Merge branch 'dev'

CHANGELOG.md

@@ -1,3 +1,10 @@
+# v5.5.2
+- [#1500](https://github.com/xmrig/xmrig/pull/1500) Removed unnecessary code from RandomX JIT compiler.
+- [#1502](https://github.com/xmrig/xmrig/pull/1502) Optimizations for AMD Bulldozer.
+- [#1508](https://github.com/xmrig/xmrig/pull/1508) Added support for BMI2 instructions.
+- [#1510](https://github.com/xmrig/xmrig/pull/1510) Optimized `CFROUND` instruction for RandomX.
+- [#1520](https://github.com/xmrig/xmrig/pull/1520) Fixed thread affinity.
+
 # v5.5.1
 - [#1469](https://github.com/xmrig/xmrig/issues/1469) Fixed build with gcc 4.8.
 - [#1473](https://github.com/xmrig/xmrig/pull/1473) Added RandomX auto-config for mobile Ryzen APUs.

src/backend/cpu/interfaces/ICpuInfo.h

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -62,6 +62,7 @@ public:
     virtual Assembly::Id assembly() const                                           = 0;
     virtual bool hasAES() const                                                     = 0;
     virtual bool hasAVX2() const                                                    = 0;
+    virtual bool hasBMI2() const                                                    = 0;
     virtual bool hasOneGbPages() const                                              = 0;
     virtual const char *backend() const                                             = 0;
     virtual const char *brand() const                                               = 0;

src/backend/cpu/platform/AdvancedCpuInfo.cpp

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -153,6 +153,7 @@ xmrig::AdvancedCpuInfo::AdvancedCpuInfo() :
     }
 
     m_avx2 = data.flags[CPU_FEATURE_AVX2] && data.flags[CPU_FEATURE_OSXSAVE];
+    m_bmi2 = data.flags[CPU_FEATURE_BMI2];
 }
 
 

src/backend/cpu/platform/AdvancedCpuInfo.h

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -43,6 +43,7 @@ protected:
     inline Assembly::Id assembly() const override   { return m_assembly; }
     inline bool hasAES() const override             { return m_aes; }
     inline bool hasAVX2() const override            { return m_avx2; }
+    inline bool hasBMI2() const override            { return m_bmi2; }
     inline bool hasOneGbPages() const override      { return m_pdpe1gb; }
     inline const char *backend() const override     { return m_backend; }
     inline const char *brand() const override       { return m_brand; }
@@ -59,6 +60,7 @@ private:
     Assembly m_assembly;
     bool m_aes            = false;
     bool m_avx2           = false;
+    bool m_bmi2           = false;
     bool m_L2_exclusive   = false;
     char m_backend[32]{};
     char m_brand[64 + 5]{};

src/backend/cpu/platform/BasicCpuInfo.cpp

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -45,6 +45,10 @@
 #   define bit_AVX2 (1 << 5)
 #endif
 
+#ifndef bit_BMI2
+#   define bit_BMI2 (1 << 8)
+#endif
+
 #ifndef bit_PDPE1GB
 #   define bit_PDPE1GB (1 << 26)
 #endif
@@ -141,6 +145,12 @@ static inline bool has_avx2()
 }
 
 
+static inline bool has_bmi2()
+{
+    return has_feature(EXTENDED_FEATURES, EBX_Reg, bit_BMI2);
+}
+
+
 static inline bool has_pdpe1gb()
 {
     return has_feature(PROCESSOR_EXT_INFO, EDX_Reg, bit_PDPE1GB);
@@ -154,6 +164,7 @@ xmrig::BasicCpuInfo::BasicCpuInfo() :
     m_threads(std::thread::hardware_concurrency()),
     m_aes(has_aes_ni()),
     m_avx2(has_avx2()),
+    m_bmi2(has_bmi2()),
     m_pdpe1gb(has_pdpe1gb())
 {
     cpu_brand_string(m_brand);

src/backend/cpu/platform/BasicCpuInfo.h

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -44,6 +44,7 @@ protected:
     inline Assembly::Id assembly() const override   { return m_assembly; }
     inline bool hasAES() const override             { return m_aes; }
     inline bool hasAVX2() const override            { return m_avx2; }
+    inline bool hasBMI2() const override            { return m_bmi2; }
     inline bool hasOneGbPages() const override      { return m_pdpe1gb; }
     inline const char *brand() const override       { return m_brand; }
     inline MsrMod msrMod() const override           { return m_msrMod; }
@@ -63,6 +64,7 @@ private:
     Assembly m_assembly     = Assembly::NONE;
     bool m_aes              = false;
     const bool m_avx2       = false;
+    const bool m_bmi2       = false;
     const bool m_pdpe1gb    = false;
     MsrMod m_msrMod         = MSR_MOD_NONE;
     Vendor m_vendor         = VENDOR_UNKNOWN;

src/base/kernel/Platform.h

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by

src/base/kernel/Platform_hwloc.cpp

@@ -6,8 +6,8 @@
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -30,6 +30,7 @@
 
 
 #include <hwloc.h>
+#include <thread>
 
 
 bool xmrig::Platform::setThreadAffinity(uint64_t cpu_id)
@@ -42,8 +43,11 @@ bool xmrig::Platform::setThreadAffinity(uint64_t cpu_id)
     }
 
     if (hwloc_set_cpubind(cpu->topology(), pu->cpuset, HWLOC_CPUBIND_THREAD | HWLOC_CPUBIND_STRICT) >= 0) {
+        std::this_thread::sleep_for(std::chrono::milliseconds(1));
         return true;
     }
 
-    return hwloc_set_cpubind(cpu->topology(), pu->cpuset, HWLOC_CPUBIND_THREAD) >= 0;
+    const bool result = (hwloc_set_cpubind(cpu->topology(), pu->cpuset, HWLOC_CPUBIND_THREAD) >= 0);
+    std::this_thread::sleep_for(std::chrono::milliseconds(1));
+    return result;
 }

src/base/kernel/Platform_mac.cpp

@@ -4,8 +4,8 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- *
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -28,6 +28,7 @@
 #include <stdlib.h>
 #include <sys/resource.h>
 #include <uv.h>
+#include <thread>
 
 
 #include "base/kernel/Platform.h"
@@ -67,7 +68,9 @@ bool xmrig::Platform::setThreadAffinity(uint64_t cpu_id)
     thread_affinity_policy_data_t policy = { static_cast<integer_t>(cpu_id) };
     mach_thread = pthread_mach_thread_np(pthread_self());
 
-    return thread_policy_set(mach_thread, THREAD_AFFINITY_POLICY, (thread_policy_t)&policy, 1) == KERN_SUCCESS;
+    const bool result = (thread_policy_set(mach_thread, THREAD_AFFINITY_POLICY, (thread_policy_t)&policy, 1) == KERN_SUCCESS);
+    std::this_thread::sleep_for(std::chrono::milliseconds(1));
+    return result;
 }
 #endif
 

src/base/kernel/Platform_unix.cpp

@@ -4,8 +4,8 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -37,6 +37,7 @@
 #include <sys/resource.h>
 #include <unistd.h>
 #include <uv.h>
+#include <thread>
 
 
 #include "base/kernel/Platform.h"
@@ -92,10 +93,13 @@ bool xmrig::Platform::setThreadAffinity(uint64_t cpu_id)
     CPU_SET(cpu_id, &mn);
 
 #   ifndef __ANDROID__
-    return pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &mn) == 0;
+    const bool result = (pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &mn) == 0);
 #   else
-    return sched_setaffinity(gettid(), sizeof(cpu_set_t), &mn) == 0;
+    const bool result = (sched_setaffinity(gettid(), sizeof(cpu_set_t), &mn) == 0);
 #   endif
+
+    std::this_thread::sleep_for(std::chrono::milliseconds(1));
+    return result;
 }
 #endif
 

src/base/kernel/Platform_win.cpp

@@ -4,9 +4,8 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2018      SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
- * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -98,7 +97,9 @@ bool xmrig::Platform::setThreadAffinity(uint64_t cpu_id)
         LOG_ERR("Unable to set affinity. Windows supports only affinity up to 63.");
     }
 
-    return SetThreadAffinityMask(GetCurrentThread(), 1ULL << cpu_id) != 0;
+    const bool result = (SetThreadAffinityMask(GetCurrentThread(), 1ULL << cpu_id) != 0);
+    Sleep(1);
+    return result;
 }
 #endif
 

src/crypto/randomx/asm/program_epilogue_store.inc

@@ -1,5 +1,5 @@
 	;# save VM register values
-	add rsp, 24
+	add rsp, 40
 	pop rcx
 	mov qword ptr [rcx+0], r8
 	mov qword ptr [rcx+8], r9

src/crypto/randomx/asm/program_loop_load.inc

@@ -1,5 +1,5 @@
 	lea rcx, [rsi+rax]
-	mov [rsp+8], rcx
+	mov [rsp+16], rcx
 	xor r8,  qword ptr [rcx+0]
 	xor r9,  qword ptr [rcx+8]
 	xor r10, qword ptr [rcx+16]
@@ -9,7 +9,7 @@
 	xor r14, qword ptr [rcx+48]
 	xor r15, qword ptr [rcx+56]
 	lea rcx, [rsi+rdx]
-	mov [rsp+16], rcx
+	mov [rsp+24], rcx
 	cvtdq2pd xmm0, qword ptr [rcx+0]
 	cvtdq2pd xmm1, qword ptr [rcx+8]
 	cvtdq2pd xmm2, qword ptr [rcx+16]
@@ -18,11 +18,11 @@
 	cvtdq2pd xmm5, qword ptr [rcx+40]
 	cvtdq2pd xmm6, qword ptr [rcx+48]
 	cvtdq2pd xmm7, qword ptr [rcx+56]
-	andps xmm4, xmm13
+	andpd xmm4, xmm13
-	andps xmm5, xmm13
+	andpd xmm5, xmm13
-	andps xmm6, xmm13
+	andpd xmm6, xmm13
-	andps xmm7, xmm13
+	andpd xmm7, xmm13
-	orps xmm4, xmm14
+	orpd xmm4, xmm14
-	orps xmm5, xmm14
+	orpd xmm5, xmm14
-	orps xmm6, xmm14
+	orpd xmm6, xmm14
-	orps xmm7, xmm14
+	orpd xmm7, xmm14

src/crypto/randomx/asm/program_loop_load_xop.inc

@@ -0,0 +1,24 @@
+	lea rcx, [rsi+rax]
+	mov [rsp+8], rcx
+	xor r8,  qword ptr [rcx+0]
+	xor r9,  qword ptr [rcx+8]
+	xor r10, qword ptr [rcx+16]
+	xor r11, qword ptr [rcx+24]
+	xor r12, qword ptr [rcx+32]
+	xor r13, qword ptr [rcx+40]
+	xor r14, qword ptr [rcx+48]
+	xor r15, qword ptr [rcx+56]
+	lea rcx, [rsi+rdx]
+	mov [rsp+16], rcx
+	cvtdq2pd xmm0, qword ptr [rcx+0]
+	cvtdq2pd xmm1, qword ptr [rcx+8]
+	cvtdq2pd xmm2, qword ptr [rcx+16]
+	cvtdq2pd xmm3, qword ptr [rcx+24]
+	cvtdq2pd xmm4, qword ptr [rcx+32]
+	cvtdq2pd xmm5, qword ptr [rcx+40]
+	cvtdq2pd xmm6, qword ptr [rcx+48]
+	cvtdq2pd xmm7, qword ptr [rcx+56]
+	vpcmov xmm4, xmm4, xmm14, xmm13
+	vpcmov xmm5, xmm5, xmm14, xmm13
+	vpcmov xmm6, xmm6, xmm14, xmm13
+	vpcmov xmm7, xmm7, xmm14, xmm13

src/crypto/randomx/asm/program_loop_store.inc

@@ -1,4 +1,4 @@
-	mov rcx, [rsp+16]
+	mov rcx, [rsp+24]
 	mov qword ptr [rcx+0], r8
 	mov qword ptr [rcx+8], r9
 	mov qword ptr [rcx+16], r10
@@ -7,7 +7,7 @@
 	mov qword ptr [rcx+40], r13
 	mov qword ptr [rcx+48], r14
 	mov qword ptr [rcx+56], r15
-	mov rcx, [rsp+8]
+	mov rcx, [rsp+16]
 	xorpd xmm0, xmm4
 	xorpd xmm1, xmm5
 	xorpd xmm2, xmm6

src/crypto/randomx/asm/program_xmm_constants.inc

@@ -1,5 +1,5 @@
 mantissaMask:
-	db 255, 255, 255, 255, 255, 255, 255, 0, 255, 255, 255, 255, 255, 255, 255, 0
+	db 0, 0, 192, 255, 255, 255, 255, 0, 0, 0, 192, 255, 255, 255, 255, 0
 exp240:
 	db 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
 scaleMask:

src/crypto/randomx/jit_compiler_x86.cpp

@@ -89,6 +89,7 @@ namespace randomx {
 	const uint8_t* codePrologue = (uint8_t*)&randomx_program_prologue;
 	const uint8_t* codeLoopBegin = (uint8_t*)&randomx_program_loop_begin;
 	const uint8_t* codeLoopLoad = (uint8_t*)&randomx_program_loop_load;
+	const uint8_t* codeLoopLoadXOP = (uint8_t*)&randomx_program_loop_load_xop;
 	const uint8_t* codeProgamStart = (uint8_t*)&randomx_program_start;
 	const uint8_t* codeReadDatasetLightSshInit = (uint8_t*)&randomx_program_read_dataset_sshash_init;
 	const uint8_t* codeReadDatasetLightSshFin = (uint8_t*)&randomx_program_read_dataset_sshash_fin;
@@ -104,7 +105,8 @@ namespace randomx {
 
 	const int32_t prefetchScratchpadSize = codePrefetchScratchpadEnd - codePrefetchScratchpad;
 	const int32_t prologueSize = codeLoopBegin - codePrologue;
-	const int32_t loopLoadSize = codeProgamStart - codeLoopLoad;
+	const int32_t loopLoadSize = codeLoopLoadXOP - codeLoopLoad;
+	const int32_t loopLoadXOPSize = codeProgamStart - codeLoopLoadXOP;
 	const int32_t readDatasetLightInitSize = codeReadDatasetLightSshFin - codeReadDatasetLightSshInit;
 	const int32_t readDatasetLightFinSize = codeLoopStore - codeReadDatasetLightSshFin;
 	const int32_t loopStoreSize = codeLoopEnd - codeLoopStore;
@@ -184,6 +186,7 @@ namespace randomx {
 	static const uint8_t REX_XOR_RAX_R64[] = { 0x49, 0x33 };
 	static const uint8_t REX_XCHG[] = { 0x4d, 0x87 };
 	static const uint8_t REX_ANDPS_XMM12[] = { 0x45, 0x0F, 0x54, 0xE5, 0x45, 0x0F, 0x56, 0xE6 };
+	static const uint8_t REX_VPCMOV_XMM12[] = { 0x8F, 0x48, 0x18, 0xA2, 0xE6, 0xD0 };
 	static const uint8_t REX_PADD[] = { 0x66, 0x44, 0x0f };
 	static const uint8_t PADD_OPCODES[] = { 0xfc, 0xfd, 0xfe, 0xd4 };
 	static const uint8_t CALL = 0xe8;
@@ -295,11 +298,23 @@ namespace randomx {
 		cpuid(1, info);
 		hasAVX = ((info[2] & (1 << 27)) != 0) && ((info[2] & (1 << 28)) != 0);
 
+		cpuid(0x80000001, info);
+		hasXOP = ((info[2] & (1 << 11)) != 0);
+
 		allocatedCode = (uint8_t*)allocExecutableMemory(CodeSize * 2);
 		// Shift code base address to improve caching - all threads will use different L2/L3 cache sets
 		code = allocatedCode + (codeOffset.fetch_add(59 * 64) % CodeSize);
 		memcpy(code, codePrologue, prologueSize);
+		if (hasXOP) {
+			memcpy(code + prologueSize, codeLoopLoadXOP, loopLoadXOPSize);
+		}
+		else {
+			memcpy(code + prologueSize, codeLoopLoad, loopLoadSize);
+		}
 		memcpy(code + epilogueOffset, codeEpilogue, epilogueSize);
+
+		codePosFirst = prologueSize + (hasXOP ? loopLoadXOPSize : loopLoadSize);
+
 #		ifdef XMRIG_FIX_RYZEN
 		mainLoopBounds.first = code + prologueSize;
 		mainLoopBounds.second = code + epilogueOffset;
@@ -317,7 +332,7 @@ namespace randomx {
 
 		uint8_t* p;
 		uint32_t n;
-		if (flags & RANDOMX_FLAG_RYZEN) {
+		if (flags & RANDOMX_FLAG_AMD) {
 			p = RandomX_CurrentConfig.codeReadDatasetRyzenTweaked;
 			n = RandomX_CurrentConfig.codeReadDatasetRyzenTweakedSize;
 		}
@@ -385,7 +400,7 @@ namespace randomx {
 		*(uint32_t*)(code + codePos + 10) = RandomX_CurrentConfig.ScratchpadL3Mask64_Calculated;
 		*(uint32_t*)(code + codePos + 20) = RandomX_CurrentConfig.ScratchpadL3Mask64_Calculated;
 		if (hasAVX) {
-			uint32_t* p = (uint32_t*)(code + codePos + 32);
+			uint32_t* p = (uint32_t*)(code + codePos + 67);
 			*p = (*p & 0xFF000000U) | 0x0077F8C5U;
 		}
 
@@ -393,10 +408,8 @@ namespace randomx {
 		xmrig::Rx::setMainLoopBounds(mainLoopBounds);
 #		endif
 
-		codePos = prologueSize;
+		memcpy(code + prologueSize - 48, &pcfg.eMask, sizeof(pcfg.eMask));
-		memcpy(code + codePos - 48, &pcfg.eMask, sizeof(pcfg.eMask));
+		codePos = codePosFirst;
-		memcpy(code + codePos, codeLoopLoad, loopLoadSize);
-		codePos += loopLoadSize;
 
 		//mark all registers as used
 		uint64_t* r = (uint64_t*)registerUsage;
@@ -708,14 +721,31 @@ namespace randomx {
 		uint8_t* const p = code;
 		int pos = codePos;
 
+		const uint32_t dst = instr.dst;
+
 		emit(REX_MOV_RR64, p, pos);
-		emitByte(0xc0 + instr.dst, p, pos);
+		emitByte(0xc0 + dst, p, pos);
 		emit(REX_MUL_R, p, pos);
 		emitByte(0xe0 + instr.src, p, pos);
 		emit(REX_MOV_R64R, p, pos);
-		emitByte(0xc2 + 8 * instr.dst, p, pos);
+		emitByte(0xc2 + 8 * dst, p, pos);
 
-		registerUsage[instr.dst] = pos;
+		registerUsage[dst] = pos;
+		codePos = pos;
+	}
+
+	void JitCompilerX86::h_IMULH_R_BMI2(const Instruction& instr) {
+		uint8_t* const p = code;
+		int pos = codePos;
+
+		const uint32_t src = instr.src;
+		const uint32_t dst = instr.dst;
+
+		*(uint32_t*)(p + pos) = 0xC4D08B49 + (dst << 16);
+		*(uint32_t*)(p + pos + 4) = 0xC0F6FB42 + (dst << 27) + (src << 24);
+		pos += 8;
+
+		registerUsage[dst] = pos;
 		codePos = pos;
 	}
 
@@ -743,6 +773,29 @@ namespace randomx {
 		codePos = pos;
 	}
 
+	void JitCompilerX86::h_IMULH_M_BMI2(const Instruction& instr) {
+		uint8_t* const p = code;
+		int pos = codePos;
+
+		const uint64_t src = instr.src;
+		const uint64_t dst = instr.dst;
+
+		if (src != dst) {
+			genAddressReg<false>(instr, p, pos);
+			*(uint32_t*)(p + pos) = static_cast<uint32_t>(0xC4D08B49 + (dst << 16));
+			*(uint64_t*)(p + pos + 4) = 0x0E04F6FB62ULL + (dst << 27);
+			pos += 9;
+		}
+		else {
+			*(uint64_t*)(p + pos) = 0x86F6FB62C4D08B49ULL + (dst << 16) + (dst << 59);
+			*(uint32_t*)(p + pos + 8) = instr.getImm32() & ScratchpadL3Mask;
+			pos += 12;
+		}
+
+		registerUsage[dst] = pos;
+		codePos = pos;
+	}
+
 	void JitCompilerX86::h_ISMULH_R(const Instruction& instr) {
 		uint8_t* const p = code;
 		int pos = codePos;
@@ -992,7 +1045,12 @@ namespace randomx {
 		const uint32_t dst = instr.dst % RegisterCountFlt;
 		genAddressReg<true>(instr, p, pos);
 		emit(REX_CVTDQ2PD_XMM12, p, pos);
-		emit(REX_ANDPS_XMM12, p, pos);
+		if (hasXOP) {
+			emit(REX_VPCMOV_XMM12, p, pos);
+		}
+		else {
+			emit(REX_ANDPS_XMM12, p, pos);
+		}
 		emit(REX_DIVPD, p, pos);
 		emitByte(0xe4 + 8 * dst, p, pos);
 
@@ -1014,18 +1072,21 @@ namespace randomx {
 		uint8_t* const p = code;
 		int pos = codePos;
 
-		emit(REX_MOV_RR64, p, pos);
+		const uint32_t src = instr.src;
-		emitByte(0xc0 + instr.src, p, pos);
+
-		int rotate = (13 - (instr.getImm32() & 63)) & 63;
+		*(uint32_t*)(p + pos) = 0x00C08B49 + (src << 16);
-		if (rotate != 0) {
+		const int rotate = (static_cast<int>(instr.getImm32() & 63) - 2) & 63;
-			emit(ROL_RAX, p, pos);
+		*(uint32_t*)(p + pos + 3) = 0x00C8C148 + (rotate << 24);
-			emitByte(rotate, p, pos);
+
-		}
+		if (vm_flags & RANDOMX_FLAG_AMD) {
-		if (vm_flags & RANDOMX_FLAG_RYZEN) {
+			*(uint64_t*)(p + pos + 7) = 0x742024443B0CE083ULL;
-			emit(AND_OR_MOV_LDMXCSR_RYZEN, p, pos);
+			*(uint8_t*)(p + pos + 15) = 8;
+			*(uint64_t*)(p + pos + 16) = 0x202444890414AE0FULL;
+			pos += 24;
 		}
 		else {
-			emit(AND_OR_MOV_LDMXCSR, p, pos);
+			*(uint64_t*)(p + pos + 7) = 0x0414AE0F0CE083ULL;
+			pos += 14;
 		}
 
 		codePos = pos;

src/crypto/randomx/jit_compiler_x86.hpp

@@ -73,10 +73,12 @@ namespace randomx {
 		std::pair<const void*, const void*> mainLoopBounds;
 #		endif
 		int32_t codePos;
+		int32_t codePosFirst;
 		uint32_t vm_flags;
 
 		static bool BranchesWithin32B;
 		bool hasAVX;
+		bool hasXOP;
 
 		static void applyTweaks();
 		void generateProgramPrologue(Program&, ProgramConfiguration&);
@@ -121,7 +123,9 @@ namespace randomx {
 		void h_IMUL_R(const Instruction&);
 		void h_IMUL_M(const Instruction&);
 		void h_IMULH_R(const Instruction&);
+		void h_IMULH_R_BMI2(const Instruction&);
 		void h_IMULH_M(const Instruction&);
+		void h_IMULH_M_BMI2(const Instruction&);
 		void h_ISMULH_R(const Instruction&);
 		void h_ISMULH_M(const Instruction&);
 		void h_IMUL_RCP(const Instruction&);

src/crypto/randomx/jit_compiler_x86_static.S

@@ -43,6 +43,7 @@
 .global DECL(randomx_program_prologue_first_load)
 .global DECL(randomx_program_loop_begin)
 .global DECL(randomx_program_loop_load)
+.global DECL(randomx_program_loop_load_xop)
 .global DECL(randomx_program_start)
 .global DECL(randomx_program_read_dataset)
 .global DECL(randomx_program_read_dataset_ryzen)
@@ -93,8 +94,12 @@ DECL(randomx_program_prologue_first_load):
 	and eax, RANDOMX_SCRATCHPAD_MASK
 	ror rdx, 32
 	and edx, RANDOMX_SCRATCHPAD_MASK
-	sub rsp, 24
+	sub rsp, 40
-	stmxcsr dword ptr [rsp]
+	mov dword ptr [rsp], 0x9FC0
+	mov dword ptr [rsp+4], 0xBFC0
+	mov dword ptr [rsp+8], 0xDFC0
+	mov dword ptr [rsp+12], 0xFFC0
+	mov dword ptr [rsp+32], -1
 	nop
 	nop
 	nop
@@ -110,6 +115,9 @@ DECL(randomx_program_loop_begin):
 DECL(randomx_program_loop_load):
 	#include "asm/program_loop_load.inc"
 
+DECL(randomx_program_loop_load_xop):
+	#include "asm/program_loop_load_xop.inc"
+
 DECL(randomx_program_start):
 	nop
 

src/crypto/randomx/jit_compiler_x86_static.asm

@@ -34,6 +34,7 @@ PUBLIC randomx_program_prologue
 PUBLIC randomx_program_prologue_first_load
 PUBLIC randomx_program_loop_begin
 PUBLIC randomx_program_loop_load
+PUBLIC randomx_program_loop_load_xop
 PUBLIC randomx_program_start
 PUBLIC randomx_program_read_dataset
 PUBLIC randomx_program_read_dataset_ryzen
@@ -81,8 +82,12 @@ randomx_program_prologue_first_load PROC
 	and eax, RANDOMX_SCRATCHPAD_MASK
 	ror rdx, 32
 	and edx, RANDOMX_SCRATCHPAD_MASK
-	sub rsp, 24
+	sub rsp, 40
-	stmxcsr dword ptr [rsp]
+	mov dword ptr [rsp], 9FC0h
+	mov dword ptr [rsp+4], 0BFC0h
+	mov dword ptr [rsp+8], 0DFC0h
+	mov dword ptr [rsp+12], 0FFC0h
+	mov dword ptr [rsp+32], -1
 	nop
 	nop
 	nop
@@ -101,6 +106,10 @@ randomx_program_loop_load PROC
 	include asm/program_loop_load.inc
 randomx_program_loop_load ENDP
 
+randomx_program_loop_load_xop PROC
+	include asm/program_loop_load_xop.inc
+randomx_program_loop_load_xop ENDP
+
 randomx_program_start PROC
 	nop
 randomx_program_start ENDP

src/crypto/randomx/jit_compiler_x86_static.hpp

@@ -35,6 +35,7 @@ extern "C" {
 	void randomx_program_prologue_first_load();
 	void randomx_program_loop_begin();
 	void randomx_program_loop_load();
+	void randomx_program_loop_load_xop();
 	void randomx_program_start();
 	void randomx_program_read_dataset();
 	void randomx_program_read_dataset_ryzen();

src/crypto/randomx/randomx.cpp

@@ -41,6 +41,8 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "crypto/randomx/jit_compiler_a64_static.hpp"
 #endif
 
+#include "backend/cpu/Cpu.h"
+
 #include <cassert>
 
 RandomX_ConfigurationWownero::RandomX_ConfigurationWownero()
@@ -235,14 +237,29 @@ void RandomX_ConfigurationBase::Apply()
 	CEIL_##x = CEIL_##prev + RANDOMX_FREQ_##x; \
 	for (; k < CEIL_##x; ++k) { JIT_HANDLE(x, prev); }
 
+#define INST_HANDLE2(x, func_name, prev) \
+	CEIL_##x = CEIL_##prev + RANDOMX_FREQ_##x; \
+	for (; k < CEIL_##x; ++k) { JIT_HANDLE(func_name, prev); }
+
 	INST_HANDLE(IADD_RS, NULL);
 	INST_HANDLE(IADD_M, IADD_RS);
 	INST_HANDLE(ISUB_R, IADD_M);
 	INST_HANDLE(ISUB_M, ISUB_R);
 	INST_HANDLE(IMUL_R, ISUB_M);
 	INST_HANDLE(IMUL_M, IMUL_R);
-	INST_HANDLE(IMULH_R, IMUL_M);
+
-	INST_HANDLE(IMULH_M, IMULH_R);
+#if defined(_M_X64) || defined(__x86_64__)
+	if (xmrig::Cpu::info()->hasBMI2()) {
+		INST_HANDLE2(IMULH_R, IMULH_R_BMI2, IMUL_M);
+		INST_HANDLE2(IMULH_M, IMULH_M_BMI2, IMULH_R);
+	}
+	else
+#endif
+	{
+		INST_HANDLE(IMULH_R, IMUL_M);
+		INST_HANDLE(IMULH_M, IMULH_R);
+	}
+
 	INST_HANDLE(ISMULH_R, IMULH_M);
 	INST_HANDLE(ISMULH_M, ISMULH_R);
 	INST_HANDLE(IMUL_RCP, ISMULH_M);

src/crypto/randomx/randomx.h

@@ -49,7 +49,7 @@ enum randomx_flags {
   RANDOMX_FLAG_FULL_MEM = 4,
   RANDOMX_FLAG_JIT = 8,
   RANDOMX_FLAG_1GB_PAGES = 16,
-  RANDOMX_FLAG_RYZEN = 64,
+  RANDOMX_FLAG_AMD = 64,
 };
 
 

src/crypto/rx/RxConfig.cpp

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by

src/crypto/rx/RxConfig.h

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by

src/crypto/rx/RxVm.cpp

@@ -7,8 +7,8 @@
  * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
  * Copyright 2018-2019 tevador     <tevador@gmail.com>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -45,8 +45,12 @@ xmrig::RxVm::RxVm(RxDataset *dataset, uint8_t *scratchpad, bool softAes, xmrig::
         m_flags |= RANDOMX_FLAG_JIT;
     }
 
-    if ((assembly == Assembly::RYZEN) || ((assembly == Assembly::AUTO) && (Cpu::info()->assembly() == Assembly::RYZEN))) {
+    if (assembly == Assembly::AUTO) {
-        m_flags |= RANDOMX_FLAG_RYZEN;
+        assembly = Cpu::info()->assembly();
+    }
+
+    if ((assembly == Assembly::RYZEN) || (assembly == Assembly::BULLDOZER)) {
+        m_flags |= RANDOMX_FLAG_AMD;
     }
 
     m_vm = randomx_create_vm(static_cast<randomx_flags>(m_flags), dataset->cache() ? dataset->cache()->get() : nullptr, dataset->get(), scratchpad);

src/crypto/rx/RxVm.h

@@ -7,8 +7,8 @@
  * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
  * Copyright 2018-2019 tevador     <tevador@gmail.com>
- * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by

src/crypto/rx/Rx_linux.cpp

@@ -7,10 +7,10 @@
  * Copyright 2017-2019 XMR-Stak                 <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett              <https://github.com/vtnerd>
  * Copyright 2018-2019 tevador                  <tevador@gmail.com>
- * Copyright 2018-2019 SChernykh                <https://github.com/SChernykh>
  * Copyright 2000      Transmeta Corporation    <https://github.com/intel/msr-tools>
  * Copyright 2004-2008 H. Peter Anvin           <https://github.com/intel/msr-tools>
- * Copyright 2016-2019 XMRig                    <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2020 SChernykh                <https://github.com/SChernykh>
+ * Copyright 2016-2020 XMRig                    <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -91,13 +91,19 @@ static MsrItem rdmsr(uint32_t reg)
 }
 
 
+static uint64_t get_masked_value(uint64_t old_value, uint64_t new_value, uint64_t mask)
+{
+    return (new_value & mask) | (old_value & ~mask);
+}
+
+
 static bool wrmsr_on_cpu(uint32_t reg, uint32_t cpu, uint64_t value, uint64_t mask)
 {
     // If a bit in mask is set to 1, use new value, otherwise use old value
     if (mask != MsrItem::kNoMask) {
         uint64_t old_value;
         if (rdmsr_on_cpu(reg, cpu, old_value)) {
-            value = (value & mask) | (old_value & ~mask);
+            value = get_masked_value(old_value, value, mask);
         }
     }
 
@@ -162,7 +168,7 @@ static bool wrmsr(const MsrItems &preset, bool save)
     if (save) {
         for (const auto &i : preset) {
             auto item = rdmsr(i.reg());
-            LOG_VERBOSE(CLEAR "%s" CYAN_BOLD("0x%08" PRIx32) CYAN(":0x%016" PRIx64) CYAN_BOLD(" -> 0x%016" PRIx64), tag, i.reg(), item.value(), i.value());
+            LOG_VERBOSE(CLEAR "%s" CYAN_BOLD("0x%08" PRIx32) CYAN(":0x%016" PRIx64) CYAN_BOLD(" -> 0x%016" PRIx64), tag, i.reg(), item.value(), get_masked_value(item.value(), i.value(), i.mask()));
 
             if (item.isValid()) {
                 savedState.emplace_back(item);

src/crypto/rx/Rx_win.cpp

@@ -10,8 +10,8 @@
  * Copyright 2000      Transmeta Corporation    <https://github.com/intel/msr-tools>
  * Copyright 2004-2008 H. Peter Anvin           <https://github.com/intel/msr-tools>
  * Copyright 2007-2009 hiyohiyo                 <https://openlibsys.org>, <hiyohiyo@crystalmark.info>
- * Copyright 2018-2019 SChernykh                <https://github.com/SChernykh>
+ * Copyright 2018-2020 SChernykh                <https://github.com/SChernykh>
- * Copyright 2016-2019 XMRig                    <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2020 XMRig                    <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -217,6 +217,12 @@ static MsrItem rdmsr(HANDLE driver, uint32_t reg)
 }
 
 
+static uint64_t get_masked_value(uint64_t old_value, uint64_t new_value, uint64_t mask)
+{
+    return (new_value & mask) | (old_value & ~mask);
+}
+
+
 static bool wrmsr(HANDLE driver, uint32_t reg, uint64_t value, uint64_t mask)
 {
     struct {
@@ -230,7 +236,7 @@ static bool wrmsr(HANDLE driver, uint32_t reg, uint64_t value, uint64_t mask)
     if (mask != MsrItem::kNoMask) {
         uint64_t old_value;
         if (rdmsr(driver, reg, old_value)) {
-            value = (value & mask) | (old_value & ~mask);
+            value = get_masked_value(old_value, value, mask);
         }
     }
 
@@ -268,7 +274,7 @@ static bool wrmsr(const MsrItems &preset, bool save)
     if (save) {
         for (const auto &i : preset) {
             auto item = rdmsr(driver, i.reg());
-            LOG_VERBOSE(CLEAR "%s" CYAN_BOLD("0x%08" PRIx32) CYAN(":0x%016" PRIx64) CYAN_BOLD(" -> 0x%016" PRIx64), tag, i.reg(), item.value(), i.value());
+            LOG_VERBOSE(CLEAR "%s" CYAN_BOLD("0x%08" PRIx32) CYAN(":0x%016" PRIx64) CYAN_BOLD(" -> 0x%016" PRIx64), tag, i.reg(), item.value(), get_masked_value(item.value(), i.value(), i.mask()));
 
             if (item.isValid()) {
                 savedState.emplace_back(item);

src/version.h

@@ -28,7 +28,7 @@
 #define APP_ID        "xmrig"
 #define APP_NAME      "XMRig"
 #define APP_DESC      "XMRig miner"
-#define APP_VERSION   "5.5.1"
+#define APP_VERSION   "5.5.2-dev"
 #define APP_DOMAIN    "xmrig.com"
 #define APP_SITE      "www.xmrig.com"
 #define APP_COPYRIGHT "Copyright (C) 2016-2020 xmrig.com"
@@ -36,7 +36,7 @@
 
 #define APP_VER_MAJOR  5
 #define APP_VER_MINOR  5
-#define APP_VER_PATCH  1
+#define APP_VER_PATCH  2
 
 #ifdef _MSC_VER
 #   if (_MSC_VER >= 1920)