early-access version 1536

2021-03-29 00:03:19 +02:00 · 2021-03-29 00:03:19 +02:00 · 92a647d0ae
parent eeca4f785c
commit 92a647d0ae
29 changed files with 1017 additions and 184 deletions
--- a/README.md
+++ b/README.md
@ -1,7 +1,7 @@
 yuzu emulator early access
 =============

-This is the source code for early-access 1535.
+This is the source code for early-access 1536.

 ## Legal Notice

--- a/externals/dynarmic/externals/xbyak/.github/workflows/main.yml
+++ b/externals/dynarmic/externals/xbyak/.github/workflows/main.yml
@ -0,0 +1,11 @@
+name: test
+on: [push]
+
+jobs:
+  build:
+    name: test
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v2
+    - run: sudo apt install nasm yasm g++-multilib tcsh
+    - run: make test
--- a/externals/dynarmic/externals/xbyak/.gitignore
+++ b/externals/dynarmic/externals/xbyak/.gitignore
@ -0,0 +1 @@
+/build* # cmake
--- a/externals/dynarmic/externals/xbyak/CMakeLists.txt
+++ b/externals/dynarmic/externals/xbyak/CMakeLists.txt
@ -1,6 +1,46 @@
-cmake_minimum_required(VERSION 2.6)
-project(xbyak)
+cmake_minimum_required(VERSION 2.6...3.0.2)
+
+project(xbyak CXX)

 file(GLOB headers xbyak/*.h)
-install(FILES ${headers} DESTINATION include/xbyak)

+if (DEFINED CMAKE_VERSION AND CMAKE_VERSION VERSION_GREATER_EQUAL 3.0.2)
+	include(GNUInstallDirs)
+	add_library(${PROJECT_NAME} INTERFACE)
+	add_library(${PROJECT_NAME}::${PROJECT_NAME} ALIAS ${PROJECT_NAME})
+	
+	target_include_directories(
+		${PROJECT_NAME} INTERFACE
+		"$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>"
+		"$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>"
+	)
+
+	install(
+		TARGETS ${PROJECT_NAME}
+		EXPORT ${PROJECT_NAME}-targets
+	)
+
+	configure_file(
+		cmake/config.cmake.in
+		${PROJECT_NAME}Config.cmake
+		@ONLY
+	)
+
+	install(
+		FILES ${CMAKE_CURRENT_BINARY_DIR}/${PROJECT_NAME}Config.cmake
+		DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME}
+	)
+
+	install(
+		EXPORT ${PROJECT_NAME}-targets
+		NAMESPACE ${PROJECT_NAME}::
+		DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME}
+	)
+elseif(NOT DEFINED CMAKE_INSTALL_INCLUDEDIR)
+	set(CMAKE_INSTALL_INCLUDEDIR "include")
+endif()
+
+install(
+	FILES ${headers}
+	DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/xbyak
+)
--- a/externals/dynarmic/externals/xbyak/cmake/config.cmake.in
+++ b/externals/dynarmic/externals/xbyak/cmake/config.cmake.in
@ -0,0 +1 @@
+include("${CMAKE_CURRENT_LIST_DIR}/@PROJECT_NAME@-targets.cmake")
--- a/externals/dynarmic/externals/xbyak/gen/gen_avx512.cpp
+++ b/externals/dynarmic/externals/xbyak/gen/gen_avx512.cpp
@ -363,12 +363,6 @@ void putX_X_XM_IMM()
 		{ 0x73, "vpshrdvd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
 		{ 0x73, "vpshrdvq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_B64, false },

-		{ 0x50, "vpdpbusd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
-		{ 0x51, "vpdpbusds", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
-
-		{ 0x52, "vpdpwssd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
-		{ 0x53, "vpdpwssds", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
-
 		{ 0x72, "vcvtne2ps2bf16", T_F2 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
 		{ 0x52, "vdpbf16ps", T_F3 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
 	};
--- a/externals/dynarmic/externals/xbyak/gen/gen_code.cpp
+++ b/externals/dynarmic/externals/xbyak/gen/gen_code.cpp
@ -1729,6 +1729,24 @@ void put()
 			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);
 		}
 	}
+	// vnni
+	{
+		const struct Tbl {
+			uint8_t code;
+			const char *name;
+			int type;
+		} tbl[] = {
+			{ 0x50, "vpdpbusd", T_66 | T_0F38 | T_YMM | T_EW0 | T_SAE_Z | T_B32},
+			{ 0x51, "vpdpbusds", T_66 | T_0F38 | T_YMM | T_EW0 | T_SAE_Z | T_B32},
+			{ 0x52, "vpdpwssd", T_66 | T_0F38 | T_YMM | T_EW0 | T_SAE_Z | T_B32},
+			{ 0x53, "vpdpwssds", T_66 | T_0F38 | T_YMM | T_EW0 | T_SAE_Z | T_B32},
+		};
+		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+			const Tbl *p = &tbl[i];
+			std::string type = type2String(p->type);
+			printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opVnni(x1, x2, op, %s, 0x%02X, encoding); }\n", p->name, type.c_str(), p->code);
+		}
+	}
 }

 void put32()
--- a/externals/dynarmic/externals/xbyak/readme.md
+++ b/externals/dynarmic/externals/xbyak/readme.md
@ -1,11 +1,14 @@
-[![Build Status](https://travis-ci.org/herumi/xbyak.png)](https://travis-ci.org/herumi/xbyak)
+[![Build Status](https://github.com/herumi/xbyak/actions/workflows/main.yml/badge.svg)](https://github.com/herumi/xbyak/actions/workflows/main.yml)

-# Xbyak 5.97 ; JIT assembler for x86(IA32), x64(AMD64, x86-64) by C++
+# Xbyak 5.991 ; JIT assembler for x86(IA32), x64(AMD64, x86-64) by C++

 ## Abstract

 Xbyak is a C++ header library that enables dynamically to assemble x86(IA32), x64(AMD64, x86-64) mnemonic.

+The pronunciation of Xbyak is `kəi-bja-k`.
+It is named from a Japanese word [開闢](https://translate.google.com/?hl=ja&sl=ja&tl=en&text=%E9%96%8B%E9%97%A2&op=translate), which means the beginning of the world.
+
 ## Feature
 * header file only
 * Intel/MASM like syntax
@ -16,6 +19,7 @@ Use `and_()`, `or_()`, ... instead of `and()`, `or()`.
 If you want to use them, then specify `-fno-operator-names` option to gcc/clang.

 ### News
+- vnni instructions such as vpdpbusd supports vex encoding.
 - (break backward compatibility) `push(byte, imm)` (resp. `push(word, imm)`) forces to cast `imm` to 8(resp. 16) bit.
 - (Windows) `#include <winsock2.h>` has been removed from xbyak.h, so add it explicitly if you need it.
 - support exception-less mode see. [Exception-less mode](#exception-less-mode)
@ -154,6 +158,10 @@ vcvtpd2dq xmm19, [eax+32]{1to4}         --> vcvtpd2dq(xmm19, yword_b [eax+32]);
 vfpclassps k5{k3}, zword [rax+64], 5    --> vfpclassps(k5|k3, zword [rax+64], 5); // specify m512
 vfpclasspd k5{k3}, [rax+64]{1to2}, 5    --> vfpclasspd(k5|k3, xword_b [rax+64], 5); // broadcast 64-bit to 128-bit
 vfpclassps k5{k3}, [rax+64]{1to4}, 5    --> vfpclassps(k5|k3, yword_b [rax+64], 5); // broadcast 64-bit to 256-bit
+
+vpdpbusd(xm0, xm1, xm2); // default encoding is EVEX
+vpdpbusd(xm0, xm1, xm2, EvexEncoding); // same as the above
+vpdpbusd(xm0, xm1, xm2, VexEncoding); // VEX encoding
 ```
 ### Remark
 * `k1`, ..., `k7` are opmask registers.
@ -339,9 +347,9 @@ public:

 ## User allocated memory

-You can make jit code on prepaired memory.
+You can make jit code on prepared memory.

-Call `setProtectModeRE` yourself to change memory mode if using the prepaired memory.
+Call `setProtectModeRE` yourself to change memory mode if using the prepared memory.

 ```
 uint8_t alignas(4096) buf[8192]; // C++11 or later
@ -438,6 +446,9 @@ modified new BSD License
 http://opensource.org/licenses/BSD-3-Clause

 ## History
+* 2020/Nov/16 ver 5.991 disable constexpr for gcc-5 with -std=c++-14
+* 2020/Oct/19 ver 5.99 support VNNI instructions(Thanks to akharito)
+* 2020/Oct/17 ver 5.98 support the form of [scale * reg]
 * 2020/Sep/08 ver 5.97 replace uint32 with uint32_t etc.
 * 2020/Aug/28 ver 5.95 some constructors of register classes support constexpr if C++14 or later
 * 2020/Aug/04 ver 5.941 `CodeGenerator::reset()` calls `ClearError()`.
--- a/externals/dynarmic/externals/xbyak/readme.txt
+++ b/externals/dynarmic/externals/xbyak/readme.txt
@ -1,5 +1,5 @@

-    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 5.97
+    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 5.991

 -----------------------------------------------------------------------------
 ◎概要
@ -163,6 +163,9 @@ vfpclassps k5{k3}, zword [rax+64], 5    --> vfpclassps(k5|k3, zword [rax+64], 5)
 vfpclasspd k5{k3}, [rax+64]{1to2}, 5    --> vfpclasspd(k5|k3, xword_b [rax+64], 5); // broadcast 64-bit to 128-bit
 vfpclassps k5{k3}, [rax+64]{1to4}, 5    --> vfpclassps(k5|k3, xword_b [rax+64], 5); // broadcast 64-bit to 256-bit

+vpdpbusd(xm0, xm1, xm2); // default encoding is EVEX
+vpdpbusd(xm0, xm1, xm2, EvexEncoding); // same as the above
+vpdpbusd(xm0, xm1, xm2, VexEncoding); // VEX encoding

 注意
 * k1, ..., k7 は新しいopmaskレジスタです。
@ -379,6 +382,9 @@ sample/{echo,hello}.bfは http://www.kmonos.net/alang/etc/brainfuck.php から
 -----------------------------------------------------------------------------
 ◎履歴

+2020/11/16 ver 5.991 g++-5のC++14でconstexpr機能の抑制
+2020/10/19 ver 5.99 VNNI命令サポート(Thanks to akharito)
+2020/10/17 ver 5.98 [scale * reg]のサポート
 2020/09/08 ver 5.97 uint32などをuint32_tに置換
 2020/08/28 ver 5.95 レジスタクラスのコンストラクタがconstexprに対応(C++14以降)
 2020/08/04 ver 5.941 `CodeGenerator::reset()`が`ClearError()`を呼ぶように変更
--- a/externals/dynarmic/externals/xbyak/sample/test_util.cpp
+++ b/externals/dynarmic/externals/xbyak/sample/test_util.cpp
@ -1,12 +1,13 @@
 #include <stdio.h>
-#define XBYAK_NO_OP_NAMES
 #include "xbyak/xbyak_util.h"

 #define NUM_OF_ARRAY(x) (sizeof(x) / sizeof(x[0]))

 struct PopCountTest : public Xbyak::CodeGenerator {
 	PopCountTest(int n)
+		: Xbyak::CodeGenerator(4096, Xbyak::DontSetProtectRWE)
 	{
+ret();
 		mov(eax, n);
 		popcnt(eax, eax);
 		ret();
@ -80,6 +81,10 @@ void putCPUinfo()
 		{ Cpu::tAVX512_VPOPCNTDQ, "avx512_vpopcntdq" },
 		{ Cpu::tAVX512_BF16, "avx512_bf16" },
 		{ Cpu::tAVX512_VP2INTERSECT, "avx512_vp2intersect" },
+		{ Cpu::tAMX_TILE, "amx(tile)" },
+		{ Cpu::tAMX_INT8, "amx(int8)" },
+		{ Cpu::tAMX_BF16, "amx(bf16)" },
+		{ Cpu::tAVX_VNNI, "avx_vnni" },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		if (cpu.has(tbl[i].type)) printf(" %s", tbl[i].str);
@ -88,12 +93,16 @@ void putCPUinfo()
 	if (cpu.has(Cpu::tPOPCNT)) {
 		const int n = 0x12345678; // bitcount = 13
 		const int ok = 13;
-		int r = PopCountTest(n).getCode<int (*)()>()();
+		PopCountTest code(n);
+		code.setProtectModeRE();
+		int (*f)() = code.getCode<int (*)()>();
+		int r = f();
 		if (r == ok) {
 			puts("popcnt ok");
 		} else {
 			printf("popcnt ng %d %d\n", r, ok);
 		}
+		code.setProtectModeRW();
 	}
 	/*
 		                displayFamily displayModel
--- a/externals/dynarmic/externals/xbyak/test/make_nm.cpp
+++ b/externals/dynarmic/externals/xbyak/test/make_nm.cpp
@ -643,6 +643,7 @@ class Test {
 		puts(isXbyak_ ? "out_(dx, al); dump();" : "out dx, al");
 		puts(isXbyak_ ? "out_(dx, ax); dump();" : "out dx, ax");
 		puts(isXbyak_ ? "out_(dx, eax); dump();" : "out dx, eax");
+		puts(isXbyak_ ? "lea(eax, ptr[edi + 4 * eax]); dump();" : "lea eax, [edi + 4 * eax]");
 	}
 	void putJmp() const
 	{
--- a/externals/dynarmic/externals/xbyak/test/misc.cpp
+++ b/externals/dynarmic/externals/xbyak/test/misc.cpp
@ -815,4 +815,32 @@ CYBOZU_TEST_AUTO(tileloadd)
 	CYBOZU_TEST_EXCEPTION(c.notSupported(), std::exception);
 	CYBOZU_TEST_EXCEPTION(c.notSupported2(), std::exception);
 }
+
+CYBOZU_TEST_AUTO(vnni)
+{
+	struct Code : Xbyak::CodeGenerator {
+		Code()
+		{
+			// default encoding is EVEX
+			vpdpbusd(xm0, xm1, xm2);
+			vpdpbusd(xm0, xm1, xm2, EvexEncoding); // EVEX
+			vpdpbusd(xm0, xm1, xm2, VexEncoding); // VEX
+		}
+		void badVex()
+		{
+			vpdpbusd(xm0, xm1, xm31, VexEncoding);
+		}
+	} c;
+	const uint8_t tbl[] = {
+		0x62, 0xF2, 0x75, 0x08, 0x50, 0xC2,
+		0x62, 0xF2, 0x75, 0x08, 0x50, 0xC2,
+		0xC4, 0xE2, 0x71, 0x50, 0xC2,
+	};
+	const size_t n = sizeof(tbl) / sizeof(tbl[0]);
+	CYBOZU_TEST_EQUAL(c.getSize(), n);
+	CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
+
+	CYBOZU_TEST_EXCEPTION(c.badVex(), std::exception);
+}
+
 #endif
--- a/externals/dynarmic/externals/xbyak/test/test_address.sh
+++ b/externals/dynarmic/externals/xbyak/test/test_address.sh
@ -12,7 +12,7 @@ g++ $CFLAGS address.cpp -o address
 ./address $1 > a.asm
 echo "asm"
 $EXE -f$OPT3 a.asm -l a.lst
-awk '{if (index($3, "-")) { conti=substr($3, 0, length($3) - 1) } else { conti = conti $3; print conti; conti = "" }} ' < a.lst | $FILTER > ok.lst
+awk '{printf "%s", sub(/-$/, "", $3) ? $3 : $3 ORS}' a.lst | $FILTER > ok.lst

 echo "xbyak"
 ./address $1 jit > nm.cpp
--- a/externals/dynarmic/externals/xbyak/test/test_avx.sh
+++ b/externals/dynarmic/externals/xbyak/test/test_avx.sh
@ -1,39 +1,44 @@
-#!/bin/tcsh
+#!/bin/sh

-set FILTER="grep -v warning"
+FILTER="grep -v warning"

-if ($1 == "Y") then
+case $1 in
+Y)
 	echo "yasm(32bit)"
-	set EXE=yasm
-	set OPT2="-DUSE_YASM -DXBYAK32"
-	set OPT3=win32
-else if ($1 == "64") then
+	EXE=yasm
+	OPT2="-DUSE_YASM -DXBYAK32"
+	OPT3=win32
+	;;
+64)
 	echo "nasm(64bit)"
-	set EXE=nasm
-	set OPT2=-DXBYAK64
-	set OPT3=win64
-	set FILTER=./normalize_prefix
-else if ($1 == "Y64") then
+	EXE=nasm
+	OPT2=-DXBYAK64
+	OPT3=win64
+	FILTER=./normalize_prefix
+	;;
+Y64)
 	echo "yasm(64bit)"
-	set EXE=yasm
-	set OPT2="-DUSE_YASM -DXBYAK64"
-	set OPT3=win64
-	set FILTER=./normalize_prefix
-else
+	EXE=yasm
+	OPT2="-DUSE_YASM -DXBYAK64"
+	OPT3=win64
+	FILTER=./normalize_prefix
+	;;
+*)
 	echo "nasm(32bit)"
-	set EXE=nasm
-	set OPT2=-DXBYAK32
-	set OPT3=win32
-endif
+	EXE=nasm
+	OPT2=-DXBYAK32
+	OPT3=win32
+	;;
+esac

-set CFLAGS="-Wall -fno-operator-names -I../ $OPT2 -DUSE_AVX"
+CFLAGS="-Wall -fno-operator-names -I../ $OPT2 -DUSE_AVX"
 echo "compile make_nm.cpp"
 g++ $CFLAGS make_nm.cpp -o make_nm

 ./make_nm > a.asm
 echo "asm"
 $EXE -f$OPT3 a.asm -l a.lst
-awk '{if (index($3, "-")) { conti=substr($3, 0, length($3) - 1) } else { conti = conti $3; print conti; conti = "" }} ' < a.lst | $FILTER | grep -v "1+1" > ok.lst
+awk '$3 != "1+1" {printf "%s", sub(/-$/, "", $3) ? $3 : $3 ORS}' a.lst | $FILTER > ok.lst

 echo "xbyak"
 ./make_nm jit > nm.cpp
--- a/externals/dynarmic/externals/xbyak/test/test_avx512.sh
+++ b/externals/dynarmic/externals/xbyak/test/test_avx512.sh
@ -1,28 +1,31 @@
-#!/bin/tcsh
+#!/bin/sh

-set FILTER="grep -v warning"
+FILTER="grep -v warning"

-if ($1 == "64") then
+case $1 in
+64)
 	echo "nasm(64bit)"
-	set EXE=nasm
-	set OPT2=-DXBYAK64
-	set OPT3=win64
-	set FILTER=./normalize_prefix
-else
+	EXE=nasm
+	OPT2=-DXBYAK64
+	OPT3=win64
+	FILTER=./normalize_prefix
+	;;
+*)
 	echo "nasm(32bit)"
-	set EXE=nasm
-	set OPT2=-DXBYAK32
-	set OPT3=win32
-endif
+	EXE=nasm
+	OPT2=-DXBYAK32
+	OPT3=win32
+	;;
+esac

-set CFLAGS="-Wall -fno-operator-names -I../ $OPT2 -DUSE_AVX512"
+CFLAGS="-Wall -fno-operator-names -I../ $OPT2 -DUSE_AVX512"
 echo "compile make_512.cpp"
 g++ $CFLAGS make_512.cpp -o make_512

 ./make_512 > a.asm
 echo "asm"
 $EXE -f$OPT3 a.asm -l a.lst
-awk '{if (index($3, "-")) { conti=substr($3, 0, length($3) - 1) } else { conti = conti $3; print conti; conti = "" }} ' < a.lst | $FILTER > ok.lst
+awk '{printf "%s", sub(/-$/, "", $3) ? $3 : $3 ORS}' a.lst | $FILTER > ok.lst

 echo "xbyak"
 ./make_512 jit > nm.cpp
--- a/externals/dynarmic/externals/xbyak/test/test_nm.sh
+++ b/externals/dynarmic/externals/xbyak/test/test_nm.sh
@ -1,50 +1,57 @@
-#!/bin/tcsh
+#!/bin/sh

-set FILTER=cat
+FILTER=cat

-if ($1 == "Y") then
+case $1 in
+Y)
 	echo "yasm(32bit)"
-	set EXE=yasm
-	set OPT2="-DUSE_YASM -DXBYAK32"
-	set OPT3=win32
-else if ($1 == "64") then
+	EXE=yasm
+	OPT2="-DUSE_YASM -DXBYAK32"
+	OPT3=win32
+	;;
+64)
 	echo "nasm(64bit)"
-	set EXE=nasm
-	set OPT2=-DXBYAK64
-	set OPT3=win64
-	set FILTER=./normalize_prefix
-else if ($1 == "Y64") then
+	EXE=nasm
+	OPT2=-DXBYAK64
+	OPT3=win64
+	FILTER=./normalize_prefix
+	;;
+Y64)
 	echo "yasm(64bit)"
-	set EXE=yasm
-	set OPT2="-DUSE_YASM -DXBYAK64"
-	set OPT3=win64
-	set FILTER=./normalize_prefix
-else if ($1 == "avx512") then
+	EXE=yasm
+	OPT2="-DUSE_YASM -DXBYAK64"
+	OPT3=win64
+	FILTER=./normalize_prefix
+	;;
+avx512)
 	echo "nasm(64bit) + avx512"
-	set EXE=nasm
-	set OPT2="-DXBYAK64 -DUSE_AVX512"
-	set OPT3=win64
-	set FILTER=./normalize_prefix
-else if ($1 == "noexcept") then
+	EXE=nasm
+	OPT2="-DXBYAK64 -DUSE_AVX512"
+	OPT3=win64
+	FILTER=./normalize_prefix
+	;;
+noexcept)
 	echo "nasm(32bit) without exception"
-	set EXE=nasm
-	set OPT2="-DXBYAK32 -DXBYAK_NO_EXCEPTION"
-	set OPT3=win32
-else
+	EXE=nasm
+	OPT2="-DXBYAK32 -DXBYAK_NO_EXCEPTION"
+	OPT3=win32
+	;;
+*)
 	echo "nasm(32bit)"
-	set EXE=nasm
-	set OPT2=-DXBYAK32
-	set OPT3=win32
-endif
+	EXE=nasm
+	OPT2=-DXBYAK32
+	OPT3=win32
+	;;
+esac

-set CFLAGS="-Wall -fno-operator-names -I../ $OPT2"
+CFLAGS="-Wall -fno-operator-names -I../ $OPT2"
 echo "compile make_nm.cpp with $CFLAGS"
 g++ $CFLAGS make_nm.cpp -o make_nm

 ./make_nm > a.asm
 echo "asm"
 $EXE -f$OPT3 a.asm -l a.lst
-awk '{if (index($3, "-")) { conti=substr($3, 0, length($3) - 1) } else { conti = conti $3; print conti; conti = "" }} ' < a.lst | $FILTER | grep -v "1+1" > ok.lst
+awk '$3 != "1+1" {printf "%s", sub(/-$/, "", $3) ? $3 : $3 ORS}' a.lst | $FILTER > ok.lst

 echo "xbyak"
 ./make_nm jit > nm.cpp
--- a/externals/dynarmic/externals/xbyak/xbyak/xbyak.h
+++ b/externals/dynarmic/externals/xbyak/xbyak/xbyak.h
@ -108,7 +108,7 @@
 	#endif
 #endif

-#if (__cplusplus >= 201103) || (_MSC_VER >= 1800)
+#if (__cplusplus >= 201103) || (defined(_MSC_VER) && _MSC_VER >= 1800)
 	#undef XBYAK_TLS
 	#define XBYAK_TLS thread_local
 	#define XBYAK_VARIADIC_TEMPLATE
@ -117,8 +117,11 @@
 	#define XBYAK_NOEXCEPT throw()
 #endif

-#if (__cplusplus >= 201402L) || (_MSC_VER >= 1910) // Visual Studio 2017 version 15.0
-	#define XBYAK_CONSTEXPR constexpr // require c++14 or later
+// require c++14 or later
+// Visual Studio 2017 version 15.0 or later
+// g++-6 or later
+#if ((__cplusplus >= 201402L) && !(!defined(__clang__) && defined(__GNUC__) && (__GNUC__ <= 5))) || (defined(_MSC_VER) && _MSC_VER >= 1910)
+	#define XBYAK_CONSTEXPR constexpr
 #else
 	#define XBYAK_CONSTEXPR
 #endif
@ -135,7 +138,7 @@ namespace Xbyak {

 enum {
 	DEFAULT_MAX_CODE_SIZE = 4096,
-	VERSION = 0x5970 /* 0xABCD = A.BC(D) */
+	VERSION = 0x5991 /* 0xABCD = A.BC(D) */
 };

 #ifndef MIE_INTEGER_TYPE_DEFINED
@ -413,16 +416,16 @@ public:
 	{
 		const size_t alignedSizeM1 = inner::ALIGN_PAGE_SIZE - 1;
 		size = (size + alignedSizeM1) & ~alignedSizeM1;
-#if defined(XBYAK_USE_MAP_JIT)
+#if defined(MAP_ANONYMOUS)
 		int mode = MAP_PRIVATE | MAP_ANONYMOUS;
-		const int mojaveVersion = 18;
-		if (util::getMacOsVersion() >= mojaveVersion) mode |= MAP_JIT;
-#elif defined(MAP_ANONYMOUS)
-		const int mode = MAP_PRIVATE | MAP_ANONYMOUS;
 #elif defined(MAP_ANON)
-		const int mode = MAP_PRIVATE | MAP_ANON;
+		int mode = MAP_PRIVATE | MAP_ANON;
 #else
 		#error "not supported"
+#endif
+#if defined(XBYAK_USE_MAP_JIT)
+		const int mojaveVersion = 18;
+		if (util::getMacOsVersion() >= mojaveVersion) mode |= MAP_JIT;
 #endif
 		void *p = mmap(NULL, size, PROT_READ | PROT_WRITE, mode, -1, 0);
 		if (p == MAP_FAILED) XBYAK_THROW_RET(ERR_CANT_ALLOC, 0)
@ -924,6 +927,10 @@ inline RegExp operator*(const Reg& r, int scale)
 {
 	return RegExp(r, scale);
 }
+inline RegExp operator*(int scale, const Reg& r)
+{
+	return r * scale;
+}
 inline RegExp operator-(const RegExp& e, size_t disp)
 {
 	RegExp ret = e;
@ -1539,6 +1546,12 @@ inline const uint8_t* Label::getAddress() const
 	return mgr->getCode() + offset;
 }

+typedef enum {
+	DefaultEncoding,
+	VexEncoding,
+	EvexEncoding
+} PreferredEncoding;
+
 class CodeGenerator : public CodeArray {
 public:
 	enum LabelType {
@ -2293,6 +2306,19 @@ private:
 		if (addr.getRegExp().getIndex().getKind() != kind) XBYAK_THROW(ERR_BAD_VSIB_ADDRESSING)
 		opVex(x, 0, addr, type, code);
 	}
+	void opVnni(const Xmm& x1, const Xmm& x2, const Operand& op, int type, int code0, PreferredEncoding encoding)
+	{
+		if (encoding == DefaultEncoding) {
+			encoding = EvexEncoding;
+		}
+		if (encoding == EvexEncoding) {
+#ifdef XBYAK_DISABLE_AVX512
+			XBYAK_THROW(ERR_EVEX_IS_INVALID)
+#endif
+			type |= T_MUST_EVEX;
+		}
+		opAVX_X_X_XM(x1, x2, op, type, code0);
+	}
 	void opInOut(const Reg& a, const Reg& d, uint8_t code)
 	{
 		if (a.getIdx() == Operand::AL && d.getIdx() == Operand::DX && d.getBit() == 16) {
--- a/externals/dynarmic/externals/xbyak/xbyak/xbyak_mnemonic.h
+++ b/externals/dynarmic/externals/xbyak/xbyak/xbyak_mnemonic.h
@ -1,4 +1,4 @@
-const char *getVersionString() const { return "5.97"; }
+const char *getVersionString() const { return "5.991"; }
 void adc(const Operand& op, uint32_t imm) { opRM_I(op, imm, 0x10, 2); }
 void adc(const Operand& op1, const Operand& op2) { opRM_RM(op1, op2, 0x10); }
 void adcx(const Reg32e& reg, const Operand& op) { opGen(reg, op, 0xF6, 0x66, isREG32_REG32orMEM, NONE, 0x38); }
@ -1180,6 +1180,10 @@ 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 vpdpbusd(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opVnni(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32, 0x50, encoding); }
+void vpdpbusds(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opVnni(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32, 0x51, encoding); }
+void vpdpwssd(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opVnni(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32, 0x52, encoding); }
+void vpdpwssds(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opVnni(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32, 0x53, encoding); }
 void vperm2f128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8_t imm) { if (!(y1.isYMM() && y2.isYMM() && op.isYMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x06, imm); }
 void vperm2i128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8_t imm) { if (!(y1.isYMM() && y2.isYMM() && op.isYMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x46, imm); }
 void vpermd(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x36); }
@ -1883,10 +1887,6 @@ void vpcompressd(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N4 |
 void vpcompressq(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x8B); }
 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 vpdpbusd(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, 0x50); }
-void vpdpbusds(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, 0x51); }
-void vpdpwssd(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, 0x52); }
-void vpdpwssds(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, 0x53); }
 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); }
--- a/externals/dynarmic/externals/xbyak/xbyak/xbyak_util.h
+++ b/externals/dynarmic/externals/xbyak/xbyak/xbyak_util.h
@ -368,6 +368,7 @@ public:
 	static const Type tAMX_TILE = uint64_t(1) << 59;
 	static const Type tAMX_INT8 = uint64_t(1) << 60;
 	static const Type tAMX_BF16 = uint64_t(1) << 61;
+	static const Type tAVX_VNNI = uint64_t(1) << 62;

 	Cpu()
 		: type_(NONE)
@ -389,19 +390,35 @@ public:
 		if (ECX == get32bitAsBE(amd)) {
 			type_ |= tAMD;
 			getCpuid(0x80000001, data);
-			if (EDX & (1U << 31)) type_ |= t3DN;
-			if (EDX & (1U << 15)) type_ |= tCMOV;
-			if (EDX & (1U << 30)) type_ |= tE3DN;
-			if (EDX & (1U << 22)) type_ |= tMMX2;
-			if (EDX & (1U << 27)) type_ |= tRDTSCP;
+			if (EDX & (1U << 31)) {
+				type_ |= t3DN;
+				// 3DNow! implies support for PREFETCHW on AMD
+				type_ |= tPREFETCHW;
+			}
+
+			if (EDX & (1U << 29)) {
+				// Long mode implies support for PREFETCHW on AMD
+				type_ |= tPREFETCHW;
+			}
 		}
 		if (ECX == get32bitAsBE(intel)) {
 			type_ |= tINTEL;
+		}
+
+		// Extended flags information
+		getCpuid(0x80000000, data);
+		if (EAX >= 0x80000001) {
 			getCpuid(0x80000001, data);
+
+			if (EDX & (1U << 31)) type_ |= t3DN;
+			if (EDX & (1U << 30)) type_ |= tE3DN;
 			if (EDX & (1U << 27)) type_ |= tRDTSCP;
+			if (EDX & (1U << 22)) type_ |= tMMX2;
+			if (EDX & (1U << 15)) type_ |= tCMOV;
 			if (ECX & (1U << 5)) type_ |= tLZCNT;
 			if (ECX & (1U << 8)) type_ |= tPREFETCHW;
 		}
+
 		getCpuid(1, data);
 		if (ECX & (1U << 0)) type_ |= tSSE3;
 		if (ECX & (1U << 9)) type_ |= tSSSE3;
@ -426,7 +443,11 @@ public:
 			if ((bv & 6) == 6) {
 				if (ECX & (1U << 28)) type_ |= tAVX;
 				if (ECX & (1U << 12)) type_ |= tFMA;
-				if (((bv >> 5) & 7) == 7) {
+				// do *not* check AVX-512 state on macOS because it has on-demand AVX-512 support
+#if !defined(__APPLE__)
+				if (((bv >> 5) & 7) == 7)
+#endif
+				{
 					getCpuidEx(7, 0, data);
 					if (EBX & (1U << 16)) type_ |= tAVX512F;
 					if (type_ & tAVX512F) {
@ -449,16 +470,12 @@ public:
 						if (EDX & (1U << 3)) type_ |= tAVX512_4FMAPS;
 						if (EDX & (1U << 8)) type_ |= tAVX512_VP2INTERSECT;
 					}
-					// EAX=07H, ECX=1
-					getCpuidEx(7, 1, data);
-					if (type_ & tAVX512F) {
-						if (EAX & (1U << 5)) type_ |= tAVX512_BF16;
-					}
 				}
 			}
 		}
 		if (maxNum >= 7) {
 			getCpuidEx(7, 0, data);
+			const uint32_t maxNumSubLeaves = EAX;
 			if (type_ & tAVX && (EBX & (1U << 5))) type_ |= tAVX2;
 			if (EBX & (1U << 3)) type_ |= tBMI1;
 			if (EBX & (1U << 8)) type_ |= tBMI2;
@ -474,6 +491,13 @@ public:
 			if (EDX & (1U << 24)) type_ |= tAMX_TILE;
 			if (EDX & (1U << 25)) type_ |= tAMX_INT8;
 			if (EDX & (1U << 22)) type_ |= tAMX_BF16;
+			if (maxNumSubLeaves >= 1) {
+				getCpuidEx(7, 1, data);
+				if (EAX & (1U << 4)) type_ |= tAVX_VNNI;
+				if (type_ & tAVX512F) {
+					if (EAX & (1U << 5)) type_ |= tAVX512_BF16;
+				}
+			}
 		}
 		setFamily();
 		setNumCores();
--- a/externals/dynarmic/src/CMakeLists.txt
+++ b/externals/dynarmic/src/CMakeLists.txt
@ -163,6 +163,7 @@ if ("A32" IN_LIST DYNARMIC_FRONTENDS)
        frontend/A32/translate/impl/thumb32_data_processing_shifted_register.cpp
        frontend/A32/translate/impl/thumb32_load_byte.cpp
        frontend/A32/translate/impl/thumb32_load_halfword.cpp
+        frontend/A32/translate/impl/thumb32_load_store_multiple.cpp
        frontend/A32/translate/impl/thumb32_load_word.cpp
        frontend/A32/translate/impl/thumb32_long_multiply.cpp
        frontend/A32/translate/impl/thumb32_misc.cpp
--- a/externals/dynarmic/src/backend/x64/emit_x64_vector.cpp
+++ b/externals/dynarmic/src/backend/x64/emit_x64_vector.cpp
@ -440,13 +440,13 @@ void EmitX64::EmitVectorAnd(EmitContext& ctx, IR::Inst* inst) {

 static void ArithmeticShiftRightByte(EmitContext& ctx, BlockOfCode& code, const Xbyak::Xmm& result, u8 shift_amount) {
    if (code.HasAVX512_Icelake()) {
-        // Do a logical shift right upon the 8x8 bit-matrix, but shift in
-        // `0x80` bytes into the matrix to repeat the most significant bit.
-        const u64 zero_extend = ~(0xFFFFFFFFFFFFFFFF << (shift_amount * 8)) & 0x8080808080808080;
-        const u64 shift_matrix = (0x0102040810204080 >> (shift_amount * 8)) | zero_extend;
+        const u64 shift_matrix = shift_amount < 8
+                               ? (0x0102040810204080 << (shift_amount * 8)) | (0x8080808080808080 >> (64 - shift_amount * 8))
+                               : 0x8080808080808080;
        code.vgf2p8affineqb(result, result, code.MConst(xword_b, shift_matrix), 0);
        return;
    }
+
    const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();

    code.punpckhbw(tmp, result);
@ -1465,20 +1465,21 @@ void EmitX64::EmitVectorLogicalShiftLeft8(EmitContext& ctx, IR::Inst* inst) {
    const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
    const u8 shift_amount = args[1].GetImmediateU8();

-    if (shift_amount == 1) {
+    if (shift_amount == 0) {
+        // do nothing
+    } else if (shift_amount >= 8) {
+        code.pxor(result, result);
+    } else if (shift_amount == 1) {
        code.paddb(result, result);
-    } else if (shift_amount > 0) {
-        if (code.HasAVX512_Icelake()) {
-            // Galois 8x8 identity matrix, bit-shifted by the shift-amount
-            const u64 shift_matrix = 0x0102040810204080 >> (shift_amount * 8);
-            code.vgf2p8affineqb(result, result, code.MConst(xword_b, shift_matrix), 0);
-        } else {
-            const u64 replicand = (0xFFULL << shift_amount) & 0xFF;
-            const u64 mask = Common::Replicate(replicand, Common::BitSize<u8>());
+    } else if (code.HasAVX512_Icelake()) {
+        const u64 shift_matrix = 0x0102040810204080 >> (shift_amount * 8);
+        code.vgf2p8affineqb(result, result, code.MConst(xword_b, shift_matrix), 0);
+    } else {
+        const u64 replicand = (0xFFULL << shift_amount) & 0xFF;
+        const u64 mask = Common::Replicate(replicand, Common::BitSize<u8>());

-            code.psllw(result, shift_amount);
-            code.pand(result, code.MConst(xword, mask, mask));
-        }
+        code.psllw(result, shift_amount);
+        code.pand(result, code.MConst(xword, mask, mask));
    }

    ctx.reg_alloc.DefineValue(inst, result);
@ -1523,18 +1524,19 @@ void EmitX64::EmitVectorLogicalShiftRight8(EmitContext& ctx, IR::Inst* inst) {
    const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
    const u8 shift_amount = args[1].GetImmediateU8();

-    if (shift_amount > 0) {
-        if (code.HasAVX512_Icelake()) {
-            // Galois 8x8 identity matrix, bit-shifted by the shift-amount
-            const u64 shift_matrix = 0x0102040810204080 << (shift_amount * 8);
-            code.vgf2p8affineqb(result, result, code.MConst(xword_b, shift_matrix), 0);
-        } else {
-            const u64 replicand = 0xFEULL >> shift_amount;
-            const u64 mask = Common::Replicate(replicand, Common::BitSize<u8>());
+    if (shift_amount == 0) {
+        // Do nothing
+    } else if (shift_amount >= 8) {
+        code.pxor(result, result);
+    } else if (code.HasAVX512_Icelake()) {
+        const u64 shift_matrix = 0x0102040810204080 << (shift_amount * 8);
+        code.vgf2p8affineqb(result, result, code.MConst(xword_b, shift_matrix), 0);
+    } else {
+        const u64 replicand = 0xFEULL >> shift_amount;
+        const u64 mask = Common::Replicate(replicand, Common::BitSize<u8>());

-            code.psrlw(result, shift_amount);
-            code.pand(result, code.MConst(xword, mask, mask));
-        }
+        code.psrlw(result, shift_amount);
+        code.pand(result, code.MConst(xword, mask, mask));
    }

    ctx.reg_alloc.DefineValue(inst, result);
@ -2768,12 +2770,8 @@ void EmitX64::EmitVectorReverseBits(EmitContext& ctx, IR::Inst* inst) {

    const Xbyak::Xmm data = ctx.reg_alloc.UseScratchXmm(args[0]);

-    if (code.HasAVX512_Icelake() && code.HasSSSE3()) {
-        // GFNI(vgf2p8affineqb) and SSSE3(pshuf)
-        // Reverse bits within bytes
+    if (code.HasAVX512_Icelake()) {
        code.vgf2p8affineqb(data, data, code.MConst(xword_b, 0x8040201008040201), 0);
-        // Reverse bytes within vector
-        code.pshufb(data, code.MConst(xword, 0x0001020304050607, 0x08090a0b0c0d0e0f));
    } else {
        const Xbyak::Xmm high_nibble_reg = ctx.reg_alloc.ScratchXmm();
        code.movdqa(high_nibble_reg, code.MConst(xword, 0xF0F0F0F0F0F0F0F0, 0xF0F0F0F0F0F0F0F0));
--- a/externals/dynarmic/src/frontend/A32/decoder/thumb32.inc
+++ b/externals/dynarmic/src/frontend/A32/decoder/thumb32.inc
@ -1,12 +1,12 @@
 // Load/Store Multiple
 //INST(thumb32_SRS_1,          "SRS",                      "1110100000-0--------------------")
 //INST(thumb32_RFE_2,          "RFE",                      "1110100000-1--------------------")
-//INST(thumb32_STMIA,          "STMIA/STMEA",              "1110100010-0--------------------")
-//INST(thumb32_POP,            "POP",                      "1110100010111101----------------")
-//INST(thumb32_LDMIA,          "LDMIA/LDMFD",              "1110100010-1--------------------")
-//INST(thumb32_PUSH,           "PUSH",                     "1110100100101101----------------")
-//INST(thumb32_STMDB,          "STMDB/STMFD",              "1110100100-0--------------------")
-//INST(thumb32_LDMDB,          "LDMDB/LDMEA",              "1110100100-1--------------------")
+INST(thumb32_STMIA,          "STMIA/STMEA",              "1110100010W0nnnn0iiiiiiiiiiiiiii")
+INST(thumb32_POP,            "POP",                      "1110100010111101iiiiiiiiiiiiiiii")
+INST(thumb32_LDMIA,          "LDMIA/LDMFD",              "1110100010W1nnnniiiiiiiiiiiiiiii")
+INST(thumb32_PUSH,           "PUSH",                     "11101001001011010iiiiiiiiiiiiiii")
+INST(thumb32_STMDB,          "STMDB/STMFD",              "1110100100W0nnnn0iiiiiiiiiiiiiii")
+INST(thumb32_LDMDB,          "LDMDB/LDMEA",              "1110100100W1nnnniiiiiiiiiiiiiiii")
 //INST(thumb32_SRS_1,          "SRS",                      "1110100110-0--------------------")
 //INST(thumb32_RFE_2,          "RFE",                      "1110100110-1--------------------")

@ -66,10 +66,10 @@ INST(thumb32_SUB_imm_1,      "SUB (imm)",                "11110v01101Snnnn0vvvdd
 INST(thumb32_RSB_imm,        "RSB (imm)",                "11110v01110Snnnn0vvvddddvvvvvvvv")

 // Data Processing (Plain Binary Immediate)
-//INST(thumb32_ADR,            "ADR",                      "11110-10000011110---------------")
+INST(thumb32_ADR_t3,         "ADR",                      "11110i10000011110iiiddddiiiiiiii")
 INST(thumb32_ADD_imm_2,      "ADD (imm)",                "11110i10000011010iiiddddiiiiiiii")
 INST(thumb32_MOVW_imm,       "MOVW (imm)",               "11110i100100iiii0iiiddddiiiiiiii")
-//INST(thumb32_ADR,            "ADR",                      "11110-10101011110---------------")
+INST(thumb32_ADR_t2,         "ADR",                      "11110i10101011110iiiddddiiiiiiii")
 INST(thumb32_SUB_imm_2,      "SUB (imm)",                "11110i10101011010iiiddddiiiiiiii")
 INST(thumb32_MOVT,           "MOVT",                     "11110i101100iiii0iiiddddiiiiiiii")
 INST(thumb32_UDF,            "Invalid decoding",         "11110011-010----0000----0001----")
@ -124,24 +124,25 @@ INST(thumb32_UDF,            "Invalid decoding",         "11110-111-------10-0--
 INST(thumb32_B_cond,         "B (cond)",                 "11110Sccccvvvvvv10i0ivvvvvvvvvvv")

 // Store Single Data Item
-//INST(thumb32_STRB_imm_1,     "STRB (imm)",               "111110000000--------1--1--------")
-//INST(thumb32_STRB_imm_2,     "STRB (imm)",               "111110000000--------1100--------")
-//INST(thumb32_STRB_imm_3,     "STRB (imm)",               "111110001000--------------------")
-//INST(thumb32_STRBT,          "STRBT",                    "111110000000--------1110--------")
+INST(thumb32_STRB_imm_1,     "STRB (imm)",               "111110000000nnnntttt1PU1iiiiiiii")
+INST(thumb32_STRB_imm_2,     "STRB (imm)",               "111110000000nnnntttt1100iiiiiiii")
+INST(thumb32_STRB_imm_3,     "STRB (imm)",               "111110001000nnnnttttiiiiiiiiiiii")
+INST(thumb32_STRBT,          "STRBT",                    "111110000000nnnntttt1110iiiiiiii")
 INST(thumb32_STRB,           "STRB (reg)",               "111110000000nnnntttt000000iimmmm")
-//INST(thumb32_STRH_imm_1,     "STRH (imm)",               "111110000010--------1--1--------")
-//INST(thumb32_STRH_imm_2,     "STRH (imm)",               "111110000010--------1100--------")
-//INST(thumb32_STRH_imm_3,     "STRH (imm)",               "111110001010--------------------")
-//INST(thumb32_STRHT,          "STRHT",                    "111110000010--------1110--------")
+INST(thumb32_STRH_imm_1,     "STRH (imm)",               "111110000010nnnntttt1PU1iiiiiiii")
+INST(thumb32_STRH_imm_2,     "STRH (imm)",               "111110000010nnnntttt1100iiiiiiii")
+INST(thumb32_STRH_imm_3,     "STRH (imm)",               "111110001010nnnnttttiiiiiiiiiiii")
+INST(thumb32_STRHT,          "STRHT",                    "111110000010nnnntttt1110iiiiiiii")
 INST(thumb32_STRH,           "STRH (reg)",               "111110000010nnnntttt000000iimmmm")
-//INST(thumb32_STR_imm_1,      "STR (imm)",                "111110000100--------1--1--------")
-//INST(thumb32_STR_imm_2,      "STR (imm)",                "111110000100--------1100--------")
-//INST(thumb32_STR_imm_3,      "STR (imm)",                "111110001100--------------------")
-//INST(thumb32_STRT,           "STRT",                     "111110000100--------1110--------")
+INST(thumb32_STR_imm_1,      "STR (imm)",                "111110000100nnnntttt1PU1iiiiiiii")
+INST(thumb32_STR_imm_2,      "STR (imm)",                "111110000100nnnntttt1100iiiiiiii")
+INST(thumb32_STR_imm_3,      "STR (imm)",                "111110001100nnnnttttiiiiiiiiiiii")
+INST(thumb32_STRT,           "STRT",                     "111110000100nnnntttt1110iiiiiiii")
 INST(thumb32_STR_reg,        "STR (reg)",                "111110000100nnnntttt000000iimmmm")

 // Load Byte and Memory Hints
 INST(thumb32_PLD_lit,        "PLD (lit)",                "11111000U00111111111iiiiiiiiiiii")
+INST(thumb32_PLD_lit,        "PLD (lit)",                "11111000U01111111111iiiiiiiiiiii")
 INST(thumb32_PLD_reg,        "PLD (reg)",                "1111100000W1nnnn1111000000iimmmm")
 INST(thumb32_PLD_imm8,       "PLD (imm8)",               "1111100000W1nnnn11111100iiiiiiii")
 INST(thumb32_PLD_imm12,      "PLD (imm12)",              "1111100010W1nnnn1111iiiiiiiiiiii")
@ -161,27 +162,27 @@ INST(thumb32_LDRSB_imm8,     "LDRSB (imm8)",             "111110010001nnnntttt1P
 INST(thumb32_LDRSB_imm12,    "LDRSB (imm12)",            "111110011001nnnnttttiiiiiiiiiiii")

 // Load Halfword and Memory Hints
-//INST(thumb32_LDRH_lit,       "LDRH (lit)",               "11111000-0111111----------------")
-//INST(thumb32_LDRH_reg,       "LDRH (reg)",               "111110000011--------000000------")
-//INST(thumb32_LDRHT,          "LDRHT",                    "111110000011--------1110--------")
-//INST(thumb32_LDRH_imm8,      "LDRH (imm8)",              "111110000011--------1-----------")
-//INST(thumb32_LDRH_imm12,     "LDRH (imm12)",             "111110001011--------------------")
-//INST(thumb32_LDRSH_lit,      "LDRSH (lit)",              "11111001-0111111----------------")
-//INST(thumb32_LDRSH_reg,      "LDRSH (reg)",              "111110010011--------000000------")
-//INST(thumb32_LDRSHT,         "LDRSHT",                   "111110010011--------1110--------")
-//INST(thumb32_LDRSH_imm8,     "LDRSH (imm8)",             "111110010011--------1-----------")
-//INST(thumb32_LDRSH_imm12,    "LDRSH (imm12)",            "111110011011--------------------")
-//INST(thumb32_NOP,            "NOP",                      "111110010011----1111000000------")
-//INST(thumb32_NOP,            "NOP",                      "111110010011----11111100--------")
-//INST(thumb32_NOP,            "NOP",                      "11111001-01111111111------------")
-//INST(thumb32_NOP,            "NOP",                      "111110011011----1111------------")
+INST(thumb32_LDRH_lit,       "LDRH (lit)",               "11111000U0111111ttttiiiiiiiiiiii")
+INST(thumb32_LDRH_reg,       "LDRH (reg)",               "111110000011nnnntttt000000iimmmm")
+INST(thumb32_LDRHT,          "LDRHT",                    "111110000011nnnntttt1110iiiiiiii")
+INST(thumb32_LDRH_imm8,      "LDRH (imm8)",              "111110000011nnnntttt1PUWiiiiiiii")
+INST(thumb32_LDRH_imm12,     "LDRH (imm12)",             "111110001011nnnnttttiiiiiiiiiiii")
+INST(thumb32_NOP,            "NOP",                      "11111001-01111111111------------")
+INST(thumb32_LDRSH_lit,      "LDRSH (lit)",              "11111001U0111111ttttiiiiiiiiiiii")
+INST(thumb32_NOP,            "NOP",                      "111110010011----1111000000------")
+INST(thumb32_LDRSH_reg,      "LDRSH (reg)",              "111110010011nnnntttt000000iimmmm")
+INST(thumb32_LDRSHT,         "LDRSHT",                   "111110010011nnnntttt1110iiiiiiii")
+INST(thumb32_NOP,            "NOP",                      "111110010011----11111100--------")
+INST(thumb32_NOP,            "NOP",                      "111110011011----1111------------")
+INST(thumb32_LDRSH_imm8,     "LDRSH (imm8)",             "111110010011nnnntttt1PUWiiiiiiii")
+INST(thumb32_LDRSH_imm12,    "LDRSH (imm12)",            "111110011011nnnnttttiiiiiiiiiiii")

 // Load Word
-//INST(thumb32_LDR_lit,        "LDR (lit)",                "11111000-1011111----------------")
-//INST(thumb32_LDRT,           "LDRT",                     "111110000101--------1110--------")
-//INST(thumb32_LDR_reg,        "LDR (reg)",                "111110000101--------000000------")
-//INST(thumb32_LDR_imm8,       "LDR (imm8)",               "111110000101--------1-----------")
-//INST(thumb32_LDR_imm12,      "LDR (imm12)",              "111110001101--------------------")
+INST(thumb32_LDR_lit,        "LDR (lit)",                "11111000U1011111ttttiiiiiiiiiiii")
+INST(thumb32_LDRT,           "LDRT",                     "111110000101nnnntttt1110iiiiiiii")
+INST(thumb32_LDR_reg,        "LDR (reg)",                "111110000101nnnntttt000000iimmmm")
+INST(thumb32_LDR_imm8,       "LDR (imm8)",               "111110000101nnnntttt1PUWiiiiiiii")
+INST(thumb32_LDR_imm12,      "LDR (imm12)",              "111110001101nnnnttttiiiiiiiiiiii")

 // Data Processing (register)
 INST(thumb32_LSL_reg,        "LSL (reg)",                "111110100000mmmm1111dddd0000ssss")
--- a/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_data_processing_plain_binary_immediate.cpp
+++ b/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_data_processing_plain_binary_immediate.cpp
@ -52,6 +52,30 @@ static bool Saturation16(ThumbTranslatorVisitor& v, Reg n, Reg d, size_t saturat
    return true;
 }

+bool ThumbTranslatorVisitor::thumb32_ADR_t2(Imm<1> imm1, Imm<3> imm3, Reg d, Imm<8> imm8) {
+    if (d == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+
+    const auto imm32 = concatenate(imm1, imm3, imm8).ZeroExtend();
+    const auto result = ir.AlignPC(4) - imm32;
+
+    ir.SetRegister(d, ir.Imm32(result));
+    return true;
+}
+
+bool ThumbTranslatorVisitor::thumb32_ADR_t3(Imm<1> imm1, Imm<3> imm3, Reg d, Imm<8> imm8) {
+    if (d == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+
+    const auto imm32 = concatenate(imm1, imm3, imm8).ZeroExtend();
+    const auto result = ir.AlignPC(4) + imm32;
+
+    ir.SetRegister(d, ir.Imm32(result));
+    return true;
+}
+
 bool ThumbTranslatorVisitor::thumb32_ADD_imm_2(Imm<1> imm1, Imm<3> imm3, Reg d, Imm<8> imm8) {
    if (d == Reg::PC) {
        return UnpredictableInstruction();
--- a/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_load_halfword.cpp
+++ b/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_load_halfword.cpp
@ -7,5 +7,135 @@

 namespace Dynarmic::A32 {

+using ExtensionFunction = IR::U32 (IREmitter::*)(const IR::U16&);
+
+static bool LoadHalfLiteral(ThumbTranslatorVisitor& v, bool U, Reg t, Imm<12> imm12,
+                            ExtensionFunction ext_fn) {
+    const auto imm32 = imm12.ZeroExtend();
+    const auto base = v.ir.AlignPC(4);
+    const auto address = U ? (base + imm32) : (base - imm32);
+    const auto data = (v.ir.*ext_fn)(v.ir.ReadMemory16(v.ir.Imm32(address)));
+
+    v.ir.SetRegister(t, data);
+    return true;
+}
+
+static bool LoadHalfRegister(ThumbTranslatorVisitor& v, Reg n, Reg t, Imm<2> imm2, Reg m,
+                             ExtensionFunction ext_fn) {
+    if (m == Reg::PC) {
+        return v.UnpredictableInstruction();
+    }
+
+    const IR::U32 reg_m = v.ir.GetRegister(m);
+    const IR::U32 reg_n = v.ir.GetRegister(n);
+    const IR::U32 offset = v.ir.LogicalShiftLeft(reg_m, v.ir.Imm8(imm2.ZeroExtend<u8>()));
+    const IR::U32 address = v.ir.Add(reg_n, offset);
+    const IR::U32 data = (v.ir.*ext_fn)(v.ir.ReadMemory16(address));
+
+    v.ir.SetRegister(t, data);
+    return true;
+}
+
+static bool LoadHalfImmediate(ThumbTranslatorVisitor& v, Reg n, Reg t, bool P, bool U, bool W,
+                              Imm<12> imm12, ExtensionFunction ext_fn) {
+    const u32 imm32 = imm12.ZeroExtend();
+    const IR::U32 reg_n = v.ir.GetRegister(n);
+    const IR::U32 offset_address = U ? v.ir.Add(reg_n, v.ir.Imm32(imm32))
+                                     : v.ir.Sub(reg_n, v.ir.Imm32(imm32));
+    const IR::U32 address = P ? offset_address
+                              : reg_n;
+    const IR::U32 data = (v.ir.*ext_fn)(v.ir.ReadMemory16(address));
+
+    if (W) {
+        v.ir.SetRegister(n, offset_address);
+    }
+
+    v.ir.SetRegister(t, data);
+    return true;
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRH_lit(bool U, Reg t, Imm<12> imm12) {
+    return LoadHalfLiteral(*this, U, t, imm12, &IREmitter::ZeroExtendHalfToWord);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRH_reg(Reg n, Reg t, Imm<2> imm2, Reg m) {
+    return LoadHalfRegister(*this, n, t, imm2, m, &IREmitter::ZeroExtendHalfToWord);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRH_imm8(Reg n, Reg t, bool P, bool U, bool W, Imm<8> imm8) {
+    if (!P && !W) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC && W) {
+        return UnpredictableInstruction();
+    }
+    if (W && n == t) {
+        return UnpredictableInstruction();
+    }
+
+    return LoadHalfImmediate(*this, n, t, P, U, W, Imm<12>{imm8.ZeroExtend()},
+                             &IREmitter::ZeroExtendHalfToWord);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRH_imm12(Reg n, Reg t, Imm<12> imm12) {
+    return LoadHalfImmediate(*this, n, t, true, true, false, imm12,
+                             &IREmitter::ZeroExtendHalfToWord);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRHT(Reg n, Reg t, Imm<8> imm8) {
+    // TODO: Add an unpredictable instruction path if this
+    //       is executed in hypervisor mode if we ever support
+    //       privileged execution levels.
+
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+
+    // Treat it as a normal LDRH, given we don't support
+    // execution levels other than EL0 currently.
+    return thumb32_LDRH_imm8(n, t, true, true, false, imm8);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRSH_lit(bool U, Reg t, Imm<12> imm12) {
+    return LoadHalfLiteral(*this, U, t, imm12, &IREmitter::SignExtendHalfToWord);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRSH_reg(Reg n, Reg t, Imm<2> imm2, Reg m) {
+    return LoadHalfRegister(*this, n, t, imm2, m, &IREmitter::SignExtendHalfToWord);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRSH_imm8(Reg n, Reg t, bool P, bool U, bool W, Imm<8> imm8) {
+    if (!P && !W) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC && W) {
+        return UnpredictableInstruction();
+    }
+    if (W && n == t) {
+        return UnpredictableInstruction();
+    }
+
+    return LoadHalfImmediate(*this, n, t, P, U, W, Imm<12>{imm8.ZeroExtend()},
+                             &IREmitter::SignExtendHalfToWord);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRSH_imm12(Reg n, Reg t, Imm<12> imm12) {
+    return LoadHalfImmediate(*this, n, t, true, true, false, imm12,
+                             &IREmitter::SignExtendHalfToWord);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRSHT(Reg n, Reg t, Imm<8> imm8) {
+    // TODO: Add an unpredictable instruction path if this
+    //       is executed in hypervisor mode if we ever support
+    //       privileged execution levels.
+
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+
+    // Treat it as a normal LDRSH, given we don't support
+    // execution levels other than EL0 currently.
+    return thumb32_LDRSH_imm8(n, t, true, true, false, imm8);
+}

 } // namespace Dynarmic::A32
--- a/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_load_store_multiple.cpp
+++ b/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_load_store_multiple.cpp
@ -0,0 +1,150 @@
+/* This file is part of the dynarmic project.
+ * Copyright (c) 2021 MerryMage
+ * SPDX-License-Identifier: 0BSD
+ */
+
+#include "common/bit_util.h"
+#include "frontend/A32/translate/impl/translate_thumb.h"
+
+namespace Dynarmic::A32 {
+static bool ITBlockCheck(const A32::IREmitter& ir) {
+    return ir.current_location.IT().IsInITBlock() && !ir.current_location.IT().IsLastInITBlock();
+}
+
+static bool LDMHelper(A32::IREmitter& ir, bool W, Reg n, u32 list,
+                      const IR::U32& start_address, const IR::U32& writeback_address) {
+    auto address = start_address;
+    for (size_t i = 0; i <= 14; i++) {
+        if (Common::Bit(i, list)) {
+            ir.SetRegister(static_cast<Reg>(i), ir.ReadMemory32(address));
+            address = ir.Add(address, ir.Imm32(4));
+        }
+    }
+    if (W && !Common::Bit(RegNumber(n), list)) {
+        ir.SetRegister(n, writeback_address);
+    }
+    if (Common::Bit<15>(list)) {
+        ir.UpdateUpperLocationDescriptor();
+        ir.LoadWritePC(ir.ReadMemory32(address));
+        if (n == Reg::R13) {
+            ir.SetTerm(IR::Term::PopRSBHint{});
+        } else {
+            ir.SetTerm(IR::Term::FastDispatchHint{});
+        }
+        return false;
+    }
+    return true;
+}
+
+static bool STMHelper(A32::IREmitter& ir, bool W, Reg n, u32 list,
+                      const IR::U32& start_address, const IR::U32& writeback_address) {
+    auto address = start_address;
+    for (size_t i = 0; i <= 14; i++) {
+        if (Common::Bit(i, list)) {
+            ir.WriteMemory32(address, ir.GetRegister(static_cast<Reg>(i)));
+            address = ir.Add(address, ir.Imm32(4));
+        }
+    }
+    if (W) {
+        ir.SetRegister(n, writeback_address);
+    }
+    return true;
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDMDB(bool W, Reg n, Imm<16> reg_list) {
+    const auto regs_imm = reg_list.ZeroExtend();
+    const auto num_regs = static_cast<u32>(Common::BitCount(regs_imm));
+
+    if (n == Reg::PC || num_regs < 2) {
+        return UnpredictableInstruction();
+    }
+    if (reg_list.Bit<15>() && reg_list.Bit<14>()) {
+        return UnpredictableInstruction();
+    }
+    if (W && Common::Bit(static_cast<size_t>(n), regs_imm)) {
+        return UnpredictableInstruction();
+    }
+    if (reg_list.Bit<13>()) {
+        return UnpredictableInstruction();
+    }
+    if (reg_list.Bit<15>() && ITBlockCheck(ir)) {
+        return UnpredictableInstruction();
+    }
+
+    // Start address is the same as the writeback address.
+    const IR::U32 start_address = ir.Sub(ir.GetRegister(n), ir.Imm32(4 * num_regs));
+    return LDMHelper(ir, W, n, regs_imm, start_address, start_address);
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDMIA(bool W, Reg n, Imm<16> reg_list) {
+    const auto regs_imm = reg_list.ZeroExtend();
+    const auto num_regs = static_cast<u32>(Common::BitCount(regs_imm));
+
+    if (n == Reg::PC || num_regs < 2) {
+        return UnpredictableInstruction();
+    }
+    if (reg_list.Bit<15>() && reg_list.Bit<14>()) {
+        return UnpredictableInstruction();
+    }
+    if (W && Common::Bit(static_cast<size_t>(n), regs_imm)) {
+        return UnpredictableInstruction();
+    }
+    if (reg_list.Bit<13>()) {
+        return UnpredictableInstruction();
+    }
+    if (reg_list.Bit<15>() && ITBlockCheck(ir)) {
+        return UnpredictableInstruction();
+    }
+
+    const auto start_address = ir.GetRegister(n);
+    const auto writeback_address = ir.Add(start_address, ir.Imm32(num_regs * 4));
+    return LDMHelper(ir, W, n, regs_imm, start_address, writeback_address);
+}
+
+bool ThumbTranslatorVisitor::thumb32_POP(Imm<16> reg_list) {
+    return thumb32_LDMIA(true, Reg::SP, reg_list);
+}
+
+bool ThumbTranslatorVisitor::thumb32_PUSH(Imm<15> reg_list) {
+    return thumb32_STMDB(true, Reg::SP, reg_list);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STMIA(bool W, Reg n, Imm<15> reg_list) {
+    const auto regs_imm = reg_list.ZeroExtend();
+    const auto num_regs = static_cast<u32>(Common::BitCount(regs_imm));
+
+    if (n == Reg::PC || num_regs < 2) {
+        return UnpredictableInstruction();
+    }
+    if (W && Common::Bit(static_cast<size_t>(n), regs_imm)) {
+        return UnpredictableInstruction();
+    }
+    if (reg_list.Bit<13>()) {
+        return UnpredictableInstruction();
+    }
+
+    const auto start_address = ir.GetRegister(n);
+    const auto writeback_address = ir.Add(start_address, ir.Imm32(num_regs * 4));
+    return STMHelper(ir, W, n, regs_imm, start_address, writeback_address);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STMDB(bool W, Reg n, Imm<15> reg_list) {
+    const auto regs_imm = reg_list.ZeroExtend();
+    const auto num_regs = static_cast<u32>(Common::BitCount(regs_imm));
+
+    if (n == Reg::PC || num_regs < 2) {
+        return UnpredictableInstruction();
+    }
+    if (W && Common::Bit(static_cast<size_t>(n), regs_imm)) {
+        return UnpredictableInstruction();
+    }
+    if (reg_list.Bit<13>()) {
+        return UnpredictableInstruction();
+    }
+
+    // Start address is the same as the writeback address.
+    const IR::U32 start_address = ir.Sub(ir.GetRegister(n), ir.Imm32(4 * num_regs));
+    return STMHelper(ir, W, n, regs_imm, start_address, start_address);
+}
+
+} // namespace Dynarmic::A32
--- a/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_load_word.cpp
+++ b/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_load_word.cpp
@ -6,6 +6,129 @@
 #include "frontend/A32/translate/impl/translate_thumb.h"

 namespace Dynarmic::A32 {
+static bool ITBlockCheck(const A32::IREmitter& ir) {
+    return ir.current_location.IT().IsInITBlock() && !ir.current_location.IT().IsLastInITBlock();
+}

+bool ThumbTranslatorVisitor::thumb32_LDR_lit(bool U, Reg t, Imm<12> imm12) {
+    if (t == Reg::PC && ITBlockCheck(ir)) {
+        return UnpredictableInstruction();
+    }
+
+    const u32 imm32 = imm12.ZeroExtend();
+    const u32 base = ir.AlignPC(4);
+    const u32 address = U ? base + imm32 : base - imm32;
+    const auto data = ir.ReadMemory32(ir.Imm32(address));
+
+    if (t == Reg::PC) {
+        ir.UpdateUpperLocationDescriptor();
+        ir.LoadWritePC(data);
+        ir.SetTerm(IR::Term::FastDispatchHint{});
+        return false;
+    }
+
+    ir.SetRegister(t, data);
+    return true;
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDR_imm8(Reg n, Reg t, bool P, bool U, bool W, Imm<8> imm8) {
+    if (!P && !W) {
+        return UndefinedInstruction();
+    }
+    if (W && n == t) {
+        return UnpredictableInstruction();
+    }
+    if (t == Reg::PC && ITBlockCheck(ir)) {
+        return UnpredictableInstruction();
+    }
+
+    const u32 imm32 = imm8.ZeroExtend();
+    const IR::U32 reg_n = ir.GetRegister(n);
+    const IR::U32 offset_address = U ? ir.Add(reg_n, ir.Imm32(imm32))
+                                     : ir.Sub(reg_n, ir.Imm32(imm32));
+    const IR::U32 address = P ? offset_address
+                              : reg_n;
+    const IR::U32 data = ir.ReadMemory32(address);
+
+    if (W) {
+        ir.SetRegister(n, offset_address);
+    }
+
+    if (t == Reg::PC) {
+        ir.UpdateUpperLocationDescriptor();
+        ir.LoadWritePC(data);
+
+        if (!P && W && n == Reg::R13) {
+            ir.SetTerm(IR::Term::PopRSBHint{});
+        } else {
+            ir.SetTerm(IR::Term::FastDispatchHint{});
+        }
+
+        return false;
+    }
+
+    ir.SetRegister(t, data);
+    return true;
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDR_imm12(Reg n, Reg t, Imm<12> imm12) {
+    if (t == Reg::PC && ITBlockCheck(ir)) {
+        return UnpredictableInstruction();
+    }
+
+    const auto imm32 = imm12.ZeroExtend();
+    const auto reg_n = ir.GetRegister(n);
+    const auto address = ir.Add(reg_n, ir.Imm32(imm32));
+    const auto data = ir.ReadMemory32(address);
+
+    if (t == Reg::PC) {
+        ir.UpdateUpperLocationDescriptor();
+        ir.LoadWritePC(data);
+        ir.SetTerm(IR::Term::FastDispatchHint{});
+        return false;
+    }
+
+    ir.SetRegister(t, data);
+    return true;
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDR_reg(Reg n, Reg t, Imm<2> imm2, Reg m) {
+    if (m == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+    if (t == Reg::PC && ITBlockCheck(ir)) {
+        return UnpredictableInstruction();
+    }
+
+    const auto reg_m = ir.GetRegister(m);
+    const auto reg_n = ir.GetRegister(n);
+    const auto offset = ir.LogicalShiftLeft(reg_m, ir.Imm8(imm2.ZeroExtend<u8>()));
+    const auto address = ir.Add(reg_n, offset);
+    const auto data = ir.ReadMemory32(address);
+
+    if (t == Reg::PC) {
+        ir.UpdateUpperLocationDescriptor();
+        ir.LoadWritePC(data);
+        ir.SetTerm(IR::Term::FastDispatchHint{});
+        return false;
+    }
+
+    ir.SetRegister(t, data);
+    return true;
+}
+
+bool ThumbTranslatorVisitor::thumb32_LDRT(Reg n, Reg t, Imm<8> imm8) {
+    // TODO: Add an unpredictable instruction path if this
+    //       is executed in hypervisor mode if we ever support
+    //       privileged execution levels.
+
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+
+    // Treat it as a normal LDR, given we don't support
+    // execution levels other than EL0 currently.
+    return thumb32_LDR_imm8(n, t, true, true, false, imm8);
+}

 } // namespace Dynarmic::A32
--- a/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_store_single_data_item.cpp
+++ b/externals/dynarmic/src/frontend/A32/translate/impl/thumb32_store_single_data_item.cpp
@ -29,18 +29,192 @@ static bool StoreRegister(ThumbTranslatorVisitor& v, Reg n, Reg t, Imm<2> imm2,
    return true;
 }

+using StoreImmFn = void (*)(ThumbTranslatorVisitor&, const IR::U32&, const IR::U32&);
+
+static void StoreImmByteFn(ThumbTranslatorVisitor& v, const IR::U32& address, const IR::U32& data) {
+    v.ir.WriteMemory8(address, v.ir.LeastSignificantByte(data));
+}
+
+static void StoreImmHalfFn(ThumbTranslatorVisitor& v, const IR::U32& address, const IR::U32& data) {
+    v.ir.WriteMemory16(address, v.ir.LeastSignificantHalf(data));
+}
+
+static void StoreImmWordFn(ThumbTranslatorVisitor& v, const IR::U32& address, const IR::U32& data) {
+    v.ir.WriteMemory32(address, data);
+}
+
+static bool StoreImmediate(ThumbTranslatorVisitor& v, Reg n, Reg t, bool P, bool U, bool W, Imm<12> imm12,
+                           StoreImmFn store_fn) {
+    const auto imm32 = imm12.ZeroExtend();
+    const auto reg_n = v.ir.GetRegister(n);
+    const auto reg_t = v.ir.GetRegister(t);
+
+    const IR::U32 offset_address = U ? v.ir.Add(reg_n, v.ir.Imm32(imm32))
+                                     : v.ir.Sub(reg_n, v.ir.Imm32(imm32));
+    const IR::U32 address = P ? offset_address
+                              : reg_n;
+
+    store_fn(v, address, reg_t);
+    if (W) {
+        v.ir.SetRegister(n, offset_address);
+    }
+
+    return true;
+}
+
+bool ThumbTranslatorVisitor::thumb32_STRB_imm_1(Reg n, Reg t, bool P, bool U, Imm<8> imm8) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC || n == t) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, P, U, true, Imm<12>{imm8.ZeroExtend()}, StoreImmByteFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STRB_imm_2(Reg n, Reg t, Imm<8> imm8) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, true, false, false, Imm<12>{imm8.ZeroExtend()}, StoreImmByteFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STRB_imm_3(Reg n, Reg t, Imm<12> imm12) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, true, true, false, imm12, StoreImmByteFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STRBT(Reg n, Reg t, Imm<8> imm8) {
+    // TODO: Add an unpredictable instruction path if this
+    //       is executed in hypervisor mode if we ever support
+    //       privileged execution levels.
+
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+
+    // Treat this as a normal STRB, given we don't support
+    // execution levels other than EL0 currently.
+    return StoreImmediate(*this, n, t, true, true, false, Imm<12>{imm8.ZeroExtend()}, StoreImmByteFn);
+}
+
 bool ThumbTranslatorVisitor::thumb32_STRB(Reg n, Reg t, Imm<2> imm2, Reg m) {
    return StoreRegister(*this, n, t, imm2, m, [this](const IR::U32& offset_address, const IR::U32& data) {
        ir.WriteMemory8(offset_address, ir.LeastSignificantByte(data));
    });
 }

+bool ThumbTranslatorVisitor::thumb32_STRH_imm_1(Reg n, Reg t, bool P, bool U, Imm<8> imm8) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC || n == t) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, P, U, true, Imm<12>{imm8.ZeroExtend()}, StoreImmHalfFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STRH_imm_2(Reg n, Reg t, Imm<8> imm8) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, true, false, false, Imm<12>{imm8.ZeroExtend()}, StoreImmHalfFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STRH_imm_3(Reg n, Reg t, Imm<12> imm12) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, true, true, false, imm12, StoreImmHalfFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STRHT(Reg n, Reg t, Imm<8> imm8) {
+    // TODO: Add an unpredictable instruction path if this
+    //       is executed in hypervisor mode if we ever support
+    //       privileged execution levels.
+
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+
+    // Treat this as a normal STRH, given we don't support
+    // execution levels other than EL0 currently.
+    return StoreImmediate(*this, n, t, true, true, false, Imm<12>{imm8.ZeroExtend()}, StoreImmHalfFn);
+}
+
 bool ThumbTranslatorVisitor::thumb32_STRH(Reg n, Reg t, Imm<2> imm2, Reg m) {
    return StoreRegister(*this, n, t, imm2, m, [this](const IR::U32& offset_address, const IR::U32& data) {
        ir.WriteMemory16(offset_address, ir.LeastSignificantHalf(data));
    });
 }

+bool ThumbTranslatorVisitor::thumb32_STR_imm_1(Reg n, Reg t, bool P, bool U, Imm<8> imm8) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC || n == t) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, P, U, true, Imm<12>{imm8.ZeroExtend()}, StoreImmWordFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STR_imm_2(Reg n, Reg t, Imm<8> imm8) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, true, false, false, Imm<12>{imm8.ZeroExtend()}, StoreImmWordFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STR_imm_3(Reg n, Reg t, Imm<12> imm12) {
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+    return StoreImmediate(*this, n, t, true, true, false, imm12, StoreImmWordFn);
+}
+
+bool ThumbTranslatorVisitor::thumb32_STRT(Reg n, Reg t, Imm<8> imm8) {
+    // TODO: Add an unpredictable instruction path if this
+    //       is executed in hypervisor mode if we ever support
+    //       privileged execution levels.
+
+    if (n == Reg::PC) {
+        return UndefinedInstruction();
+    }
+    if (t == Reg::PC) {
+        return UnpredictableInstruction();
+    }
+
+    // Treat this as a normal STR, given we don't support
+    // execution levels other than EL0 currently.
+    return StoreImmediate(*this, n, t, true, true, false, Imm<12>{imm8.ZeroExtend()}, StoreImmWordFn);
+}
+
 bool ThumbTranslatorVisitor::thumb32_STR_reg(Reg n, Reg t, Imm<2> imm2, Reg m) {
    return StoreRegister(*this, n, t, imm2, m, [this](const IR::U32& offset_address, const IR::U32& data) {
        ir.WriteMemory32(offset_address, data);
--- a/externals/dynarmic/src/frontend/A32/translate/impl/translate_thumb.h
+++ b/externals/dynarmic/src/frontend/A32/translate/impl/translate_thumb.h
@ -171,6 +171,14 @@ struct ThumbTranslatorVisitor final {
    bool thumb16_B_t1(Cond cond, Imm<8> imm8);
    bool thumb16_B_t2(Imm<11> imm11);

+    // thumb32 load/store multiple instructions
+    bool thumb32_LDMDB(bool W, Reg n, Imm<16> reg_list);
+    bool thumb32_LDMIA(bool W, Reg n, Imm<16> reg_list);
+    bool thumb32_POP(Imm<16> reg_list);
+    bool thumb32_PUSH(Imm<15> reg_list);
+    bool thumb32_STMIA(bool W, Reg n, Imm<15> reg_list);
+    bool thumb32_STMDB(bool W, Reg n, Imm<15> reg_list);
+
    // thumb32 data processing (shifted register) instructions
    bool thumb32_TST_reg(Reg n, Imm<3> imm3, Imm<2> imm2, ShiftType type, Reg m);
    bool thumb32_AND_reg(bool S, Reg n, Imm<3> imm3, Reg d, Imm<2> imm2, ShiftType type, Reg m);
@ -209,6 +217,8 @@ struct ThumbTranslatorVisitor final {
    bool thumb32_RSB_imm(Imm<1> i, bool S, Reg n, Imm<3> imm3, Reg d, Imm<8> imm8);

    // thumb32 data processing (plain binary immediate) instructions.
+    bool thumb32_ADR_t2(Imm<1> imm1, Imm<3> imm3, Reg d, Imm<8> imm8);
+    bool thumb32_ADR_t3(Imm<1> imm1, Imm<3> imm3, Reg d, Imm<8> imm8);
    bool thumb32_ADD_imm_2(Imm<1> imm1, Imm<3> imm3, Reg d, Imm<8> imm8);
    bool thumb32_BFC(Imm<3> imm3, Reg d, Imm<2> imm2, Imm<5> msb);
    bool thumb32_BFI(Reg n, Imm<3> imm3, Reg d, Imm<2> imm2, Imm<5> msb);
@ -243,8 +253,20 @@ struct ThumbTranslatorVisitor final {
    bool thumb32_B_cond(Imm<1> S, Cond cond, Imm<6> hi, Imm<1> j1, Imm<1> j2, Imm<11> lo);

    // thumb32 store single data item instructions
+    bool thumb32_STRB_imm_1(Reg n, Reg t, bool P, bool U, Imm<8> imm8);
+    bool thumb32_STRB_imm_2(Reg n, Reg t, Imm<8> imm8);
+    bool thumb32_STRB_imm_3(Reg n, Reg t, Imm<12> imm12);
+    bool thumb32_STRBT(Reg n, Reg t, Imm<8> imm8);
    bool thumb32_STRB(Reg n, Reg t, Imm<2> imm2, Reg m);
+    bool thumb32_STRH_imm_1(Reg n, Reg t, bool P, bool U, Imm<8> imm8);
+    bool thumb32_STRH_imm_2(Reg n, Reg t, Imm<8> imm8);
+    bool thumb32_STRH_imm_3(Reg n, Reg t, Imm<12> imm12);
+    bool thumb32_STRHT(Reg n, Reg t, Imm<8> imm8);
    bool thumb32_STRH(Reg n, Reg t, Imm<2> imm2, Reg m);
+    bool thumb32_STR_imm_1(Reg n, Reg t, bool P, bool U, Imm<8> imm8);
+    bool thumb32_STR_imm_2(Reg n, Reg t, Imm<8> imm8);
+    bool thumb32_STR_imm_3(Reg n, Reg t, Imm<12> imm12);
+    bool thumb32_STRT(Reg n, Reg t, Imm<8> imm8);
    bool thumb32_STR_reg(Reg n, Reg t, Imm<2> imm2, Reg m);

    // thumb32 load byte and memory hints
@ -267,6 +289,25 @@ struct ThumbTranslatorVisitor final {
    bool thumb32_LDRSB_imm12(Reg n, Reg t, Imm<12> imm12);
    bool thumb32_LDRSBT(Reg n, Reg t, Imm<8> imm8);

+    // thumb32 load halfword instructions
+    bool thumb32_LDRH_lit(bool U, Reg t, Imm<12> imm12);
+    bool thumb32_LDRH_reg(Reg n, Reg t, Imm<2> imm2, Reg m);
+    bool thumb32_LDRH_imm8(Reg n, Reg t, bool P, bool U, bool W, Imm<8> imm8);
+    bool thumb32_LDRH_imm12(Reg n, Reg t, Imm<12> imm12);
+    bool thumb32_LDRHT(Reg n, Reg t, Imm<8> imm8);
+    bool thumb32_LDRSH_lit(bool U, Reg t, Imm<12> imm12);
+    bool thumb32_LDRSH_reg(Reg n, Reg t, Imm<2> imm2, Reg m);
+    bool thumb32_LDRSH_imm8(Reg n, Reg t, bool P, bool U, bool W, Imm<8> imm8);
+    bool thumb32_LDRSH_imm12(Reg n, Reg t, Imm<12> imm12);
+    bool thumb32_LDRSHT(Reg n, Reg t, Imm<8> imm8);
+
+    // thumb32 load word instructions
+    bool thumb32_LDR_lit(bool U, Reg t, Imm<12> imm12);
+    bool thumb32_LDR_reg(Reg n, Reg t, Imm<2> imm2, Reg m);
+    bool thumb32_LDR_imm8(Reg n, Reg t, bool P, bool U, bool W, Imm<8> imm8);
+    bool thumb32_LDR_imm12(Reg n, Reg t, Imm<12> imm12);
+    bool thumb32_LDRT(Reg n, Reg t, Imm<8> imm8);
+
    // thumb32 data processing (register) instructions
    bool thumb32_ASR_reg(Reg m, Reg d, Reg s);
    bool thumb32_LSL_reg(Reg m, Reg d, Reg s);
--- a/externals/dynarmic/src/frontend/A64/decoder/a64.inc
+++ b/externals/dynarmic/src/frontend/A64/decoder/a64.inc
@ -61,6 +61,8 @@ INST(WFI,                    "WFI",                                       "11010
 INST(SEV,                    "SEV",                                       "11010101000000110010000010011111")
 INST(SEVL,                   "SEVL",                                      "11010101000000110010000010111111")
 //INST(DGH,                    "DGH",                                       "11010101000000110010000011011111") // v8.6
+//INST(WFET,                   "WFET",                                      "110101010000001100010000000ddddd") // v8.7
+//INST(WFIT,                   "WFIT",                                      "110101010000001100010000001ddddd") // v8.7
 //INST(XPAC_1,                 "XPACD, XPACI, XPACLRI",                     "110110101100000101000D11111ddddd")
 //INST(XPAC_2,                 "XPACD, XPACI, XPACLRI",                     "11010101000000110010000011111111")
 //INST(PACIA_1,                "PACIA, PACIA1716, PACIASP, PACIAZ, PACIZA", "110110101100000100Z000nnnnnddddd")
@ -267,6 +269,10 @@ INST(LDTRSW,                 "LDTRSW",                                    "10111
 //INST(LDUMIN,                 "LDUMIN, LDUMINA, LDUMINAL, LDUMINL",        "1-111000AR1sssss011100nnnnnttttt")
 //INST(SWP,                    "SWP, SWPA, SWPAL, SWPL",                    "1-111000AR1sssss100000nnnnnttttt")
 //INST(LDAPR,                  "LDAPR",                                     "1-11100010111111110000nnnnnttttt")
+//INST(LD64B,                  "LD64B",                                     "1111100000111111110100nnnnnttttt") // v8.7
+//INST(ST64B,                  "ST64B",                                     "1111100000111111100100nnnnnttttt") // v8.7
+//INST(ST64BV,                 "ST64BV",                                    "11111000001sssss101100nnnnnttttt") // v8.7
+//INST(ST64BV0,                "ST64BV0",                                   "11111000001sssss101000nnnnnttttt") // v8.7

 // Loads and stores - Load/Store register (register offset)
 INST(STRx_reg,               "STRx (register)",                           "zz111000o01mmmmmxxxS10nnnnnttttt")
				`@ -0,0 +1 @@`
				`include("${CMAKE_CURRENT_LIST_DIR}/@PROJECT_NAME@-targets.cmake")`