30% faster JIT compiler

makefile

@@ -95,7 +95,8 @@ $(OBJDIR)/jit_compiler_x86_static.o: $(SRCDIR)/jit_compiler_x86_static.S \
  $(SRCDIR)/asm/program_read_dataset_sshash_fin.inc \
  $(SRCDIR)/asm/program_loop_store.inc $(SRCDIR)/asm/program_epilogue_linux.inc \
  $(SRCDIR)/asm/program_epilogue_store.inc $(SRCDIR)/asm/program_sshash_load.inc \
- $(SRCDIR)/asm/program_sshash_prefetch.inc $(SRCDIR)/asm/program_sshash_constants.inc
+ $(SRCDIR)/asm/program_sshash_prefetch.inc $(SRCDIR)/asm/program_sshash_constants.inc \
+ $(SRCDIR)/asm/randomx_reciprocal.inc
 $(OBJDIR)/soft_aes.o: $(SRCDIR)/soft_aes.cpp $(SRCDIR)/soft_aes.h $(SRCDIR)/intrin_portable.h
 $(OBJDIR)/virtual_memory.o: $(SRCDIR)/virtual_memory.cpp $(SRCDIR)/virtual_memory.hpp
 $(OBJDIR)/vm_interpreted.o: $(SRCDIR)/vm_interpreted.cpp $(SRCDIR)/vm_interpreted.hpp \

randomx.sln

@@ -23,6 +23,8 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "code-generator", "vcxproj\c
 EndProject
 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "scratchpad-entropy", "vcxproj\scratchpad-entropy.vcxproj", "{FF8BD408-AFD8-43C6-BE98-4D03B37E840B}"
 EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "jit-performance", "vcxproj\jit-performance.vcxproj", "{535F2111-FA81-4C76-A354-EDD2F9AA00E3}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|x64 = Debug|x64
@@ -103,6 +105,14 @@ Global
 		{FF8BD408-AFD8-43C6-BE98-4D03B37E840B}.Release|x64.Build.0 = Release|x64
 		{FF8BD408-AFD8-43C6-BE98-4D03B37E840B}.Release|x86.ActiveCfg = Release|Win32
 		{FF8BD408-AFD8-43C6-BE98-4D03B37E840B}.Release|x86.Build.0 = Release|Win32
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3}.Debug|x64.ActiveCfg = Debug|x64
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3}.Debug|x64.Build.0 = Debug|x64
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3}.Debug|x86.ActiveCfg = Debug|Win32
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3}.Debug|x86.Build.0 = Debug|Win32
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3}.Release|x64.ActiveCfg = Release|x64
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3}.Release|x64.Build.0 = Release|x64
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3}.Release|x86.ActiveCfg = Release|Win32
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3}.Release|x86.Build.0 = Release|Win32
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE
@@ -116,6 +126,7 @@ Global
 		{44947B9C-E6B1-4C06-BD01-F8EF43B59223} = {4A4A689F-86AF-41C0-A974-1080506D0923}
 		{3E490DEC-1874-43AA-92DA-1AC57C217EAC} = {4A4A689F-86AF-41C0-A974-1080506D0923}
 		{FF8BD408-AFD8-43C6-BE98-4D03B37E840B} = {4A4A689F-86AF-41C0-A974-1080506D0923}
+		{535F2111-FA81-4C76-A354-EDD2F9AA00E3} = {4A4A689F-86AF-41C0-A974-1080506D0923}
 	EndGlobalSection
 	GlobalSection(ExtensibilityGlobals) = postSolution
 		SolutionGuid = {4EBC03DB-AE37-4141-8147-692F16E0ED02}

src/asm/randomx_reciprocal.inc

@@ -0,0 +1,7 @@
+	mov edx, 1
+	mov r8, rcx
+	xor eax, eax
+	bsr rcx, rcx
+	shl rdx, cl
+	div r8
+	ret

src/jit_compiler_x86.cpp

@@ -616,7 +616,7 @@ namespace randomx {
 		if (!isPowerOf2(divisor)) {
 			registerUsage[instr.dst].lastUsed = i;
 			emit(MOV_RAX_I);
-			emit64(randomx_reciprocal(divisor));
+			emit64(randomx_reciprocal_fast(divisor));
 			emit(REX_IMUL_RM);
 			emitByte(0xc0 + 8 * instr.dst);
 		}

src/jit_compiler_x86.hpp

@@ -78,23 +78,13 @@ namespace randomx {
 		}
 
 		void emit32(uint32_t val) {
-			code[codePos + 0] = val;
+			memcpy(code + codePos, &val, sizeof val);
-			code[codePos + 1] = val >> 8;
+			codePos += sizeof val;
-			code[codePos + 2] = val >> 16;
-			code[codePos + 3] = val >> 24;
-			codePos += 4;
 		}
 
 		void emit64(uint64_t val) {
-			code[codePos + 0] = val;
+			memcpy(code + codePos, &val, sizeof val);
-			code[codePos + 1] = val >> 8;
+			codePos += sizeof val;
-			code[codePos + 2] = val >> 16;
-			code[codePos + 3] = val >> 24;
-			code[codePos + 4] = val >> 32;
-			code[codePos + 5] = val >> 40;
-			code[codePos + 6] = val >> 48;
-			code[codePos + 7] = val >> 56;
-			codePos += 8;
 		}
 
 		template<size_t N>

src/jit_compiler_x86_static.S

@@ -42,6 +42,7 @@
 .global DECL(randomx_sshash_end)
 .global DECL(randomx_sshash_init)
 .global DECL(randomx_program_end)
+.global DECL(randomx_reciprocal_fast)
 
 #define db .byte
 
@@ -158,3 +159,7 @@ DECL(randomx_sshash_init):
 .balign 64
 DECL(randomx_program_end):
 	nop
+
+DECL(randomx_reciprocal_fast):
+	mov rcx, rdi
+	#include "asm/randomx_reciprocal.inc"

src/jit_compiler_x86_static.asm

@@ -35,6 +35,7 @@ PUBLIC randomx_sshash_prefetch
 PUBLIC randomx_sshash_end
 PUBLIC randomx_sshash_init
 PUBLIC randomx_program_end
+PUBLIC randomx_reciprocal_fast
 
 ALIGN 64
 randomx_program_prologue PROC
@@ -169,6 +170,10 @@ randomx_program_end PROC
 	nop
 randomx_program_end ENDP
 
+randomx_reciprocal_fast PROC
+	include asm/randomx_reciprocal.inc
+randomx_reciprocal_fast ENDP
+
 _RANDOMX_JITX86_STATIC ENDS
 
 ENDIF

src/reciprocal.h

@@ -26,6 +26,7 @@ extern "C" {
 #endif
 
 uint64_t randomx_reciprocal(uint64_t);
+uint64_t randomx_reciprocal_fast(uint64_t);
 
 #if defined(__cplusplus)
 }

src/tests/jit-performance.cpp

@@ -0,0 +1,44 @@
+#include "../aes_hash.hpp"
+#include "../jit_compiler_x86.hpp"
+#include "../program.hpp"
+#include "utility.hpp"
+#include "stopwatch.hpp"
+#include "../blake2/blake2.h"
+#include "../reciprocal.h"
+
+int main(int argc, char** argv) {
+	int count;
+	readInt(argc, argv, count, 1000000);
+
+	const char seed[] = "JIT performance test seed";
+	uint8_t hash[64];
+
+	blake2b(&hash, sizeof hash, &seed, sizeof seed, nullptr, 0);
+
+	randomx::ProgramConfiguration config;
+
+	randomx::Program program;
+	randomx::JitCompilerX86 jit;
+
+	std::cout << "Compiling " << count << " programs..." << std::endl;
+
+	Stopwatch sw(true);
+
+	for (int i = 0; i < count; ++i) {
+		fillAes1Rx4<false>(hash, sizeof(program), &program);
+		auto addressRegisters = program.getEntropy(12);
+		config.readReg0 = 0 + (addressRegisters & 1);
+		addressRegisters >>= 1;
+		config.readReg1 = 2 + (addressRegisters & 1);
+		addressRegisters >>= 1;
+		config.readReg2 = 4 + (addressRegisters & 1);
+		addressRegisters >>= 1;
+		config.readReg3 = 6 + (addressRegisters & 1);
+		jit.generateProgram(program, config);
+	}
+
+	std::cout << "Elapsed: " << sw.getElapsed() << " s" << std::endl;
+
+	dump((const char*)jit.getProgramFunc(), randomx::CodeSize, "program.bin");
+	return 0;
+}

vcxproj/jit-performance.vcxproj

@@ -0,0 +1,128 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="15.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug|Win32">
+      <Configuration>Debug</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|Win32">
+      <Configuration>Release</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|x64">
+      <Configuration>Debug</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|x64">
+      <Configuration>Release</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <VCProjectVersion>15.0</VCProjectVersion>
+    <ProjectGuid>{535F2111-FA81-4C76-A354-EDD2F9AA00E3}</ProjectGuid>
+    <RootNamespace>jitperformance</RootNamespace>
+    <WindowsTargetPlatformVersion>10.0.17763.0</WindowsTargetPlatformVersion>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v141</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v141</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v141</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v141</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <ImportGroup Label="ExtensionSettings">
+  </ImportGroup>
+  <ImportGroup Label="Shared">
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <PropertyGroup />
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <SDLCheck>true</SDLCheck>
+      <ConformanceMode>true</ConformanceMode>
+    </ClCompile>
+    <Link>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <SDLCheck>true</SDLCheck>
+      <ConformanceMode>true</ConformanceMode>
+    </ClCompile>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <SDLCheck>true</SDLCheck>
+      <ConformanceMode>true</ConformanceMode>
+    </ClCompile>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <SDLCheck>true</SDLCheck>
+      <ConformanceMode>true</ConformanceMode>
+    </ClCompile>
+    <Link>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemGroup>
+    <ClCompile Include="..\src\tests\jit-performance.cpp" />
+  </ItemGroup>
+  <ItemGroup>
+    <ProjectReference Include="..\vcxproj\randomx.vcxproj">
+      <Project>{3346a4ad-c438-4324-8b77-47a16452954b}</Project>
+    </ProjectReference>
+  </ItemGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ImportGroup Label="ExtensionTargets">
+  </ImportGroup>
+</Project>

vcxproj/jit-performance.vcxproj.filters

@@ -0,0 +1,22 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup>
+    <Filter Include="Source Files">
+      <UniqueIdentifier>{4FC737F1-C7A5-4376-A066-2A32D752A2FF}</UniqueIdentifier>
+      <Extensions>cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx</Extensions>
+    </Filter>
+    <Filter Include="Header Files">
+      <UniqueIdentifier>{93995380-89BD-4b04-88EB-625FBE52EBFB}</UniqueIdentifier>
+      <Extensions>h;hh;hpp;hxx;hm;inl;inc;ipp;xsd</Extensions>
+    </Filter>
+    <Filter Include="Resource Files">
+      <UniqueIdentifier>{67DA6AB6-F800-4c08-8B7A-83BB121AAD01}</UniqueIdentifier>
+      <Extensions>rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe;resx;tiff;tif;png;wav;mfcribbon-ms</Extensions>
+    </Filter>
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="..\src\tests\jit-performance.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+  </ItemGroup>
+</Project>