pdep, pext support apx

herumi · Nov 10, 2023 · 3a85aad · 3a85aad
1 parent 16f1a5d
commit 3a85aad
Show file tree

Hide file tree

Showing 3 changed files with 19 additions and 11 deletions.
diff --git a/gen/gen_code.cpp b/gen/gen_code.cpp
@@ -1734,8 +1734,8 @@ void put()
 		} tbl[] = {
 			{ "andn", T_0F38 | T_VEX | T_NF, 0xF2 },
 			{ "mulx", T_F2 | T_0F38 | T_VEX , 0xF6 },
-			{ "pdep", T_F2 | T_0F38, 0xF5 },
-			{ "pext", T_F3 | T_0F38, 0xF5 },
+			{ "pdep", T_F2 | T_0F38 | T_VEX, 0xF5 },
+			{ "pext", T_F3 | T_0F38 | T_VEX, 0xF5 },
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];

diff --git a/test/apx.cpp b/test/apx.cpp
@@ -464,24 +464,32 @@ CYBOZU_TEST_AUTO(andn_etc)
 		{
 			andn(rax, rcx, r8); // legacy
 			andn(rax|T_nf, rcx, r8); // with NF
-			andn(eax, ecx, r8d); // 32bit
+			andn(eax, ecx, r17d); // 32bit
 			andn(r29, r30, r31); // 64bit
 			andn(r29, r30, ptr [r31+r20*4]);
 
-			mulx(rax, rcx, r8);
-			mulx(eax, ecx, r8d);
+			mulx(eax, ecx, r17d);
 			mulx(r29, r30, r31);
 			mulx(r29, r30, ptr [r31+r20*4]);
+			pdep(eax, ecx, r17d);
+			pdep(r29, r30, r31);
+			pdep(r29, r30, ptr [r31+r20*4]);
+			pext(eax, ecx, r17d);
+			pext(r29, r30, r31);
+			pext(r29, r30, ptr [r31+r20*4]);
+
 		}
 	} c;
 	const uint8_t tbl[] = {
 		// andn
-		0xc4, 0xc2, 0xf0, 0xf2, 0xc0, 0x62, 0xd2, 0xf4, 0x08, 0xf2, 0xc0, 0xc4, 0xc2, 0x70, 0xf2, 0xc0,
+		0xc4, 0xc2, 0xf0, 0xf2, 0xc0,
+		0x62, 0xd2, 0xf4, 0x08, 0xf2, 0xc0,
+		0x62, 0xfa, 0x74, 0x08, 0xf2, 0xc1,
 		0x62, 0x4a, 0x8c, 0x00, 0xf2, 0xef, 0x62, 0x4a, 0x88, 0x00, 0xf2, 0x2c, 0xa7,
 
-		// mulx
-		0xc4, 0xc2, 0xf3, 0xf6, 0xc0, 0xc4, 0xc2, 0x73, 0xf6, 0xc0, 0x62, 0x4a, 0x8f, 0x00, 0xf6, 0xef,
-0x62, 0x4a, 0x8b, 0x00, 0xf6, 0x2c, 0xa7,
+		0x62, 0xfa, 0x77, 0x08, 0xf6, 0xc1, 0x62, 0x4a, 0x8f, 0x00, 0xf6, 0xef, 0x62, 0x4a, 0x8b, 0x00, 0xf6, 0x2c, 0xa7,
+		0x62, 0xfa, 0x77, 0x08, 0xf5, 0xc1, 0x62, 0x4a, 0x8f, 0x00, 0xf5, 0xef, 0x62, 0x4a, 0x8b, 0x00, 0xf5, 0x2c, 0xa7,
+		0x62, 0xfa, 0x76, 0x08, 0xf5, 0xc1, 0x62, 0x4a, 0x8e, 0x00, 0xf5, 0xef, 0x62, 0x4a, 0x8a, 0x00, 0xf5, 0x2c, 0xa7,
 	};
 	const size_t n = sizeof(tbl);
 	CYBOZU_TEST_EQUAL(c.getSize(), n);

diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h
@@ -616,8 +616,8 @@ void pcmpgtq(const Xmm& xmm, const Operand& op) { opSSE(xmm, op, T_66 | T_0F38,
 void pcmpgtw(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x65); }
 void pcmpistri(const Xmm& xmm, const Operand& op, uint8_t imm) { opSSE(xmm, op, T_66|T_0F3A, 0x63, isXMM_XMMorMEM, imm); }
 void pcmpistrm(const Xmm& xmm, const Operand& op, uint8_t imm) { opSSE(xmm, op, T_66|T_0F3A, 0x62, isXMM_XMMorMEM, imm); }
-void pdep(const Reg& r1, const Reg32e& r2, const Operand& op) { opRRO(r1, r2, op, T_F2|T_0F38, 0xf5); }
-void pext(const Reg& r1, const Reg32e& r2, const Operand& op) { opRRO(r1, r2, op, T_F3|T_0F38, 0xf5); }
+void pdep(const Reg& r1, const Reg32e& r2, const Operand& op) { opRRO(r1, r2, op, T_VEX|T_F2|T_0F38, 0xf5); }
+void pext(const Reg& r1, const Reg32e& r2, const Operand& op) { opRRO(r1, r2, op, T_VEX|T_F3|T_0F38, 0xf5); }
 void pextrb(const Operand& op, const Xmm& xmm, uint8_t imm) { opExt(op, xmm, 0x14, imm); }
 void pextrd(const Operand& op, const Xmm& xmm, uint8_t imm) { opExt(op, xmm, 0x16, imm); }
 void pextrw(const Operand& op, const Mmx& xmm, uint8_t imm) { opExt(op, xmm, 0x15, imm, true); }