remove mutable in Address

8 files changed, 104 insertions, 84 deletions

diff --git a/gen/avx_type.hpp b/gen/avx_type.hpp
index 244a2c0..6f51166 100644
--- a/gen/avx_type.hpp
+++ b/gen/avx_type.hpp
@@ -36,6 +36,7 @@
 		T_B32 = 1 << 26, // m32bcst
 		T_B64 = 1 << 27, // m64bcst
 		T_M_K = 1 << 28, // mem{k}
+		T_VSIB = 1 << 29,
 		T_XXX
 	};
 
@@ -156,5 +157,9 @@ std::string type2String(int type)
 		if (!str.empty()) str += " | ";
 		str += "T_M_K";
 	}
+	if (type & T_VSIB) {
+		if (!str.empty()) str += " | ";
+		str += "T_VSIB";
+	}
 	return str;
 }
diff --git a/gen/gen_avx512.cpp b/gen/gen_avx512.cpp
index 526877e..29b95c1 100644
--- a/gen/gen_avx512.cpp
+++ b/gen/gen_avx512.cpp
@@ -533,7 +533,7 @@ void putGather()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl& p = tbl[i];
-		std::string type = type2String(p.type);
+		std::string type = type2String(p.type | T_VSIB);
 		printf("void %s(const Xmm& x, const Address& addr) { opGather2(x, addr, %s, 0x%02X, %d); }\n", p.name, type.c_str(), p.code, p.mode);
 	}
 }
@@ -557,7 +557,7 @@ void putScatter()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl& p = tbl[i];
-		std::string type = type2String(p.type);
+		std::string type = type2String(p.type | T_VSIB);
 		printf("void %s(const Address& addr, const Xmm& x) { opGather2(x, addr, %s, 0x%02X, %d); }\n", p.name, type.c_str(), p.code, p.mode);
 	}
 }
@@ -669,26 +669,41 @@ void putMisc()
 {
 	puts("void vpbroadcastmb2q(const Xmm& x, const Opmask& k) { opVex(x, 0, k, T_F3 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1, 0x2A); }");
 	puts("void vpbroadcastmw2d(const Xmm& x, const Opmask& k) { opVex(x, 0, k, T_F3 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0, 0x3A); }");
-
-	puts("void vgatherpf0dps(const Address& addr) { opGatherFetch(addr, zm1, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC6, Operand::ZMM); }");
-	puts("void vgatherpf0qps(const Address& addr) { opGatherFetch(addr, zm1, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC7, Operand::ZMM); }");
-	puts("void vgatherpf0dpd(const Address& addr) { opGatherFetch(addr, zm1, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC6, Operand::YMM); }");
-	puts("void vgatherpf0qpd(const Address& addr) { opGatherFetch(addr, zm1, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC7, Operand::ZMM); }");
-
-	puts("void vgatherpf1dps(const Address& addr) { opGatherFetch(addr, zm2, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC6, Operand::ZMM); }");
-	puts("void vgatherpf1qps(const Address& addr) { opGatherFetch(addr, zm2, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC7, Operand::ZMM); }");
-	puts("void vgatherpf1dpd(const Address& addr) { opGatherFetch(addr, zm2, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC6, Operand::YMM); }");
-	puts("void vgatherpf1qpd(const Address& addr) { opGatherFetch(addr, zm2, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC7, Operand::ZMM); }");
-
-	puts("void vscatterpf0dps(const Address& addr) { opGatherFetch(addr, zm5, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC6, Operand::ZMM); }");
-	puts("void vscatterpf0qps(const Address& addr) { opGatherFetch(addr, zm5, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC7, Operand::ZMM); }");
-	puts("void vscatterpf0dpd(const Address& addr) { opGatherFetch(addr, zm5, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC6, Operand::YMM); }");
-	puts("void vscatterpf0qpd(const Address& addr) { opGatherFetch(addr, zm5, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC7, Operand::ZMM); }");
-
-	puts("void vscatterpf1dps(const Address& addr) { opGatherFetch(addr, zm6, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC6, Operand::ZMM); }");
-	puts("void vscatterpf1qps(const Address& addr) { opGatherFetch(addr, zm6, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC7, Operand::ZMM); }");
-	puts("void vscatterpf1dpd(const Address& addr) { opGatherFetch(addr, zm6, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC6, Operand::YMM); }");
-	puts("void vscatterpf1qpd(const Address& addr) { opGatherFetch(addr, zm6, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC7, Operand::ZMM); }");
+	{
+		const struct Tbl {
+			const char *name;
+			int zm;
+			int type;
+			uint8 code;
+			bool isZmm;
+		} tbl[] = {
+			{ "vgatherpf0dps", 1, T_EW0 | T_N4, 0xC6, true },
+			{ "vgatherpf0qps", 1, T_EW0 | T_N4, 0xC7, true },
+			{ "vgatherpf0dpd", 1, T_EW1 | T_N8, 0xC6, false },
+			{ "vgatherpf0qpd", 1, T_EW1 | T_N8, 0xC7, true },
+
+			{ "vgatherpf1dps", 2, T_EW0 | T_N4, 0xC6, true },
+			{ "vgatherpf1qps", 2, T_EW0 | T_N4, 0xC7, true },
+			{ "vgatherpf1dpd", 2, T_EW1 | T_N8, 0xC6, false },
+			{ "vgatherpf1qpd", 2, T_EW1 | T_N8, 0xC7, true },
+
+			{ "vscatterpf0dps", 5, T_EW0 | T_N4, 0xC6, true },
+			{ "vscatterpf0qps", 5, T_EW0 | T_N4, 0xC7, true },
+			{ "vscatterpf0dpd", 5, T_EW1 | T_N8, 0xC6, false },
+			{ "vscatterpf0qpd", 5, T_EW1 | T_N8, 0xC7, true },
+
+			{ "vscatterpf1dps", 6, T_EW0 | T_N4, 0xC6, true },
+			{ "vscatterpf1qps", 6, T_EW0 | T_N4, 0xC7, true },
+			{ "vscatterpf1dpd", 6, T_EW1 | T_N8, 0xC6, false },
+			{ "vscatterpf1qpd", 6, T_EW1 | T_N8, 0xC7, true },
+		};
+		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+			const Tbl& p = tbl[i];
+			std::string type = type2String(p.type | T_66 | T_0F38 | T_MUST_EVEX | T_M_K | T_VSIB);
+			printf("void %s(const Address& addr) { opGatherFetch(addr, zm%d, %s, 0x%2X, Operand::%s); }\n"
+				, p.name, p.zm, type.c_str(), p.code, p.isZmm ? "ZMM" : "YMM");
+		}
+	}
 
 	puts("void vfpclasspd(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(128|256|512)) throw Error(ERR_BAD_MEM_SIZE); Reg x = k; x.setBit(op.getBit()); opVex(x, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW1 | T_B64, 0x66, imm); }");
 	puts("void vfpclassps(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(128|256|512)) throw Error(ERR_BAD_MEM_SIZE); Reg x = k; x.setBit(op.getBit()); opVex(x, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32, 0x66, imm); }");
diff --git a/gen/gen_code.cpp b/gen/gen_code.cpp
index 500d3e3..37877bf 100644
--- a/gen/gen_code.cpp
+++ b/gen/gen_code.cpp
@@ -1652,7 +1652,7 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			printf("void %s(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W%d, 0x%x, %d); }\n", p.name, p.w, p.code, p.mode);
+			printf("void %s(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W%d, 0x%x, %d); }\n", p.name, p.w, p.code, p.mode);
 		}
 	}
 }
diff --git a/readme.md b/readme.md
index 5e217ce..507faaf 100644
--- a/readme.md
+++ b/readme.md
@@ -1,5 +1,5 @@
 
-Xbyak 5.661 ; JIT assembler for x86(IA32), x64(AMD64, x86-64) by C++
+Xbyak 5.67 ; JIT assembler for x86(IA32), x64(AMD64, x86-64) by C++
 =============
 
 Abstract
@@ -333,6 +333,7 @@ The header files under xbyak/ are independent of cybozulib.
 
 History
 -------------
+* 2018/Aug/14 ver 5.67 remove mutable in Address ; fix setCacheHierarchy for cloud vm
 * 2018/Jul/26 ver 5.661 support mingw64
 * 2018/Jul/24 ver 5.66 add CodeArray::PROTECT_RE to mode of protect()
 * 2018/Jun/26 ver 5.65 fix push(qword [mem])
diff --git a/readme.txt b/readme.txt
index 83038eb..ca38c61 100644
--- a/readme.txt
+++ b/readme.txt
@@ -1,5 +1,5 @@
 

-    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 5.661

+    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 5.67

 

 -----------------------------------------------------------------------------

 ◎概要

@@ -343,6 +343,7 @@ cybozulibは単体テストでのみ利用されていて、xbyak/ディレク�
 -----------------------------------------------------------------------------

 ◎履歴

 

+2018/08/14 ver 5.67 Addressクラス内のmutableを削除 ; fix setCacheHierarchy for cloud vm

 2018/07/26 ver 5.661 mingw64対応

 2018/07/24 ver 5.66 protect()のmodeにCodeArray::PROTECT_REを追加

 2018/06/26 ver 5.65 fix push(qword [mem])

diff --git a/test/bad_address.cpp b/test/bad_address.cpp
index 3cac3fa..a74dd99 100644
--- a/test/bad_address.cpp
+++ b/test/bad_address.cpp
@@ -27,6 +27,8 @@ struct Code : Xbyak::CodeGenerator {
 		TEST_EXCEPTION(mov(eax, ptr [eax + eax + eax]));
 		TEST_EXCEPTION(mov(eax, ptr [eax * 2 + ecx * 4]));
 		TEST_EXCEPTION(mov(eax, ptr [eax * 2 + ecx * 4]));
+		TEST_EXCEPTION(mov(eax, ptr [xmm0]));
+		TEST_EXCEPTION(fld(dword [xmm0]));
 		TEST_EXCEPTION(vgatherdpd(xmm0, ptr [eax * 2], ymm3));
 		TEST_EXCEPTION(vgatherdpd(xmm0, ptr [xmm0 + xmm1], ymm3));
 #ifdef XBYAK64
diff --git a/xbyak/xbyak.h b/xbyak/xbyak.h
index aa4ce87..87d8519 100644
--- a/xbyak/xbyak.h
+++ b/xbyak/xbyak.h
@@ -105,7 +105,7 @@ namespace Xbyak {
 
 enum {
 	DEFAULT_MAX_CODE_SIZE = 4096,
-	VERSION = 0x5661 /* 0xABCD = A.BC(D) */
+	VERSION = 0x5670 /* 0xABCD = A.BC(D) */
 };
 
 #ifndef MIE_INTEGER_TYPE_DEFINED
@@ -1021,46 +1021,43 @@ public:
 		M_ripAddr
 	};
 	Address(uint32 sizeBit, bool broadcast, const RegExp& e)
-		: Operand(0, MEM, sizeBit), e_(e), label_(0), mode_(M_ModRM), permitVsib_(false), broadcast_(broadcast)
+		: Operand(0, MEM, sizeBit), e_(e), label_(0), mode_(M_ModRM), broadcast_(broadcast)
 	{
 		e_.verify();
 	}
 #ifdef XBYAK64
 	explicit Address(size_t disp)
-		: Operand(0, MEM, 64), e_(disp), label_(0), mode_(M_64bitDisp), permitVsib_(false), broadcast_(false){ }
+		: Operand(0, MEM, 64), e_(disp), label_(0), mode_(M_64bitDisp), broadcast_(false){ }
 	Address(uint32 sizeBit, bool broadcast, const RegRip& addr)
-		: Operand(0, MEM, sizeBit), e_(addr.disp_), label_(addr.label_), mode_(addr.isAddr_ ? M_ripAddr : M_rip), permitVsib_(false), broadcast_(broadcast) { }
+		: Operand(0, MEM, sizeBit), e_(addr.disp_), label_(addr.label_), mode_(addr.isAddr_ ? M_ripAddr : M_rip), broadcast_(broadcast) { }
 #endif
-	void permitVsib() const { permitVsib_ = true; }
 	RegExp getRegExp(bool optimize = true) const
 	{
 		return optimize ? e_.optimize() : e_;
 	}
 	Mode getMode() const { return mode_; }
-	bool is32bit() const { verify(); return e_.getBase().getBit() == 32 || e_.getIndex().getBit() == 32; }
-	bool isOnlyDisp() const { verify(); return !e_.getBase().getBit() && !e_.getIndex().getBit(); } // for mov eax
-	size_t getDisp() const { verify(); return e_.getDisp(); }
+	bool is32bit() const { return e_.getBase().getBit() == 32 || e_.getIndex().getBit() == 32; }
+	bool isOnlyDisp() const { return !e_.getBase().getBit() && !e_.getIndex().getBit(); } // for mov eax
+	size_t getDisp() const { return e_.getDisp(); }
 	uint8 getRex() const
 	{
-		verify();
 		if (mode_ != M_ModRM) return 0;
 		return getRegExp().getRex();
 	}
-	bool is64bitDisp() const { verify(); return mode_ == M_64bitDisp; } // for moffset
+	bool is64bitDisp() const { return mode_ == M_64bitDisp; } // for moffset
 	bool isBroadcast() const { return broadcast_; }
 	const Label* getLabel() const { return label_; }
 	bool operator==(const Address& rhs) const
 	{
-		return getBit() == rhs.getBit() && e_ == rhs.e_ && label_ == rhs.label_ && mode_ == rhs.mode_ && permitVsib_ == rhs.permitVsib_ && broadcast_ == rhs.broadcast_;
+		return getBit() == rhs.getBit() && e_ == rhs.e_ && label_ == rhs.label_ && mode_ == rhs.mode_ && broadcast_ == rhs.broadcast_;
 	}
 	bool operator!=(const Address& rhs) const { return !operator==(rhs); }
+	bool isVsib() const { return e_.isVsib(); }
 private:
 	RegExp e_;
 	const Label* label_;
 	Mode mode_;
-	mutable bool permitVsib_;
 	bool broadcast_;
-	void verify() const { if (e_.isVsib() && !permitVsib_) throw Error(ERR_BAD_VSIB_ADDRESSING); }
 };
 
 inline const Address& Operand::getAddress() const
@@ -1465,6 +1462,7 @@ private:
 		T_B32 = 1 << 26, // m32bcst
 		T_B64 = 1 << 27, // m64bcst
 		T_M_K = 1 << 28, // mem{k}
+		T_VSIB = 1 << 29,
 		T_XXX
 	};
 	void vex(const Reg& reg, const Reg& base, const Operand *v, int type, int code, bool x = false)
@@ -1691,8 +1689,9 @@ private:
 	// reg is reg field of ModRM
 	// immSize is the size for immediate value
 	// disp8N = 0(normal), disp8N = 1(force disp32), disp8N = {2, 4, 8} ; compressed displacement
-	void opAddr(const Address &addr, int reg, int immSize = 0, int disp8N = 0)
+	void opAddr(const Address &addr, int reg, int immSize = 0, int disp8N = 0, bool permitVisb = false)
 	{
+		if (!permitVisb && addr.isVsib()) throw Error(ERR_BAD_VSIB_ADDRESSING);
 		if (addr.getMode() == Address::M_ModRM) {
 			setSIB(addr.getRegExp(), reg, disp8N);
 		} else if (addr.getMode() == Address::M_rip || addr.getMode() == Address::M_ripAddr) {
@@ -1952,7 +1951,7 @@ private:
 			} else {
 				vex(r, base, p1, type, code, x);
 			}
-			opAddr(addr, r.getIdx(), (imm8 != NONE) ? 1 : 0, disp8N);
+			opAddr(addr, r.getIdx(), (imm8 != NONE) ? 1 : 0, disp8N, (type & T_VSIB) != 0);
 		} else {
 			const Reg& base = op2.getReg();
 			if ((type & T_MUST_EVEX) || r.hasEvex() || (p1 && p1->hasEvex()) || base.hasEvex()) {
@@ -2053,8 +2052,7 @@ private:
 			}
 			if (!isOK) throw Error(ERR_BAD_VSIB_ADDRESSING);
 		}
-		addr.permitVsib();
-		opAVX_X_X_XM(isAddrYMM ? Ymm(x1.getIdx()) : x1, isAddrYMM ? Ymm(x2.getIdx()) : x2, addr, type | T_YMM, code);
+		opAVX_X_X_XM(isAddrYMM ? Ymm(x1.getIdx()) : x1, isAddrYMM ? Ymm(x2.getIdx()) : x2, addr, type, code);
 	}
 	enum {
 		xx_yy_zz = 0,
@@ -2078,7 +2076,6 @@ private:
 	{
 		if (x.hasZero()) throw Error(ERR_INVALID_ZERO);
 		checkGather2(x, addr.getRegExp().getIndex(), mode);
-		addr.permitVsib();
 		opVex(x, 0, addr, type, code);
 	}
 	/*
@@ -2098,7 +2095,6 @@ private:
 	{
 		if (addr.hasZero()) throw Error(ERR_INVALID_ZERO);
 		if (addr.getRegExp().getIndex().getKind() != kind) throw Error(ERR_BAD_VSIB_ADDRESSING);
-		addr.permitVsib();
 		opVex(x, 0, addr, type, code);
 	}
 public:
diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h
index d71683e..cbd3eba 100644
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@@ -1010,10 +1010,10 @@ void vfnmsub231pd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_X
 void vfnmsub231ps(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0xBE); }
 void vfnmsub231sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_W1 | T_EW1 | T_EVEX | T_ER_X, 0xBF); }
 void vfnmsub231ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F38 | T_W0 | T_EW0 | T_EVEX | T_ER_X, 0xBF); }
-void vgatherdpd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W1, 0x92, 0); }
-void vgatherdps(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W0, 0x92, 1); }
-void vgatherqpd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W1, 0x93, 1); }
-void vgatherqps(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W0, 0x93, 2); }
+void vgatherdpd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W1, 0x92, 0); }
+void vgatherdps(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W0, 0x92, 1); }
+void vgatherqpd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W1, 0x93, 1); }
+void vgatherqps(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W0, 0x93, 2); }
 void vgf2p8affineinvqb(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W1 | T_EW1 | T_YMM | T_EVEX | T_SAE_Z | T_B64, 0xCF, imm); }
 void vgf2p8affineqb(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W1 | T_EW1 | T_YMM | T_EVEX | T_SAE_Z | T_B64, 0xCE, imm); }
 void vgf2p8mulb(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_SAE_Z, 0xCF); }
@@ -1144,10 +1144,10 @@ void vpextrb(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(8|16|
 void vpextrd(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(32) || op.isMEM()) && x.isXMM())) throw Error(ERR_BAD_COMBINATION); opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N4, 0x16, imm); }
 void vpextrq(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(64) || op.isMEM()) && x.isXMM())) throw Error(ERR_BAD_COMBINATION); opVex(x, 0, op, T_0F3A | T_66 | T_W1 | T_EVEX | T_EW1 | T_N8, 0x16, imm); }
 void vpextrw(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(16|i32e) || op.isMEM()) && x.isXMM())) throw Error(ERR_BAD_COMBINATION); if (op.isREG() && x.getIdx() < 16) { opAVX_X_X_XM(Xmm(op.getIdx()), xm0, x, T_0F | T_66, 0xC5, imm); } else { opVex(x, 0, op, T_0F3A | T_66 | T_EVEX | T_N2, 0x15, imm); } }
-void vpgatherdd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W0, 0x90, 1); }
-void vpgatherdq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W1, 0x90, 0); }
-void vpgatherqd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W0, 0x91, 2); }
-void vpgatherqq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W1, 0x91, 1); }
+void vpgatherdd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W0, 0x90, 1); }
+void vpgatherdq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W1, 0x90, 0); }
+void vpgatherqd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W0, 0x91, 2); }
+void vpgatherqq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W1, 0x91, 1); }
 void vphaddd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x02); }
 void vphaddsw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x03); }
 void vphaddw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x01); }
@@ -1729,18 +1729,18 @@ void vfpclasspd(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(1
 void vfpclassps(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(128|256|512)) throw Error(ERR_BAD_MEM_SIZE); Reg x = k; x.setBit(op.getBit()); opVex(x, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32, 0x66, imm); }
 void vfpclasssd(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isXMEM()) throw Error(ERR_BAD_MEM_SIZE); opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_N8, 0x67, imm); }
 void vfpclassss(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isXMEM()) throw Error(ERR_BAD_MEM_SIZE); opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_N4, 0x67, imm); }
-void vgatherdpd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x92, 1); }
-void vgatherdps(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x92, 0); }
-void vgatherpf0dpd(const Address& addr) { opGatherFetch(addr, zm1, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC6, Operand::YMM); }
-void vgatherpf0dps(const Address& addr) { opGatherFetch(addr, zm1, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC6, Operand::ZMM); }
-void vgatherpf0qpd(const Address& addr) { opGatherFetch(addr, zm1, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC7, Operand::ZMM); }
-void vgatherpf0qps(const Address& addr) { opGatherFetch(addr, zm1, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC7, Operand::ZMM); }
-void vgatherpf1dpd(const Address& addr) { opGatherFetch(addr, zm2, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC6, Operand::YMM); }
-void vgatherpf1dps(const Address& addr) { opGatherFetch(addr, zm2, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC6, Operand::ZMM); }
-void vgatherpf1qpd(const Address& addr) { opGatherFetch(addr, zm2, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC7, Operand::ZMM); }
-void vgatherpf1qps(const Address& addr) { opGatherFetch(addr, zm2, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC7, Operand::ZMM); }
-void vgatherqpd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x93, 0); }
-void vgatherqps(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x93, 2); }
+void vgatherdpd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_VSIB, 0x92, 1); }
+void vgatherdps(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_VSIB, 0x92, 0); }
+void vgatherpf0dpd(const Address& addr) { opGatherFetch(addr, zm1, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::YMM); }
+void vgatherpf0dps(const Address& addr) { opGatherFetch(addr, zm1, T_N4 | T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::ZMM); }
+void vgatherpf0qpd(const Address& addr) { opGatherFetch(addr, zm1, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC7, Operand::ZMM); }
+void vgatherpf0qps(const Address& addr) { opGatherFetch(addr, zm1, T_N4 | T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC7, Operand::ZMM); }
+void vgatherpf1dpd(const Address& addr) { opGatherFetch(addr, zm2, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::YMM); }
+void vgatherpf1dps(const Address& addr) { opGatherFetch(addr, zm2, T_N4 | T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::ZMM); }
+void vgatherpf1qpd(const Address& addr) { opGatherFetch(addr, zm2, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC7, Operand::ZMM); }
+void vgatherpf1qps(const Address& addr) { opGatherFetch(addr, zm2, T_N4 | T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC7, Operand::ZMM); }
+void vgatherqpd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_VSIB, 0x93, 0); }
+void vgatherqps(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_VSIB, 0x93, 2); }
 void vgetexppd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x42); }
 void vgetexpps(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x42); }
 void vgetexpsd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_SAE_X | T_MUST_EVEX, 0x43); }
@@ -1827,10 +1827,10 @@ void vpexpandb(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N1 | T
 void vpexpandd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x89); }
 void vpexpandq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x89); }
 void vpexpandw(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N2 | T_66 | T_0F38 | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX, 0x62); }
-void vpgatherdd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x90, 0); }
-void vpgatherdq(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x90, 1); }
-void vpgatherqd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x91, 2); }
-void vpgatherqq(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x91, 0); }
+void vpgatherdd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_VSIB, 0x90, 0); }
+void vpgatherdq(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_VSIB, 0x90, 1); }
+void vpgatherqd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_VSIB, 0x91, 2); }
+void vpgatherqq(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_VSIB, 0x91, 0); }
 void vplzcntd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_B32, 0x44); }
 void vplzcntq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_B64, 0x44); }
 void vpmadd52huq(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_B64, 0xB5); }
@@ -1881,10 +1881,10 @@ void vprord(const Xmm& x, const Operand& op, uint8 imm) { opAVX_X_X_XM(Xmm(x.get
 void vprorq(const Xmm& x, const Operand& op, uint8 imm) { opAVX_X_X_XM(Xmm(x.getKind(), 0), x, op, T_66 | T_0F | T_EW1 | T_YMM | T_MUST_EVEX | T_B64, 0x72, imm); }
 void vprorvd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_B32, 0x14); }
 void vprorvq(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_B64, 0x14); }
-void vpscatterdd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0xA0, 0); }
-void vpscatterdq(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K, 0xA0, 1); }
-void vpscatterqd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0xA1, 2); }
-void vpscatterqq(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K, 0xA1, 0); }
+void vpscatterdd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA0, 0); }
+void vpscatterdq(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA0, 1); }
+void vpscatterqd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA1, 2); }
+void vpscatterqq(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA1, 0); }
 void vpshldd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x71, imm); }
 void vpshldq(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x71, imm); }
 void vpshldvd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x71); }
@@ -1948,18 +1948,18 @@ void vscalefpd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x
 void vscalefps(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0x2C); }
 void vscalefsd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_ER_X | T_MUST_EVEX, 0x2D); }
 void vscalefss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x2D); }
-void vscatterdpd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K, 0xA2, 1); }
-void vscatterdps(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0xA2, 0); }
-void vscatterpf0dpd(const Address& addr) { opGatherFetch(addr, zm5, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC6, Operand::YMM); }
-void vscatterpf0dps(const Address& addr) { opGatherFetch(addr, zm5, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC6, Operand::ZMM); }
-void vscatterpf0qpd(const Address& addr) { opGatherFetch(addr, zm5, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC7, Operand::ZMM); }
-void vscatterpf0qps(const Address& addr) { opGatherFetch(addr, zm5, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC7, Operand::ZMM); }
-void vscatterpf1dpd(const Address& addr) { opGatherFetch(addr, zm6, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC6, Operand::YMM); }
-void vscatterpf1dps(const Address& addr) { opGatherFetch(addr, zm6, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC6, Operand::ZMM); }
-void vscatterpf1qpd(const Address& addr) { opGatherFetch(addr, zm6, T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_N8 | T_M_K, 0xC7, Operand::ZMM); }
-void vscatterpf1qps(const Address& addr) { opGatherFetch(addr, zm6, T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_N4 | T_M_K, 0xC7, Operand::ZMM); }
-void vscatterqpd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K, 0xA3, 0); }
-void vscatterqps(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0xA3, 2); }
+void vscatterdpd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA2, 1); }
+void vscatterdps(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA2, 0); }
+void vscatterpf0dpd(const Address& addr) { opGatherFetch(addr, zm5, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::YMM); }
+void vscatterpf0dps(const Address& addr) { opGatherFetch(addr, zm5, T_N4 | T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::ZMM); }
+void vscatterpf0qpd(const Address& addr) { opGatherFetch(addr, zm5, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC7, Operand::ZMM); }
+void vscatterpf0qps(const Address& addr) { opGatherFetch(addr, zm5, T_N4 | T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC7, Operand::ZMM); }
+void vscatterpf1dpd(const Address& addr) { opGatherFetch(addr, zm6, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::YMM); }
+void vscatterpf1dps(const Address& addr) { opGatherFetch(addr, zm6, T_N4 | T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::ZMM); }
+void vscatterpf1qpd(const Address& addr) { opGatherFetch(addr, zm6, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC7, Operand::ZMM); }
+void vscatterpf1qps(const Address& addr) { opGatherFetch(addr, zm6, T_N4 | T_66 | T_0F38 | T_EW0 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC7, Operand::ZMM); }
+void vscatterqpd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA3, 0); }
+void vscatterqps(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA3, 2); }
 void vshuff32x4(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, 0x23, imm); }
 void vshuff64x2(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, 0x23, imm); }
 void vshufi32x4(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, 0x43, imm); }