diff --git a/Makefile.am b/Makefile.am
index c863232..8950273 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -15,7 +15,7 @@ bin_PROGRAMS	= minerd
 
 minerd_SOURCES	= elist.h miner.h compat.h \
 		  cpu-miner.c util.c \
-		  sha2.c sha2-x64.S \
+		  sha2.c sha2-x86.S sha2-x64.S \
 		  scrypt.c scrypt-x86.S scrypt-x64.S
 minerd_LDFLAGS	= $(PTHREAD_FLAGS)
 minerd_LDADD	= @LIBCURL@ @JANSSON_LIBS@ @PTHREAD_LIBS@ @WS2_LIBS@
diff --git a/miner.h b/miner.h
index cbf18f8..31f464d 100644
--- a/miner.h
+++ b/miner.h
@@ -112,7 +112,7 @@ static inline void le32enc(void *pp, uint32_t x)
 void sha256_init(uint32_t *state);
 void sha256_transform(uint32_t *state, const uint32_t *block, int swap);
 
-#if defined(__x86_64__)
+#if defined(__i386__) || defined(__x86_64__)
 #define HAVE_SHA256_4WAY 1
 int sha256_use_4way();
 void sha256_init_4way(uint32_t *state);
diff --git a/sha2-x86.S b/sha2-x86.S
new file mode 100644
index 0000000..0f55ba0
--- /dev/null
+++ b/sha2-x86.S
@@ -0,0 +1,1196 @@
+/*
+ * Copyright 2012 pooler@litecoinpool.org
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.  See COPYING for more details.
+ */
+
+#include "cpuminer-config.h"
+
+#if defined(__linux__) && defined(__ELF__)
+	.section .note.GNU-stack,"",%progbits
+#endif
+
+#if defined(__i386__)
+
+	.data
+	.p2align 7
+sha256_4h:
+	.long 0x6a09e667, 0x6a09e667, 0x6a09e667, 0x6a09e667
+	.long 0xbb67ae85, 0xbb67ae85, 0xbb67ae85, 0xbb67ae85
+	.long 0x3c6ef372, 0x3c6ef372, 0x3c6ef372, 0x3c6ef372
+	.long 0xa54ff53a, 0xa54ff53a, 0xa54ff53a, 0xa54ff53a
+	.long 0x510e527f, 0x510e527f, 0x510e527f, 0x510e527f
+	.long 0x9b05688c, 0x9b05688c, 0x9b05688c, 0x9b05688c
+	.long 0x1f83d9ab, 0x1f83d9ab, 0x1f83d9ab, 0x1f83d9ab
+	.long 0x5be0cd19, 0x5be0cd19, 0x5be0cd19, 0x5be0cd19
+
+	.data
+	.p2align 7
+sha256_4k:
+	.long 0x428a2f98, 0x428a2f98, 0x428a2f98, 0x428a2f98
+	.long 0x71374491, 0x71374491, 0x71374491, 0x71374491
+	.long 0xb5c0fbcf, 0xb5c0fbcf, 0xb5c0fbcf, 0xb5c0fbcf
+	.long 0xe9b5dba5, 0xe9b5dba5, 0xe9b5dba5, 0xe9b5dba5
+	.long 0x3956c25b, 0x3956c25b, 0x3956c25b, 0x3956c25b
+	.long 0x59f111f1, 0x59f111f1, 0x59f111f1, 0x59f111f1
+	.long 0x923f82a4, 0x923f82a4, 0x923f82a4, 0x923f82a4
+	.long 0xab1c5ed5, 0xab1c5ed5, 0xab1c5ed5, 0xab1c5ed5
+	.long 0xd807aa98, 0xd807aa98, 0xd807aa98, 0xd807aa98
+	.long 0x12835b01, 0x12835b01, 0x12835b01, 0x12835b01
+	.long 0x243185be, 0x243185be, 0x243185be, 0x243185be
+	.long 0x550c7dc3, 0x550c7dc3, 0x550c7dc3, 0x550c7dc3
+	.long 0x72be5d74, 0x72be5d74, 0x72be5d74, 0x72be5d74
+	.long 0x80deb1fe, 0x80deb1fe, 0x80deb1fe, 0x80deb1fe
+	.long 0x9bdc06a7, 0x9bdc06a7, 0x9bdc06a7, 0x9bdc06a7
+	.long 0xc19bf174, 0xc19bf174, 0xc19bf174, 0xc19bf174
+	.long 0xe49b69c1, 0xe49b69c1, 0xe49b69c1, 0xe49b69c1
+	.long 0xefbe4786, 0xefbe4786, 0xefbe4786, 0xefbe4786
+	.long 0x0fc19dc6, 0x0fc19dc6, 0x0fc19dc6, 0x0fc19dc6
+	.long 0x240ca1cc, 0x240ca1cc, 0x240ca1cc, 0x240ca1cc
+	.long 0x2de92c6f, 0x2de92c6f, 0x2de92c6f, 0x2de92c6f
+	.long 0x4a7484aa, 0x4a7484aa, 0x4a7484aa, 0x4a7484aa
+	.long 0x5cb0a9dc, 0x5cb0a9dc, 0x5cb0a9dc, 0x5cb0a9dc
+	.long 0x76f988da, 0x76f988da, 0x76f988da, 0x76f988da
+	.long 0x983e5152, 0x983e5152, 0x983e5152, 0x983e5152
+	.long 0xa831c66d, 0xa831c66d, 0xa831c66d, 0xa831c66d
+	.long 0xb00327c8, 0xb00327c8, 0xb00327c8, 0xb00327c8
+	.long 0xbf597fc7, 0xbf597fc7, 0xbf597fc7, 0xbf597fc7
+	.long 0xc6e00bf3, 0xc6e00bf3, 0xc6e00bf3, 0xc6e00bf3
+	.long 0xd5a79147, 0xd5a79147, 0xd5a79147, 0xd5a79147
+	.long 0x06ca6351, 0x06ca6351, 0x06ca6351, 0x06ca6351
+	.long 0x14292967, 0x14292967, 0x14292967, 0x14292967
+	.long 0x27b70a85, 0x27b70a85, 0x27b70a85, 0x27b70a85
+	.long 0x2e1b2138, 0x2e1b2138, 0x2e1b2138, 0x2e1b2138
+	.long 0x4d2c6dfc, 0x4d2c6dfc, 0x4d2c6dfc, 0x4d2c6dfc
+	.long 0x53380d13, 0x53380d13, 0x53380d13, 0x53380d13
+	.long 0x650a7354, 0x650a7354, 0x650a7354, 0x650a7354
+	.long 0x766a0abb, 0x766a0abb, 0x766a0abb, 0x766a0abb
+	.long 0x81c2c92e, 0x81c2c92e, 0x81c2c92e, 0x81c2c92e
+	.long 0x92722c85, 0x92722c85, 0x92722c85, 0x92722c85
+	.long 0xa2bfe8a1, 0xa2bfe8a1, 0xa2bfe8a1, 0xa2bfe8a1
+	.long 0xa81a664b, 0xa81a664b, 0xa81a664b, 0xa81a664b
+	.long 0xc24b8b70, 0xc24b8b70, 0xc24b8b70, 0xc24b8b70
+	.long 0xc76c51a3, 0xc76c51a3, 0xc76c51a3, 0xc76c51a3
+	.long 0xd192e819, 0xd192e819, 0xd192e819, 0xd192e819
+	.long 0xd6990624, 0xd6990624, 0xd6990624, 0xd6990624
+	.long 0xf40e3585, 0xf40e3585, 0xf40e3585, 0xf40e3585
+	.long 0x106aa070, 0x106aa070, 0x106aa070, 0x106aa070
+	.long 0x19a4c116, 0x19a4c116, 0x19a4c116, 0x19a4c116
+	.long 0x1e376c08, 0x1e376c08, 0x1e376c08, 0x1e376c08
+	.long 0x2748774c, 0x2748774c, 0x2748774c, 0x2748774c
+	.long 0x34b0bcb5, 0x34b0bcb5, 0x34b0bcb5, 0x34b0bcb5
+	.long 0x391c0cb3, 0x391c0cb3, 0x391c0cb3, 0x391c0cb3
+	.long 0x4ed8aa4a, 0x4ed8aa4a, 0x4ed8aa4a, 0x4ed8aa4a
+	.long 0x5b9cca4f, 0x5b9cca4f, 0x5b9cca4f, 0x5b9cca4f
+	.long 0x682e6ff3, 0x682e6ff3, 0x682e6ff3, 0x682e6ff3
+	.long 0x748f82ee, 0x748f82ee, 0x748f82ee, 0x748f82ee
+	.long 0x78a5636f, 0x78a5636f, 0x78a5636f, 0x78a5636f
+	.long 0x84c87814, 0x84c87814, 0x84c87814, 0x84c87814
+	.long 0x8cc70208, 0x8cc70208, 0x8cc70208, 0x8cc70208
+	.long 0x90befffa, 0x90befffa, 0x90befffa, 0x90befffa
+	.long 0xa4506ceb, 0xa4506ceb, 0xa4506ceb, 0xa4506ceb
+	.long 0xbef9a3f7, 0xbef9a3f7, 0xbef9a3f7, 0xbef9a3f7
+	.long 0xc67178f2, 0xc67178f2, 0xc67178f2, 0xc67178f2
+
+	.data
+	.p2align 6
+sha256d_4preext2_15:
+	.long 0x00000100, 0x00000100, 0x00000100, 0x00000100
+sha256d_4preext2_17:
+	.long 0x00a00000, 0x00a00000, 0x00a00000, 0x00a00000
+sha256d_4preext2_23:
+	.long 0x11002000, 0x11002000, 0x11002000, 0x11002000
+sha256d_4preext2_24:
+	.long 0x80000000, 0x80000000, 0x80000000, 0x80000000
+sha256d_4preext2_30:
+	.long 0x00400022, 0x00400022, 0x00400022, 0x00400022
+
+
+	.text
+	.p2align 5
+	.globl sha256_init_4way
+	.globl _sha256_init_4way
+sha256_init_4way:
+_sha256_init_4way:
+	pushl	%edi
+	movl	8(%esp), %edi
+	movdqa	sha256_4h+0, %xmm0
+	movdqa	sha256_4h+16, %xmm1
+	movdqa	sha256_4h+32, %xmm2
+	movdqa	sha256_4h+48, %xmm3
+	movdqu	%xmm0, 0(%edi)
+	movdqu	%xmm1, 16(%edi)
+	movdqu	%xmm2, 32(%edi)
+	movdqu	%xmm3, 48(%edi)
+	movdqa	sha256_4h+64, %xmm0
+	movdqa	sha256_4h+80, %xmm1
+	movdqa	sha256_4h+96, %xmm2
+	movdqa	sha256_4h+112, %xmm3
+	movdqu	%xmm0, 64(%edi)
+	movdqu	%xmm1, 80(%edi)
+	movdqu	%xmm2, 96(%edi)
+	movdqu	%xmm3, 112(%edi)
+	popl	%edi
+	ret
+
+
+.macro sha256_sse2_extend_round i
+	movdqa	(\i-15)*16(%eax), %xmm0
+	movdqa	%xmm0, %xmm2
+	psrld	$3, %xmm0
+	movdqa	%xmm0, %xmm1
+	pslld	$14, %xmm2
+	psrld	$4, %xmm1
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	psrld	$11, %xmm1
+	pslld	$11, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	paddd	(\i-16)*16(%eax), %xmm0
+	paddd	(\i-7)*16(%eax), %xmm0
+
+	movdqa	%xmm3, %xmm2
+	psrld	$10, %xmm3
+	pslld	$13, %xmm2
+	movdqa	%xmm3, %xmm1
+	psrld	$7, %xmm1
+	pxor	%xmm1, %xmm3
+	pxor	%xmm2, %xmm3
+	psrld	$2, %xmm1
+	pslld	$2, %xmm2
+	pxor	%xmm1, %xmm3
+	pxor	%xmm2, %xmm3
+	paddd	%xmm0, %xmm3
+	movdqa	%xmm3, \i*16(%eax)
+.endm
+
+.macro sha256_sse2_extend_doubleround i
+	movdqa	(\i-15)*16(%eax), %xmm0
+	movdqa	(\i-14)*16(%eax), %xmm4
+	movdqa	%xmm0, %xmm2
+	movdqa	%xmm4, %xmm6
+	psrld	$3, %xmm0
+	psrld	$3, %xmm4
+	movdqa	%xmm0, %xmm1
+	movdqa	%xmm4, %xmm5
+	pslld	$14, %xmm2
+	pslld	$14, %xmm6
+	psrld	$4, %xmm1
+	psrld	$4, %xmm5
+	pxor	%xmm1, %xmm0
+	pxor	%xmm5, %xmm4
+	psrld	$11, %xmm1
+	psrld	$11, %xmm5
+	pxor	%xmm2, %xmm0
+	pxor	%xmm6, %xmm4
+	pslld	$11, %xmm2
+	pslld	$11, %xmm6
+	pxor	%xmm1, %xmm0
+	pxor	%xmm5, %xmm4
+	pxor	%xmm2, %xmm0
+	pxor	%xmm6, %xmm4
+
+	paddd	(\i-16)*16(%eax), %xmm0
+	paddd	(\i-15)*16(%eax), %xmm4
+
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+
+	paddd	(\i-7)*16(%eax), %xmm0
+	paddd	(\i-6)*16(%eax), %xmm4
+
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+
+	paddd	%xmm0, %xmm3
+	paddd	%xmm4, %xmm7
+	movdqa	%xmm3, \i*16(%eax)
+	movdqa	%xmm7, (\i+1)*16(%eax)
+.endm
+
+.macro sha256_sse2_main_round i
+	movdqa	16*(\i)(%eax), %xmm6
+	paddd	16*(\i)+sha256_4k, %xmm6
+	paddd	32(%esp), %xmm6
+
+	movdqa	%xmm0, %xmm1
+	movdqa	16(%esp), %xmm2
+	pandn	%xmm2, %xmm1
+
+	movdqa	%xmm2, 32(%esp)
+	movdqa	0(%esp), %xmm2
+	movdqa	%xmm2, 16(%esp)
+
+	pand	%xmm0, %xmm2
+	pxor	%xmm2, %xmm1
+	movdqa	%xmm0, 0(%esp)
+
+	paddd	%xmm1, %xmm6
+
+	movdqa	%xmm0, %xmm1
+	psrld	$6, %xmm0
+	movdqa	%xmm0, %xmm2
+	pslld	$7, %xmm1
+	psrld	$5, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	pslld	$14, %xmm1
+	psrld	$14, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	pslld	$5, %xmm1
+	pxor	%xmm1, %xmm0
+	paddd	%xmm0, %xmm6
+
+	movdqa	%xmm3, %xmm0
+	paddd	%xmm6, %xmm0
+
+	movdqa	%xmm5, %xmm1
+	movdqa	%xmm4, %xmm3
+	movdqa	%xmm4, %xmm2
+	pand	%xmm5, %xmm2
+	pand	%xmm7, %xmm4
+	pand	%xmm7, %xmm1
+	pxor	%xmm4, %xmm1
+	movdqa	%xmm5, %xmm4
+	movdqa	%xmm7, %xmm5
+	pxor	%xmm2, %xmm1
+	paddd	%xmm1, %xmm6
+
+	movdqa	%xmm7, %xmm2
+	psrld	$2, %xmm7
+	movdqa	%xmm7, %xmm1
+	pslld	$10, %xmm2
+	psrld	$11, %xmm1
+	pxor	%xmm2, %xmm7
+	pxor	%xmm1, %xmm7
+	pslld	$9, %xmm2
+	psrld	$9, %xmm1
+	pxor	%xmm2, %xmm7
+	pxor	%xmm1, %xmm7
+	pslld	$11, %xmm2
+	pxor	%xmm2, %xmm7
+	paddd	%xmm6, %xmm7
+.endm
+
+.macro sha256_sse2_main_quadround i
+	sha256_sse2_main_round \i+0
+	sha256_sse2_main_round \i+1
+	sha256_sse2_main_round \i+2
+	sha256_sse2_main_round \i+3
+.endm
+
+
+.macro p2bswap_esi_esp i
+	movdqu	\i*16(%esi), %xmm0
+	movdqu	(\i+1)*16(%esi), %xmm2
+	pshuflw	$0xb1, %xmm0, %xmm0
+	pshuflw	$0xb1, %xmm2, %xmm2
+	pshufhw	$0xb1, %xmm0, %xmm0
+	pshufhw	$0xb1, %xmm2, %xmm2
+	movdqa	%xmm0, %xmm1
+	movdqa	%xmm2, %xmm3
+	psrlw	$8, %xmm1
+	psrlw	$8, %xmm3
+	psllw	$8, %xmm0
+	psllw	$8, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm3, %xmm2
+	movdqa	%xmm0, (\i+3)*16(%esp)
+	movdqa	%xmm2, (\i+4)*16(%esp)
+.endm
+
+	.text
+	.p2align 5
+	.globl sha256_transform_4way
+	.globl _sha256_transform_4way
+sha256_transform_4way:
+_sha256_transform_4way:
+	pushl	%edi
+	pushl	%esi
+	movl	12(%esp), %edi
+	movl	16(%esp), %esi
+	movl	20(%esp), %ecx
+	movl	%esp, %edx
+	subl	$67*16, %esp
+	andl	$-128, %esp
+	
+	testl	%ecx, %ecx
+	jnz sha256_transform_4way_swap
+	
+	movdqu	0*16(%esi), %xmm0
+	movdqu	1*16(%esi), %xmm1
+	movdqu	2*16(%esi), %xmm2
+	movdqu	3*16(%esi), %xmm3
+	movdqu	4*16(%esi), %xmm4
+	movdqu	5*16(%esi), %xmm5
+	movdqu	6*16(%esi), %xmm6
+	movdqu	7*16(%esi), %xmm7
+	movdqa	%xmm0, 3*16(%esp)
+	movdqa	%xmm1, 4*16(%esp)
+	movdqa	%xmm2, 5*16(%esp)
+	movdqa	%xmm3, 6*16(%esp)
+	movdqa	%xmm4, 7*16(%esp)
+	movdqa	%xmm5, 8*16(%esp)
+	movdqa	%xmm6, 9*16(%esp)
+	movdqa	%xmm7, 10*16(%esp)
+	movdqu	8*16(%esi), %xmm0
+	movdqu	9*16(%esi), %xmm1
+	movdqu	10*16(%esi), %xmm2
+	movdqu	11*16(%esi), %xmm3
+	movdqu	12*16(%esi), %xmm4
+	movdqu	13*16(%esi), %xmm5
+	movdqu	14*16(%esi), %xmm6
+	movdqu	15*16(%esi), %xmm7
+	movdqa	%xmm0, 11*16(%esp)
+	movdqa	%xmm1, 12*16(%esp)
+	movdqa	%xmm2, 13*16(%esp)
+	movdqa	%xmm3, 14*16(%esp)
+	movdqa	%xmm4, 15*16(%esp)
+	movdqa	%xmm5, 16*16(%esp)
+	movdqa	%xmm6, 17*16(%esp)
+	movdqa	%xmm7, 18*16(%esp)
+	jmp sha256_transform_4way_extend
+	
+	.p2align 5
+sha256_transform_4way_swap:
+	p2bswap_esi_esp 0
+	p2bswap_esi_esp 2
+	p2bswap_esi_esp 4
+	p2bswap_esi_esp 6
+	p2bswap_esi_esp 8
+	p2bswap_esi_esp 10
+	p2bswap_esi_esp 12
+	p2bswap_esi_esp 14
+	
+sha256_transform_4way_extend:
+	leal	19*16(%esp), %ecx
+	leal	48*16(%ecx), %eax
+	movdqa	-2*16(%ecx), %xmm3
+	movdqa	-1*16(%ecx), %xmm7
+sha256_transform_4way_extend_loop:
+	movdqa	-15*16(%ecx), %xmm0
+	movdqa	-14*16(%ecx), %xmm4
+	movdqa	%xmm0, %xmm2
+	movdqa	%xmm4, %xmm6
+	psrld	$3, %xmm0
+	psrld	$3, %xmm4
+	movdqa	%xmm0, %xmm1
+	movdqa	%xmm4, %xmm5
+	pslld	$14, %xmm2
+	pslld	$14, %xmm6
+	psrld	$4, %xmm1
+	psrld	$4, %xmm5
+	pxor	%xmm1, %xmm0
+	pxor	%xmm5, %xmm4
+	psrld	$11, %xmm1
+	psrld	$11, %xmm5
+	pxor	%xmm2, %xmm0
+	pxor	%xmm6, %xmm4
+	pslld	$11, %xmm2
+	pslld	$11, %xmm6
+	pxor	%xmm1, %xmm0
+	pxor	%xmm5, %xmm4
+	pxor	%xmm2, %xmm0
+	pxor	%xmm6, %xmm4
+
+	paddd	-16*16(%ecx), %xmm0
+	paddd	-15*16(%ecx), %xmm4
+
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+
+	paddd	-7*16(%ecx), %xmm0
+	paddd	-6*16(%ecx), %xmm4
+
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+
+	paddd	%xmm0, %xmm3
+	paddd	%xmm4, %xmm7
+	movdqa	%xmm3, (%ecx)
+	movdqa	%xmm7, 16(%ecx)
+	addl	$2*16, %ecx
+	cmpl	%ecx, %eax
+	jne sha256_transform_4way_extend_loop
+	
+	movdqu	0(%edi), %xmm7
+	movdqu	16(%edi), %xmm5
+	movdqu	32(%edi), %xmm4
+	movdqu	48(%edi), %xmm3
+	movdqu	64(%edi), %xmm0
+	movdqu	80(%edi), %xmm1
+	movdqu	96(%edi), %xmm2
+	movdqu	112(%edi), %xmm6
+	movdqa	%xmm1, 0(%esp)
+	movdqa	%xmm2, 16(%esp)
+	movdqa	%xmm6, 32(%esp)
+	
+	xorl	%eax, %eax
+sha256_transform_4way_main_loop:
+	movdqa	3*16(%esp, %eax), %xmm6
+	paddd	sha256_4k(%eax), %xmm6
+	paddd	32(%esp), %xmm6
+
+	movdqa	%xmm0, %xmm1
+	movdqa	16(%esp), %xmm2
+	pandn	%xmm2, %xmm1
+
+	movdqa	%xmm2, 32(%esp)
+	movdqa	0(%esp), %xmm2
+	movdqa	%xmm2, 16(%esp)
+
+	pand	%xmm0, %xmm2
+	pxor	%xmm2, %xmm1
+	movdqa	%xmm0, 0(%esp)
+
+	paddd	%xmm1, %xmm6
+
+	movdqa	%xmm0, %xmm1
+	psrld	$6, %xmm0
+	movdqa	%xmm0, %xmm2
+	pslld	$7, %xmm1
+	psrld	$5, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	pslld	$14, %xmm1
+	psrld	$14, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	pslld	$5, %xmm1
+	pxor	%xmm1, %xmm0
+	paddd	%xmm0, %xmm6
+
+	movdqa	%xmm3, %xmm0
+	paddd	%xmm6, %xmm0
+
+	movdqa	%xmm5, %xmm1
+	movdqa	%xmm4, %xmm3
+	movdqa	%xmm4, %xmm2
+	pand	%xmm5, %xmm2
+	pand	%xmm7, %xmm4
+	pand	%xmm7, %xmm1
+	pxor	%xmm4, %xmm1
+	movdqa	%xmm5, %xmm4
+	movdqa	%xmm7, %xmm5
+	pxor	%xmm2, %xmm1
+	paddd	%xmm1, %xmm6
+
+	movdqa	%xmm7, %xmm2
+	psrld	$2, %xmm7
+	movdqa	%xmm7, %xmm1
+	pslld	$10, %xmm2
+	psrld	$11, %xmm1
+	pxor	%xmm2, %xmm7
+	pxor	%xmm1, %xmm7
+	pslld	$9, %xmm2
+	psrld	$9, %xmm1
+	pxor	%xmm2, %xmm7
+	pxor	%xmm1, %xmm7
+	pslld	$11, %xmm2
+	pxor	%xmm2, %xmm7
+	paddd	%xmm6, %xmm7
+	
+	addl	$16, %eax
+	cmpl	$16*64, %eax
+	jne sha256_transform_4way_main_loop
+	
+	movdqu	0(%edi), %xmm1
+	movdqu	16(%edi), %xmm2
+	paddd	%xmm1, %xmm7
+	paddd	%xmm2, %xmm5
+	movdqu	32(%edi), %xmm1
+	movdqu	48(%edi), %xmm2
+	paddd	%xmm1, %xmm4
+	paddd	%xmm2, %xmm3
+	
+	movdqu	%xmm7, 0(%edi)
+	movdqu	%xmm5, 16(%edi)
+	movdqu	%xmm4, 32(%edi)
+	movdqu	%xmm3, 48(%edi)
+	
+	movdqu	64(%edi), %xmm1
+	movdqu	80(%edi), %xmm2
+	movdqu	96(%edi), %xmm6
+	movdqu	112(%edi), %xmm7
+	paddd	%xmm1, %xmm0
+	paddd	0(%esp), %xmm2
+	paddd	16(%esp), %xmm6
+	paddd	32(%esp), %xmm7
+	
+	movdqu	%xmm0, 64(%edi)
+	movdqu	%xmm2, 80(%edi)
+	movdqu	%xmm6, 96(%edi)
+	movdqu	%xmm7, 112(%edi)
+	
+	movl	%edx, %esp
+	popl	%esi
+	popl	%edi
+	ret
+
+
+	.text
+	.p2align 5
+	.globl sha256d_ms_4way
+	.globl _sha256d_ms_4way
+sha256d_ms_4way:
+_sha256d_ms_4way:
+	pushl	%edi
+	pushl	%esi
+	pushl	%ebp
+	movl	16(%esp), %edi
+	movl	20(%esp), %esi
+	movl	24(%esp), %edx
+	movl	28(%esp), %ecx
+	movl	%esp, %ebp
+	subl	$67*16, %esp
+	andl	$-128, %esp
+	
+	leal	256(%esi), %eax
+	
+sha256d_ms_4way_extend_loop1:
+	movdqa	3*16(%esi), %xmm0
+	movdqa	2*16(%eax), %xmm3
+	movdqa	3*16(%eax), %xmm7
+	movdqa	%xmm3, 5*16(%esp)
+	movdqa	%xmm7, 6*16(%esp)
+	movdqa	%xmm0, %xmm2
+	paddd	%xmm0, %xmm7
+	psrld	$3, %xmm0
+	movdqa	%xmm0, %xmm1
+	pslld	$14, %xmm2
+	psrld	$4, %xmm1
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	psrld	$11, %xmm1
+	pslld	$11, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	paddd	%xmm0, %xmm3
+	movdqa	%xmm3, 2*16(%eax)
+	movdqa	%xmm7, 3*16(%eax)
+	
+	movdqa	4*16(%eax), %xmm0
+	movdqa	%xmm0, 7*16(%esp)
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	%xmm0, %xmm3
+	movdqa	%xmm3, 4*16(%eax)
+	movdqa	%xmm7, 5*16(%eax)
+	
+	movdqa	6*16(%eax), %xmm0
+	movdqa	7*16(%eax), %xmm4
+	movdqa	%xmm0, 9*16(%esp)
+	movdqa	%xmm4, 10*16(%esp)
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	%xmm0, %xmm3
+	paddd	%xmm4, %xmm7
+	movdqa	%xmm3, 6*16(%eax)
+	movdqa	%xmm7, 7*16(%eax)
+	
+	movdqa	8*16(%eax), %xmm0
+	movdqa	2*16(%eax), %xmm4
+	movdqa	%xmm0, 11*16(%esp)
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	%xmm0, %xmm3
+	paddd	%xmm4, %xmm7
+	movdqa	%xmm3, 8*16(%eax)
+	movdqa	%xmm7, 9*16(%eax)
+	
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	3*16(%eax), %xmm3
+	paddd	4*16(%eax), %xmm7
+	movdqa	%xmm3, 10*16(%eax)
+	movdqa	%xmm7, 11*16(%eax)
+	
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	5*16(%eax), %xmm3
+	paddd	6*16(%eax), %xmm7
+	movdqa	%xmm3, 12*16(%eax)
+	movdqa	%xmm7, 13*16(%eax)
+	
+	movdqa	14*16(%eax), %xmm0
+	movdqa	15*16(%eax), %xmm4
+	movdqa	%xmm0, 17*16(%esp)
+	movdqa	%xmm4, 18*16(%esp)
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	paddd	7*16(%eax), %xmm0
+	paddd	8*16(%eax), %xmm4
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	%xmm0, %xmm3
+	paddd	%xmm4, %xmm7
+	movdqa	%xmm3, 14*16(%eax)
+	movdqa	%xmm7, 15*16(%eax)
+	
+sha256d_ms_4way_extend_loop2:
+	sha256_sse2_extend_doubleround 16
+	sha256_sse2_extend_doubleround 18
+	sha256_sse2_extend_doubleround 20
+	sha256_sse2_extend_doubleround 22
+	sha256_sse2_extend_doubleround 24
+	sha256_sse2_extend_doubleround 26
+	sha256_sse2_extend_doubleround 28
+	sha256_sse2_extend_doubleround 30
+	sha256_sse2_extend_doubleround 32
+	sha256_sse2_extend_doubleround 34
+	sha256_sse2_extend_doubleround 36
+	sha256_sse2_extend_doubleround 38
+	sha256_sse2_extend_doubleround 40
+	sha256_sse2_extend_doubleround 42
+	jz sha256d_ms_4way_extend_coda2
+	sha256_sse2_extend_doubleround 44
+	sha256_sse2_extend_doubleround 46
+	
+	movdqa	0(%ecx), %xmm3
+	movdqa	16(%ecx), %xmm0
+	movdqa	32(%ecx), %xmm1
+	movdqa	48(%ecx), %xmm2
+	movdqa	64(%ecx), %xmm6
+	movdqa	80(%ecx), %xmm7
+	movdqa	96(%ecx), %xmm5
+	movdqa	112(%ecx), %xmm4
+	movdqa	%xmm1, 0(%esp)
+	movdqa	%xmm2, 16(%esp)
+	movdqa	%xmm6, 32(%esp)
+	
+	movl	%esi, %eax
+	jmp sha256d_ms_4way_main_loop1
+	
+sha256d_ms_4way_main_loop2:
+	sha256_sse2_main_round 0
+	sha256_sse2_main_round 1
+	sha256_sse2_main_round 2
+sha256d_ms_4way_main_loop1:
+	sha256_sse2_main_round 3
+	sha256_sse2_main_quadround 4
+	sha256_sse2_main_quadround 8
+	sha256_sse2_main_quadround 12
+	sha256_sse2_main_quadround 16
+	sha256_sse2_main_quadround 20
+	sha256_sse2_main_quadround 24
+	sha256_sse2_main_quadround 28
+	sha256_sse2_main_quadround 32
+	sha256_sse2_main_quadround 36
+	sha256_sse2_main_quadround 40
+	sha256_sse2_main_quadround 44
+	sha256_sse2_main_quadround 48
+	sha256_sse2_main_quadround 52
+	sha256_sse2_main_round 56
+	jz sha256d_ms_4way_finish
+	sha256_sse2_main_round 57
+	sha256_sse2_main_round 58
+	sha256_sse2_main_round 59
+	sha256_sse2_main_quadround 60
+	
+	movdqa	5*16(%esp), %xmm1
+	movdqa	6*16(%esp), %xmm2
+	movdqa	7*16(%esp), %xmm6
+	movdqa	%xmm1, 18*16(%esi)
+	movdqa	%xmm2, 19*16(%esi)
+	movdqa	%xmm6, 20*16(%esi)
+	movdqa	9*16(%esp), %xmm1
+	movdqa	10*16(%esp), %xmm2
+	movdqa	11*16(%esp), %xmm6
+	movdqa	%xmm1, 22*16(%esi)
+	movdqa	%xmm2, 23*16(%esi)
+	movdqa	%xmm6, 24*16(%esi)
+	movdqa	17*16(%esp), %xmm1
+	movdqa	18*16(%esp), %xmm2
+	movdqa	%xmm1, 30*16(%esi)
+	movdqa	%xmm2, 31*16(%esi)
+	
+	movdqa	0(%esp), %xmm1
+	movdqa	16(%esp), %xmm2
+	movdqa	32(%esp), %xmm6
+	paddd	0(%edx), %xmm7
+	paddd	16(%edx), %xmm5
+	paddd	32(%edx), %xmm4
+	paddd	48(%edx), %xmm3
+	paddd	64(%edx), %xmm0
+	paddd	80(%edx), %xmm1
+	paddd	96(%edx), %xmm2
+	paddd	112(%edx), %xmm6
+	
+	movdqa	%xmm7, 48+0(%esp)
+	movdqa	%xmm5, 48+16(%esp)
+	movdqa	%xmm4, 48+32(%esp)
+	movdqa	%xmm3, 48+48(%esp)
+	movdqa	%xmm0, 48+64(%esp)
+	movdqa	%xmm1, 48+80(%esp)
+	movdqa	%xmm2, 48+96(%esp)
+	movdqa	%xmm6, 48+112(%esp)
+	
+	movdqa	sha256d_4preext2_15, %xmm1
+	movdqa	sha256d_4preext2_24, %xmm2
+	pxor	%xmm0, %xmm0
+	movdqa	%xmm2, 48+128(%esp)
+	movdqa	%xmm0, 48+144(%esp)
+	movdqa	%xmm0, 48+160(%esp)
+	movdqa	%xmm0, 48+176(%esp)
+	movdqa	%xmm0, 48+192(%esp)
+	movdqa	%xmm0, 48+208(%esp)
+	movdqa	%xmm0, 48+224(%esp)
+	movdqa	%xmm1, 48+240(%esp)
+	
+	leal	19*16(%esp), %eax
+	cmpl	%eax, %eax
+	
+	movdqa	-15*16(%eax), %xmm0
+	movdqa	-14*16(%eax), %xmm4
+	movdqa	%xmm0, %xmm2
+	movdqa	%xmm4, %xmm6
+	psrld	$3, %xmm0
+	psrld	$3, %xmm4
+	movdqa	%xmm0, %xmm1
+	movdqa	%xmm4, %xmm5
+	pslld	$14, %xmm2
+	pslld	$14, %xmm6
+	psrld	$4, %xmm1
+	psrld	$4, %xmm5
+	pxor	%xmm1, %xmm0
+	pxor	%xmm5, %xmm4
+	psrld	$11, %xmm1
+	psrld	$11, %xmm5
+	pxor	%xmm2, %xmm0
+	pxor	%xmm6, %xmm4
+	pslld	$11, %xmm2
+	pslld	$11, %xmm6
+	pxor	%xmm1, %xmm0
+	pxor	%xmm5, %xmm4
+	pxor	%xmm2, %xmm0
+	pxor	%xmm6, %xmm4
+	paddd	-16*16(%eax), %xmm0
+	paddd	-15*16(%eax), %xmm4
+	paddd	sha256d_4preext2_17, %xmm4
+	movdqa	%xmm0, %xmm3
+	movdqa	%xmm4, %xmm7
+	movdqa	%xmm3, 0*16(%eax)
+	movdqa	%xmm7, 1*16(%eax)
+	
+	sha256_sse2_extend_doubleround 2
+	sha256_sse2_extend_doubleround 4
+	
+	movdqa	-9*16(%eax), %xmm0
+	movdqa	sha256d_4preext2_23, %xmm4
+	movdqa	%xmm0, %xmm2
+	psrld	$3, %xmm0
+	movdqa	%xmm0, %xmm1
+	pslld	$14, %xmm2
+	psrld	$4, %xmm1
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	psrld	$11, %xmm1
+	pslld	$11, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	paddd	-10*16(%eax), %xmm0
+	paddd	-9*16(%eax), %xmm4
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	paddd	-1*16(%eax), %xmm0
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	paddd	0*16(%eax), %xmm4
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	%xmm0, %xmm3
+	paddd	%xmm4, %xmm7
+	movdqa	%xmm3, 6*16(%eax)
+	movdqa	%xmm7, 7*16(%eax)
+	
+	movdqa	sha256d_4preext2_24, %xmm0
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	paddd	1*16(%eax), %xmm0
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	%xmm0, %xmm3
+	paddd	2*16(%eax), %xmm7
+	movdqa	%xmm3, 8*16(%eax)
+	movdqa	%xmm7, 9*16(%eax)
+	
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	3*16(%eax), %xmm3
+	paddd	4*16(%eax), %xmm7
+	movdqa	%xmm3, 10*16(%eax)
+	movdqa	%xmm7, 11*16(%eax)
+	
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	5*16(%eax), %xmm3
+	paddd	6*16(%eax), %xmm7
+	movdqa	%xmm3, 12*16(%eax)
+	movdqa	%xmm7, 13*16(%eax)
+	
+	movdqa	sha256d_4preext2_30, %xmm0
+	movdqa	0*16(%eax), %xmm4
+	movdqa	%xmm4, %xmm6
+	psrld	$3, %xmm4
+	movdqa	%xmm4, %xmm5
+	pslld	$14, %xmm6
+	psrld	$4, %xmm5
+	pxor	%xmm5, %xmm4
+	pxor	%xmm6, %xmm4
+	psrld	$11, %xmm5
+	pslld	$11, %xmm6
+	pxor	%xmm5, %xmm4
+	pxor	%xmm6, %xmm4
+	paddd	-1*16(%eax), %xmm4
+	movdqa	%xmm3, %xmm2
+	movdqa	%xmm7, %xmm6
+	psrld	$10, %xmm3
+	psrld	$10, %xmm7
+	movdqa	%xmm3, %xmm1
+	movdqa	%xmm7, %xmm5
+	paddd	7*16(%eax), %xmm0
+	pslld	$13, %xmm2
+	pslld	$13, %xmm6
+	psrld	$7, %xmm1
+	psrld	$7, %xmm5
+	paddd	8*16(%eax), %xmm4
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	psrld	$2, %xmm1
+	psrld	$2, %xmm5
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	pslld	$2, %xmm2
+	pslld	$2, %xmm6
+	pxor	%xmm1, %xmm3
+	pxor	%xmm5, %xmm7
+	pxor	%xmm2, %xmm3
+	pxor	%xmm6, %xmm7
+	paddd	%xmm0, %xmm3
+	paddd	%xmm4, %xmm7
+	movdqa	%xmm3, 14*16(%eax)
+	movdqa	%xmm7, 15*16(%eax)
+	
+	jmp sha256d_ms_4way_extend_loop2
+	
+sha256d_ms_4way_extend_coda2:
+	sha256_sse2_extend_round 44
+	
+	movdqa	sha256_4h+0, %xmm7
+	movdqa	sha256_4h+16, %xmm5
+	movdqa	sha256_4h+32, %xmm4
+	movdqa	sha256_4h+48, %xmm3
+	movdqa	sha256_4h+64, %xmm0
+	movdqa	sha256_4h+80, %xmm1
+	movdqa	sha256_4h+96, %xmm2
+	movdqa	sha256_4h+112, %xmm6
+	movdqa	%xmm1, 0(%esp)
+	movdqa	%xmm2, 16(%esp)
+	movdqa	%xmm6, 32(%esp)
+	
+	leal	48(%esp), %eax
+	jmp sha256d_ms_4way_main_loop2
+
+.macro sha256_sse2_main_round_red i, r7
+	movdqa	16*(\i)(%eax), %xmm6
+	paddd	16*(\i)+sha256_4k, %xmm6
+	paddd	32(%esp), %xmm6
+	movdqa	%xmm0, %xmm1
+	movdqa	16(%esp), %xmm2
+	paddd	\r7, %xmm6
+	pandn	%xmm2, %xmm1
+	movdqa	%xmm2, 32(%esp)
+	movdqa	0(%esp), %xmm2
+	movdqa	%xmm2, 16(%esp)
+	pand	%xmm0, %xmm2
+	pxor	%xmm2, %xmm1
+	movdqa	%xmm0, 0(%esp)
+	paddd	%xmm1, %xmm6
+	movdqa	%xmm0, %xmm1
+	psrld	$6, %xmm0
+	movdqa	%xmm0, %xmm2
+	pslld	$7, %xmm1
+	psrld	$5, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	pslld	$14, %xmm1
+	psrld	$14, %xmm2
+	pxor	%xmm1, %xmm0
+	pxor	%xmm2, %xmm0
+	pslld	$5, %xmm1
+	pxor	%xmm1, %xmm0
+	paddd	%xmm6, %xmm0
+.endm
+
+sha256d_ms_4way_finish:
+	sha256_sse2_main_round_red 57, %xmm3
+	sha256_sse2_main_round_red 58, %xmm4
+	sha256_sse2_main_round_red 59, %xmm5
+	sha256_sse2_main_round_red 60, %xmm7
+	
+	paddd	sha256_4h+112, %xmm0
+	movdqa	%xmm0, 112(%edi)
+	
+	movl	%ebp, %esp
+	popl	%ebp
+	popl	%esi
+	popl	%edi
+	ret
+
+
+	.text
+	.p2align 5
+	.globl sha256_use_4way
+	.globl _sha256_use_4way
+sha256_use_4way:
+_sha256_use_4way:
+	pushl	%ebx
+	
+	# Check for SSE2 availability
+	movl	$1, %eax
+	cpuid
+	andl	$0x04000000, %edx
+	jnz sha256_use_4way_sse2
+	xorl	%eax, %eax
+	popl	%ebx
+	ret
+	
+sha256_use_4way_sse2:
+	movl	$1, %eax
+	popl	%ebx
+	ret
+
+#endif
diff --git a/sha2.c b/sha2.c
index 9ef8acb..0e612b4 100644
--- a/sha2.c
+++ b/sha2.c
@@ -418,68 +418,89 @@ static inline void sha256d_ms(uint32_t *hash, uint32_t *W,
 }
 
 #ifdef HAVE_SHA256_4WAY
-#define SHA256D_MAX_WAYS 4
+
 void sha256d_ms_4way(uint32_t *hash,  uint32_t *data,
 	const uint32_t *midstate, const uint32_t *prehash);
-#else
-#define SHA256D_MAX_WAYS 1
-#endif
 
-int scanhash_sha256d(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
-	uint32_t max_nonce, unsigned long *hashes_done)
+static inline int scanhash_sha256d_4way(int thr_id, uint32_t *pdata,
+	const uint32_t *ptarget, uint32_t max_nonce, unsigned long *hashes_done)
 {
-	uint32_t data[SHA256D_MAX_WAYS * 64] __attribute__((aligned(128)));
-	uint32_t hash[SHA256D_MAX_WAYS * 8] __attribute__((aligned(32)));
-	uint32_t midstate[SHA256D_MAX_WAYS * 8] __attribute__((aligned(32)));
-	uint32_t prehash[SHA256D_MAX_WAYS * 8] __attribute__((aligned(32)));
+	uint32_t data[4 * 64] __attribute__((aligned(128)));
+	uint32_t hash[4 * 8] __attribute__((aligned(32)));
+	uint32_t midstate[4 * 8] __attribute__((aligned(32)));
+	uint32_t prehash[4 * 8] __attribute__((aligned(32)));
 	uint32_t n = pdata[19] - 1;
 	const uint32_t first_nonce = pdata[19];
 	const uint32_t Htarg = ptarget[7];
-#ifdef HAVE_SHA256_4WAY
-	const int ways = sha256_use_4way() ? 4 : 1;
-#else
-	const int ways = 1;
-#endif
 	int i, j;
 	
 	memcpy(data, pdata + 16, 64);
 	sha256d_preextend(data);
 	for (i = 31; i >= 0; i--)
-		for (j = 0; j < ways; j++)
-			data[i * ways + j] = data[i];
+		for (j = 0; j < 4; j++)
+			data[i * 4 + j] = data[i];
 	
 	sha256_init(midstate);
 	sha256_transform(midstate, pdata, 0);
 	memcpy(prehash, midstate, 32);
 	sha256d_prehash(prehash, pdata + 16);
 	for (i = 7; i >= 0; i--) {
-		for (j = 0; j < ways; j++) {
-			midstate[i * ways + j] = midstate[i];
-			prehash[i * ways + j] = prehash[i];
+		for (j = 0; j < 4; j++) {
+			midstate[i * 4 + j] = midstate[i];
+			prehash[i * 4 + j] = prehash[i];
 		}
 	}
 	
-#ifdef HAVE_SHA256_4WAY
-	if (ways == 4)
-		do {
-			for (i = 0; i < 4; i++)
-				data[4 * 3 + i] = ++n;
-			
-			sha256d_ms_4way(hash, data, midstate, prehash);
-			
-			for (i = 0; i < 4; i++) {
-				if (hash[4 * 7 + i] <= Htarg) {
-					pdata[19] = data[4 * 3 + i];
-					sha256d(hash, pdata);
-					if (fulltest(hash, ptarget)) {
-						*hashes_done = n - first_nonce + 1;
-						return 1;
-					}
+	do {
+		for (i = 0; i < 4; i++)
+			data[4 * 3 + i] = ++n;
+		
+		sha256d_ms_4way(hash, data, midstate, prehash);
+		
+		for (i = 0; i < 4; i++) {
+			if (hash[4 * 7 + i] <= Htarg) {
+				pdata[19] = data[4 * 3 + i];
+				sha256d(hash, pdata);
+				if (fulltest(hash, ptarget)) {
+					*hashes_done = n - first_nonce + 1;
+					return 1;
 				}
 			}
-		} while (n < max_nonce && !work_restart[thr_id].restart);
-	else
+		}
+	} while (n < max_nonce && !work_restart[thr_id].restart);
+	
+	*hashes_done = n - first_nonce + 1;
+	pdata[19] = n;
+	return 0;
+}
+
+#endif /* HAVE_SHA256_4WAY */
+
+int scanhash_sha256d(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
+	uint32_t max_nonce, unsigned long *hashes_done)
+{
+	uint32_t data[64] __attribute__((aligned(128)));
+	uint32_t hash[8] __attribute__((aligned(32)));
+	uint32_t midstate[8] __attribute__((aligned(32)));
+	uint32_t prehash[8] __attribute__((aligned(32)));
+	uint32_t n = pdata[19] - 1;
+	const uint32_t first_nonce = pdata[19];
+	const uint32_t Htarg = ptarget[7];
+	
+#ifdef HAVE_SHA256_4WAY
+	if (sha256_use_4way())
+		return scanhash_sha256d_4way(thr_id, pdata, ptarget,
+			max_nonce, hashes_done);
 #endif
+	
+	memcpy(data, pdata + 16, 64);
+	sha256d_preextend(data);
+	
+	sha256_init(midstate);
+	sha256_transform(midstate, pdata, 0);
+	memcpy(prehash, midstate, 32);
+	sha256d_prehash(prehash, pdata + 16);
+	
 	do {
 		data[3] = ++n;
 		sha256d_ms(hash, data, midstate, prehash);