Finally figured out how to add an autoconf option. Added the ability …

…to disable AES-NI, and split cryptonight.c into four files - a header, functions common to both implementations, the AES-NI implementation, and the slow implementation. Also moved AES-NI specific asm to its own file.

Makefile.am

@@ -13,7 +13,7 @@ EXTRA_DIST	= example-cfg.json nomacro.pl
 SUBDIRS		= compat
 
 AM_CPPFLAGS	= $(PTHREAD_FLAGS) $(JANSSON_INCLUDES)
-AM_CFLAGS	= -O3 -Ofast -flto -fuse-linker-plugin -falign-loops=16 -falign-functions=16 -falign-jumps=16 -falign-loops=16 -falign-labels=16 -maes
+AM_CFLAGS	= -O3 -Ofast -flto -fuse-linker-plugin -falign-loops=16 -falign-functions=16 -falign-jumps=16 -falign-loops=16 -falign-labels=16
 
 bin_PROGRAMS	= minerd
 
@@ -32,7 +32,7 @@ minerd_SOURCES	= elist.h \
 		  skein.c \
 		  ink.c \
 		  blake.c \
-		  cryptonight.c \
+		  cryptonight_common.c \
 		  x11.c \
 		  sha3/sph_keccak.c \
 		  sha3/sph_hefty1.c \
@@ -60,6 +60,13 @@ endif
 if ARCH_x86_64
 minerd_SOURCES += sha2-x64.S scrypt-x64.S aesb-x64.S
 endif
+if USE_LOBOTOMIZED_AES
+minerd_SOURCES += cryptonight_lobotomized.c
+else
+minerd_SOURCES += cryptonight_aesni.c aesni.S
+AM_CFLAGS += -maes
+endif
+
 if ARCH_ARM
 minerd_SOURCES += sha2-arm.S scrypt-arm.S
 endif

aesb-x64.S

@@ -4,56 +4,6 @@
     .section .note.GNU-stack,"",%progbits
 #endif
 
-    .text
-    .p2align 6
-    .globl fast_aesb_single_round
-    .globl _fast_aesb_single_round
-fast_aesb_single_round:
-_fast_aesb_single_round:
-#if defined(_WIN64) || defined(__CYGWIN__)
-    movdqa (%rcx), %xmm1
-    aesenc (%r8), %xmm1
-    movdqa %xmm1, (%rdx)
-#else
-    movdqa (%rdi), %xmm1
-    aesenc (%rdx), %xmm1
-    movdqa %xmm1, (%rsi)
-#endif
-    ret
-
-    .text
-    .p2align 6
-    .globl fast_aesb_pseudo_round_mut
-    .globl _fast_aesb_pseudo_round_mut
-fast_aesb_pseudo_round_mut:
-_fast_aesb_pseudo_round_mut:
-#if defined(_WIN64) || defined(__CYGWIN__)
-    mov %rdx, %r9
-    add $0xA0, %r9
-    movdqa (%rcx), %xmm1
-
-    .LOOP:
-            aesenc (%rdx), %xmm1
-            add $0x10, %rdx
-			cmp %r9, %rdx
-            jl .LOOP
-
-    movdqa %xmm1, (%rcx)
-#else
-    mov %rsi, %r9
-    add $0xA0, %r9
-    movdqa (%rdi), %xmm1
-
-    .LOOP:
-            aesenc (%rsi), %xmm1
-            add $0x10, %rsi
-            cmp %r9, %rsi
-            jl .LOOP
-
-    movdqa %xmm1, (%rdi)
-#endif
-    ret
-
     .text
     .globl mul128
     .globl _mul128
@@ -70,158 +20,3 @@ _mul128:
 	mov %rdx, (%r8)
 #endif
 	ret
-
-	.text
-	.p2align 4
-	.globl aesni_parallel_noxor
-# void aesni_parallel_noxor(void *output, uint8_t *input, uint8_t *expkey)
-aesni_parallel_noxor:
-	mov $10, %r9
-	movdqa (%rsi), %xmm0
-	movdqa 0x10(%rsi), %xmm1
-	movdqa 0x20(%rsi), %xmm2
-	movdqa 0x30(%rsi), %xmm3
-	movdqa 0x40(%rsi), %xmm4
-	movdqa 0x50(%rsi), %xmm5
-	movdqa 0x60(%rsi), %xmm6
-	movdqa 0x70(%rsi), %xmm7
-
-	.ENCRYPT:
-		aesenc (%rdx), %xmm0
-		aesenc (%rdx), %xmm1
-		aesenc (%rdx), %xmm2
-		aesenc (%rdx), %xmm3
-		aesenc (%rdx), %xmm4
-		aesenc (%rdx), %xmm5
-		aesenc (%rdx), %xmm6
-		aesenc (%rdx), %xmm7
-		add $0x10, %rdx
-		dec %r9
-		jnz .ENCRYPT
-
-	movdqa %xmm0, (%rdi)
-	movdqa %xmm1, 0x10(%rdi)
-	movdqa %xmm2, 0x20(%rdi)
-	movdqa %xmm3, 0x30(%rdi)
-	movdqa %xmm4, 0x40(%rdi)
-	movdqa %xmm5, 0x50(%rdi)
-	movdqa %xmm6, 0x60(%rdi)
-	movdqa %xmm7, 0x70(%rdi)
-
-	movdqa %xmm0, (%rsi)
-	movdqa %xmm1, 0x10(%rsi)
-	movdqa %xmm2, 0x20(%rsi)
-	movdqa %xmm3, 0x30(%rsi)
-	movdqa %xmm4, 0x40(%rsi)
-	movdqa %xmm5, 0x50(%rsi)
-	movdqa %xmm6, 0x60(%rsi)
-	movdqa %xmm7, 0x70(%rsi)
-
-	ret
-
-	.text
-	.p2align 4
-	.globl aesni_parallel_xor
-# void aesni_parallel_xor(void *state, uint8_t *expkey, uint8_t *xorval)
-aesni_parallel_xor:
-	mov $10, %r9
-	movdqa (%rdi), %xmm0
-	movdqa 0x10(%rdi), %xmm1
-	movdqa 0x20(%rdi), %xmm2
-	movdqa 0x30(%rdi), %xmm3
-	movdqa 0x40(%rdi), %xmm4
-	movdqa 0x50(%rdi), %xmm5
-	movdqa 0x60(%rdi), %xmm6
-	movdqa 0x70(%rdi), %xmm7
-
-	pxor (%rdx), %xmm0
-	pxor 0x10(%rdx), %xmm1
-	pxor 0x20(%rdx), %xmm2
-	pxor 0x30(%rdx), %xmm3
-	pxor 0x40(%rdx), %xmm4 
-	pxor 0x50(%rdx), %xmm5
-	pxor 0x60(%rdx), %xmm6 
-	pxor 0x70(%rdx), %xmm7
-
-	.ENCRYPT2:
-		aesenc (%rsi), %xmm0
-		aesenc (%rsi), %xmm1
-		aesenc (%rsi), %xmm2
-		aesenc (%rsi), %xmm3
-		aesenc (%rsi), %xmm4
-		aesenc (%rsi), %xmm5
-		aesenc (%rsi), %xmm6
-		aesenc (%rsi), %xmm7
-		add $0x10, %rsi
-		dec %r9
-		jnz .ENCRYPT2
-
-	movdqa %xmm0, (%rdi)
-	movdqa %xmm1, 0x10(%rdi)
-	movdqa %xmm2, 0x20(%rdi)
-	movdqa %xmm3, 0x30(%rdi)
-	movdqa %xmm4, 0x40(%rdi)
-	movdqa %xmm5, 0x50(%rdi)
-	movdqa %xmm6, 0x60(%rdi)
-	movdqa %xmm7, 0x70(%rdi)
-
-	ret
-
-	.text
-	.p2align 4
-	.globl that_fucking_loop
-# void that_fucking_loop(uint64_t *a, uint64_t *b, uint8_t *long_state)
-that_fucking_loop:
-	mov $0x80000, %r11
-	movdqa (%rdi), %xmm1
-	movdqa (%rsi), %xmm2
-	# b == xmm2 & a == xmm1 from now on
-
-	.p2align 4
-	.HUGELOOP:
-		movq %xmm1, %r9
-
-		and $0x1FFFF0, %r9
-		add %rdx, %r9
-
-		movdqa (%r9), %xmm3
-		aesenc %xmm1, %xmm3
-
-		pxor %xmm3, %xmm2
-
-		movdqa %xmm2, (%r9)
-
-		movq %xmm3, %r9
-
-		and $0x1FFFF0, %r9
-		add %rdx, %r9
-
-		movdqa (%r9), %xmm4
-		push %rdx
-
-		movq %xmm3, %rax
-		mulq (%r9)
-
-		sub $16, %rsp
-		movdqa %xmm1, (%rsp)
-		add %rdx, (%rsp)
-		add %rax, 8(%rsp)
-
-		movdqa (%rsp), %xmm1
-		add $16, %rsp
-		pop %rdx
-
-		movdqa %xmm1, (%r9)
-
-		pxor %xmm4, %xmm1
-		movdqa %xmm3, %xmm2
-
-		# This is a branch prediction hint. Don't fuck with it.
-		dec %r11
-		cmp $0, %r11
-		setne %al
-		movzx %al, %eax
-		test %rax, %rax
-		jne .HUGELOOP
-
-	ret