RandomXHFPI

scripts/generate_cl.js

@@ -60,6 +60,7 @@ function rx()
         'randomx_constants_wow.h',
         'randomx_constants_loki.h',
         'randomx_constants_arqma.h',
+        'randomx_constants_klaro.h',
         'aes.cl',
         'blake2b.cl',
         'randomx_vm.cl',

src/backend/opencl/cl/cn/algorithm.cl

@@ -25,6 +25,7 @@
 #define ALGO_AR2_CHUKWA     24
 #define ALGO_AR2_WRKZ       25
 
+
 #define FAMILY_UNKNOWN      0
 #define FAMILY_CN           1
 #define FAMILY_CN_LITE      2

src/backend/opencl/cl/rx/randomx_constants_klaro.h

@@ -0,0 +1,97 @@
+/*
+Copyright (c) 2019 SChernykh
+
+This file is part of RandomX OpenCL.
+
+RandomX OpenCL is free software: you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation, either version 3 of the License, or
+(at your option) any later version.
+
+RandomX OpenCL is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with RandomX OpenCL. If not, see <http://www.gnu.org/licenses/>.
+*/
+
+//Dataset base size in bytes. Must be a power of 2.
+#define RANDOMX_DATASET_BASE_SIZE  2147483648
+
+//Dataset extra size. Must be divisible by 64.
+#define RANDOMX_DATASET_EXTRA_SIZE 33554368
+
+//Scratchpad L3 size in bytes. Must be a power of 2.
+#define RANDOMX_SCRATCHPAD_L3      2097152
+
+//Scratchpad L2 size in bytes. Must be a power of two and less than or equal to RANDOMX_SCRATCHPAD_L3.
+#define RANDOMX_SCRATCHPAD_L2      262144
+
+//Scratchpad L1 size in bytes. Must be a power of two (minimum 64) and less than or equal to RANDOMX_SCRATCHPAD_L2.
+#define RANDOMX_SCRATCHPAD_L1      16384
+
+//Jump condition mask size in bits.
+#define RANDOMX_JUMP_BITS          8
+
+//Jump condition mask offset in bits. The sum of RANDOMX_JUMP_BITS and RANDOMX_JUMP_OFFSET must not exceed 16.
+#define RANDOMX_JUMP_OFFSET        8
+
+//Integer instructions
+#define RANDOMX_FREQ_IADD_RS       15
+#define RANDOMX_FREQ_IADD_M         2
+#define RANDOMX_FREQ_ISUB_R        16
+#define RANDOMX_FREQ_ISUB_M         2
+#define RANDOMX_FREQ_IMUL_R        16
+#define RANDOMX_FREQ_IMUL_M         4
+#define RANDOMX_FREQ_IMULH_R        4
+#define RANDOMX_FREQ_IMULH_M        1
+#define RANDOMX_FREQ_ISMULH_R       4
+#define RANDOMX_FREQ_ISMULH_M       1
+#define RANDOMX_FREQ_IMUL_RCP       8
+#define RANDOMX_FREQ_INEG_R         2
+#define RANDOMX_FREQ_IXOR_R         5
+#define RANDOMX_FREQ_IXOR_M         2
+#define RANDOMX_FREQ_IROR_R         2
+#define RANDOMX_FREQ_IROL_R         2
+#define RANDOMX_FREQ_ISWAP_R        4
+
+//Floating point instructions
+#define RANDOMX_FREQ_FSWAP_R        4
+#define RANDOMX_FREQ_FADD_R        20
+#define RANDOMX_FREQ_FADD_M        10
+#define RANDOMX_FREQ_FSUB_R        20
+#define RANDOMX_FREQ_FSUB_M        10
+#define RANDOMX_FREQ_FSCAL_R        6
+#define RANDOMX_FREQ_FMUL_R        20
+#define RANDOMX_FREQ_FMUL2I_R      20
+#define RANDOMX_FREQ_FDIV_M         7
+#define RANDOMX_FREQ_FSQRT_R       16
+
+//Control instructions
+#define RANDOMX_FREQ_CBRANCH       16
+#define RANDOMX_FREQ_CFROUND        1
+
+//Store instruction
+#define RANDOMX_FREQ_ISTORE        16
+
+//No-op instruction
+#define RANDOMX_FREQ_NOP            0
+
+#define RANDOMX_DATASET_ITEM_SIZE 64
+
+#define RANDOMX_PROGRAM_SIZE 256
+
+#define HASH_SIZE 32
+#define ENTROPY_SIZE (128 + RANDOMX_PROGRAM_SIZE * 8)
+#define REGISTERS_SIZE 256
+#define IMM_BUF_SIZE (RANDOMX_PROGRAM_SIZE * 4 - REGISTERS_SIZE)
+#define IMM_INDEX_COUNT ((IMM_BUF_SIZE / 4) - 2)
+#define VM_STATE_SIZE (REGISTERS_SIZE + IMM_BUF_SIZE + RANDOMX_PROGRAM_SIZE * 4)
+#define ROUNDING_MODE (RANDOMX_FREQ_CFROUND ? -1 : 0)
+
+// Scratchpad L1/L2/L3 bits
+#define LOC_L1 (32 - 14)
+#define LOC_L2 (32 - 18)
+#define LOC_L3 (32 - 21)

src/crypto/common/Algorithm.cpp

@@ -118,6 +118,8 @@ static AlgoName const algorithm_names[] = {
     { "RandomARQ",                 nullptr,            Algorithm::RX_ARQ          },
     { "randomx/sfx",               "rx/sfx",           Algorithm::RX_SFX          },
     { "RandomSFX",                 nullptr,            Algorithm::RX_SFX          },
+    { "randomx/klr",               "rx/klaro",         Algorithm::RX_KLARO        },
+    { "RandomXHFPI",                nullptr,           Algorithm::RX_KLARO        },
 #   endif
 #   ifdef XMRIG_ALGO_ARGON2
     { "argon2/chukwa",             nullptr,            Algorithm::AR2_CHUKWA      },

src/crypto/common/Algorithm.h

@@ -69,6 +69,7 @@ public:
         RX_LOKI,       // "rx/loki"          RandomXL (Loki).
         RX_ARQ,        // "rx/arq"           RandomARQ (Arqma).
         RX_SFX,        // "rx/sfx"           RandomSFX (Safex Cash).
+        RX_KLARO,      // "rx/klaro"         RandomHFPI (Klaro Electronic Money).
         AR2_CHUKWA,    // "argon2/chukwa"    Argon2id (Chukwa).
         AR2_WRKZ,      // "argon2/wrkz"      Argon2id (WRKZ)
         MAX

src/crypto/randomx/bytecode_machine.cpp

@@ -63,6 +63,7 @@ namespace randomx {
 			INSTR_CASE(FSUB_M)
 			INSTR_CASE(FSCAL_R)
 			INSTR_CASE(FMUL_R)
+			INSTR_CASE(FMUL2I_R)
 			INSTR_CASE(FDIV_M)
 			INSTR_CASE(FSQRT_R)
 			INSTR_CASE(CBRANCH)
@@ -414,6 +415,17 @@ namespace randomx {
 			return;
 		}
 
+		if (opcode < RandomX_CurrentConfig.CEIL_FMUL2I_R) {
+			auto dst = instr.dst % RegisterCountFlt;
+			int reg_a = (instr.getImm32() & 0xff) % RegistersCount;
+			int reg_b = (instr.getImm32()>>8 & 0xff) % RegistersCount;
+			ibc.type = InstructionType::FMUL2I_R;
+			ibc.fdst = &nreg->e[dst];
+			ibc.isrc = &nreg->r[reg_a];
+			ibc.isrc2 = &nreg->r[reg_b];
+			return;
+		}
+
 		if (opcode < RandomX_CurrentConfig.CEIL_FDIV_M) {
 			auto dst = instr.dst % RegisterCountFlt;
 			auto src = instr.src % RegistersCount;

src/crypto/randomx/bytecode_machine.hpp

@@ -52,6 +52,7 @@ namespace randomx {
 			const int_reg_t* isrc;
 			const rx_vec_f128* fsrc;
 		};
+        const int_reg_t* isrc2;
 		union {
 			uint64_t imm;
 			int64_t simm;
@@ -205,6 +206,14 @@ namespace randomx {
 			*ibc.fdst = rx_mul_vec_f128(*ibc.fdst, *ibc.fsrc);
 		}
 
+		static void exe_FMUL2I_R(RANDOMX_EXE_ARGS) {
+			 const rx_vec_f128 a = _mm_cvtepi32_pd(_mm_set_epi64x(0, *ibc.isrc)); 
+			*ibc.fdst = rx_mul_vec_f128(*ibc.fdst, a);
+			 const rx_vec_f128 b = _mm_cvtepi32_pd(_mm_set_epi64x(0, *ibc.isrc2));
+			*ibc.fdst = rx_mul_vec_f128(*ibc.fdst, b);
+			*ibc.fdst = rx_sqrt_vec_f128(*ibc.fdst);
+		}
+
 		static void exe_FDIV_M(RANDOMX_EXE_ARGS) {
 			rx_vec_f128 fsrc = maskRegisterExponentMantissa(
 				config,
@@ -277,6 +286,7 @@ namespace randomx {
 		void gen_FSUB_M(RANDOMX_GEN_ARGS);
 		void gen_FSCAL_R(RANDOMX_GEN_ARGS);
 		void gen_FMUL_R(RANDOMX_GEN_ARGS);
+		void gen_FMUL2I_R(RANDOMX_GEN_ARGS);
 		void gen_FDIV_M(RANDOMX_GEN_ARGS);
 		void gen_FSQRT_R(RANDOMX_GEN_ARGS);
 		void gen_CBRANCH(RANDOMX_GEN_ARGS);

src/crypto/randomx/instruction.hpp

@@ -61,12 +61,13 @@ namespace randomx {
 		FSUB_M = 21,
 		FSCAL_R = 22,
 		FMUL_R = 23,
-		FDIV_M = 24,
+		FMUL2I_R = 24,
-		FSQRT_R = 25,
+		FDIV_M = 25,
-		CBRANCH = 26,
+		FSQRT_R = 26,
-		CFROUND = 27,
+		CBRANCH = 27,
-		ISTORE = 28,
+		CFROUND = 28,
-		NOP = 29,
+		ISTORE = 29,
+		NOP = 30,
 	};
 
 	class Instruction {

src/crypto/randomx/jit_compiler_x86.cpp

@@ -1038,6 +1038,33 @@ namespace randomx {
 		codePos = pos;
 	}
 
+    static const uint8_t MOVQ_XMM12_REX[] = { 0x66, 0x4d, 0x0f, 0x6e };
+	static const uint8_t CVTDQ2PD_XMM12[] = { 0xf3, 0x45, 0x0f, 0xe6, 0xe4 };
+	static const uint8_t MULPD_XMM12[] = { 0x66, 0x41, 0x0f, 0x59 };
+    
+  	void JitCompilerX86::h_FMUL2I_R(const Instruction& instr) {
+        uint8_t* const p = code;
+		int pos = codePos;
+        
+		const uint32_t dst = instr.dst % RegisterCountFlt;
+		int reg_a = (instr.getImm32() & 0xff) % RegistersCount;
+		int reg_b = (instr.getImm32()>>8 & 0xff) % RegistersCount;
+		emit(MOVQ_XMM12_REX, p, pos);
+		emitByte(0xe0 + reg_a, p, pos);
+		emit(CVTDQ2PD_XMM12, p, pos);
+		emit(MULPD_XMM12, p, pos);
+		emitByte(0xc4 + 8 * (dst+4), p, pos);
+		emit(MOVQ_XMM12_REX, p, pos);
+		emitByte(0xe0 + reg_b, p, pos);
+		emit(CVTDQ2PD_XMM12, p, pos);
+		emit(MULPD_XMM12, p, pos);
+		emitByte(0xc4 + 8 * (dst+4), p, pos);
+		emit(SQRTPD, p, pos);
+		emitByte(0xe4 + 9 * dst, p, pos);
+        
+        codePos = pos;
+	}
+
 	void JitCompilerX86::h_FDIV_M(const Instruction& instr) {
 		uint8_t* const p = code;
 		int pos = codePos;

src/crypto/randomx/jit_compiler_x86.hpp

@@ -142,6 +142,7 @@ namespace randomx {
 		void h_FSUB_M(const Instruction&);
 		void h_FSCAL_R(const Instruction&);
 		void h_FMUL_R(const Instruction&);
+		void h_FMUL2I_R(const Instruction&);
 		void h_FDIV_M(const Instruction&);
 		void h_FSQRT_R(const Instruction&);
 		void h_CBRANCH(const Instruction&);

src/crypto/randomx/randomx.cpp

@@ -99,6 +99,27 @@ RandomX_ConfigurationSafex::RandomX_ConfigurationSafex()
 	ArgonSalt = "RandomSFX\x01";
 }
 
+RandomX_ConfigurationKlaro::RandomX_ConfigurationKlaro()
+{
+RANDOMX_FREQ_IADD_RS = 15;
+RANDOMX_FREQ_IADD_M = 2;
+RANDOMX_FREQ_ISUB_M = 2;
+RANDOMX_FREQ_IXOR_R = 5;
+RANDOMX_FREQ_IXOR_M = 2;
+RANDOMX_FREQ_IROR_R = 2;
+
+RANDOMX_FREQ_FADD_R = 20;
+RANDOMX_FREQ_FADD_M = 10;
+RANDOMX_FREQ_FSUB_R = 20;
+RANDOMX_FREQ_FSUB_M = 10;
+RANDOMX_FREQ_FMUL_R = 20;
+RANDOMX_FREQ_FMUL2I_R = 20;
+RANDOMX_FREQ_FDIV_M = 7;
+RANDOMX_FREQ_FSQRT_R = 16;
+RANDOMX_FREQ_CBRANCH = 16;
+
+}
+
 RandomX_ConfigurationBase::RandomX_ConfigurationBase()
 	: ArgonMemory(262144)
 	, ArgonIterations(3)
@@ -276,7 +297,8 @@ void RandomX_ConfigurationBase::Apply()
 	INST_HANDLE(FSUB_M, FSUB_R);
 	INST_HANDLE(FSCAL_R, FSUB_M);
 	INST_HANDLE(FMUL_R, FSCAL_R);
-	INST_HANDLE(FDIV_M, FMUL_R);
+	INST_HANDLE(FMUL2I_R, FMUL_R);
+	INST_HANDLE(FDIV_M, FMUL2I_R);
 	INST_HANDLE(FSQRT_R, FDIV_M);
 	INST_HANDLE(CBRANCH, FSQRT_R);
 	INST_HANDLE(CFROUND, CBRANCH);
@@ -290,6 +312,7 @@ RandomX_ConfigurationWownero RandomX_WowneroConfig;
 RandomX_ConfigurationLoki RandomX_LokiConfig;
 RandomX_ConfigurationArqma RandomX_ArqmaConfig;
 RandomX_ConfigurationSafex RandomX_SafexConfig;
+RandomX_ConfigurationKlaro RandomX_KlaroConfig;
 
 RandomX_ConfigurationBase RandomX_CurrentConfig;
 

src/crypto/randomx/randomx.h

@@ -109,6 +109,7 @@ struct RandomX_ConfigurationBase
 	uint32_t RANDOMX_FREQ_FSUB_M;
 	uint32_t RANDOMX_FREQ_FSCAL_R;
 	uint32_t RANDOMX_FREQ_FMUL_R;
+	uint32_t RANDOMX_FREQ_FMUL2I_R;
 	uint32_t RANDOMX_FREQ_FDIV_M;
 	uint32_t RANDOMX_FREQ_FSQRT_R;
 	uint32_t RANDOMX_FREQ_CBRANCH;
@@ -170,6 +171,7 @@ struct RandomX_ConfigurationBase
 	int CEIL_FSUB_M;
 	int CEIL_FSCAL_R;
 	int CEIL_FMUL_R;
+	int CEIL_FMUL2I_R;
 	int CEIL_FDIV_M;
 	int CEIL_FSQRT_R;
 	int CEIL_CBRANCH;
@@ -183,12 +185,14 @@ struct RandomX_ConfigurationWownero : public RandomX_ConfigurationBase { RandomX
 struct RandomX_ConfigurationLoki : public RandomX_ConfigurationBase { RandomX_ConfigurationLoki(); };
 struct RandomX_ConfigurationArqma : public RandomX_ConfigurationBase { RandomX_ConfigurationArqma(); };
 struct RandomX_ConfigurationSafex : public RandomX_ConfigurationBase { RandomX_ConfigurationSafex(); };
+struct RandomX_ConfigurationKlaro : public RandomX_ConfigurationBase { RandomX_ConfigurationKlaro(); };
 
 extern RandomX_ConfigurationMonero RandomX_MoneroConfig;
 extern RandomX_ConfigurationWownero RandomX_WowneroConfig;
 extern RandomX_ConfigurationLoki RandomX_LokiConfig;
 extern RandomX_ConfigurationArqma RandomX_ArqmaConfig;
 extern RandomX_ConfigurationSafex RandomX_SafexConfig;
+extern RandomX_ConfigurationKlaro RandomX_KlaroConfig;
 
 extern RandomX_ConfigurationBase RandomX_CurrentConfig;
 

src/crypto/rx/RxAlgo.cpp

@@ -52,6 +52,9 @@ const RandomX_ConfigurationBase *xmrig::RxAlgo::base(Algorithm::Id algorithm)
     case Algorithm::RX_SFX:
         return &RandomX_SafexConfig;
 
+    case Algorithm::RX_KLARO:
+        return &RandomX_KlaroConfig;
+
     default:
         break;
     }