x86_64: New md5 assembly code.

ChangeLog

+2013-12-03  Niels Möller  <nisse@lysator.liu.se>
+
+	* x86_64/md5-compress.asm: New file, similar to the x86 version.
+	35% speedup.
+
 2013-11-25  Niels Möller  <nisse@lysator.liu.se>
 
 	* testsuite/testutils.c (test_dsa_sign, test_dsa_verify): New

x86_64/md5-compress.asm

+C nettle, low-level cryptographics library
+C
+C Copyright (C) 2005, 2013 Niels Möller
+C
+C The nettle library is free software; you can redistribute it and/or modify
+C it under the terms of the GNU Lesser General Public License as published by
+C the Free Software Foundation; either version 2.1 of the License, or (at your
+C option) any later version.
+C
+C The nettle library is distributed in the hope that it will be useful, but
+C WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+C or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
+C License for more details.
+C
+C You should have received a copy of the GNU Lesser General Public License
+C along with the nettle library; see the file COPYING.LIB.  If not, write to
+C the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
+C MA 02111-1301, USA.
+
+C Registers:
+
+define(<STATE>, <%rdi>)
+define(<INPUT>, <%rsi>)
+define(<SA>,<%rax>)
+define(<SB>,<%rbx>)
+define(<SC>,<%rcx>)
+define(<SD>,<%rbp>)
+define(<TMP>, <%r8>)
+
+C F1(x,y,z) = (z ^ (x & (y ^ z)))
+define(<F1>, <
+	movl	XREG($3), XREG(TMP)
+	xorl	XREG($2), XREG(TMP)
+	andl	XREG($1), XREG(TMP)
+	xorl	XREG($3), XREG(TMP)>)
+
+define(<F2>,<F1($3, $1, $2)>)
+
+C F3(x,y,z) = x ^ y ^ z
+define(<F3>,<
+	movl	XREG($1), XREG(TMP)
+	xorl	XREG($2), XREG(TMP)
+	xorl	XREG($3), XREG(TMP)>)
+
+C F4(x,y,z) = y ^ (x | ~z)
+define(<F4>,<
+	movl	XREG($3), XREG(TMP)
+	notl	XREG(TMP)
+	orl	XREG($1), XREG(TMP)
+	xorl	XREG($2), XREG(TMP)>)
+
+C Index to 4*i, or to the empty string if zero
+define(<REF>,<ifelse($1,0,,eval(4*$1))(INPUT)>)
+
+C ROUND(f, w, x, y, z, k, data, s):
+C	w += f(x,y,z) + data + k
+C	w <<< s
+C	w += x
+define(<ROUND>,<
+	addl	<$>$7, XREG($2)
+	$1($3, $4, $5)
+	addl	$6, XREG($2)
+	addl	XREG(TMP), XREG($2)
+	roll	<$>$8, XREG($2)
+	addl	XREG($3), XREG($2)>)
+
+	.file "md5-compress.asm"
+
+	C _nettle_md5_compress(uint32_t *state, uint8_t *input)
+	.text
+	ALIGN(16)
+PROLOGUE(_nettle_md5_compress)
+	W64_ENTRY(2,0)
+	C save all registers that need to be saved
+	push	%rbp
+	push	%rbx
+
+	C load the state vector
+	movl	(STATE),   XREG(SA)
+	movl	4(STATE),  XREG(SB)
+	movl	8(STATE),  XREG(SC)
+	movl	12(STATE), XREG(SD)
+
+	ROUND(<F1>, SA, SB, SC, SD, REF( 0), 0xd76aa478, 7)
+	ROUND(<F1>, SD, SA, SB, SC, REF( 1), 0xe8c7b756, 12)
+	ROUND(<F1>, SC, SD, SA, SB, REF( 2), 0x242070db, 17)
+	ROUND(<F1>, SB, SC, SD, SA, REF( 3), 0xc1bdceee, 22)
+	ROUND(<F1>, SA, SB, SC, SD, REF( 4), 0xf57c0faf, 7)
+	ROUND(<F1>, SD, SA, SB, SC, REF( 5), 0x4787c62a, 12)
+	ROUND(<F1>, SC, SD, SA, SB, REF( 6), 0xa8304613, 17)
+	ROUND(<F1>, SB, SC, SD, SA, REF( 7), 0xfd469501, 22)
+	ROUND(<F1>, SA, SB, SC, SD, REF( 8), 0x698098d8, 7)
+	ROUND(<F1>, SD, SA, SB, SC, REF( 9), 0x8b44f7af, 12)
+	ROUND(<F1>, SC, SD, SA, SB, REF(10), 0xffff5bb1, 17)
+	ROUND(<F1>, SB, SC, SD, SA, REF(11), 0x895cd7be, 22)
+	ROUND(<F1>, SA, SB, SC, SD, REF(12), 0x6b901122, 7)
+	ROUND(<F1>, SD, SA, SB, SC, REF(13), 0xfd987193, 12)
+	ROUND(<F1>, SC, SD, SA, SB, REF(14), 0xa679438e, 17)
+	ROUND(<F1>, SB, SC, SD, SA, REF(15), 0x49b40821, 22)
+
+	ROUND(<F2>, SA, SB, SC, SD, REF( 1), 0xf61e2562, 5)
+	ROUND(<F2>, SD, SA, SB, SC, REF( 6), 0xc040b340, 9)
+	ROUND(<F2>, SC, SD, SA, SB, REF(11), 0x265e5a51, 14)
+	ROUND(<F2>, SB, SC, SD, SA, REF( 0), 0xe9b6c7aa, 20)
+	ROUND(<F2>, SA, SB, SC, SD, REF( 5), 0xd62f105d, 5)
+	ROUND(<F2>, SD, SA, SB, SC, REF(10), 0x02441453, 9)
+	ROUND(<F2>, SC, SD, SA, SB, REF(15), 0xd8a1e681, 14)
+	ROUND(<F2>, SB, SC, SD, SA, REF( 4), 0xe7d3fbc8, 20)
+	ROUND(<F2>, SA, SB, SC, SD, REF( 9), 0x21e1cde6, 5)
+	ROUND(<F2>, SD, SA, SB, SC, REF(14), 0xc33707d6, 9)
+	ROUND(<F2>, SC, SD, SA, SB, REF( 3), 0xf4d50d87, 14)
+	ROUND(<F2>, SB, SC, SD, SA, REF( 8), 0x455a14ed, 20)
+	ROUND(<F2>, SA, SB, SC, SD, REF(13), 0xa9e3e905, 5)
+	ROUND(<F2>, SD, SA, SB, SC, REF( 2), 0xfcefa3f8, 9)
+	ROUND(<F2>, SC, SD, SA, SB, REF( 7), 0x676f02d9, 14)
+	ROUND(<F2>, SB, SC, SD, SA, REF(12), 0x8d2a4c8a, 20)
+
+	ROUND(<F3>, SA, SB, SC, SD, REF( 5), 0xfffa3942, 4)
+	ROUND(<F3>, SD, SA, SB, SC, REF( 8), 0x8771f681, 11)
+	ROUND(<F3>, SC, SD, SA, SB, REF(11), 0x6d9d6122, 16)
+	ROUND(<F3>, SB, SC, SD, SA, REF(14), 0xfde5380c, 23)
+	ROUND(<F3>, SA, SB, SC, SD, REF( 1), 0xa4beea44, 4)
+	ROUND(<F3>, SD, SA, SB, SC, REF( 4), 0x4bdecfa9, 11)
+	ROUND(<F3>, SC, SD, SA, SB, REF( 7), 0xf6bb4b60, 16)
+	ROUND(<F3>, SB, SC, SD, SA, REF(10), 0xbebfbc70, 23)
+	ROUND(<F3>, SA, SB, SC, SD, REF(13), 0x289b7ec6, 4)
+	ROUND(<F3>, SD, SA, SB, SC, REF( 0), 0xeaa127fa, 11)
+	ROUND(<F3>, SC, SD, SA, SB, REF( 3), 0xd4ef3085, 16)
+	ROUND(<F3>, SB, SC, SD, SA, REF( 6), 0x04881d05, 23)
+	ROUND(<F3>, SA, SB, SC, SD, REF( 9), 0xd9d4d039, 4)
+	ROUND(<F3>, SD, SA, SB, SC, REF(12), 0xe6db99e5, 11)
+	ROUND(<F3>, SC, SD, SA, SB, REF(15), 0x1fa27cf8, 16)
+	ROUND(<F3>, SB, SC, SD, SA, REF( 2), 0xc4ac5665, 23)
+
+	ROUND(<F4>, SA, SB, SC, SD, REF( 0), 0xf4292244, 6)
+	ROUND(<F4>, SD, SA, SB, SC, REF( 7), 0x432aff97, 10)
+	ROUND(<F4>, SC, SD, SA, SB, REF(14), 0xab9423a7, 15)
+	ROUND(<F4>, SB, SC, SD, SA, REF( 5), 0xfc93a039, 21)
+	ROUND(<F4>, SA, SB, SC, SD, REF(12), 0x655b59c3, 6)
+	ROUND(<F4>, SD, SA, SB, SC, REF( 3), 0x8f0ccc92, 10)
+	ROUND(<F4>, SC, SD, SA, SB, REF(10), 0xffeff47d, 15)
+	ROUND(<F4>, SB, SC, SD, SA, REF( 1), 0x85845dd1, 21)
+	ROUND(<F4>, SA, SB, SC, SD, REF( 8), 0x6fa87e4f, 6)
+	ROUND(<F4>, SD, SA, SB, SC, REF(15), 0xfe2ce6e0, 10)
+	ROUND(<F4>, SC, SD, SA, SB, REF( 6), 0xa3014314, 15)
+	ROUND(<F4>, SB, SC, SD, SA, REF(13), 0x4e0811a1, 21)
+	ROUND(<F4>, SA, SB, SC, SD, REF( 4), 0xf7537e82, 6)
+	ROUND(<F4>, SD, SA, SB, SC, REF(11), 0xbd3af235, 10)
+	ROUND(<F4>, SC, SD, SA, SB, REF( 2), 0x2ad7d2bb, 15)
+	ROUND(<F4>, SB, SC, SD, SA, REF( 9), 0xeb86d391, 21)
+
+	C Update the state vector
+	addl	XREG(SA), (STATE)
+	addl	XREG(SB), 4(STATE)
+	addl	XREG(SC), 8(STATE)
+	addl	XREG(SD), 12(STATE)
+
+	pop	%rbx
+	pop	%rbp
+	W64_EXIT(2,0)
+
+	ret
+EPILOGUE(_nettle_md5_compress)