libbb/sha1: x86_64 version: generate from a script, optimize a bit

function                                             old     new   delta
sha1_process_block64                                3569    3502     -67

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

3 files changed, 474 insertions, 267 deletions

diff --git a/libbb/Config.src b/libbb/Config.src
index 42a2283..c80bee2 100644
--- a/libbb/Config.src
+++ b/libbb/Config.src
@@ -59,7 +59,7 @@ config SHA1_SMALL
 	Trade binary size versus speed for the sha1 algorithm.
 	                throughput MB/s   size of sha1_process_block64
 	value           486  x86-64       486   x86-64
-	0               367  367          3657  3570
+	0               367  375          3657  3502
 	1               224  229           654   732
 	2,3             200  195           358   380
 
diff --git a/libbb/hash_md5_sha_x86-64.S b/libbb/hash_md5_sha_x86-64.S
index 466cd9a..3e1c4b4 100644
--- a/libbb/hash_md5_sha_x86-64.S
+++ b/libbb/hash_md5_sha_x86-64.S
@@ -1,23 +1,27 @@
 ### Generated by hash_md5_sha_x86-64.S.sh ###
-#if defined(__GNUC__) && defined(__x86_64__)
+
+#if CONFIG_SHA1_SMALL == 0 && defined(__GNUC__) && defined(__x86_64__)
 	.section	.text.sha1_process_block64,"ax",@progbits
-        .globl  sha1_process_block64
-        .hidden sha1_process_block64
+	.globl  sha1_process_block64
+	.hidden sha1_process_block64
 	.type	sha1_process_block64, @function
+
+	.balign	8	# allow decoders to fetch at least 4 first insns
 sha1_process_block64:
-	pushq	%r15		#
-	pushq	%r14		#
-	pushq	%r13		#
-	pushq	%r12		#
-	pushq	%rbp		#
-	pushq	%rbx		#
-	pushq	%rdi		# we need ctx at the end
+	pushq	%r15	#
+	pushq	%r14	#
+	pushq	%r13	#
+	pushq	%r12	#
+	pushq	%rbp	#
+	pushq	%rbx	#
+	pushq	%rdi	# we need ctx at the end
 
 #Register and stack use:
 # eax..edx: a..d
 # ebp: e
 # esi,edi: temps
 # -32+4*n(%rsp),r8...r15: W[0..7,8..15]
+# (TODO: actually W[0..7] are used a bit more often, put _thme_ into r8..r15?)
 
 	movq	4*8(%rdi), %r8
 	bswapq	%r8
@@ -253,7 +257,7 @@ sha1_process_block64:
 	xorl	%ecx, %edi		# ^d
 	andl	%eax, %edi		# &b
 	xorl	%ecx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rdx,%rsi),%edx # e += RCONST + W[n]
+	leal	0x5A827999(%rdx,%rsi), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (((c ^ d) & b) ^ d)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -270,7 +274,7 @@ sha1_process_block64:
 	xorl	%ebx, %edi		# ^d
 	andl	%ebp, %edi		# &b
 	xorl	%ebx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rcx,%rsi),%ecx # e += RCONST + W[n]
+	leal	0x5A827999(%rcx,%rsi), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (((c ^ d) & b) ^ d)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -287,7 +291,7 @@ sha1_process_block64:
 	xorl	%eax, %edi		# ^d
 	andl	%edx, %edi		# &b
 	xorl	%eax, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rbx,%rsi),%ebx # e += RCONST + W[n]
+	leal	0x5A827999(%rbx,%rsi), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (((c ^ d) & b) ^ d)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -304,7 +308,7 @@ sha1_process_block64:
 	xorl	%ebp, %edi		# ^d
 	andl	%ecx, %edi		# &b
 	xorl	%ebp, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rax,%rsi),%eax # e += RCONST + W[n]
+	leal	0x5A827999(%rax,%rsi), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (((c ^ d) & b) ^ d)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -320,7 +324,7 @@ sha1_process_block64:
 	movl	%ecx, %edi		# c
 	xorl	%edx, %edi		# ^d
 	xorl	%ebx, %edi		# ^b
-	leal	0x6ED9EBA1(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rbp,%rsi), %ebp # e += RCONST + W[n & 15]
 	addl	%edi, %ebp		# e += (c ^ d ^ b)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -336,7 +340,7 @@ sha1_process_block64:
 	movl	%ebx, %edi		# c
 	xorl	%ecx, %edi		# ^d
 	xorl	%eax, %edi		# ^b
-	leal	0x6ED9EBA1(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rdx,%rsi), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (c ^ d ^ b)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -352,7 +356,7 @@ sha1_process_block64:
 	movl	%eax, %edi		# c
 	xorl	%ebx, %edi		# ^d
 	xorl	%ebp, %edi		# ^b
-	leal	0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (c ^ d ^ b)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -368,135 +372,119 @@ sha1_process_block64:
 	movl	%ebp, %edi		# c
 	xorl	%eax, %edi		# ^d
 	xorl	%edx, %edi		# ^b
-	leal	0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (c ^ d ^ b)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebx		# e += rotl32(a,5)
 	rorl	$2, %edx		# b = rotl32(b,30)
 # 24
-	movl	-32+4*5(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*0(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r10d, %esi		# ^W[(n+2) & 15]
-	xorl	%r8d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r8d		# store to W[n & 15]
+	xorl	-32+4*5(%rsp), %r8d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*0(%rsp), %r8d	# ^W[(n+8) & 15]
+	xorl	%r10d, %r8d	# ^W[(n+2) & 15]
+	roll	%r8d		#
 	movl	%edx, %edi		# c
 	xorl	%ebp, %edi		# ^d
 	xorl	%ecx, %edi		# ^b
-	leal	0x6ED9EBA1(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rax,%r8), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (c ^ d ^ b)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %eax		# e += rotl32(a,5)
 	rorl	$2, %ecx		# b = rotl32(b,30)
 # 25
-	movl	-32+4*6(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*1(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r11d, %esi		# ^W[(n+2) & 15]
-	xorl	%r9d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r9d		# store to W[n & 15]
+	xorl	-32+4*6(%rsp), %r9d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*1(%rsp), %r9d	# ^W[(n+8) & 15]
+	xorl	%r11d, %r9d	# ^W[(n+2) & 15]
+	roll	%r9d		#
 	movl	%ecx, %edi		# c
 	xorl	%edx, %edi		# ^d
 	xorl	%ebx, %edi		# ^b
-	leal	0x6ED9EBA1(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rbp,%r9), %ebp # e += RCONST + W[n & 15]
 	addl	%edi, %ebp		# e += (c ^ d ^ b)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebp		# e += rotl32(a,5)
 	rorl	$2, %ebx		# b = rotl32(b,30)
 # 26
-	movl	-32+4*7(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*2(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r12d, %esi		# ^W[(n+2) & 15]
-	xorl	%r10d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r10d		# store to W[n & 15]
+	xorl	-32+4*7(%rsp), %r10d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*2(%rsp), %r10d	# ^W[(n+8) & 15]
+	xorl	%r12d, %r10d	# ^W[(n+2) & 15]
+	roll	%r10d		#
 	movl	%ebx, %edi		# c
 	xorl	%ecx, %edi		# ^d
 	xorl	%eax, %edi		# ^b
-	leal	0x6ED9EBA1(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rdx,%r10), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (c ^ d ^ b)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %edx		# e += rotl32(a,5)
 	rorl	$2, %eax		# b = rotl32(b,30)
 # 27
-	movl	%r8d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*3(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r13d, %esi		# ^W[(n+2) & 15]
-	xorl	%r11d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r11d		# store to W[n & 15]
+	xorl	%r8d, %r11d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*3(%rsp), %r11d	# ^W[(n+8) & 15]
+	xorl	%r13d, %r11d	# ^W[(n+2) & 15]
+	roll	%r11d		#
 	movl	%eax, %edi		# c
 	xorl	%ebx, %edi		# ^d
 	xorl	%ebp, %edi		# ^b
-	leal	0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rcx,%r11), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (c ^ d ^ b)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ecx		# e += rotl32(a,5)
 	rorl	$2, %ebp		# b = rotl32(b,30)
 # 28
-	movl	%r9d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*4(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r14d, %esi		# ^W[(n+2) & 15]
-	xorl	%r12d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r12d		# store to W[n & 15]
+	xorl	%r9d, %r12d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*4(%rsp), %r12d	# ^W[(n+8) & 15]
+	xorl	%r14d, %r12d	# ^W[(n+2) & 15]
+	roll	%r12d		#
 	movl	%ebp, %edi		# c
 	xorl	%eax, %edi		# ^d
 	xorl	%edx, %edi		# ^b
-	leal	0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rbx,%r12), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (c ^ d ^ b)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebx		# e += rotl32(a,5)
 	rorl	$2, %edx		# b = rotl32(b,30)
 # 29
-	movl	%r10d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*5(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r15d, %esi		# ^W[(n+2) & 15]
-	xorl	%r13d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r13d		# store to W[n & 15]
+	xorl	%r10d, %r13d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*5(%rsp), %r13d	# ^W[(n+8) & 15]
+	xorl	%r15d, %r13d	# ^W[(n+2) & 15]
+	roll	%r13d		#
 	movl	%edx, %edi		# c
 	xorl	%ebp, %edi		# ^d
 	xorl	%ecx, %edi		# ^b
-	leal	0x6ED9EBA1(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rax,%r13), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (c ^ d ^ b)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %eax		# e += rotl32(a,5)
 	rorl	$2, %ecx		# b = rotl32(b,30)
 # 30
-	movl	%r11d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*6(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	-32+4*0(%rsp), %esi		# ^W[(n+2) & 15]
-	xorl	%r14d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r14d		# store to W[n & 15]
+	xorl	%r11d, %r14d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*6(%rsp), %r14d	# ^W[(n+8) & 15]
+	xorl	-32+4*0(%rsp), %r14d	# ^W[(n+2) & 15]
+	roll	%r14d		#
 	movl	%ecx, %edi		# c
 	xorl	%edx, %edi		# ^d
 	xorl	%ebx, %edi		# ^b
-	leal	0x6ED9EBA1(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rbp,%r14), %ebp # e += RCONST + W[n & 15]
 	addl	%edi, %ebp		# e += (c ^ d ^ b)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebp		# e += rotl32(a,5)
 	rorl	$2, %ebx		# b = rotl32(b,30)
 # 31
-	movl	%r12d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*7(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	-32+4*1(%rsp), %esi		# ^W[(n+2) & 15]
-	xorl	%r15d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r15d		# store to W[n & 15]
+	xorl	%r12d, %r15d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*7(%rsp), %r15d	# ^W[(n+8) & 15]
+	xorl	-32+4*1(%rsp), %r15d	# ^W[(n+2) & 15]
+	roll	%r15d		#
 	movl	%ebx, %edi		# c
 	xorl	%ecx, %edi		# ^d
 	xorl	%eax, %edi		# ^b
-	leal	0x6ED9EBA1(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rdx,%r15), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (c ^ d ^ b)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -512,7 +500,7 @@ sha1_process_block64:
 	movl	%eax, %edi		# c
 	xorl	%ebx, %edi		# ^d
 	xorl	%ebp, %edi		# ^b
-	leal	0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (c ^ d ^ b)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -528,7 +516,7 @@ sha1_process_block64:
 	movl	%ebp, %edi		# c
 	xorl	%eax, %edi		# ^d
 	xorl	%edx, %edi		# ^b
-	leal	0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (c ^ d ^ b)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -544,7 +532,7 @@ sha1_process_block64:
 	movl	%edx, %edi		# c
 	xorl	%ebp, %edi		# ^d
 	xorl	%ecx, %edi		# ^b
-	leal	0x6ED9EBA1(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rax,%rsi), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (c ^ d ^ b)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -560,7 +548,7 @@ sha1_process_block64:
 	movl	%ecx, %edi		# c
 	xorl	%edx, %edi		# ^d
 	xorl	%ebx, %edi		# ^b
-	leal	0x6ED9EBA1(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rbp,%rsi), %ebp # e += RCONST + W[n & 15]
 	addl	%edi, %ebp		# e += (c ^ d ^ b)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -576,7 +564,7 @@ sha1_process_block64:
 	movl	%ebx, %edi		# c
 	xorl	%ecx, %edi		# ^d
 	xorl	%eax, %edi		# ^b
-	leal	0x6ED9EBA1(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rdx,%rsi), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (c ^ d ^ b)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -592,7 +580,7 @@ sha1_process_block64:
 	movl	%eax, %edi		# c
 	xorl	%ebx, %edi		# ^d
 	xorl	%ebp, %edi		# ^b
-	leal	0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (c ^ d ^ b)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -608,7 +596,7 @@ sha1_process_block64:
 	movl	%ebp, %edi		# c
 	xorl	%eax, %edi		# ^d
 	xorl	%edx, %edi		# ^b
-	leal	0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (c ^ d ^ b)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -624,7 +612,7 @@ sha1_process_block64:
 	movl	%edx, %edi		# c
 	xorl	%ebp, %edi		# ^d
 	xorl	%ecx, %edi		# ^b
-	leal	0x6ED9EBA1(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	0x6ED9EBA1(%rax,%rsi), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (c ^ d ^ b)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -637,14 +625,12 @@ sha1_process_block64:
 	andl	%ecx, %esi		# si: b & c
 	andl	%edx, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	-32+4*5(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*0(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r10d, %esi		# ^W[(n+2) & 15]
-	xorl	%r8d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r8d		# store to W[n & 15]
+	xorl	-32+4*5(%rsp), %r8d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*0(%rsp), %r8d	# ^W[(n+8) & 15]
+	xorl	%r10d, %r8d	# ^W[(n+2) & 15]
+	roll	%r8d		#
 	addl	%edi, %ebp		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	-0x70E44324(%rbp,%r8), %ebp # e += RCONST + W[n & 15]
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebp		# e += rotl32(a,5)
@@ -656,14 +642,12 @@ sha1_process_block64:
 	andl	%ebx, %esi		# si: b & c
 	andl	%ecx, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	-32+4*6(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*1(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r11d, %esi		# ^W[(n+2) & 15]
-	xorl	%r9d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r9d		# store to W[n & 15]
+	xorl	-32+4*6(%rsp), %r9d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*1(%rsp), %r9d	# ^W[(n+8) & 15]
+	xorl	%r11d, %r9d	# ^W[(n+2) & 15]
+	roll	%r9d		#
 	addl	%edi, %edx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rdx,%r9), %edx # e += RCONST + W[n & 15]
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %edx		# e += rotl32(a,5)
@@ -675,14 +659,12 @@ sha1_process_block64:
 	andl	%eax, %esi		# si: b & c
 	andl	%ebx, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	-32+4*7(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*2(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r12d, %esi		# ^W[(n+2) & 15]
-	xorl	%r10d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r10d		# store to W[n & 15]
+	xorl	-32+4*7(%rsp), %r10d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*2(%rsp), %r10d	# ^W[(n+8) & 15]
+	xorl	%r12d, %r10d	# ^W[(n+2) & 15]
+	roll	%r10d		#
 	addl	%edi, %ecx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rcx,%r10), %ecx # e += RCONST + W[n & 15]
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ecx		# e += rotl32(a,5)
@@ -694,14 +676,12 @@ sha1_process_block64:
 	andl	%ebp, %esi		# si: b & c
 	andl	%eax, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	%r8d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*3(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r13d, %esi		# ^W[(n+2) & 15]
-	xorl	%r11d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r11d		# store to W[n & 15]
+	xorl	%r8d, %r11d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*3(%rsp), %r11d	# ^W[(n+8) & 15]
+	xorl	%r13d, %r11d	# ^W[(n+2) & 15]
+	roll	%r11d		#
 	addl	%edi, %ebx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rbx,%r11), %ebx # e += RCONST + W[n & 15]
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebx		# e += rotl32(a,5)
@@ -713,14 +693,12 @@ sha1_process_block64:
 	andl	%edx, %esi		# si: b & c
 	andl	%ebp, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	%r9d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*4(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r14d, %esi		# ^W[(n+2) & 15]
-	xorl	%r12d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r12d		# store to W[n & 15]
+	xorl	%r9d, %r12d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*4(%rsp), %r12d	# ^W[(n+8) & 15]
+	xorl	%r14d, %r12d	# ^W[(n+2) & 15]
+	roll	%r12d		#
 	addl	%edi, %eax		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	-0x70E44324(%rax,%r12), %eax # e += RCONST + W[n & 15]
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %eax		# e += rotl32(a,5)
@@ -732,14 +710,12 @@ sha1_process_block64:
 	andl	%ecx, %esi		# si: b & c
 	andl	%edx, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	%r10d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*5(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r15d, %esi		# ^W[(n+2) & 15]
-	xorl	%r13d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r13d		# store to W[n & 15]
+	xorl	%r10d, %r13d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*5(%rsp), %r13d	# ^W[(n+8) & 15]
+	xorl	%r15d, %r13d	# ^W[(n+2) & 15]
+	roll	%r13d		#
 	addl	%edi, %ebp		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	-0x70E44324(%rbp,%r13), %ebp # e += RCONST + W[n & 15]
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebp		# e += rotl32(a,5)
@@ -751,14 +727,12 @@ sha1_process_block64:
 	andl	%ebx, %esi		# si: b & c
 	andl	%ecx, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	%r11d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*6(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	-32+4*0(%rsp), %esi		# ^W[(n+2) & 15]
-	xorl	%r14d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r14d		# store to W[n & 15]
+	xorl	%r11d, %r14d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*6(%rsp), %r14d	# ^W[(n+8) & 15]
+	xorl	-32+4*0(%rsp), %r14d	# ^W[(n+2) & 15]
+	roll	%r14d		#
 	addl	%edi, %edx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rdx,%r14), %edx # e += RCONST + W[n & 15]
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %edx		# e += rotl32(a,5)
@@ -770,14 +744,12 @@ sha1_process_block64:
 	andl	%eax, %esi		# si: b & c
 	andl	%ebx, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	%r12d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*7(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	-32+4*1(%rsp), %esi		# ^W[(n+2) & 15]
-	xorl	%r15d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r15d		# store to W[n & 15]
+	xorl	%r12d, %r15d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*7(%rsp), %r15d	# ^W[(n+8) & 15]
+	xorl	-32+4*1(%rsp), %r15d	# ^W[(n+2) & 15]
+	roll	%r15d		#
 	addl	%edi, %ecx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rcx,%r15), %ecx # e += RCONST + W[n & 15]
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ecx		# e += rotl32(a,5)
@@ -796,7 +768,7 @@ sha1_process_block64:
 	roll	%esi			#
 	movl	%esi, -32+4*0(%rsp)		# store to W[n & 15]
 	addl	%edi, %ebx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rbx,%rsi), %ebx # e += RCONST + W[n & 15]
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebx		# e += rotl32(a,5)
@@ -815,7 +787,7 @@ sha1_process_block64:
 	roll	%esi			#
 	movl	%esi, -32+4*1(%rsp)		# store to W[n & 15]
 	addl	%edi, %eax		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	-0x70E44324(%rax,%rsi), %eax # e += RCONST + W[n & 15]
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %eax		# e += rotl32(a,5)
@@ -834,7 +806,7 @@ sha1_process_block64:
 	roll	%esi			#
 	movl	%esi, -32+4*2(%rsp)		# store to W[n & 15]
 	addl	%edi, %ebp		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	-0x70E44324(%rbp,%rsi), %ebp # e += RCONST + W[n & 15]
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebp		# e += rotl32(a,5)
@@ -853,7 +825,7 @@ sha1_process_block64:
 	roll	%esi			#
 	movl	%esi, -32+4*3(%rsp)		# store to W[n & 15]
 	addl	%edi, %edx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rdx,%rsi), %edx # e += RCONST + W[n & 15]
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %edx		# e += rotl32(a,5)
@@ -872,7 +844,7 @@ sha1_process_block64:
 	roll	%esi			#
 	movl	%esi, -32+4*4(%rsp)		# store to W[n & 15]
 	addl	%edi, %ecx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rcx,%rsi), %ecx # e += RCONST + W[n & 15]
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ecx		# e += rotl32(a,5)
@@ -891,7 +863,7 @@ sha1_process_block64:
 	roll	%esi			#
 	movl	%esi, -32+4*5(%rsp)		# store to W[n & 15]
 	addl	%edi, %ebx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rbx,%rsi), %ebx # e += RCONST + W[n & 15]
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebx		# e += rotl32(a,5)
@@ -910,7 +882,7 @@ sha1_process_block64:
 	roll	%esi			#
 	movl	%esi, -32+4*6(%rsp)		# store to W[n & 15]
 	addl	%edi, %eax		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	-0x70E44324(%rax,%rsi), %eax # e += RCONST + W[n & 15]
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %eax		# e += rotl32(a,5)
@@ -929,7 +901,7 @@ sha1_process_block64:
 	roll	%esi			#
 	movl	%esi, -32+4*7(%rsp)		# store to W[n & 15]
 	addl	%edi, %ebp		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	-0x70E44324(%rbp,%rsi), %ebp # e += RCONST + W[n & 15]
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebp		# e += rotl32(a,5)
@@ -941,14 +913,12 @@ sha1_process_block64:
 	andl	%ebx, %esi		# si: b & c
 	andl	%ecx, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	-32+4*5(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*0(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r10d, %esi		# ^W[(n+2) & 15]
-	xorl	%r8d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r8d		# store to W[n & 15]
+	xorl	-32+4*5(%rsp), %r8d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*0(%rsp), %r8d	# ^W[(n+8) & 15]
+	xorl	%r10d, %r8d	# ^W[(n+2) & 15]
+	roll	%r8d		#
 	addl	%edi, %edx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rdx,%r8), %edx # e += RCONST + W[n & 15]
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %edx		# e += rotl32(a,5)
@@ -960,14 +930,12 @@ sha1_process_block64:
 	andl	%eax, %esi		# si: b & c
 	andl	%ebx, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	-32+4*6(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*1(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r11d, %esi		# ^W[(n+2) & 15]
-	xorl	%r9d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r9d		# store to W[n & 15]
+	xorl	-32+4*6(%rsp), %r9d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*1(%rsp), %r9d	# ^W[(n+8) & 15]
+	xorl	%r11d, %r9d	# ^W[(n+2) & 15]
+	roll	%r9d		#
 	addl	%edi, %ecx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rcx,%r9), %ecx # e += RCONST + W[n & 15]
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ecx		# e += rotl32(a,5)
@@ -979,14 +947,12 @@ sha1_process_block64:
 	andl	%ebp, %esi		# si: b & c
 	andl	%eax, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	-32+4*7(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*2(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r12d, %esi		# ^W[(n+2) & 15]
-	xorl	%r10d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r10d		# store to W[n & 15]
+	xorl	-32+4*7(%rsp), %r10d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*2(%rsp), %r10d	# ^W[(n+8) & 15]
+	xorl	%r12d, %r10d	# ^W[(n+2) & 15]
+	roll	%r10d		#
 	addl	%edi, %ebx		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	-0x70E44324(%rbx,%r10), %ebx # e += RCONST + W[n & 15]
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebx		# e += rotl32(a,5)
@@ -998,77 +964,67 @@ sha1_process_block64:
 	andl	%edx, %esi		# si: b & c
 	andl	%ebp, %edi		# di: (b | c) & d
 	orl	%esi, %edi		# ((b | c) & d) | (b & c)
-	movl	%r8d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*3(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r13d, %esi		# ^W[(n+2) & 15]
-	xorl	%r11d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r11d		# store to W[n & 15]
+	xorl	%r8d, %r11d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*3(%rsp), %r11d	# ^W[(n+8) & 15]
+	xorl	%r13d, %r11d	# ^W[(n+2) & 15]
+	roll	%r11d		#
 	addl	%edi, %eax		# += ((b | c) & d) | (b & c)
-	leal	-0x70e44324(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	-0x70E44324(%rax,%r11), %eax # e += RCONST + W[n & 15]
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %eax		# e += rotl32(a,5)
 	rorl	$2, %ecx		# b = rotl32(b,30)
 # 60
-	movl	%r9d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*4(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r14d, %esi		# ^W[(n+2) & 15]
-	xorl	%r12d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r12d		# store to W[n & 15]
+	xorl	%r9d, %r12d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*4(%rsp), %r12d	# ^W[(n+8) & 15]
+	xorl	%r14d, %r12d	# ^W[(n+2) & 15]
+	roll	%r12d		#
 	movl	%ecx, %edi		# c
 	xorl	%edx, %edi		# ^d
 	xorl	%ebx, %edi		# ^b
-	leal	-0x359d3e2a(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rbp,%r12), %ebp # e += RCONST + W[n & 15]
 	addl	%edi, %ebp		# e += (c ^ d ^ b)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebp		# e += rotl32(a,5)
 	rorl	$2, %ebx		# b = rotl32(b,30)
 # 61
-	movl	%r10d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*5(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r15d, %esi		# ^W[(n+2) & 15]
-	xorl	%r13d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r13d		# store to W[n & 15]
+	xorl	%r10d, %r13d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*5(%rsp), %r13d	# ^W[(n+8) & 15]
+	xorl	%r15d, %r13d	# ^W[(n+2) & 15]
+	roll	%r13d		#
 	movl	%ebx, %edi		# c
 	xorl	%ecx, %edi		# ^d
 	xorl	%eax, %edi		# ^b
-	leal	-0x359d3e2a(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rdx,%r13), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (c ^ d ^ b)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %edx		# e += rotl32(a,5)
 	rorl	$2, %eax		# b = rotl32(b,30)
 # 62
-	movl	%r11d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*6(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	-32+4*0(%rsp), %esi		# ^W[(n+2) & 15]
-	xorl	%r14d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r14d		# store to W[n & 15]
+	xorl	%r11d, %r14d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*6(%rsp), %r14d	# ^W[(n+8) & 15]
+	xorl	-32+4*0(%rsp), %r14d	# ^W[(n+2) & 15]
+	roll	%r14d		#
 	movl	%eax, %edi		# c
 	xorl	%ebx, %edi		# ^d
 	xorl	%ebp, %edi		# ^b
-	leal	-0x359d3e2a(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rcx,%r14), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (c ^ d ^ b)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ecx		# e += rotl32(a,5)
 	rorl	$2, %ebp		# b = rotl32(b,30)
 # 63
-	movl	%r12d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*7(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	-32+4*1(%rsp), %esi		# ^W[(n+2) & 15]
-	xorl	%r15d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r15d		# store to W[n & 15]
+	xorl	%r12d, %r15d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*7(%rsp), %r15d	# ^W[(n+8) & 15]
+	xorl	-32+4*1(%rsp), %r15d	# ^W[(n+2) & 15]
+	roll	%r15d		#
 	movl	%ebp, %edi		# c
 	xorl	%eax, %edi		# ^d
 	xorl	%edx, %edi		# ^b
-	leal	-0x359d3e2a(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rbx,%r15), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (c ^ d ^ b)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -1084,7 +1040,7 @@ sha1_process_block64:
 	movl	%edx, %edi		# c
 	xorl	%ebp, %edi		# ^d
 	xorl	%ecx, %edi		# ^b
-	leal	-0x359d3e2a(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rax,%rsi), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (c ^ d ^ b)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -1100,7 +1056,7 @@ sha1_process_block64:
 	movl	%ecx, %edi		# c
 	xorl	%edx, %edi		# ^d
 	xorl	%ebx, %edi		# ^b
-	leal	-0x359d3e2a(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rbp,%rsi), %ebp # e += RCONST + W[n & 15]
 	addl	%edi, %ebp		# e += (c ^ d ^ b)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -1116,7 +1072,7 @@ sha1_process_block64:
 	movl	%ebx, %edi		# c
 	xorl	%ecx, %edi		# ^d
 	xorl	%eax, %edi		# ^b
-	leal	-0x359d3e2a(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rdx,%rsi), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (c ^ d ^ b)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -1132,7 +1088,7 @@ sha1_process_block64:
 	movl	%eax, %edi		# c
 	xorl	%ebx, %edi		# ^d
 	xorl	%ebp, %edi		# ^b
-	leal	-0x359d3e2a(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rcx,%rsi), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (c ^ d ^ b)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -1148,7 +1104,7 @@ sha1_process_block64:
 	movl	%ebp, %edi		# c
 	xorl	%eax, %edi		# ^d
 	xorl	%edx, %edi		# ^b
-	leal	-0x359d3e2a(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rbx,%rsi), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (c ^ d ^ b)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -1164,7 +1120,7 @@ sha1_process_block64:
 	movl	%edx, %edi		# c
 	xorl	%ebp, %edi		# ^d
 	xorl	%ecx, %edi		# ^b
-	leal	-0x359d3e2a(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rax,%rsi), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (c ^ d ^ b)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -1180,7 +1136,7 @@ sha1_process_block64:
 	movl	%ecx, %edi		# c
 	xorl	%edx, %edi		# ^d
 	xorl	%ebx, %edi		# ^b
-	leal	-0x359d3e2a(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rbp,%rsi), %ebp # e += RCONST + W[n & 15]
 	addl	%edi, %ebp		# e += (c ^ d ^ b)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -1196,135 +1152,119 @@ sha1_process_block64:
 	movl	%ebx, %edi		# c
 	xorl	%ecx, %edi		# ^d
 	xorl	%eax, %edi		# ^b
-	leal	-0x359d3e2a(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rdx,%rsi), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (c ^ d ^ b)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %edx		# e += rotl32(a,5)
 	rorl	$2, %eax		# b = rotl32(b,30)
 # 72
-	movl	-32+4*5(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*0(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r10d, %esi		# ^W[(n+2) & 15]
-	xorl	%r8d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r8d		# store to W[n & 15]
+	xorl	-32+4*5(%rsp), %r8d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*0(%rsp), %r8d	# ^W[(n+8) & 15]
+	xorl	%r10d, %r8d	# ^W[(n+2) & 15]
+	roll	%r8d		#
 	movl	%eax, %edi		# c
 	xorl	%ebx, %edi		# ^d
 	xorl	%ebp, %edi		# ^b
-	leal	-0x359d3e2a(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rcx,%r8), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (c ^ d ^ b)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ecx		# e += rotl32(a,5)
 	rorl	$2, %ebp		# b = rotl32(b,30)
 # 73
-	movl	-32+4*6(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*1(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r11d, %esi		# ^W[(n+2) & 15]
-	xorl	%r9d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r9d		# store to W[n & 15]
+	xorl	-32+4*6(%rsp), %r9d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*1(%rsp), %r9d	# ^W[(n+8) & 15]
+	xorl	%r11d, %r9d	# ^W[(n+2) & 15]
+	roll	%r9d		#
 	movl	%ebp, %edi		# c
 	xorl	%eax, %edi		# ^d
 	xorl	%edx, %edi		# ^b
-	leal	-0x359d3e2a(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rbx,%r9), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (c ^ d ^ b)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebx		# e += rotl32(a,5)
 	rorl	$2, %edx		# b = rotl32(b,30)
 # 74
-	movl	-32+4*7(%rsp), %esi	# W[(n+13) & 15]
-	xorl	-32+4*2(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r12d, %esi		# ^W[(n+2) & 15]
-	xorl	%r10d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r10d		# store to W[n & 15]
+	xorl	-32+4*7(%rsp), %r10d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*2(%rsp), %r10d	# ^W[(n+8) & 15]
+	xorl	%r12d, %r10d	# ^W[(n+2) & 15]
+	roll	%r10d		#
 	movl	%edx, %edi		# c
 	xorl	%ebp, %edi		# ^d
 	xorl	%ecx, %edi		# ^b
-	leal	-0x359d3e2a(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rax,%r10), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (c ^ d ^ b)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %eax		# e += rotl32(a,5)
 	rorl	$2, %ecx		# b = rotl32(b,30)
 # 75
-	movl	%r8d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*3(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r13d, %esi		# ^W[(n+2) & 15]
-	xorl	%r11d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r11d		# store to W[n & 15]
+	xorl	%r8d, %r11d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*3(%rsp), %r11d	# ^W[(n+8) & 15]
+	xorl	%r13d, %r11d	# ^W[(n+2) & 15]
+	roll	%r11d		#
 	movl	%ecx, %edi		# c
 	xorl	%edx, %edi		# ^d
 	xorl	%ebx, %edi		# ^b
-	leal	-0x359d3e2a(%rbp,%rsi), %ebp # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rbp,%r11), %ebp # e += RCONST + W[n & 15]
 	addl	%edi, %ebp		# e += (c ^ d ^ b)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebp		# e += rotl32(a,5)
 	rorl	$2, %ebx		# b = rotl32(b,30)
 # 76
-	movl	%r9d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*4(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r14d, %esi		# ^W[(n+2) & 15]
-	xorl	%r12d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	movl	%esi, %r12d		# store to W[n & 15]
+	xorl	%r9d, %r12d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*4(%rsp), %r12d	# ^W[(n+8) & 15]
+	xorl	%r14d, %r12d	# ^W[(n+2) & 15]
+	roll	%r12d		#
 	movl	%ebx, %edi		# c
 	xorl	%ecx, %edi		# ^d
 	xorl	%eax, %edi		# ^b
-	leal	-0x359d3e2a(%rdx,%rsi), %edx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rdx,%r12), %edx # e += RCONST + W[n & 15]
 	addl	%edi, %edx		# e += (c ^ d ^ b)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %edx		# e += rotl32(a,5)
 	rorl	$2, %eax		# b = rotl32(b,30)
 # 77
-	movl	%r10d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*5(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	%r15d, %esi		# ^W[(n+2) & 15]
-	xorl	%r13d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	# store to W[n & 15] - unused, not done
+	xorl	%r10d, %r13d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*5(%rsp), %r13d	# ^W[(n+8) & 15]
+	xorl	%r15d, %r13d	# ^W[(n+2) & 15]
+	roll	%r13d		#
 	movl	%eax, %edi		# c
 	xorl	%ebx, %edi		# ^d
 	xorl	%ebp, %edi		# ^b
-	leal	-0x359d3e2a(%rcx,%rsi), %ecx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rcx,%r13), %ecx # e += RCONST + W[n & 15]
 	addl	%edi, %ecx		# e += (c ^ d ^ b)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ecx		# e += rotl32(a,5)
 	rorl	$2, %ebp		# b = rotl32(b,30)
 # 78
-	movl	%r11d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*6(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	-32+4*0(%rsp), %esi		# ^W[(n+2) & 15]
-	xorl	%r14d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	# store to W[n & 15] - unused, not done
+	xorl	%r11d, %r14d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*6(%rsp), %r14d	# ^W[(n+8) & 15]
+	xorl	-32+4*0(%rsp), %r14d	# ^W[(n+2) & 15]
+	roll	%r14d		#
 	movl	%ebp, %edi		# c
 	xorl	%eax, %edi		# ^d
 	xorl	%edx, %edi		# ^b
-	leal	-0x359d3e2a(%rbx,%rsi), %ebx # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rbx,%r14), %ebx # e += RCONST + W[n & 15]
 	addl	%edi, %ebx		# e += (c ^ d ^ b)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
 	addl	%esi, %ebx		# e += rotl32(a,5)
 	rorl	$2, %edx		# b = rotl32(b,30)
 # 79
-	movl	%r12d, %esi	# W[(n+13) & 15]
-	xorl	-32+4*7(%rsp), %esi		# ^W[(n+8) & 15]
-	xorl	-32+4*1(%rsp), %esi		# ^W[(n+2) & 15]
-	xorl	%r15d, %esi		# ^W[n & 15]
-	roll	%esi			#
-	# store to W[n & 15] - unused, not done
+	xorl	%r12d, %r15d	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	-32+4*7(%rsp), %r15d	# ^W[(n+8) & 15]
+	xorl	-32+4*1(%rsp), %r15d	# ^W[(n+2) & 15]
+	roll	%r15d		#
 	movl	%edx, %edi		# c
 	xorl	%ebp, %edi		# ^d
 	xorl	%ecx, %edi		# ^b
-	leal	-0x359d3e2a(%rax,%rsi), %eax # e += RCONST + mixed_W
+	leal	-0x359D3E2A(%rax,%r15), %eax # e += RCONST + W[n & 15]
 	addl	%edi, %eax		# e += (c ^ d ^ b)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
diff --git a/libbb/hash_md5_sha_x86-64.S.sh b/libbb/hash_md5_sha_x86-64.S.sh
new file mode 100755
index 0000000..931c0f0
--- /dev/null
+++ b/libbb/hash_md5_sha_x86-64.S.sh
@@ -0,0 +1,267 @@
+#!/bin/sh
+
+# We don't regenerate it on every "make" invocation - only by hand.
+# The reason is that the changes to generated code are difficult
+# to visualize by looking only at this script, it helps when the commit
+# also contains the diff of the generated file.
+exec >hash_md5_sha_x86-64.S
+
+echo \
+'### Generated by hash_md5_sha_x86-64.S.sh ###
+
+#if CONFIG_SHA1_SMALL == 0 && defined(__GNUC__) && defined(__x86_64__)
+	.section	.text.sha1_process_block64,"ax",@progbits
+	.globl  sha1_process_block64
+	.hidden sha1_process_block64
+	.type	sha1_process_block64, @function
+
+	.balign	8	# allow decoders to fetch at least 4 first insns
+sha1_process_block64:
+	pushq	%r15	#
+	pushq	%r14	#
+	pushq	%r13	#
+	pushq	%r12	#
+	pushq	%rbp	#
+	pushq	%rbx	#
+	pushq	%rdi	# we need ctx at the end
+
+#Register and stack use:
+# eax..edx: a..d
+# ebp: e
+# esi,edi: temps
+# -32+4*n(%rsp),r8...r15: W[0..7,8..15]
+# (TODO: actually W[0..7] are used a bit more often, put _thme_ into r8..r15?)
+
+	movq	4*8(%rdi), %r8
+	bswapq	%r8
+	movl	%r8d, %r9d
+	shrq	$32, %r8
+	movq	4*10(%rdi), %r10
+	bswapq	%r10
+	movl	%r10d, %r11d
+	shrq	$32, %r10
+	movq	4*12(%rdi), %r12
+	bswapq	%r12
+	movl	%r12d, %r13d
+	shrq	$32, %r12
+	movq	4*14(%rdi), %r14
+	bswapq	%r14
+	movl	%r14d, %r15d
+	shrq	$32, %r14
+
+	movl	$3, %eax
+1:
+	movq	(%rdi,%rax,8), %rsi
+	bswapq	%rsi
+	rolq	$32, %rsi
+	movq	%rsi, -32(%rsp,%rax,8)
+	decl	%eax
+	jns	1b
+	movl	80(%rdi), %eax		# a = ctx->hash[0]
+	movl	84(%rdi), %ebx		# b = ctx->hash[1]
+	movl	88(%rdi), %ecx		# c = ctx->hash[2]
+	movl	92(%rdi), %edx		# d = ctx->hash[3]
+	movl	96(%rdi), %ebp		# e = ctx->hash[4]
+'
+W32() {
+test "$1" || exit 1
+test "$1" -lt 0 && exit 1
+test "$1" -gt 15 && exit 1
+test "$1" -lt 8 && echo "-32+4*$1(%rsp)"
+test "$1" -ge 8 && echo "%r${1}d"
+}
+
+RD1A() {
+local a=$1;local b=$2;local c=$3;local d=$4;local e=$5
+local n=$(($6))
+echo "# $n"
+test $n = 0 && echo "
+	# W[0], already in %esi
+";test $n != 0 && test $n -lt 8 && echo "
+	movl	`W32 $n`, %esi		# W[n]
+";test $n -ge 8 && echo "
+	# W[n], in %r$n
+";echo "
+	movl	%e$c, %edi		# c
+	xorl	%e$d, %edi		# ^d
+	andl	%e$b, %edi		# &b
+	xorl	%e$d, %edi		# (((c ^ d) & b) ^ d)
+";test $n -lt 8 && echo "
+	leal	$RCONST(%r$e,%rsi),%e$e # e += RCONST + W[n]
+";test $n -ge 8 && echo "
+	leal	$RCONST(%r$e,%r$n),%e$e # e += RCONST + W[n]
+";echo "
+	addl	%edi, %e$e		# e += (((c ^ d) & b) ^ d)
+	movl	%e$a, %esi		#
+	roll	\$5, %esi		# rotl32(a,5)
+	addl	%esi, %e$e		# e += rotl32(a,5)
+	rorl	\$2, %e$b		# b = rotl32(b,30)
+"
+}
+RD1B() {
+local a=$1;local b=$2;local c=$3;local d=$4;local e=$5
+local n=$(($6))
+local n13=$(((n+13) & 15))
+local n8=$(((n+8) & 15))
+local n2=$(((n+2) & 15))
+local n0=$(((n+0) & 15))
+echo "
+# $n
+";test $n0 -lt 8 && echo "
+	movl	`W32 $n13`, %esi	# W[(n+13) & 15]
+	xorl	`W32 $n8`, %esi		# ^W[(n+8) & 15]
+	xorl	`W32 $n2`, %esi		# ^W[(n+2) & 15]
+	xorl	`W32 $n0`, %esi		# ^W[n & 15]
+	roll	%esi			#
+	movl	%esi, `W32 $n0`		# store to W[n & 15]
+";test $n0 -ge 8 && echo "
+	xorl	`W32 $n13`, `W32 $n0`	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	`W32 $n8`, `W32 $n0`	# ^W[(n+8) & 15]
+	xorl	`W32 $n2`, `W32 $n0`	# ^W[(n+2) & 15]
+	roll	`W32 $n0`		#
+"; echo "
+	movl	%e$c, %edi		# c
+	xorl	%e$d, %edi		# ^d
+	andl	%e$b, %edi		# &b
+	xorl	%e$d, %edi		# (((c ^ d) & b) ^ d)
+";test $n0 -lt 8 && echo "
+	leal	$RCONST(%r$e,%rsi), %e$e # e += RCONST + W[n & 15]
+";test $n0 -ge 8 && echo "
+	leal	$RCONST(%r$e,%r$n0), %e$e # e += RCONST + W[n & 15]
+";echo "
+	addl	%edi, %e$e		# e += (((c ^ d) & b) ^ d)
+	movl	%e$a, %esi		#
+	roll	\$5, %esi		# rotl32(a,5)
+	addl	%esi, %e$e		# e += rotl32(a,5)
+	rorl	\$2, %e$b		# b = rotl32(b,30)
+"
+}
+{
+RCONST=0x5A827999
+RD1A ax bx cx dx bp  0; RD1A bp ax bx cx dx  1; RD1A dx bp ax bx cx  2; RD1A cx dx bp ax bx  3; RD1A bx cx dx bp ax  4
+RD1A ax bx cx dx bp  5; RD1A bp ax bx cx dx  6; RD1A dx bp ax bx cx  7; RD1A cx dx bp ax bx  8; RD1A bx cx dx bp ax  9
+RD1A ax bx cx dx bp 10; RD1A bp ax bx cx dx 11; RD1A dx bp ax bx cx 12; RD1A cx dx bp ax bx 13; RD1A bx cx dx bp ax 14
+RD1A ax bx cx dx bp 15; RD1B bp ax bx cx dx 16; RD1B dx bp ax bx cx 17; RD1B cx dx bp ax bx 18; RD1B bx cx dx bp ax 19
+} | grep -v '^$'
+
+RD2() {
+local a=$1;local b=$2;local c=$3;local d=$4;local e=$5
+local n=$(($6))
+local n13=$(((n+13) & 15))
+local n8=$(((n+8) & 15))
+local n2=$(((n+2) & 15))
+local n0=$(((n+0) & 15))
+echo "
+# $n
+";test $n0 -lt 8 && echo "
+	movl	`W32 $n13`, %esi	# W[(n+13) & 15]
+	xorl	`W32 $n8`, %esi		# ^W[(n+8) & 15]
+	xorl	`W32 $n2`, %esi		# ^W[(n+2) & 15]
+	xorl	`W32 $n0`, %esi		# ^W[n & 15]
+	roll	%esi			#
+	movl	%esi, `W32 $n0`		# store to W[n & 15]
+";test $n0 -ge 8 && echo "
+	xorl	`W32 $n13`, `W32 $n0`	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	`W32 $n8`, `W32 $n0`	# ^W[(n+8) & 15]
+	xorl	`W32 $n2`, `W32 $n0`	# ^W[(n+2) & 15]
+	roll	`W32 $n0`		#
+"; echo "
+	movl	%e$c, %edi		# c
+	xorl	%e$d, %edi		# ^d
+	xorl	%e$b, %edi		# ^b
+";test $n0 -lt 8 && echo "
+	leal	$RCONST(%r$e,%rsi), %e$e # e += RCONST + W[n & 15]
+";test $n0 -ge 8 && echo "
+	leal	$RCONST(%r$e,%r$n0), %e$e # e += RCONST + W[n & 15]
+";echo "
+	addl	%edi, %e$e		# e += (c ^ d ^ b)
+	movl	%e$a, %esi		#
+	roll	\$5, %esi		# rotl32(a,5)
+	addl	%esi, %e$e		# e += rotl32(a,5)
+	rorl	\$2, %e$b		# b = rotl32(b,30)
+"
+}
+{
+RCONST=0x6ED9EBA1
+RD2 ax bx cx dx bp 20; RD2 bp ax bx cx dx 21; RD2 dx bp ax bx cx 22; RD2 cx dx bp ax bx 23; RD2 bx cx dx bp ax 24
+RD2 ax bx cx dx bp 25; RD2 bp ax bx cx dx 26; RD2 dx bp ax bx cx 27; RD2 cx dx bp ax bx 28; RD2 bx cx dx bp ax 29
+RD2 ax bx cx dx bp 30; RD2 bp ax bx cx dx 31; RD2 dx bp ax bx cx 32; RD2 cx dx bp ax bx 33; RD2 bx cx dx bp ax 34
+RD2 ax bx cx dx bp 35; RD2 bp ax bx cx dx 36; RD2 dx bp ax bx cx 37; RD2 cx dx bp ax bx 38; RD2 bx cx dx bp ax 39
+} | grep -v '^$'
+
+RD3() {
+local a=$1;local b=$2;local c=$3;local d=$4;local e=$5
+local n=$(($6))
+local n13=$(((n+13) & 15))
+local n8=$(((n+8) & 15))
+local n2=$(((n+2) & 15))
+local n0=$(((n+0) & 15))
+echo "
+# $n
+	movl	%e$b, %edi		# di: b
+	movl	%e$b, %esi		# si: b
+	orl	%e$c, %edi		# di: b | c
+	andl	%e$c, %esi		# si: b & c
+	andl	%e$d, %edi		# di: (b | c) & d
+	orl	%esi, %edi		# ((b | c) & d) | (b & c)
+";test $n0 -lt 8 && echo "
+	movl	`W32 $n13`, %esi	# W[(n+13) & 15]
+	xorl	`W32 $n8`, %esi		# ^W[(n+8) & 15]
+	xorl	`W32 $n2`, %esi		# ^W[(n+2) & 15]
+	xorl	`W32 $n0`, %esi		# ^W[n & 15]
+	roll	%esi			#
+	movl	%esi, `W32 $n0`		# store to W[n & 15]
+";test $n0 -ge 8 && echo "
+	xorl	`W32 $n13`, `W32 $n0`	# W[n & 15] ^= W[(n+13) & 15]
+	xorl	`W32 $n8`, `W32 $n0`	# ^W[(n+8) & 15]
+	xorl	`W32 $n2`, `W32 $n0`	# ^W[(n+2) & 15]
+	roll	`W32 $n0`		#
+"; echo "
+	addl	%edi, %e$e		# += ((b | c) & d) | (b & c)
+";test $n0 -lt 8 && echo "
+	leal	$RCONST(%r$e,%rsi), %e$e # e += RCONST + W[n & 15]
+";test $n0 -ge 8 && echo "
+	leal	$RCONST(%r$e,%r$n0), %e$e # e += RCONST + W[n & 15]
+";echo "
+	movl	%e$a, %esi		#
+	roll	\$5, %esi		# rotl32(a,5)
+	addl	%esi, %e$e		# e += rotl32(a,5)
+	rorl	\$2, %e$b		# b = rotl32(b,30)
+"
+}
+{
+#RCONST=0x8F1BBCDC "out of range for signed 32bit displacement"
+RCONST=-0x70E44324
+RD3 ax bx cx dx bp 40; RD3 bp ax bx cx dx 41; RD3 dx bp ax bx cx 42; RD3 cx dx bp ax bx 43; RD3 bx cx dx bp ax 44
+RD3 ax bx cx dx bp 45; RD3 bp ax bx cx dx 46; RD3 dx bp ax bx cx 47; RD3 cx dx bp ax bx 48; RD3 bx cx dx bp ax 49
+RD3 ax bx cx dx bp 50; RD3 bp ax bx cx dx 51; RD3 dx bp ax bx cx 52; RD3 cx dx bp ax bx 53; RD3 bx cx dx bp ax 54
+RD3 ax bx cx dx bp 55; RD3 bp ax bx cx dx 56; RD3 dx bp ax bx cx 57; RD3 cx dx bp ax bx 58; RD3 bx cx dx bp ax 59
+} | grep -v '^$'
+
+# Round 4 has the same logic as round 2, only n and RCONST are different
+{
+#RCONST=0xCA62C1D6 "out of range for signed 32bit displacement"
+RCONST=-0x359D3E2A
+RD2 ax bx cx dx bp 60; RD2 bp ax bx cx dx 61; RD2 dx bp ax bx cx 62; RD2 cx dx bp ax bx 63; RD2 bx cx dx bp ax 64
+RD2 ax bx cx dx bp 65; RD2 bp ax bx cx dx 66; RD2 dx bp ax bx cx 67; RD2 cx dx bp ax bx 68; RD2 bx cx dx bp ax 69
+RD2 ax bx cx dx bp 70; RD2 bp ax bx cx dx 71; RD2 dx bp ax bx cx 72; RD2 cx dx bp ax bx 73; RD2 bx cx dx bp ax 74
+RD2 ax bx cx dx bp 75; RD2 bp ax bx cx dx 76; RD2 dx bp ax bx cx 77; RD2 cx dx bp ax bx 78; RD2 bx cx dx bp ax 79
+} | grep -v '^$'
+
+echo "
+	popq	%rdi		#
+	addl	%eax, 80(%rdi)  # ctx->hash[0] += a
+	addl	%ebx, 84(%rdi)  # ctx->hash[1] += b
+	addl	%ecx, 88(%rdi)  # ctx->hash[2] += c
+	addl	%edx, 92(%rdi)  # ctx->hash[3] += d
+	addl	%ebp, 96(%rdi)  # ctx->hash[4] += e
+	popq	%rbx		#
+	popq	%rbp		#
+	popq	%r12		#
+	popq	%r13		#
+	popq	%r14		#
+	popq	%r15		#
+
+	ret
+	.size	sha1_process_block64, .-sha1_process_block64
+#endif"