add pclmullqlqdq

gen/gen_code.cpp

@@ -767,6 +767,21 @@ void put()
 			printf("void %s(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0x%02X, 0x66, isXMM_XMMorMEM, static_cast<uint8>(imm), 0x3A); }\n", p->name, p->code);
 		}
 	}
+	{
+		const struct Tbl {
+			const char *name;
+			uint8 code;
+		} tbl[] = {
+			{ "pclmullqlqdq", 0 },
+			{ "pclmulhqlqdq", 1 },
+			{ "pclmullqhdq", 0x10 },
+			{ "pclmulhqhdq", 0x11 },
+		};
+		for (int i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+			const Tbl *p = &tbl[i];
+			printf("void %s(const Xmm& xmm, const Operand& op) { pclmulqdq(xmm, op, 0x%02X); }\n", p->name, p->code);
+		}
+	}
 	{
 		const struct Tbl {
 			uint8 code;

readme.md

@@ -197,7 +197,7 @@ License
 -------------
 
 modified new BSD License
-http://www.opensource.org/licenses/bsd-license.php
+http://opensource.org/licenses/BSD-3-Clause
 
 History
 -------------
@@ -252,5 +252,5 @@ Author
 MITSUNARI Shigeo(herumi at nifty dot com)
 
 ---
-$Revision: 1.15 $
+$Revision: 1.16 $
-$Date: 2011/11/09 02:50:21 $
+$Date: 2011/12/22 01:43:13 $

test/make_nm.cpp

@@ -1096,6 +1096,18 @@ class Test {
 			const char *p = tbl[i];
 			put(p, XMM, XMM|MEM, IMM);
 		}
+		{
+			const char tbl[][16] = {
+				"pclmullqlqdq",
+				"pclmulhqlqdq",
+//				"pclmullqhdq", // QQQ : not supported by nasm/yasm
+//				"pclmulhqhdq",
+			};
+			for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+				const char *p = tbl[i];
+				put(p, XMM, XMM|MEM);
+			}
+		}
 		put("extractps", REG32e|MEM, XMM, IMM);
 		put("pextrw", REG32e|MEM, XMM, IMM); // pextrw for REG32 is for MMX2
 		put("pextrb", REG32e|MEM, XMM, IMM);

xbyak/xbyak.h

@@ -5,14 +5,14 @@
 	@file xbyak.h
 	@brief Xbyak ; JIT assembler for x86(IA32)/x64 by C++
 	@author herumi
-	@version $Revision: 1.256 $
+	@version $Revision: 1.260 $
 	@url http://homepage1.nifty.com/herumi/soft/xbyak.html
-	@date $Date: 2011/11/09 05:06:37 $
+	@date $Date: 2012/01/05 00:34:07 $
 	@note modified new BSD license
-	http://www.opensource.org/licenses/bsd-license.php
+	http://opensource.org/licenses/BSD-3-Clause
 */
 #if not +0
-	#error "use -fno-operator-names"
+	#error "use -fno-operator-names to use 'not', 'xor, 'and' as function names"
 #endif
 
 #include <stdio.h> // for debug print
@@ -54,7 +54,7 @@ namespace Xbyak {
 
 enum {
 	DEFAULT_MAX_CODE_SIZE = 4096,
-	VERSION = 0x3050, /* 0xABCD = A.BC(D) */
+	VERSION = 0x3060, /* 0xABCD = A.BC(D) */
 };
 
 #ifndef MIE_INTEGER_TYPE_DEFINED
@@ -1498,7 +1498,7 @@ public:
 #include "xbyak_mnemonic.h"
 	void align(int x = 16)
 	{
-		if (x != 4 && x != 8 && x != 16 && x != 32) throw ERR_BAD_ALIGN;
+		if (x < 2 || (x & (x - 1))) throw ERR_BAD_ALIGN;
 		while (size_t(getCurr()) % x) {
 			nop();
 		}

xbyak/xbyak_mnemonic.h

@@ -1,4 +1,4 @@
-const char *getVersionString() const { return "3.05"; }
+const char *getVersionString() const { return "3.06"; }
 void packssdw(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x6B); }
 void packsswb(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x63); }
 void packuswb(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x67); }
@@ -478,6 +478,10 @@ void pcmpistrm(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0x62
 void pcmpistri(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0x63, 0x66, isXMM_XMMorMEM, static_cast<uint8>(imm), 0x3A); }
 void pclmulqdq(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0x44, 0x66, isXMM_XMMorMEM, static_cast<uint8>(imm), 0x3A); }
 void aeskeygenassist(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0xDF, 0x66, isXMM_XMMorMEM, static_cast<uint8>(imm), 0x3A); }
+void pclmullqlqdq(const Xmm& xmm, const Operand& op) { pclmulqdq(xmm, op, 0x00); }
+void pclmulhqlqdq(const Xmm& xmm, const Operand& op) { pclmulqdq(xmm, op, 0x01); }
+void pclmullqhdq(const Xmm& xmm, const Operand& op) { pclmulqdq(xmm, op, 0x10); }
+void pclmulhqhdq(const Xmm& xmm, const Operand& op) { pclmulqdq(xmm, op, 0x11); }
 void ldmxcsr(const Address& addr) { opModM(addr, Reg32(2), 0x0F, 0xAE); }
 void stmxcsr(const Address& addr) { opModM(addr, Reg32(3), 0x0F, 0xAE); }
 void clflush(const Address& addr) { opModM(addr, Reg32(7), 0x0F, 0xAE); }