- port x64 assembly code to MASM

- improve stack unwindability on x64 for GCC by not modifying RBP/RSP registers in inline assembly

git-svn-id: svn://svn.code.sf.net/p/cryptopp/code/trunk/c5@396 57ff6487-cd31-0410-9ec3-f628ee90f5f0

1 files changed, 116 insertions, 5 deletions

diff --git a/cpu.h b/cpu.h
index bec63fd..2ef55c3 100755
--- a/cpu.h
+++ b/cpu.h
@@ -1,6 +1,15 @@
 #ifndef CRYPTOPP_CPU_H
 #define CRYPTOPP_CPU_H
 
+#ifdef CRYPTOPP_GENERATE_X64_MASM
+
+#define CRYPTOPP_X86_ASM_AVAILABLE
+#define CRYPTOPP_BOOL_X64 1
+#define CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE 1
+#define NAMESPACE_END
+
+#else
+
 #include "config.h"
 
 #ifdef CRYPTOPP_MSVC6PP_OR_LATER
@@ -98,7 +107,18 @@ inline bool HasMMX()	{return false;}
 
 #endif		// #ifdef CRYPTOPP_X86_ASM_AVAILABLE || _MSC_VER >= 1400
 
-#if defined(__GNUC__)
+#endif
+
+#ifdef CRYPTOPP_GENERATE_X64_MASM
+	#define AS1(x) x*newline*
+	#define AS2(x, y) x, y*newline*
+	#define AS3(x, y, z) x, y, z*newline*
+	#define ASS(x, y, a, b, c, d) x, y, a*64+b*16+c*4+d*newline*
+	#define ASL(x) label##x:*newline*
+	#define ASJ(x, y, z) x label##y*newline*
+	#define ASC(x, y) x label##y*newline*
+	#define AS_HEX(y) y##h
+#elif defined(__GNUC__)
 	// define these in two steps to allow arguments to be expanded
 	#define GNU_AS1(x) #x ";"
 	#define GNU_AS2(x, y) #x ", " #y ";"
@@ -113,6 +133,7 @@ inline bool HasMMX()	{return false;}
 	#define ASJ(x, y, z) GNU_ASJ(x, y, z)
 	#define ASC(x, y) #x " " #y ";"
 	#define CRYPTOPP_NAKED
+	#define AS_HEX(y) 0x##y
 #else
 	#define AS1(x) __asm {x}
 	#define AS2(x, y) __asm {x, y}
@@ -122,25 +143,115 @@ inline bool HasMMX()	{return false;}
 	#define ASJ(x, y, z) __asm {x label##y}
 	#define ASC(x, y) __asm {x label##y}
 	#define CRYPTOPP_NAKED __declspec(naked)
+	#define AS_HEX(y) 0x##y
 #endif
 
+#ifdef CRYPTOPP_GENERATE_X64_MASM
+#define ASM_MOD(x, y) ((x) MOD (y))
+#else
 // GNU assembler doesn't seem to have mod operator
 #define ASM_MOD(x, y) ((x)-((x)/(y))*(y))
+#endif
 
 #if CRYPTOPP_BOOL_X86
+	#define AS_REG_1 ecx
+	#define AS_REG_2 edx
+	#define AS_REG_3 esi
+	#define AS_REG_4 edi
+	#define AS_REG_5 eax
+	#define AS_REG_6 ebx
+	#define AS_REG_7 ebp
+	#define AS_REG_1d ecx
+	#define AS_REG_2d edx
+	#define AS_REG_3d esi
+	#define AS_REG_4d edi
+	#define AS_REG_5d eax
+	#define AS_REG_6d ebx
+	#define AS_REG_7d ebp
 	#define WORD_SZ 4
 	#define WORD_REG(x)	e##x
 	#define WORD_PTR DWORD PTR
-	#define AS_PUSH(x) AS1(push e##x)
-	#define AS_POP(x) AS1(pop e##x)
+	#define AS_PUSH_IF86(x) AS1(push e##x)
+	#define AS_POP_IF86(x) AS1(pop e##x)
+	#define AS_JCXZ jecxz
 #elif CRYPTOPP_BOOL_X64
+	#ifdef CRYPTOPP_GENERATE_X64_MASM
+		#define AS_REG_1 rcx
+		#define AS_REG_2 rdx
+		#define AS_REG_3 r8
+		#define AS_REG_4 r9
+		#define AS_REG_5 rax
+		#define AS_REG_6 r10
+		#define AS_REG_7 r11
+		#define AS_REG_1d ecx
+		#define AS_REG_2d edx
+		#define AS_REG_3d r8d
+		#define AS_REG_4d r9d
+		#define AS_REG_5d eax
+		#define AS_REG_6d r10d
+		#define AS_REG_7d r11d
+	#else
+		#define AS_REG_1 rdi
+		#define AS_REG_2 rsi
+		#define AS_REG_3 rdx
+		#define AS_REG_4 rcx
+		#define AS_REG_5 r8
+		#define AS_REG_6 r9
+		#define AS_REG_7 r10
+		#define AS_REG_1d edi
+		#define AS_REG_2d esi
+		#define AS_REG_3d edx
+		#define AS_REG_4d ecx
+		#define AS_REG_5d r8d
+		#define AS_REG_6d r9d
+		#define AS_REG_7d r10d
+	#endif
 	#define WORD_SZ 8
 	#define WORD_REG(x)	r##x
 	#define WORD_PTR QWORD PTR
-	#define AS_PUSH(x) AS1(pushq r##x)
-	#define AS_POP(x) AS1(popq r##x)
+	#define AS_PUSH_IF86(x)
+	#define AS_POP_IF86(x)
+	#define AS_JCXZ jrcxz
 #endif
 
+// helper macro for stream cipher output
+#define AS_XMM_OUTPUT4(labelPrefix, inputPtr, outputPtr, x0, x1, x2, x3, t, p0, p1, p2, p3, increment)\
+	AS2(	test	inputPtr, inputPtr)\
+	ASC(	jz,		labelPrefix##3)\
+	AS2(	test	inputPtr, 15)\
+	ASC(	jnz,	labelPrefix##7)\
+	AS2(	pxor	xmm##x0, [inputPtr+p0*16])\
+	AS2(	pxor	xmm##x1, [inputPtr+p1*16])\
+	AS2(	pxor	xmm##x2, [inputPtr+p2*16])\
+	AS2(	pxor	xmm##x3, [inputPtr+p3*16])\
+	AS2(	add		inputPtr, increment*16)\
+	ASC(	jmp,	labelPrefix##3)\
+	ASL(labelPrefix##7)\
+	AS2(	movdqu	xmm##t, [inputPtr+p0*16])\
+	AS2(	pxor	xmm##x0, xmm##t)\
+	AS2(	movdqu	xmm##t, [inputPtr+p1*16])\
+	AS2(	pxor	xmm##x1, xmm##t)\
+	AS2(	movdqu	xmm##t, [inputPtr+p2*16])\
+	AS2(	pxor	xmm##x2, xmm##t)\
+	AS2(	movdqu	xmm##t, [inputPtr+p3*16])\
+	AS2(	pxor	xmm##x3, xmm##t)\
+	AS2(	add		inputPtr, increment*16)\
+	ASL(labelPrefix##3)\
+	AS2(	test	outputPtr, 15)\
+	ASC(	jnz,	labelPrefix##8)\
+	AS2(	movdqa	[outputPtr+p0*16], xmm##x0)\
+	AS2(	movdqa	[outputPtr+p1*16], xmm##x1)\
+	AS2(	movdqa	[outputPtr+p2*16], xmm##x2)\
+	AS2(	movdqa	[outputPtr+p3*16], xmm##x3)\
+	ASC(	jmp,	labelPrefix##9)\
+	ASL(labelPrefix##8)\
+	AS2(	movdqu	[outputPtr+p0*16], xmm##x0)\
+	AS2(	movdqu	[outputPtr+p1*16], xmm##x1)\
+	AS2(	movdqu	[outputPtr+p2*16], xmm##x2)\
+	AS2(	movdqu	[outputPtr+p3*16], xmm##x3)\
+	ASL(labelPrefix##9)\
+	AS2(	add		outputPtr, increment*16)
+
 NAMESPACE_END
 
 #endif