add vpdpbssd for avx10.2

4 files changed, 47 insertions, 4 deletions

diff --git a/gen/gen_avx512.cpp b/gen/gen_avx512.cpp
index 9840844..9159a64 100644
--- a/gen/gen_avx512.cpp
+++ b/gen/gen_avx512.cpp
@@ -454,7 +454,37 @@ void putX_X_XM_IMM()
 		printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op%s) { opAVX_X_X_XM(x1, x2, op, %s, 0x%02X%s); }\n"
 			, p->name, p->hasIMM ? ", uint8_t imm" : "", s.c_str(), p->code, p->hasIMM ? ", imm" : "");
 	}
-	puts("void vmpsadbw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, T_0F3A | T_YMM, 0x42, encoding, imm, T_66 | T_W0 | T_YMM, T_F3 | T_0F3A | T_EW0 | T_B32, 1); }");
+}
+
+void putX_X_XM_IMM_AVX10()
+{
+	const struct Tbl {
+		uint8_t code;
+		const char *name;
+		uint64_t type;
+		uint64_t typeVex;
+		uint64_t typeEvex;
+		int sel;
+		bool hasIMM;
+	} tbl[] = {
+		{ 0x50, "vpdpbssd", T_F2|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+#if 0
+		{ 0x51, "vpdpbssds", T_MUST_EVEX | T_YMM | T_F2 | T_0F38 | T_EW0 | T_B32, false },
+		{ 0x50, "vpdpbsud", T_MUST_EVEX | T_YMM | T_F3 | T_0F38 | T_EW0 | T_B32, false },
+		{ 0x51, "vpdpbsuds", T_MUST_EVEX | T_YMM | T_F3 | T_0F38 | T_EW0 | T_B32, false },
+		{ 0x50, "vpdpbuud", T_MUST_EVEX | T_YMM | T_0F38 | T_EW0 | T_B32, false },
+		{ 0x51, "vpdpbuuds", T_MUST_EVEX | T_YMM | T_0F38 | T_EW0 | T_B32, false },
+#endif
+		{ 0x42, "vmpsadbw", T_0F3A|T_YMM, T_66|T_W0|T_YMM, T_F3|T_0F3A|T_EW0|T_B32, 1, true },
+	};
+	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+		const Tbl *p = &tbl[i];
+		std::string s = type2String(p->type);
+		std::string sVex = type2String(p->typeVex);
+		std::string sEvex = type2String(p->typeEvex);
+		printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op%s, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, %s, 0x%02X, encoding, %s, %s, %s, %d); }\n"
+			, p->name, p->hasIMM ? ", uint8_t imm" : "", s.c_str(), p->code, p->hasIMM ? "imm" : "NONE", sVex.c_str(), sEvex.c_str(), p->sel);
+	}
 }
 
 void putShift()
@@ -1059,6 +1089,7 @@ int main(int argc, char *[])
 	putM_X();
 	putXM_X();
 	putX_X_XM_IMM();
+	putX_X_XM_IMM_AVX10();
 	putShift();
 	putExtractInsert();
 	putCvt();
diff --git a/gen/gen_code.cpp b/gen/gen_code.cpp
index 58c176a..caa9e79 100644
--- a/gen/gen_code.cpp
+++ b/gen/gen_code.cpp
@@ -1907,7 +1907,7 @@ void put()
 			const char *name;
 			uint64_t type;
 		} tbl[] = {
-			{ 0x50, "vpdpbssd", T_F2 | T_0F38 | T_W0 | T_YMM },
+//			{ 0x50, "vpdpbssd", T_F2 | T_0F38 | T_W0 | T_YMM },
 			{ 0x51, "vpdpbssds", T_F2 | T_0F38 | T_W0 | T_YMM },
 			{ 0x50, "vpdpbsud", T_F3 | T_0F38 | T_W0 | T_YMM },
 			{ 0x51, "vpdpbsuds", T_F3 | T_0F38 | T_W0 | T_YMM },
diff --git a/test/avx10/misc.txt b/test/avx10/misc.txt
index 5c39e81..8993107 100644
--- a/test/avx10/misc.txt
+++ b/test/avx10/misc.txt
@@ -18,3 +18,15 @@ vmpsadbw(ym1, ym4, ptr[rax+128], 5);
 
 vmpsadbw(zm1|k4, zm3, zm15, 3);
 vmpsadbw(zm1, zm4, ptr[rax+128], 5);
+
+vpdpbssd(xm1, xm2, xm3);
+vpdpbssd(xm1, xm2, ptr[rax+128]);
+vpdpbssd(xm1, xm2, ptr_b[rax+128]);
+
+vpdpbssd(ym1, ym2, ym3);
+vpdpbssd(ym1, ym2, ptr[rax+128]);
+vpdpbssd(ym1, ym2, ptr_b[rax+128]);
+
+vpdpbssd(zm1, zm2, zm3);
+vpdpbssd(zm1, zm2, ptr[rax+128]);
+vpdpbssd(zm1, zm2, ptr_b[rax+128]);
diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h
index 8515e41..dbe52e9 100644
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@@ -1419,7 +1419,6 @@ void vpcmpgtq(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1
 void vpcmpgtw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66|T_0F|T_YMM, 0x65); }
 void vpcmpistri(const Xmm& xm, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(xm, op, T_66|T_0F3A, 0x63, imm); }
 void vpcmpistrm(const Xmm& xm, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(xm, op, T_66|T_0F3A, 0x62, imm); }
-void vpdpbssd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F2|T_0F38|T_W0|T_YMM, 0x50); }
 void vpdpbssds(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F2|T_0F38|T_W0|T_YMM, 0x51); }
 void vpdpbsud(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3|T_0F38|T_W0|T_YMM, 0x50); }
 void vpdpbsuds(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3|T_0F38|T_W0|T_YMM, 0x51); }
@@ -2407,7 +2406,7 @@ void vmovsh(const Xmm& x1, const Xmm& x2, const Xmm& x3) { opAVX_X_X_XM(x1, x2,
 void vmovw(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_N2|T_66|T_MAP5|T_MUST_EVEX, 0x7E); }
 void vmovw(const Reg32e& r, const Xmm& x) { opAVX_X_X_XM(x, xm0, r, T_N2|T_66|T_MAP5|T_MUST_EVEX, 0x7E); }
 void vmovw(const Xmm& x, const Operand& op) { if (!op.isREG(32|64) && !op.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, xm0, op, T_N2|T_66|T_MAP5|T_MUST_EVEX, 0x6E); }
-void vmpsadbw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, T_0F3A | T_YMM, 0x42, encoding, imm, T_66 | T_W0 | T_YMM, T_F3 | T_0F3A | T_EW0 | T_B32, 1); }
+void vmpsadbw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, T_0F3A|T_YMM, 0x42, encoding, imm, T_66|T_W0|T_YMM, T_F3|T_0F3A|T_EW0|T_B32, 1); }
 void vmulnepbf16(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66|T_MAP5|T_EW0|T_YMM|T_MUST_EVEX|T_B16, 0x59); }
 void vmulph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x59); }
 void vmulsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x59); }
@@ -2451,6 +2450,7 @@ void vpcompressq(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N8|T
 void vpcompressw(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N2|T_66|T_0F38|T_EW1|T_YMM|T_MUST_EVEX, 0x63); }
 void vpconflictd(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, 0xC4); }
 void vpconflictq(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, 0xC4); }
+void vpdpbssd(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, T_F2|T_0F38|T_YMM, 0x50, encoding, NONE, T_W0, T_EW0|T_B32, 1); }
 void vpermb(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, 0x8D); }
 void vpermi2b(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, 0x75); }
 void vpermi2d(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, 0x76); }