn  rndn[ K58 ]rndn  rndn[ K59 ]rndn  
> 
>   rndn[ K60 ]rndn  rndn[ K61 ]rndn  rndn[ K62 ]rndn  rndn[ K63 ]rndn  
> 
>   rndn[ K64 ]rndn  rndn[ K65 ]rndn  rndn[ K66 ]rndn  rndn[ K67 ]rndn  
> 
>   rndn[ K68 ]rndn  rndn[ K69 ]rndn  rndn[ K70 ]rndn  rndn[ K71 ]rndn  
> 
>   rndn[ K72 ]rndn  rndn[ K73 ]rndn  rndn[ K74 ]rndn  rndn[ K75 ]rndn  
> 
>   rndn[ K76 ]rndn  rndn[ K77 ]rndn  rndn[ K78 ]rndn  rndn[ K79 ]rndn ~rndn 
> 
> EndMake 
> 
> 
> 
> : storelen      D2* D2* D2* ( bytes->bits)  W #112 CHARS + !  W #120 CHARS + ! ; PRIVATE ( lo hi -- ) 
> 
> : setlen  	SHAlen  storelen ; PRIVATE ( -- )
> 
> : cellsreverse	0 ?DO  DUP  @  BSWAP  OVER !  CELL+  LOOP  DROP ; PRIVATE ( addr n -- ) 
> 
> : endian16 	DUP  #16 cellsreverse ; PRIVATE ( addr1 -- addr2 )  
> 
> : endian14 	DUP  #14 cellsreverse ; PRIVATE ( addr1 -- addr2 ) 
> 
> 
> 
> -- Do all 128 byte blocks leaving remainder block
> 
> : hashfullblocks ( addr1 dcount -- addr2 dcount )	      \ dcount is double number: lo hi
> 
> 	SWAP  DUP >R  7 RSHIFT               ( addr1 hi lo* ) \ Store lo on return, lo*=lo/128
> 
> 	OVER CELLSIZE 7 -  LSHIFT OR >R      ( addr1 hi     ) \ Return is now: :R lo lo'
> 
> 	( hi) 7 RSHIFT 
> 
> 	0 ?DO                                       ( addr1 ) \ Do if hi'= hi/128 > 0
> 
> 		0 0 DO  DUP endian16 SHA512 #128 +  
> 
> 	 	  LOOP 			     	    ( addr' ) \ Hash for 2^cellsize full blocks
> 
> 	 LOOP                                	    ( addr' ) \ Hash for hi'*2^cellsize full blocks
> 
> 	R> 0 ?DO  DUP endian16 SHA512 #128 +  LOOP  ( addr' ) \ Hash for lo' count full 128 byte blocks
> 
> 	R> ( lo) #127 AND ; PRIVATE             ( addr2 cnt ) \ Leave address and count for partial block
> 
> 
> 
> : hashfinal ( addr count -- )  		      \ Hash partial and/or last block
> 
> 	DUP >R  W  SWAP  CMOVE                \ Move bytes into block W array
> 
> 	W  R@ +  #128 OVER  C!   ( addr     ) \ Put 80h after last message byte
> 
> 	CHAR+  #111 R@ -         ( addr #   ) \ Compute tentative 0 byte FILL count
> 
> 	R> #111 >                ( addr # ? ) \ Is partial block byte count > 111 ?
> 
> 	   IF   #16 + ERASE            ( -- ) \ If yes, FILL rest of block w/zeroes
> 
> 		W  endian16  SHA512    ( -- ) \ Endian adjust block if required, then hash
> 
> 		W  #112          ( addr 112 ) \ Now setup last block containing bit count
> 
> 	ENDIF                    ( addr #   )
> 
> 	ERASE setlen  W  endian14  SHA512 ;   \ Zero FILL last block, set message bit count
> 
>  	PRIVATE 			      \ Endian adjust, except bit count, then hash
> 
> 
> 
> -- Compute SHA512 from a counted buffer of text
> 
> : SHAbuffer ( addr dcount -- ) SHAinit  2DUP TO SHAlen  hashfullblocks  hashfinal ;
> 
> 
> 
> -- ===============  Hash string display wordset  ===============
> 
> -- Array of digits 0123456789abcdef
> 
> : digit$      ( -- addr )  S" 0123456789abcdef" DROP ; PRIVATE 
> 
> : intdigits   ( -- )  PAD 0! ; PRIVATE 
> 
> : savedigit   ( n -- )  PAD C@ 1+  DUP PAD C!  PAD +  C! ; PRIVATE 
> 
> : bytedigits  ( n -- )  DUP 4 RSHIFT digit$ + C@ savedigit #15 AND digit$ + C@ savedigit ; PRIVATE 
> 
> : celldigits  ( addr -- )  DUP 7 + DO  I C@ bytedigits  -1 +LOOP  BL savedigit ; PRIVATE 
> 
> : SHAstring   ( -- addr u ) intdigits  SHAval 7 CELLS +   8 0 DO  DUP  celldigits  CELL-  LOOP  DROP  PAD COUNT ;
> 
> : HASH.       ( -- ) CR  SHAstring  TYPE SPACE ; -- Display SHA-512 hash value in hex
> 
> : QuoteString ( addr cnt -- ) CR CR &" EMIT TYPE &" EMIT ; PRIVATE 
> 
> 
> 
> -- ====================  File hash wordset  ====================
> 
> 0 VALUE rfileid PRIVATE -- Holds fileid of input file
> 
> : bytes@    ( addr n -- )  rfileid  READ-FILE  2DROP ; PRIVATE
> 
> : block@    ( -- )  W #128 bytes@ ; PRIVATE 
> 
> 
> 
> : getpartial ( cnt  -- W' cnt2 ? )
> 
> 	W 2DUP  SWAP  DUP >R  bytes@               	    ( cnt1 addr1  )
> 
> 	+ #128 OVER C! CHAR+ #111 R@ - R> #111 > ; PRIVATE  ( addr2 cnt2 ? )
> 
> 
> 
> : @SHAfile ( c-addr u -- )
> 
> 	R/O BIN OPEN-FILE  SWAP  TO rfileid  ( ior) ?FILE
> 
> 	SHAinit 
> 
> 	rfileid FILE-SIZE  DROP ( ud )  	       		\ Get bytesize of input file
> 
> 	0. D-                                         		\ Adjust to hash subset of file
> 
> 	CR ." Bytesize: " 2DUP D.                    		\ Display hash size to screen
> 
> 	2DUP  2>R                                     		\ ( lo  hi ) Save file byte cnt on RETURN
> 
> 	OVER  7 RSHIFT OVER					\ ( lo  hi lo* hi )
> 
> 	CELLSIZE 7 -  LSHIFT OR  SWAP  7 RSHIFT       		\ ( lo  lo' hi') lo' hi' now full block count
> 
> 	0 ?DO  0 0 DO  block@  W endian16 SHA512  LOOP LOOP  	\ Hash hi*2^cellsize full blocks
> 
> 	0 ?DO  block@  W endian16 SHA512  LOOP               	\ Hash lo count full 128 byte blocks
> 
> 	( lo) #127 AND ( rembytes) getpartial ( addr cnt ? ) 	\ Read remaining bytes
> 
> 	  IF  #16 + ERASE  W endian16 SHA512  W #112  ENDIF    	\ Do if rembytes > 111
> 
> 	ERASE  2R> storelen  W endian14  SHA512   	    	\ Do last block
> 
> 	CR  ." SHA-512 : "  CR SHAstring TYPE CR      		\ Show SHA-512 hash for file
> 
> 	rfileid CLOSE-FILE ?FILE ; 	            	   	\ Close the input file
> 
> 
> 
> NESTING @ 1 = 
> 
>   [IF]
> 
> 
> 
> -- ====================  SHA-512 Test Suite  ================================================================================================================
> 
> -- Load W array with data on stack
> 
> : WLoad	W #15 CELLS +  ( d0..d15 W[15] )  #16 0 DO  TUCK ! CELL-  LOOP DROP ; PRIVATE ( d0..d15 -- ) 
> 
> 
> 
> -- ----------------------------------------------------------------------------------------------------------------------------------------------------------
> 
> -- EXAMPLE 0: from Wikipedia
> 
> : EX0	S" " 2DUP QuoteString  
> 
> 	U>D ( addr dcount) SHAbuffer HASH.  
> 
> 	CR S" cf83e1357eefb8bd f1542850d66d8007 d620e4050b5715dc 83f4a921d36ce9ce 47d0d13c5d85f2b0 ff8318d2877eec2f 63b931bd47417a81 a538327af927da3e" TYPE ;
> 
> -- ----------------------------------------------------------------------------------------------------------------------------------------------------------
> 
> -- EXAMPLE 1: from FIPS PUB
> 
> : EX1	S" abc" 2DUP QuoteString 
> 
> 	U>D ( addr dcount) SHAbuffer HASH.  
> 
> 	CR S" ddaf35a193617aba cc417349ae204131 12e6fa4e89a97ea2 0a9eeee64b55d39a 2192992a274fc1a8 36ba3c23a3feebbd 454d4423643ce80e 2a9ac94fa54ca49f" TYPE ;
> 
> -- ----------------------------------------------------------------------------------------------------------------------------------------------------------
> 
> -- EXAMPLE 2: from FIPS PUB
> 
> : EX2a	S" abcdefghbcdefghicdefghijdefghijkefghijklfghijklmghijklmnhijklmnoijklmnopjklmnopqklmnopqrlmnopqrsmnopqrstnopqrstu" ;
> 
> : EX2	EX2a 2DUP QuoteString 
> 
> 	U>D ( addr dcount)  SHAbuffer HASH. 
> 
> 	CR S" 8e959b75dae313da 8cf4f72814fc143f 8f7779c6eb9f7fa1 7299aeadb6889018 501d289e4900f7e4 331b99dec4b5433a c7d329eeb6dd2654 5e96e55b874be909" TYPE ;
> 
> 
> 
> : EX5	S" The quick brown fox jumps over the lazy dog" 2DUP QuoteString 
> 
> 	U>D ( addr dcount)  SHAbuffer HASH. 
> 
> 	CR S" 07e547d9586f6a73 f73fbac0435ed769 51218fb7d0c8d788 a309d785436bbb64 2e93a252a954f239 12547d1e8a3b5ed6 e1bfd7097821233f a0538f3db854fee6" TYPE ;
> 
> -- ----------------------------------------------------------------------------------------------------------------------------------------------------------
> 
> -- EXAMPLE 3:
> 
> -- Message: 1 million copies of 'a' (61h), (8 million bits)
> 
> -- Hash = ?
> 
> -- Load block of all 'a's (61h), must hash 15,625 times
> 
> : EX3a	W #128 'a'  FILL  ;
> 
> -- Last message block: 1st bit a '1', bit-count = 16 million
> 
> : EX3b	$8000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 #16000000 WLoad ;
> 
> -- Do hash for message of 2 million copies of ASCII 'a' (61h)
> 
> : EX3	SHAinit  EX3a  #15625 0 DO  W SHA512  LOOP  EX3b  W SHA512 HASH. ;
> 
> -- -------------------------------------------------------------
> 
> -- EXAMPLE 4:
> 
> -- Message: 400,000 SPACES 'BL' (20h), (3,200,000 bits)
> 
> -- Hash = ?
> 
> -- Load block of all "BL' (20h), hash 156 full blocks + 16 bytes
> 
> : EX4a	W #128  BLANK ;
> 
> -- Last message block: 1st bit a '1', bit-count = 3,200,000
> 
> : EX4b	$8000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0  0 #3200000 WLoad ;
> 
> -- Do hash for message of 400,000 SPACES 'BL' (20h)
> 
> : EX4	SHAinit  EX4a  #3125 0 DO  W SHA512  LOOP  EX4b  W SHA512 HASH. ;
> 
> 
> 
> : SHATest ( -- )
> 
> 	CR ." SHA-512 test suite:"
> 
> 	EX0  EX1  EX2  EX5
> 
> 	CR CR S" 2 million copies of ASCII 'a' (61h)" TYPE EX3
> 
> 	CR CR S" 400,000 copies of ASCII BL (20h)"    TYPE EX4 CR ;
> 
> 
> 
> : [EX1]	S" abc" U>D ( addr dcount) SHAbuffer ; PRIVATE 
> 
> : [EX2]	EX2a    U>D  SHAbuffer  ; PRIVATE  
> 
> : [EX3] SHAinit EX3a  #15625 0 DO W SHA512 LOOP EX3b W SHA512 ; PRIVATE 
> 
> : test1 CR ." SHA-512 test for EX1, 1000 loops: "  TIMER-RESET  #1000 0 DO  [EX1]  LOOP  .ELAPSED ;
> 
> : test2 CR ." SHA-512 test for EX2, 1000 loops: "  TIMER-RESET  #1000 0 DO  [EX2]  LOOP  .ELAPSED ;
> 
> : test3 CR ." SHA-512 test for EX3, 1000 loops: "  TIMER-RESET  #1000 0 DO  [EX3]  LOOP  .ELAPSED ;
> 
> : SHAbm ( -- ) test1 test2 test3 ;
> 
> 
> 
> :ABOUT	CR ." Try: SHAtest                     -- test SHA-512 (roughly)"
> 
> 	CR ."      SHAbm                       -- speed benchmark" 
> 
> 	CR ." ( c-addr ud -- ) SHAbuffer HASH. -- print hash value of string" 
> 
> 	CR ." ( c-addr u -- ) @SHAfile         -- print hash value of file." ;
> 
> 
> 
> [ELSE]
> 
> 
> 
> :ABOUT	CR ." Try: ( c-addr ud -- ) SHAbuffer HASH. -- print hash value of string" 
> 
> 	CR ."      ( c-addr u -- )  @SHAfile        -- print hash value of file." ;
> 
> 
> 
> [THEN]
> 
> 
> 
> NESTING @ 1 = [IF]	.ABOUT -sha-512 CR  [THEN]
> 
> 			DEPRIVE
> 
> 
> 
>                               (* End of Source *)

You may get a 3-5% speed increase using these better versions of 'UpDateHash.  Use the one the works the best. With VFX the first version takes 85 bytes and the 2nd 77 bytes, but the first one is faster. 

If you want even more speed unroll the loop and do it explicitly.

------------------------------  New UpDateHash -----------------------------
: UpDateHash ( -)  \ Update hash values: SHAsh(i) = SHAval(i) = SHAval(i-1) + H[H](i-1)
  H[H] 8 0 DO  DUP @  I CELLS DUP >R SHAval + TUCK  +! \ Compute updated hash subvalue
               @  SHAsh R> +  !  CELL+  LOOP  DROP     \ Store updated hash subvalue
  SHAsh  TO  H[H]                                      \ Point H[H] to updated SHAsh(0)=h
;

: UpDateHash ( -)  \ Update hash values: SHAsh(i) = SHAval(i) = SHAval(i-1) + H[H](i-1)
  8 0 DO H[H] I CELLS DUP >R + @ SHAval R@ + TUCK +!   \ Compute updated hash subvalue
               @  SHAsh R> +  !  LOOP                  \ Store updated hash subvalue
  SHAsh  TO  H[H]                                      \ Point H[H] to updated SHAsh(0)=h
;
----------------------------------------------------------------------------

jz

[toc] | [prev] | [next] | [standalone]


#18042

Fromjzakiya@gmail.com
Date2012-12-16 22:35 -0800
Message-ID<c9c0448b-2eb3-4d2a-b5d4-b4aa8e02da4d@googlegroups.com>
In reply to#18040
Found a website which lists SHA512s of iso you can check against.

http://ftp.netbsd.org/pub/NetBSD/iso/5.1.2/SHA512

jz

[toc] | [prev] | [next] | [standalone]


#18043

Frommhx@iae.nl (Marcel Hendrix)
Date2012-12-17 10:14 +0200
Message-ID<90899318918435@frunobulax.edu>
In reply to#18042
jzakiya@gmail.com writes Re: SHA-512

> Found a website which lists SHA512s of iso you can check against.
> http://ftp.netbsd.org/pub/NetBSD/iso/5.1.2/SHA512

SHA512 (sourcecd-5.1.2.iso) = 
c809baa170082aadea5255c4239f61d2ebc5042ad9bb67eb8bb134d7ab7327d0
a6218b31bff56a3709bae6b99fcd73db157eca7b83f6695c7801b570b216a82d

FORTH> S" C:\Users\marcel\Desktop\sourcecd-5.1.2.iso" @SHAfile
Bytesize: 324468736
SHA-512 :
c809baa170082aad ea5255c4239f61d2 ebc5042ad9bb67eb 8bb134d7ab7327d0 
a6218b31bff56a37 09bae6b99fcd73db 157eca7b83f6695c 7801b570b216a82d

So, it apperas to work correctly for up to 300MB. I think (looking at 
the code) that files up to 18,446,744,073,709,551,616 bytes will be OK. 
An actual test with larger values will have to wait until the release 
of Windows 9,

-marcel

[toc] | [prev] | [next] | [standalone]


#18044

Frommhx@iae.nl (Marcel Hendrix)
Date2012-12-17 10:51 +0200
Message-ID<91849318918435@frunobulax.edu>
In reply to#18040
jzakiya@gmail.com writes Re: SHA-512

> On Sunday, December 16, 2012 9:00:07 AM UTC-5, Marcel Hendrix wrote:
> mhx@iae.nl (Marcel Hendrix) writes Re: SHA-512
>> 
>> > jzakiya@gmail.com writes Re: SHA-512
>> [..]
>> > I have looked at my own (PICK-less) idea of SHA-512 again, and it couldn't work. 
>> > However, I found a way to improve your algorithm (literalize H[x]). ATM  test3  
>> > runs in 11.048 seconds on my 2.66 GHz i7 system. SHAfile needs 2.741 seconds 
>> > to process 385,241,618 bytes (64.9.. 50 MB/s/GHz). Hopefully your fix of 15 PICK 
>> > will create additional room for speedup.
>> [..]

> You may get a 3-5% speed increase using these better versions of 'UpDateHash.  
> Use the one the works the best. With VFX the first version takes 85 bytes and the 
> 2nd 77 bytes, but the first one is faster. 

> If you want even more speed unroll the loop and do it explicitly.

The speed variation between runs is much larger than the improvement. 
If there is any, it is probably around 1%. The code is indeed much shorter.

-marcel

-- ------------------------------------------------------------------
: UpDateHash ( a b c -- )  \ Update hash values and load arrays with new values
	8 0 DO  DUP >R @  SWAP  DUP >R @  +  DUP        \ Compute updated hash subvalue
	  	R@ !  OVER !  CELL+  R> CELL+  R> CELL+ \ Store updated hash subvalue
	  LOOP  3DROP ;        				\ Clear stack when done

FORTH> ' updatehash idis
$0140BA00  : [trashed]
$0140BA0A  mov           rcx, 8 d#
$0140BA11  xor           rbx, rbx
$0140BA14  call          (DO) offset NEAR
$0140BA1E  nop
$0140BA1F  nop
$0140BA20  pop           rdi
$0140BA21  mov           rax, [rbx] qword
$0140BA24  add           rax, [rdi] qword
$0140BA27  mov           [rdi] qword, rax
$0140BA2A  pop           rdx
$0140BA2B  mov           [rdx] qword, rax
$0140BA2E  lea           rax, [rdx 8 +] qword
$0140BA32  push          rax
$0140BA33  lea           rdi, [rdi 8 +] qword
$0140BA37  push          rdi
$0140BA38  lea           rbx, [rbx 8 +] qword
$0140BA3C  add           [rbp 0 +] qword, 1 b#
$0140BA41  add           [rbp 8 +] qword, 1 b#
$0140BA46  jno           $0140BA20 offset NEAR
$0140BA4C  add           rbp, #24 b#
$0140BA50  pop           rdi
$0140BA51  pop           rax
$0140BA52  ;

: UpDateHash ( h[h] -)  \ Update hash values: SHAsh(i) = SHAval(i) = SHAval(i-1) + H[H](i-1)
  8 0 DO  DUP @  I CELLS DUP >R SHAval + TUCK  +! \ Compute updated hash subvalue
              @  SHAsh R> +  !  CELL+  LOOP  DROP \ Store updated hash subvalue
;

FORTH> ' updatehash idis
$0140BA00  : [trashed]
$0140BA0A  mov           rcx, 8 d#
$0140BA11  xor           rbx, rbx
$0140BA14  call          (DO) offset NEAR
$0140BA1E  nop
$0140BA1F  nop
$0140BA20  mov           rdi, [rbp 0 +] qword
$0140BA24  mov           rax, [rbx] qword
$0140BA27  add           [rdi*8 $01401720 +] qword, rax
$0140BA2F  mov           rax, [rdi*8 $01401720 +] qword
$0140BA37  mov           [rdi*8 $01401770 +] qword, rax
$0140BA3F  lea           rbx, [rbx 8 +] qword
$0140BA43  add           [rbp 0 +] qword, 1 b#
$0140BA48  add           [rbp 8 +] qword, 1 b#
$0140BA4D  jno           $0140BA20 offset NEAR
$0140BA53  add           rbp, #24 b#
$0140BA57  ;

[toc] | [prev] | [next] | [standalone]


#18047

Fromjzakiya@gmail.com
Date2012-12-17 09:13 -0800
Message-ID<bc497962-8b67-4b1a-a8e3-ef1311baf16b@googlegroups.com>
In reply to#18044
On Monday, December 17, 2012 3:51:12 AM UTC-5, Marcel Hendrix wrote:
> jzakiya@gmail.com writes Re: SHA-512
> 
> 
> 
> > On Sunday, December 16, 2012 9:00:07 AM UTC-5, Marcel Hendrix wrote:
> 
> > mhx@iae.nl (Marcel Hendrix) writes Re: SHA-512
> 
> >> 
> 
> >> > jzakiya@gmail.com writes Re: SHA-512
> 
> >> [..]
> 
> >> > I have looked at my own (PICK-less) idea of SHA-512 again, and it couldn't work. 
> 
> >> > However, I found a way to improve your algorithm (literalize H[x]). ATM  test3  
> 
> >> > runs in 11.048 seconds on my 2.66 GHz i7 system. SHAfile needs 2.741 seconds 
> 
> >> > to process 385,241,618 bytes (64.9.. 50 MB/s/GHz). Hopefully your fix of 15 PICK 
> 
> >> > will create additional room for speedup.
> 
> >> [..]
> 
> 
> 
> > You may get a 3-5% speed increase using these better versions of 'UpDateHash.  
> 
> > Use the one the works the best. With VFX the first version takes 85 bytes and the 
> 
> > 2nd 77 bytes, but the first one is faster. 
> 
> 
> 
> > If you want even more speed unroll the loop and do it explicitly.
> 
> 
> 
> The speed variation between runs is much larger than the improvement. 
> 
> If there is any, it is probably around 1%. The code is indeed much shorter.
> 
> 
> 
> -marcel
> 
> 
> 
> -- ------------------------------------------------------------------
> 
> : UpDateHash ( a b c -- )  \ Update hash values and load arrays with new values
> 
> 	8 0 DO  DUP >R @  SWAP  DUP >R @  +  DUP        \ Compute updated hash subvalue
> 
> 	  	R@ !  OVER !  CELL+  R> CELL+  R> CELL+ \ Store updated hash subvalue
> 
> 	  LOOP  3DROP ;        				\ Clear stack when done
> 
> 
> 
> FORTH> ' updatehash idis
> 
> $0140BA00  : [trashed]
> 
> $0140BA0A  mov           rcx, 8 d#
> 
> $0140BA11  xor           rbx, rbx
> 
> $0140BA14  call          (DO) offset NEAR
> 
> $0140BA1E  nop
> 
> $0140BA1F  nop
> 
> $0140BA20  pop           rdi
> 
> $0140BA21  mov           rax, [rbx] qword
> 
> $0140BA24  add           rax, [rdi] qword
> 
> $0140BA27  mov           [rdi] qword, rax
> 
> $0140BA2A  pop           rdx
> 
> $0140BA2B  mov           [rdx] qword, rax
> 
> $0140BA2E  lea           rax, [rdx 8 +] qword
> 
> $0140BA32  push          rax
> 
> $0140BA33  lea           rdi, [rdi 8 +] qword
> 
> $0140BA37  push          rdi
> 
> $0140BA38  lea           rbx, [rbx 8 +] qword
> 
> $0140BA3C  add           [rbp 0 +] qword, 1 b#
> 
> $0140BA41  add           [rbp 8 +] qword, 1 b#
> 
> $0140BA46  jno           $0140BA20 offset NEAR
> 
> $0140BA4C  add           rbp, #24 b#
> 
> $0140BA50  pop           rdi
> 
> $0140BA51  pop           rax
> 
> $0140BA52  ;
> 
> 
> 
> : UpDateHash ( h[h] -)  \ Update hash values: SHAsh(i) = SHAval(i) = SHAval(i-1) + H[H](i-1)
> 
>   8 0 DO  DUP @  I CELLS DUP >R SHAval + TUCK  +! \ Compute updated hash subvalue
> 
>               @  SHAsh R> +  !  CELL+  LOOP  DROP \ Store updated hash subvalue
> 
> ;
> 
> 
> 
> FORTH> ' updatehash idis
> 
> $0140BA00  : [trashed]
> 
> $0140BA0A  mov           rcx, 8 d#
> 
> $0140BA11  xor           rbx, rbx
> 
> $0140BA14  call          (DO) offset NEAR
> 
> $0140BA1E  nop
> 
> $0140BA1F  nop
> 
> $0140BA20  mov           rdi, [rbp 0 +] qword
> 
> $0140BA24  mov           rax, [rbx] qword
> 
> $0140BA27  add           [rdi*8 $01401720 +] qword, rax
> 
> $0140BA2F  mov           rax, [rdi*8 $01401720 +] qword
> 
> $0140BA37  mov           [rdi*8 $01401770 +] qword, rax
> 
> $0140BA3F  lea           rbx, [rbx 8 +] qword
> 
> $0140BA43  add           [rbp 0 +] qword, 1 b#
> 
> $0140BA48  add           [rbp 8 +] qword, 1 b#
> 
> $0140BA4D  jno           $0140BA20 offset NEAR
> 
> $0140BA53  add           rbp, #24 b#
> 
> $0140BA57  ;

Unrolling UpDateHash on VFX reduced time from 16.405 secs to 16.030.

: UpDateHash ( -)  \ Update hash values: SHAsh(i) = SHAval(i) = SHAval(i-1) + H[H](i-1)
  H[H] DUP @ [ SHAval 0 CELLS+ ]L TUCK +! @ [ SHAsh 0 CELLS+ ]L ! CELL+
       DUP @ [ SHAval 1 CELLS+ ]L TUCK +! @ [ SHAsh 1 CELLS+ ]L ! CELL+
       DUP @ [ SHAval 2 CELLS+ ]L TUCK +! @ [ SHAsh 2 CELLS+ ]L ! CELL+
       DUP @ [ SHAval 3 CELLS+ ]L TUCK +! @ [ SHAsh 3 CELLS+ ]L ! CELL+
       DUP @ [ SHAval 4 CELLS+ ]L TUCK +! @ [ SHAsh 4 CELLS+ ]L ! CELL+
       DUP @ [ SHAval 5 CELLS+ ]L TUCK +! @ [ SHAsh 5 CELLS+ ]L ! CELL+
       DUP @ [ SHAval 6 CELLS+ ]L TUCK +! @ [ SHAsh 6 CELLS+ ]L ! CELL+
           @ [ SHAval 7 CELLS+ ]L TUCK +! @ [ SHAsh 7 CELLS+ ]L !
   SHAsh  TO  H[H]
;

jz

[toc] | [prev] | [next] | [standalone]


#18053

Frommhx@iae.nl (Marcel Hendrix)
Date2012-12-17 22:24 +0200
Message-ID<65111318918435@frunobulax.edu>
In reply to#18047
jzakiya@gmail.com writes Re: SHA-512
[..]
> Unrolling UpDateHash on VFX reduced time from 16.405 secs to 16.030.
[..]

I can't measure the improvement that gives, but I added it anyway.

Today I backported SHA-512 to SHA-256. It took only a few minutes.

Unfortunately, SHA-256 runs only on iForth32 as ROR for iForth64 
rotates 64 bits, which is not suitable. Fixing this to let SHA-256 
run on a 64bit Forth is rather inefficient.

I see that SHA256's test3 only processes 1 million characters. 
The test3 for SHA512 is doing 2 million characters. Take care
when comparing times for SHA-256 with SHA-512!

SHA256, iForth32, i7 920, 2.77 GHz, test3: 8.621 seconds or 41.58 MB/s/GHz.

It seems VFX is doing rather poorly on SHA256 if it needs 16.030 seconds?

-marcel

[toc] | [prev] | [next] | [standalone]


#18836

Fromjzakiya@gmail.com
Date2013-01-15 20:38 -0800
Message-ID<c0fd0525-52e8-4a2f-b0ec-918fcf65cd6f@googlegroups.com>
In reply to#18053
On Monday, December 17, 2012 3:24:38 PM UTC-5, Marcel Hendrix wrote:
> jzakiya@gmail.com writes Re: SHA-512
> 
> [..]
> 
> > Unrolling UpDateHash on VFX reduced time from 16.405 secs to 16.030.
> 
> [..]
> 
> 
> 
> I can't measure the improvement that gives, but I added it anyway.
> 
> 
> 
> Today I backported SHA-512 to SHA-256. It took only a few minutes.
> 
> 
> 
> Unfortunately, SHA-256 runs only on iForth32 as ROR for iForth64 
> 
> rotates 64 bits, which is not suitable. Fixing this to let SHA-256 
> 
> run on a 64bit Forth is rather inefficient.
> 
> 
> 
> I see that SHA256's test3 only processes 1 million characters. 
> 
> The test3 for SHA512 is doing 2 million characters. Take care
> 
> when comparing times for SHA-256 with SHA-512!
> 
> 
> 
> SHA256, iForth32, i7 920, 2.77 GHz, test3: 8.621 seconds or 41.58 MB/s/GHz.
> 
> 
> 
> It seems VFX is doing rather poorly on SHA256 if it needs 16.030 seconds?
> 
> 
> 
> -marcel

Hi Marcel,

OK, I went on a coding rampage.
I implemented all four 1024-bit block Secure Hash Algorithms
SHA-384, SHA-512, SHA-512/256, and SHA-512/224 in one file,
and you just select the one you want to use when you load it.

Through **extensive** test with VFX with SHA-256 and SHA-224
the phrasings of the hashing words in the code optimize performance
in VFX (32-bit), but under IForth different phrasing may produce
better performance.

The versions I posted implemented the rounds within loops (versus
unrolling the round loops because it gave either better (VFX) or no
worse performance than the unrolled loop versions. Your mileage may vary.

The file, SHA-384||512xxx.F is in my 4shared.com repository here:
www.4shared.com/dir/TcMrUvTB/sharing.htm

Or at http://gist.gitgub.com/jzakiya

or directly at
https://gist.github.com/4544478

[toc] | [prev] | [next] | [standalone]


#18837

Fromjzakiya@gmail.com
Date2013-01-15 20:54 -0800
Message-ID<bc280052-b791-4b44-999f-c712616b30d4@googlegroups.com>
In reply to#18053
On Monday, December 17, 2012 3:24:38 PM UTC-5, Marcel Hendrix wrote:
> jzakiya@gmail.com writes Re: SHA-512
> 
> [..]
> 
> > Unrolling UpDateHash on VFX reduced time from 16.405 secs to 16.030.
> 
> [..]
> 
> 
> 
> I can't measure the improvement that gives, but I added it anyway.
> 
> 
> 
> Today I backported SHA-512 to SHA-256. It took only a few minutes.
> 
> 
> 
> Unfortunately, SHA-256 runs only on iForth32 as ROR for iForth64 
> 
> rotates 64 bits, which is not suitable. Fixing this to let SHA-256 
> 
> run on a 64bit Forth is rather inefficient.
> 
> 
> 
> I see that SHA256's test3 only processes 1 million characters. 
> 
> The test3 for SHA512 is doing 2 million characters. Take care
> 
> when comparing times for SHA-256 with SHA-512!
> 
> 
> 
> SHA256, iForth32, i7 920, 2.77 GHz, test3: 8.621 seconds or 41.58 MB/s/GHz.
> 
> 
> 
> It seems VFX is doing rather poorly on SHA256 if it needs 16.030 seconds?
> 
> 
> 
> -marcel

Hi Marcel,

I went on a coding rampage since my last post on this.

I implemented all four 1024-bit block Secure Hash Algorithms
SHA-384, SHA-512, SHA-512/256, and SHA-512/224 in one file,
that you select the one you want to use when you load it.

Through **extensive** test with VFX using SHA-256 and SHA-224,
the phrasings of the hashing words optimize performance in VFX (32-bit),
but under IForth different phrasing may produce better performance.

The versions I posted perform the rounds within loops (versus
unrolling the round loops) because it gave either better (VFX) or no
worse performance than the unrolled loop versions. Your mileage may vary.

The file SHA-384||512xxx.F is in my 4shared.com repository here:

https://www.4shared.com/folder/4o-httdZ/Forth.html

or thru here:      https://gist.github.com/jzakiya

or directly here:  https://gist.github.com/4544478

Jabari

[toc] | [prev] | [next] | [standalone]


#17999

FromHowerd <howerdo@yahoo.co.uk>
Date2012-12-13 13:06 -0800
Message-ID<740e9412-edbb-4452-a157-ff91a094b72b@googlegroups.com>
In reply to#17996
On Thursday, December 13, 2012 3:30:07 PM UTC+1, jza...@gmail.com wrote:
> On Monday, December 10, 2012 4:17:24 PM UTC-5, Marcel Hendrix wrote:
> 
> > jzaxxx@gmail.com writes Re: SHA-512
> 
> > 
> 
> > 
> 
> > 
> 
> > > On Monday, December 10, 2012 6:53:44 AM UTC-5, m.a.m....@tue.nl wrote:
> 
> > 
> 
> > >> On Monday, December 10, 2012 2:29:04 AM UTC+1, jza...@gmail.com wrote:
> 
> > 
> 
> > >>
> 
> > 
> 
> > >> [..]
> 
> > 
> 
> > >>
> 
> > 
> 
> > >> [ Egg on my face, should not have touched the DECIMAL/HEX issues yet. ]
> 
> > 
> 
> > >
> 
> > 
> 
> > >> Yes, the code now runs and gives the correct results for S" abc".
> 
> > 
> 
> > >>
> 
> > 
> 
> > >> This would be the first working SHA-512 in 64-bit Forth!
> 
> > 
> 
> > 
> 
> > 
> 
> > > Hey GREAT! :-)
> 
> > 
> 
> > 
> 
> > 
> 
> > > Be sure to do the longer example also to make sure there are no 
> 
> > 
> 
> > > other problems (which I don't forsee).
> 
> > 
> 
> > 
> 
> > 
> 
> > Unfortunately, the longer string does not hash properly. To avoid 
> 
> > 
> 
> > more egg on my face, I downloaded your latest code from 4share. 
> 
> > 
> 
> > To make it work in iForth64 I needed the patch that I have appended
> 
> > 
> 
> > (Enable PLACE and rewrite timer-start and ms?, typos in SHA256 and 
> 
> > 
> 
> > SHAinit, extra stack item in EX2a dropped.)
> 
> > 
> 
> > 
> 
> > 
> 
> > [..]
> 
> > 
> 
> > 
> 
> > 
> 
> > > I ALWAYS start from the algorithm description to do Forth versions, even if
> 
> > 
> 
> > > I look at other existing language implementations, because I think so much
> 
> > 
> 
> > > differently when I create in Forth.
> 
> > 
> 
> > 
> 
> > 
> 
> > Ok. Please forget about macro's, ]L and EVALUATE for the time being. Also,
> 
> > 
> 
> > the 15 PICKs you may get rid off without any ill effects :-)
> 
> > 
> 
> > 
> 
> > 
> 
> > Here are the current results with your patched code:
> 
> > 
> 
> > 
> 
> > 
> 
> > FORTH> in
> 
> > 
> 
> > Redefining split-at-char
> 
> > 
> 
> > Redefining MACRO
> 
> > 
> 
> > Redefining MS?  ok
> 
> > 
> 
> > FORTH> SHAtest
> 
> > 
> 
> > SHA-512 test suite:
> 
> > 
> 
> > cf83e1357eefb8bdf1542850d66d8007d620e4050b5715dc83f4a921d36ce9ce47d0d13c5d85f2b0ff8318d2877eec2f63b931bd47417a81a538327af927da3e ""
> 
> > 
> 
> > ddaf35a193617abacc417349ae20413112e6fa4e89a97ea20a9eeee64b55d39a2192992a274fc1a836ba3c23a3feebbd454d4423643ce80e2a9ac94fa54ca49f "abc"
> 
> > 
> 
> > 8cc1fec83f21176a675306d9f80b26597469eff277ab9d0f03a725b15d57d33a2e3aadea4727cf632a3e5ec3c744bcc0eb2183c41c717cc4a84fa6dbf8244243 "abcdefghbcdefghicdefghijdefghijkefghijklfghijklmghijklmnhijklmnoijklmnopjklmnopqklmnopqrlmnopqrsmnopqrstnopqrstu"
> 
> > 
> 
> > 9bc68759247e3332bec1c79d128d28a8931d0c9f96c8aa975731b563475fdddddf7f873c25086908effe270e23c5a01e5dfb3289bf5d091d8fb454b1bcf98dda 2 million copies of ASCII 'a' (61h)
> 
> > 
> 
> > 19ef4876e03c93476e00b486c62108d4d7136f6d4782817b449bacf44d0963e03f5ab474cb6822c2963a979e8e15298cd60f22841c51f6161620c0e031c65f4a 400,000 copies of ASCII BL (20h)
> 
> > 
> 
> >  ok
> 
> > 
> 
> > 
> 
> > 
> 
> > -marcel
> 
> > 
> 
> > 
> 
> > 
> 
> > -- ----------
> 
> > 
> 
> > Compare: (<)C:\Users\marcel\Desktop\SHA-512.frt
> 
> > 
> 
> >    with: (>)C:\Users\marcel\Desktop\SHA-512VFX.F
> 
> > 
> 
> > 
> 
> > 
> 
> > 1d1
> 
> > 
> 
> > < ANEW -sha-512
> 
> > 
> 
> > 19c18,19
> 
> > 
> 
> > < : PLACE  ( caddr n addr -)  2DUP  C!  CHAR+  SWAP  CHARS  MOVE ;
> 
> > 
> 
> > ---
> 
> > 
> 
> > > \ Not needed for SwiftForth v 2.00.3, needed for Win32Forth V 4.10
> 
> > 
> 
> > > \ : PLACE  ( caddr n addr -)  2DUP  C!  CHAR+  SWAP  CHARS  MOVE ;
> 
> > 
> 
> > 161c161
> 
> > 
> 
> > < : SHA512 ( Wadr - )
> 
> > 
> 
> > ---
> 
> > 
> 
> > > : SHA256  ( Wadr - )
> 
> > 
> 
> > 233c233
> 
> > 
> 
> > <   SHAinit  2DUP  SHAlen 2!  hashfullblocks  hashfinal
> 
> > 
> 
> > ---
> 
> > 
> 
> > >   SHAinit  2DUP  SHAlen 2 !  hashfullblocks  hashfinal
> 
> > 
> 
> > 375,376c375,376
> 
> > 
> 
> > <   EX1  S" abc"   QuoteString
> 
> > 
> 
> > <   EX2  EX2a DROP QuoteString
> 
> > 
> 
> > ---
> 
> > 
> 
> > >   EX1  S" abc"  QuoteString
> 
> > 
> 
> > >   EX2  EX2a     QuoteString
> 
> > 
> 
> > 381,384c381,391
> 
> > 
> 
> > <   VARIABLE  start-ms
> 
> > 
> 
> > < 
> 
> > 
> 
> > < : TIMER-START  ( -- )  ?MS  start-ms  ! ;
> 
> > 
> 
> > < : MS?  ( -- u )  ?MS  start-ms  @  - ( abs ) ;
> 
> > 
> 
> > ---
> 
> > 
> 
> > > \ ===========  VFX Forth specific performance test ===========
> 
> > 
> 
> > > 
> 
> > 
> 
> > >   [undefined] GetTickCount
> 
> > 
> 
> > >   [IF] extern: DWORD PASCAL GetTickCount( void ) [THEN]
> 
> > 
> 
> > > 
> 
> > 
> 
> > >   VARIABLE  start-ms
> 
> > 
> 
> > > 
> 
> > 
> 
> > > : TIMER-START  ( -- )  GetTickCount  start-ms  ! ;
> 
> > 
> 
> > > 
> 
> > 
> 
> > > : MS?  ( -- u )  GetTickCount  start-ms  @  - ( abs ) ;
> 
> > 
> 
> > >
> 
> 
> 
> Marcel, could you run  EX2  by itself, and see what the output is.
> 
> 
> 
> I think it may have to do with my making the count from the string word EX2a return a double number count instead of a single, so I changed it as below to make it consistent with 'abc'
> 
> 
> 
> : EX2a S" abcdefghbcdefghicdefghijdefghijkefghijklfghijklmghijklmnhijklmnoijklmnopjklmnopqklmnopqrlmnopqrsmnopqrstnopqrstu" ;
> 
> : EX2  EX2a U>D ( adr ud)  SHAbuffer  HASH.  
> 
> 
> 
> Also, do you know the words the current SwiftForth uses for timing.
> 
> I'm running SwiftForth i386-Win32 3.4.5 03-Oct-2012 under WINE on Linux and the old cold that uses  ucounter and utimer bombs on this version.
> 
> 
> 
> You may be more pleased with the newer code I'll release on all the SHA-xxx algorithms, which will make it a little more upto date from the code I did circa 2001, after all the issues with SHA512 are fixed.  :-)

Hi Jabari,

Thanks for doing all of this work on the SHA series, and sorry that I thought that you had copied the C versions - it was a false assumption...
Please let us all know when you have both versions polished :-)

Best regards,
Howerd

[toc] | [prev] | [next] | [standalone]


#17998

FromHowerd <howerdo@yahoo.co.uk>
Date2012-12-13 13:03 -0800
Message-ID<95cfefe2-1a05-4ca3-8955-1cdc76e7ebee@googlegroups.com>
In reply to#17967
On Monday, December 10, 2012 10:17:24 PM UTC+1, Marcel Hendrix wrote:
> jzaxxxx@gmail.com writes Re: SHA-512
> 
> 
> 
> > On Monday, December 10, 2012 6:53:44 AM UTC-5, m.a.m....@tue.nl wrote:
> 
> >> On Monday, December 10, 2012 2:29:04 AM UTC+1, jza...@gmail.com wrote:
> 
> >>
> 
> >> [..]
> 
> >>
> 
> >> [ Egg on my face, should not have touched the DECIMAL/HEX issues yet. ]
> 
> >
> 
> >> Yes, the code now runs and gives the correct results for S" abc".
> 
> >>
> 
> >> This would be the first working SHA-512 in 64-bit Forth!
> 
> 
> 
> > Hey GREAT! :-)
> 
> 
> 
> > Be sure to do the longer example also to make sure there are no 
> 
> > other problems (which I don't forsee).
> 
> 
> 
> Unfortunately, the longer string does not hash properly. To avoid 
> 
> more egg on my face, I downloaded your latest code from 4share. 
> 
> To make it work in iForth64 I needed the patch that I have appended
> 
> (Enable PLACE and rewrite timer-start and ms?, typos in SHA256 and 
> 
> SHAinit, extra stack item in EX2a dropped.)
> 
> 
> 
> [..]
> 
> 
> 
> > I ALWAYS start from the algorithm description to do Forth versions, even if
> 
> > I look at other existing language implementations, because I think so much
> 
> > differently when I create in Forth.
> 
> 
> 
> Ok. Please forget about macro's, ]L and EVALUATE for the time being. Also,
> 
> the 15 PICKs you may get rid off without any ill effects :-)
> 
> 
> 
> Here are the current results with your patched code:
> 
> 
> 
> FORTH> in
> 
> Redefining split-at-char
> 
> Redefining MACRO
> 
> Redefining MS?  ok
> 
> FORTH> SHAtest
> 
> SHA-512 test suite:
> 
> cf83e1357eefb8bdf1542850d66d8007d620e4050b5715dc83f4a921d36ce9ce47d0d13c5d85f2b0ff8318d2877eec2f63b931bd47417a81a538327af927da3e ""
> 
> ddaf35a193617abacc417349ae20413112e6fa4e89a97ea20a9eeee64b55d39a2192992a274fc1a836ba3c23a3feebbd454d4423643ce80e2a9ac94fa54ca49f "abc"
> 
> 8cc1fec83f21176a675306d9f80b26597469eff277ab9d0f03a725b15d57d33a2e3aadea4727cf632a3e5ec3c744bcc0eb2183c41c717cc4a84fa6dbf8244243 "abcdefghbcdefghicdefghijdefghijkefghijklfghijklmghijklmnhijklmnoijklmnopjklmnopqklmnopqrlmnopqrsmnopqrstnopqrstu"
> 
> 9bc68759247e3332bec1c79d128d28a8931d0c9f96c8aa975731b563475fdddddf7f873c25086908effe270e23c5a01e5dfb3289bf5d091d8fb454b1bcf98dda 2 million copies of ASCII 'a' (61h)
> 
> 19ef4876e03c93476e00b486c62108d4d7136f6d4782817b449bacf44d0963e03f5ab474cb6822c2963a979e8e15298cd60f22841c51f6161620c0e031c65f4a 400,000 copies of ASCII BL (20h)
> 
>  ok
> 
> 
> 
> -marcel
> 
> 
> 
> -- ----------
> 
> Compare: (<)C:\Users\marcel\Desktop\SHA-512.frt
> 
>    with: (>)C:\Users\marcel\Desktop\SHA-512VFX.F
> 
> 
> 
> 1d1
> 
> < ANEW -sha-512
> 
> 19c18,19
> 
> < : PLACE  ( caddr n addr -)  2DUP  C!  CHAR+  SWAP  CHARS  MOVE ;
> 
> ---
> 
> > \ Not needed for SwiftForth v 2.00.3, needed for Win32Forth V 4.10
> 
> > \ : PLACE  ( caddr n addr -)  2DUP  C!  CHAR+  SWAP  CHARS  MOVE ;
> 
> 161c161
> 
> < : SHA512 ( Wadr - )
> 
> ---
> 
> > : SHA256  ( Wadr - )
> 
> 233c233
> 
> <   SHAinit  2DUP  SHAlen 2!  hashfullblocks  hashfinal
> 
> ---
> 
> >   SHAinit  2DUP  SHAlen 2 !  hashfullblocks  hashfinal
> 
> 375,376c375,376
> 
> <   EX1  S" abc"   QuoteString
> 
> <   EX2  EX2a DROP QuoteString
> 
> ---
> 
> >   EX1  S" abc"  QuoteString
> 
> >   EX2  EX2a     QuoteString
> 
> 381,384c381,391
> 
> <   VARIABLE  start-ms
> 
> < 
> 
> < : TIMER-START  ( -- )  ?MS  start-ms  ! ;
> 
> < : MS?  ( -- u )  ?MS  start-ms  @  - ( abs ) ;
> 
> ---
> 
> > \ ===========  VFX Forth specific performance test ===========
> 
> > 
> 
> >   [undefined] GetTickCount
> 
> >   [IF] extern: DWORD PASCAL GetTickCount( void ) [THEN]
> 
> > 
> 
> >   VARIABLE  start-ms
> 
> > 
> 
> > : TIMER-START  ( -- )  GetTickCount  start-ms  ! ;
> 
> > 
> 
> > : MS?  ( -- u )  GetTickCount  start-ms  @  - ( abs ) ;
> 
> >

Hi Marcel,

> Also, the 15 PICKs you may get rid off without any ill effects :-) 
I'm curious - how do you do this?
It would be great to remove them from the SHA-256 code too :-)

Best regards,
Howerd

[toc] | [prev] | [next] | [standalone]


#17951

Frommhx@iae.nl (Marcel Hendrix)
Date2012-12-08 22:08 +0200
Message-ID<58951395918435@frunobulax.edu>
In reply to#17919
jzakiya@gmail.com writes Re: SHA-512

[..]
> I've corrected the SHA-256 code, and written a SHA-512 version for 64-bit cpus/OSs. 
> Since I don't have a 64-bit forth I haven't run it but I'm pretty 
> sure it works, as the architecture is the same as SHA-256, with just different 
> sized constants and shifts/rotates.

It runs with minor edits (see below). I removed ]L and EVALUATE as it is unnecessary
and slows down the code.

Unfortunately, the results are wrong.
The speed is OK, 2Mbytes*1000/14.417s/2.67GHz = 51.9 MB/s/GHz

FORTH> SHAtest
SHA-512 test suite:
1ca51eba68fd0ca4284ec2b2bd024de70a231b14b914411996269e0f5c61ec7f02a2054b467965def1d526c52c525a1103c8c3aea8cc5e406f60e740bcd05969 ""
3e9e4c5b0b4eacc5f38516a9afb1c79ea5a959679b9bc6e5de04eedbaa41a198b227138994715fa52fdcb64b3689061926496b9b9d197613c0d35eafc269a023 "abc"
914d0fc46d27cabc63a868fba6b8eb0a4996553265854a33344481c3a0e00536b36c2bf0df78eb059fcaadcc25687d7eb95cbaa07591fa9e6f08839f0e7bc626 "abcdefghbcdefghicdefghijdefghijkefghijklfghijklmghijklmnhijklmnoijklmnopjklmnopqklmnopqrlmnopqrsmnopqrstnopqrstu"
7e4256b1fcf2c09b867e71bc8189f631999302f27f8c5c2c27ef6c39e1d3016afd6b1bdfa452847695ca18905429a365fa19d071f30a440b0c548a4d4f99fe49 2 million copies of ASCII 'a' (61h)
f4568c36ddc5a2fa022a5af467ba59b25bde16a098ea3fb83c879104c9473795a18df1e4d407379f0fe7d266dc7cfcb3d205770c35c90e8d174d4c7ed1dc0f37 400,000 copies of ASCII BL (20h)
 ok
FORTH> test1 test2 test3
SHA-512 test for EX1 for 1000 loops in milliseconds is 2
SHA-512 test for EX2 for 1000 loops in milliseconds is 4
SHA-512 test for EX3 for 1000 loops in milliseconds is 14417  ok

-marcel

--
ANEW -sha-512

\ Forth code for Secure Hash Algorithm 512 (SHA-512)
\ NIST spec at:  http://csrc.nist.gov/encryption/tkhash.html
\ For Little/Big Endian byte addressable CPUs, e.g. Intel/Power PC
\ DEPENDENCIES: CORE EXT WORDSET ; COMMON USAGE 3DROP ?DO CELL-
\ Use of this code is free subject to acknowledgment of copyright.
\ Copyright (c) 2012 Jabari Zakiya -- jzakiya@mail.com  12/07/2012

\ ======================= MACRO Wordset ========================
\ MACRO wordset from Wil Baden's Tool Belt series in
\ Forth Dimensions (FD) Vol. 19, No. 2, July/August 1997
\ Original code has been modified to make more efficient
\ MACRO allows insertion of parameters following the macro
\ "\" represents place where parameter is inserted
\ Example:  MACRO  ??  " IF  \  THEN "
\ : FOO .. ?? EXIT .... ;  ?? compiles to -- IF EXIT THEN

\ PLACE and STRING for system if needed
\ Not needed for SwiftForth v 2.00.3, needed for Win32Forth V 4.10
\ : PLACE  ( caddr n addr -)  2DUP  C!  CHAR+  SWAP  CHARS  MOVE ;
: SSTRING ( char "ccc" -) WORD COUNT HERE OVER 1+ CHARS ALLOT PACK DROP ;

\ Versions of /STRING and ANEW if system doesn't have them
\ : /STRING ( a n k - a+k n-k) ( OVER MIN) TUCK - >R CHARS + R> ;
\ : ANEW  >IN @ BL WORD FIND IF EXECUTE ELSE DROP THEN >IN ! MARKER ;

\ : split-at-char  ( a  n  char  -  a  k  a+k  n-k)
\  >R  2DUP  BEGIN  DUP  WHILE  OVER  C@  R@  -
\            WHILE  1 /STRING  REPEAT  THEN
\            R> DROP  TUCK  2>R  -  2R>
\ ;

: DOES>MACRO  \ Compile the macro, including external parameters
  DOES> COUNT  BEGIN [CHAR]  \ split-at-char  2>R  EVALUATE  R@
               WHILE BL WORD COUNT EVALUATE 2R>  1 /STRING REPEAT
               R> DROP   R> DROP
;

\ Macro creation word which allows parameter insertion
: MACRO  CREATE  IMMEDIATE  CHAR  SSTRING  DOES>MACRO  ;

\ ====================  Start SHA-256 Code ====================
  DECIMAL
  64 CONSTANT CELLSIZE          \ CPU bitsize

  2VARIABLE  SHAlen             \ Holds byte length of string < 2^128 bits|2^125 bytes
  CREATE SHAval  8 CELLS ALLOT  \ Holds hash after each block
  CREATE SHAsh  88 CELLS ALLOT  \ Fully extended hash array
  CREATE W      16 CELLS ALLOT  \ Holds message block
  1 W !                         \ For compile time endian testing

  HEX

\ SHA-512 round constants
428a2f98d728ae22  CONSTANT  K0    7137449123ef65cd  CONSTANT  K1
b5c0fbcfec4d3b2f  CONSTANT  K2    e9b5dba58189dbbc  CONSTANT  K3
3956c25bf348b538  CONSTANT  K4    59f111f1b605d019  CONSTANT  K5
923f82a4af194f9b  CONSTANT  K6    ab1c5ed5da6d8118  CONSTANT  K7
d807aa98a3030242  CONSTANT  K8    12835b0145706fbe  CONSTANT  K9
243185be4ee4b28c  CONSTANT  K10   550c7dc3d5ffb4e2  CONSTANT  K11
72be5d74f27b896f  CONSTANT  K12   80deb1fe3b1696b1  CONSTANT  K13
9bdc06a725c71235  CONSTANT  K14   c19bf174cf692694  CONSTANT  K15
e49b69c19ef14ad2  CONSTANT  K16   efbe4786384f25e3  CONSTANT  K17
0fc19dc68b8cd5b5  CONSTANT  K18   240ca1cc77ac9c65  CONSTANT  K19
2de92c6f592b0275  CONSTANT  K20   4a7484aa6ea6e483  CONSTANT  K21
5cb0a9dcbd41fbd4  CONSTANT  K22   76f988da831153b5  CONSTANT  K23
983e5152ee66dfab  CONSTANT  K24   a831c66d2db43210  CONSTANT  K25
b00327c898fb213f  CONSTANT  K26   bf597fc7beef0ee4  CONSTANT  K27
c6e00bf33da88fc2  CONSTANT  K28   d5a79147930aa725  CONSTANT  K29
06ca6351e003826f  CONSTANT  K30   142929670a0e6e70  CONSTANT  K31
27b70a8546d22ffc  CONSTANT  K32   2e1b21385c26c926  CONSTANT  K33
4d2c6dfc5ac42aed  CONSTANT  K34   53380d139d95b3df  CONSTANT  K35
650a73548baf63de  CONSTANT  K36   766a0abb3c77b2a8  CONSTANT  K37
81c2c92e47edaee6  CONSTANT  K38   92722c851482353b  CONSTANT  K39
a2bfe8a14cf10364  CONSTANT  K40   a81a664bbc423001  CONSTANT  K41
c24b8b70d0f89791  CONSTANT  K42   c76c51a30654be30  CONSTANT  K43
d192e819d6ef5218  CONSTANT  K44   d69906245565a910  CONSTANT  K45
f40e35855771202a  CONSTANT  K46   106aa07032bbd1b8  CONSTANT  K47
19a4c116b8d2d0c8  CONSTANT  K48   1e376c085141ab53  CONSTANT  K49
2748774cdf8eeb99  CONSTANT  K50   34b0bcb5e19b48a8  CONSTANT  K51
391c0cb3c5c95a63  CONSTANT  K52   4ed8aa4ae3418acb  CONSTANT  K53
5b9cca4f7763e373  CONSTANT  K54   682e6ff3d6b2b8a3  CONSTANT  K55
748f82ee5defb2fc  CONSTANT  K56   78a5636f43172f60  CONSTANT  K57
84c87814a1f0ab72  CONSTANT  K58   8cc702081a6438ec  CONSTANT  K59
90befffa23631e28  CONSTANT  K60   a4506cebde82bde9  CONSTANT  K61
bef9a3f7b2c67915  CONSTANT  K62   c67178f2e372532b  CONSTANT  K63
ca273eceea26619c  CONSTANT  K64   d186b8c721c0c207  CONSTANT  K65
eada7dd6cde0eb1e  CONSTANT  K66   f57d4f7fee6ed178  CONSTANT  K67
06f067aa72176fba  CONSTANT  K68   0a637dc5a2c898a6  CONSTANT  K69
113f9804bef90dae  CONSTANT  K70   1b710b35131c471b  CONSTANT  K71
28db77f523047d84  CONSTANT  K72   32caab7b40c72493  CONSTANT  K73
3c9ebe0a15c9bebc  CONSTANT  K74   431d67c49c100d4c  CONSTANT  K75
4cc5d4becb3e42b6  CONSTANT  K76   597f299cfc657e2a  CONSTANT  K77
5fcb6fab3ad6faec  CONSTANT  K78   6c44198c4a475817  CONSTANT  K79

  DECIMAL

  0 VALUE H[H]  \ Pointer to addr of hash value H for each round

: H[G]  H[H]  1 CELLS + ; \ Return G adr
: H[F]  H[H]  2 CELLS + ; \ Return F adr
: H[E]  H[H]  3 CELLS + ; \ Return E adr
: H[D]  H[H]  4 CELLS + ; \ Return D adr
: H[C]  H[H]  5 CELLS + ; \ Return C adr
: H[A]  H[H]  7 CELLS + ; \ Return A adr

: SHAinit ( -)  \ Load initial hash values H0 - H7
  [ HEX ] 6a09e667f3bcc908 ( H0)  bb67ae8584caa73b ( H1)
          3c6ef372fe94f82b ( H2)  a54ff53a5f1d36f1 ( H3)
          510e527fade682d1 ( H4)  9b05688c2b3e6c1f ( H5)
          1f83d9abfb41bd6b ( H6)  5be0cd19137e2179 ( H7)
  [ DECIMAL ]
  SHAsh  7 0 DO  TUCK  !  CELL+  LOOP  !   \ Put initial hash in SHAsh array
  SHAsh  SHAval  8 CELLS  CMOVE            \ Put copy in SHAval array
  SHAsh  TO  H[H]                          \ Init pointer to last hash value
;

: UpDateHash ( -)  \ Update hash values and load arrays with new values
  SHAsh  SHAval  H[H]                                  \ Place array addresses on stack
  8 0 DO  DUP >R   @   SWAP  DUP >R  @  +  DUP         \ Compute updated hash subvalue
          R@  !  OVER  !  CELL+  R>  CELL+  R>  CELL+  \ Store updated hash subvalue
  LOOP  3DROP                                          \ Clear stack when done
  SHAsh  TO  H[H]                                      \ Init pointer to last subvalue
;

\ ( - n )  n = (E AND F) XOR (~E AND G)
: Ch  H[F] 2@  OVER  AND  SWAP  INVERT  H[G]  @  AND  XOR ; 

\ ( - n )  n = (A AND B) XOR (A AND C) XOR (B AND C)
: Maj H[C]  DUP >R  CELL+  2@  OVER  AND  SWAP  R@ @  AND XOR  R> 2@ AND  XOR ; 

\ ( - n )  T1x = Ch(e,f,g) + Sig1(e) + h
: T1x  Ch  H[E] @  DUP >R  14 ror  R@  18 ror  XOR  R>  41 ror  XOR  +  H[H] @  +  ;

\ ( - n )  T2 = Maj(a,b,c) + Sig0(a)
: T2  Maj  H[A] @  DUP >R  28 ror  R@  34 ror  XOR  R>  39 ror  XOR  +  ;

\ ( x - n )  n = ROR1(X)   XOR  ROR8(X)  XOR  SHR7(X)
: sig0  ( x - n )  DUP  DUP   1 ROR  SWAP  8 ROR  XOR  SWAP  7 RSHIFT  XOR  ;

\ ( x - n )  n = ROR19(X)  XOR  ROR61(X)  XOR  SHR6(X)
: sig1  ( x - n )  DUP  DUP  19 ROR  SWAP  61 ROR  XOR  SWAP 6 RSHIFT  XOR  ;

\ Put two copies of original Wi on stack, keep its address
: Wi@ ( [Wi] - wi [Wi] wi) DUP  @  TUCK ; 

\ Create 2 copies of new Wi' from Wi on stack  ( ..Wi -..Wi' Wi')
: Wi  15 PICK  15 PICK  sig0  +  7 PICK  +  2 PICK  sig1  +  DUP ;

\ Drop 80 Wi cells from stack ( W0..W79 - )
: WiDROP  5 0 DO  2DROP 2DROP 2DROP 2DROP 2DROP 2DROP 2DROP 2DROP LOOP ;

: subrnd  DUP  H[D]  +!  T2  +  H[G] TO H[H]  H[A] !  ;

MACRO rndi\  " Wi@  T1x  +  \  +  subrnd  CELL+"
MACRO rndn\  " Wi   T1x  +  \  +  subrnd "

: SHA512  ( Wadr - )
  rndi\  K0    rndi\  K1    rndi\  K2    rndi\  K3   \ Wi = Mi for 1st 16 rounds
  rndi\  K4    rndi\  K5    rndi\  K6    rndi\  K7
  rndi\  K8    rndi\  K9    rndi\  K10   rndi\  K11
  rndi\  K12   rndi\  K13   rndi\  K14   rndi\  K15  DROP  ( W0..W15 )
  rndn\  K16   rndn\  K17   rndn\  K18   rndn\  K19  \ Wj now function of Wi
  rndn\  K20   rndn\  K21   rndn\  K22   rndn\  K23
  rndn\  K24   rndn\  K25   rndn\  K26   rndn\  K27
  rndn\  K28   rndn\  K29   rndn\  K30   rndn\  K31
  rndn\  K32   rndn\  K33   rndn\  K34   rndn\  K35
  rndn\  K36   rndn\  K37   rndn\  K38   rndn\  K39
  rndn\  K40   rndn\  K41   rndn\  K42   rndn\  K43
  rndn\  K44   rndn\  K45   rndn\  K46   rndn\  K47
  rndn\  K48   rndn\  K49   rndn\  K50   rndn\  K51
  rndn\  K52   rndn\  K53   rndn\  K54   rndn\  K55
  rndn\  K56   rndn\  K57   rndn\  K58   rndn\  K59
  rndn\  K60   rndn\  K61   rndn\  K62   rndn\  K63
  rndn\  K64   rndn\  K65   rndn\  K66   rndn\  K67
  rndn\  K68   rndn\  K69   rndn\  K70   rndn\  K71
  rndn\  K72   rndn\  K73   rndn\  K74   rndn\  K75
  rndn\  K76   rndn\  K77   rndn\  K78   rndn\  K79  WiDROP  ( - )
  UpDateHash
;

: setlen  ( -- )  \ Store bit count into last two cells
  SHAlen 2@  D2* D2* D2* ( bytes->bits) W 112 CHARS + !  W 120 CHARS + ! ;

: bytes>< ( m -- w )  \ Reverse cell bytes: 1234567890abcdef <-> efcdab9078563412
  [ HEX ]  DUP >R  38 LSHIFT  R@ FF00 AND  28 LSHIFT OR
  R@ FF0000 AND 18 LSHIFT OR  R@ FF000000 AND 8 LSHIFT OR
  R@ 20 RSHIFT FF AND OR   R@ 18 RSHIFT FF00 AND OR   R@  10 RSHIFT FF0000 AND OR
  R>  8 RSHIFT FF000000 AND OR [ DECIMAL ]
;

: cellsreverse  ( adr n -- )  \ Reverse bytes of n cells in array
  0 DO  DUP  @  bytes><  OVER !  CELL+  LOOP  DROP
;

W C@ [IF]  \ if little ENDIAN, e.g. Intel/AMD
      : endian16 ( adr -- adr ) DUP  16  cellsreverse ;
      : endian14 ( adr -- adr ) DUP  14  cellsreverse ;
[ELSE]      \ if big ENDIAN, e.g. Macs
      : endian16 ( adr -- adr ) ; \ Do nothing
      : endian14 ( adr -- adr ) ; \ Do nothing
[THEN]

\ Do all 128 byte blocks leaving remainder block
: hashfullblocks ( adr1 dcount -- adr2 count )         \ dcount is double number: lo hi
  SWAP  DUP >R  7 RSHIFT               ( adr1 hi lo* ) \ Store lo on return, do lo*=lo/128
  OVER CELLSIZE 7 - LSHIFT OR >R       ( adr1 hi     ) \ Return is now: :R lo lo'
  ( hi) 7 RSHIFT 0 ?DO                 ( adr1        ) \ Do if hi'= hi/128 > 0
    0 0 DO DUP endian16 SHA512 128 + LOOP ( lo' adr' ) \ Hash for 2^cellsize full blocks
  LOOP                                 ( adr'        ) \ Hash for hi'*2^cellsize full blocks
  R> 0 ?DO DUP endian16 SHA512 128 + LOOP ( adr'     ) \ Hash block for lo count full 128 byte blocks
  R> ( lo) 127 AND                    ( adr2 cnt2    ) \ Leave address and count for partial block
;

: hashfinal ( addr count -- )  \ Hash partial and/or last block
  DUP >R  W  SWAP  CMOVE                \ Move bytes into block W array
  W  R@ +  128  OVER  C!   ( adr     )  \ Put 80h after last message byte
  CHAR+  111 R@ -          ( adr #   )  \ Compute tentative 0 byte FILL count
  R> 111 >                 ( adr # ? )  \ Is partial block byte count > 111 ?
  IF    16 + 0  FILL            ( -- )  \ If yes, FILL rest of block w/zeroes
        W  endian16  SHA512     ( -- )  \ Endian adjust block if required, then hash
        W  112             ( adr 112 )  \ Now setup last block containing bit count
  THEN                     ( adr #   )
  0 FILL  setlen  W  endian14  SHA512   \ Zero FILL last block, set message bit count
;                               ( -- )  \ Endian adjust, except bit count, then hash

\ Compute SHA512 from a counted buffer of text
: SHAbuffer ( addr dcount -- )
  SHAinit  2DUP  SHAlen 2!  hashfullblocks  hashfinal ;

\ ===============  Hash string display wordset  ===============
  DECIMAL

\ Array of digits 0123456789abcdef
: digit$  ( -- adr )  S" 0123456789abcdef"  DROP  ;

: intdigits ( -- )  0 PAD  ! ;
: savedigit ( n -- )  PAD  C@  1+  DUP  PAD  C!  PAD  +  C!  ;
: bytedigits ( n1 -- )
  DUP 4 RSHIFT digit$ + C@ savedigit 15 AND digit$ + C@ savedigit
;

  W C@ [IF] \ little ENDIAN
: celldigits ( a1 -- )  DUP 7 + DO I C@ bytedigits  -1 +LOOP ;
  [ELSE]    \ big ENDIAN
: celldigits ( a1 -- )  DUP 8 + SWAP DO I C@ bytedigits LOOP ;
  [THEN]

: SHAstring ( -- adr count )  \ Return counted SHA-512 string array
  intdigits  SHAval 7 CELLS +  8 0 DO  DUP  celldigits  CELL-  LOOP  DROP  PAD  COUNT ;

\ Display SHA-512 hash value in hex ( A B C D E F G H )
: HASH. CR  SHAstring  TYPE  SPACE  ;

: QuoteString ( adr cnt --)  [CHAR] " EMIT  TYPE  [CHAR] " EMIT ;


\ ====================  File hash wordset  ====================
  VARIABLE  rfileid     \ Holds fileid number of input file

: InputFileName  ( -- ior)
  CR  CR  ." Filename: "  PAD  DUP  80  ACCEPT ( adr #)
  R/O  OPEN-FILE  SWAP  rfileid !  ( ior)
;

: TryAgain?  ( -- ?)
  CR  CR ." Invalid iput file, try again? (Y/N)"
  KEY  DUP  EMIT  DUP [CHAR] N =  SWAP [CHAR] n = OR
;

\ Read n bytes from input file, store at addr array
: bytes@  ( adr n - )  rfileid @  READ-FILE  2DROP ;

: storelen  ( lo hi - )  \ Store bit count into last two cells
  D2* D2* D2* ( bytes->bits) W 112 CHARS +  !  W 120 CHARS + !
;

: getpartial ( cnt  -- W'  cnt2 ?)
  W  2DUP  SWAP  DUP >R  bytes@          ( cnt1 adr1  )
  + 128 OVER C! CHAR+ 111 R@ - R> 111 >  ( adr2 cnt2 ?)
;

: block@  W 128 bytes@ ;

: SHAfile ( -- )
  BEGIN  InputFileName  ( ior)                  \ Enter filename
  WHILE  TryAgain? IF  EXIT  THEN               \ Not valid, try (not) again
  REPEAT SHAinit                                \ Valid file, init transform
  rfileid @  FILE-SIZE  DROP  ( ud )            \ Get bytesize of input file
  2 0  D-                                       \ Dec cnt by 2 for CR|LF EOF
  CR ." Bytesize: " 2DUP  D.                    \ Display filesize to screen
  2DUP  2>R                                     \ ( lo  hi ) Save file byte cnt on RETURN
  OVER  7 RSHIFT OVER                           \ ( lo  hi lo* hi )
  CELLSIZE 7 - LSHIFT OR  SWAP  7 RSHIFT        \ ( lo  lo' hi') full block count
  0 ?DO 0 0 DO block@  W endian16 SHA512 LOOP LOOP  \ Hash hi*2^cellsize full blocks
  0 ?DO block@  W endian16 SHA512 LOOP          \ Hash lo count full 128 byte blocks
  ( lo) 127 AND ( rembytes) getpartial ( adr cnt ?) \ Read remaining bytes
  IF 16 + 0 FILL  W endian16 SHA512  W 112 THEN \ Do if rembytes > 111
  0 FILL  2R> storelen  W endian14  SHA512      \ Do last block
  CR  ." SHA-512 : "  SHAstring  TYPE  CR       \ Show SHA-512 hash for file
  rfileid @  CLOSE-FILE  DROP                   \ Close the input file
;

\ ====================  SHA-512 Test Suite  =====================
  DECIMAL

\ Load W array with data on stack
: WLoad ( d0..d15 -- )  W 15 CELLS + ( d0..d15 W[15] )
  16 0 DO  TUCK  !  CELL-  LOOP  DROP
;

\ -------------------------------------------------------------
\ EXAMPLE 1: from FIPS PUB
\ Message: ASCII string 'abc'
\ Hash = DDAF35A1 93617ABA CC417349 AE204131 12E6FA4E 89A97EA2 0A9EEEE6 4B55D39A
\        2192992A 274FC1A8 36BA3C23 A3FEEBBD 454D4423 643CE80E 2A9AC94F A54CA49F

\ Compute and display hash for ASCII string 'abc'
: EX1  S" abc" U>D ( adr dcount) SHAbuffer  HASH.  ;

\ -------------------------------------------------------------
\ EXAMPLE 2: from FIPS PUB
\ Message:"abcdefghbcdefghicdefghijdefghijkefghijklfghijklmghijklmnhijklmnoijklmnopjklmnopqklmnopqrlmnopqrsmnopqrstnopqrstu"
\ Hash = 8E959B75 DAE313DA 8CF4F728 14FC143F 8F7779C6 EB9F7FA1 7299AEAD B6889018
\        501D289E 4900F7E4 331B99DE C4B5433A C7D329EE B6DD2654 5E96E55B 874BE909

: EX2a S" abcdefghbcdefghicdefghijdefghijkefghijklfghijklmghijklmnhijklmnoijklmnopjklmnopqklmnopqrlmnopqrsmnopqrstnopqrstu" U>D ( adr dcount) ;
: EX2  EX2a  SHAbuffer  HASH.  ;

\ -------------------------------------------------------------
\ EXAMPLE 3:
\ Message: 2 million copies of 'a' (61h), (16 million bits)
\ Hash =

\ Load block of all 'a's (61h), must hash 15,625 times
: EX3a  W  128  [CHAR] a  FILL  ;

\ Last message block: 1st bit a '1', bit-count = 16 million
: EX3b  [ HEX ] 8000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0
        [ DECIMAL ] 0 16000000 WLoad
;
\ Do hash for message of 2 million copies of ASCII 'a' (61h)
: EX3  SHAinit  EX3a  15625 0 DO W SHA512 LOOP  EX3b  W SHA512 HASH. ;

\ -------------------------------------------------------------
\ EXAMPLE 4:
\ Message: 400,000 SPACES 'BL' (20h), (3,200,000 bits)
\ Hash =

\ Load block of all "BL' (20h), hash 156 full blocks + 16 bytes
: EX4a  W  128  BL  FILL  ;

\ Last message block: 1st bit a '1', bit-count = 3,200,000
: EX4b  [ HEX ] 8000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0
        [ DECIMAL ] 0 3200000 WLoad
;

\ Do hash for message of 400,000 SPACES 'BL' (20h)
: EX4  SHAinit  EX4a  3125 0 DO  W SHA512  LOOP  EX4b  W SHA512 HASH. ;

\ -------------------------------------------------------------
\ Message: blank string ''
\ Hash =

: SHATest ( -- )
  CR ." SHA-512 test suite:"
  S" " U>D ( adr dcount)  SHAbuffer  HASH.  S" "  QuoteString
  EX1  S" abc"   QuoteString
  EX2  EX2a DROP QuoteString
  EX3  S" 2 million copies of ASCII 'a' (61h)" TYPE
  EX4  S" 400,000 copies of ASCII BL (20h)" TYPE  CR ;

\ ===========  Forth specific performance test ===========

  VARIABLE  start-ms

: TIMER-START  ( -- )  ?MS  start-ms  ! ;

: MS?  ( -- u )  ?MS  start-ms  @  - ( abs ) ;

: .### ( -) BASE @ >R DECIMAL MS? 0 <# # # # [CHAR] . HOLD #S #> R> BASE ! TYPE ;

  DECIMAL 1000 VALUE N#

: [EX1]  S" abc" U>D ( adr dcount) SHAbuffer  ;
: [EX2]  EX2a     SHAbuffer  ;
: [EX3]  SHAinit  EX3a  15625 0 DO W SHA512 LOOP EX3b W SHA512 ;

: test1  [ DECIMAL ]
  cr ." SHA-512 test for EX1 for " N# . ." loops in milliseconds is "
  TIMER-START  N# 0 DO  [EX1]  LOOP  MS?  U.
;

: test2  [ DECIMAL ]
  cr ." SHA-512 test for EX2 for " N# . ." loops in milliseconds is "
  TIMER-START  N# 0 DO  [EX2]  LOOP  MS?  U.
;

: test3  [ DECIMAL ]
  cr ." SHA-512 test for EX3 for " N# . ." loops in milliseconds is "
  TIMER-START  N# 0 DO  [EX3]  LOOP  MS?  U.
;

[toc] | [prev] | [standalone]


Page 2 of 2 — ← Prev page 1 [2]

Back to top | Article view | comp.lang.forth


csiph-web