Changed IADD_RS to use mod.mem

doc/program.asm

@@ -56,7 +56,7 @@ randomx_isn_14:
 	sqrtpd xmm6, xmm6
 randomx_isn_15:
 	; IADD_RS r6, r2, LSH 1
-	lea r14, [r14+r10*8]
+	lea r14, [r14+r10*2]
 randomx_isn_16:
 	; FSUB_M f2, L1[r1-1890725713]
 	lea eax, [r9d-1890725713]
@@ -145,7 +145,7 @@ randomx_isn_35:
 	imul r14, 835132161
 randomx_isn_36:
 	; IADD_RS r3, r4, LSH 2
-	lea r11, [r11+r12*2]
+	lea r11, [r11+r12*4]
 randomx_isn_37:
 	; ISUB_M r6, L2[r4+1885029796]
 	lea eax, [r12d+1885029796]
@@ -179,7 +179,7 @@ randomx_isn_45:
 	mov qword ptr [rsi+rax], r13
 randomx_isn_46:
 	; IADD_RS r0, r7, LSH 0
-	lea r8, [r8+r15*8]
+	lea r8, [r8+r15*1]
 randomx_isn_47:
 	; IXOR_R r5, r2
 	xor r13, r10
@@ -227,7 +227,7 @@ randomx_isn_57:
 	imul r13, r9
 randomx_isn_58:
 	; IADD_RS r5, r1, -999103579, LSH 0
-	lea r13, [r13+r9*8-999103579]
+	lea r13, [r13+r9*1-999103579]
 randomx_isn_59:
 	; FMUL_R e2, a2
 	mulpd xmm6, xmm10
@@ -237,7 +237,7 @@ randomx_isn_60:
 	ror r10, cl
 randomx_isn_61:
 	; IADD_RS r0, r3, LSH 1
-	lea r8, [r8+r11*1]
+	lea r8, [r8+r11*2]
 randomx_isn_62:
 	; FSQRT_R e3
 	sqrtpd xmm7, xmm7
@@ -260,7 +260,7 @@ randomx_isn_66:
 	sub r12, 841292629
 randomx_isn_67:
 	; IADD_RS r4, r6, LSH 2
-	lea r12, [r12+r14*1]
+	lea r12, [r12+r14*4]
 randomx_isn_68:
 	; FSUB_M f3, L1[r4+613549729]
 	lea eax, [r12d+613549729]
@@ -427,7 +427,7 @@ randomx_isn_107:
 	mov r14, rdx
 randomx_isn_108:
 	; IADD_RS r7, r0, LSH 1
-	lea r15, [r15+r8*4]
+	lea r15, [r15+r8*2]
 randomx_isn_109:
 	; IMUL_R r6, r5
 	imul r14, r13
@@ -442,13 +442,13 @@ randomx_isn_111:
 	addpd xmm2, xmm12
 randomx_isn_112:
 	; IADD_RS r0, r3, LSH 0
-	lea r8, [r8+r11*2]
+	lea r8, [r8+r11*1]
 randomx_isn_113:
 	; IADD_RS r3, r4, LSH 1
 	lea r11, [r11+r12*2]
 randomx_isn_114:
 	; IADD_RS r2, r4, LSH 2
-	lea r10, [r10+r12*8]
+	lea r10, [r10+r12*4]
 randomx_isn_115:
 	; IMUL_M r7, L1[r2-106928748]
 	lea eax, [r10d-106928748]
@@ -462,7 +462,7 @@ randomx_isn_117:
 	subpd xmm2, xmm10
 randomx_isn_118:
 	; IADD_RS r2, r2, LSH 0
-	lea r10, [r10+r10*2]
+	lea r10, [r10+r10*1]
 randomx_isn_119:
 	; ISUB_R r7, -342152774
 	sub r15, -342152774
@@ -471,7 +471,7 @@ randomx_isn_120:
 	lea r12, [r12+r9*2]
 randomx_isn_121:
 	; IADD_RS r4, r7, LSH 2
-	lea r12, [r12+r15*1]
+	lea r12, [r12+r15*4]
 randomx_isn_122:
 	; FSUB_R f0, a1
 	subpd xmm0, xmm9
@@ -502,7 +502,7 @@ randomx_isn_128:
 	subpd xmm3, xmm9
 randomx_isn_129:
 	; IADD_RS r1, r2, LSH 2
-	lea r9, [r9+r10*2]
+	lea r9, [r9+r10*4]
 randomx_isn_130:
 	; FSUB_R f1, a1
 	subpd xmm1, xmm9
@@ -529,7 +529,7 @@ randomx_isn_136:
 	sub r11, r14
 randomx_isn_137:
 	; IADD_RS r4, r1, LSH 0
-	lea r12, [r12+r9*8]
+	lea r12, [r12+r9*1]
 randomx_isn_138:
 	; ISTORE L1[r0+56684410], r0
 	lea eax, [r8d+56684410]
@@ -571,10 +571,10 @@ randomx_isn_145:
 	sub r13, r11
 randomx_isn_146:
 	; IADD_RS r0, r3, LSH 1
-	lea r8, [r8+r11*4]
+	lea r8, [r8+r11*2]
 randomx_isn_147:
 	; IADD_RS r1, r3, LSH 1
-	lea r9, [r9+r11*1]
+	lea r9, [r9+r11*2]
 randomx_isn_148:
 	; FSQRT_R e1
 	sqrtpd xmm5, xmm5
@@ -623,7 +623,7 @@ randomx_isn_158:
 	mov qword ptr [rsi+rax], r12
 randomx_isn_159:
 	; IADD_RS r7, r2, LSH 3
-	lea r15, [r15+r10*4]
+	lea r15, [r15+r10*8]
 randomx_isn_160:
 	; IMUL_RCP r7, 2040763167
 	mov rax, 9705702723791900149
@@ -715,7 +715,7 @@ randomx_isn_182:
 	mulpd xmm6, xmm10
 randomx_isn_183:
 	; IADD_RS r6, r2, LSH 0
-	lea r14, [r14+r10*8]
+	lea r14, [r14+r10*1]
 randomx_isn_184:
 	; FADD_R f2, a3
 	addpd xmm2, xmm11
@@ -727,7 +727,7 @@ randomx_isn_186:
 	xorps xmm3, xmm15
 randomx_isn_187:
 	; IADD_RS r6, r6, LSH 3
-	lea r14, [r14+r14*4]
+	lea r14, [r14+r14*8]
 randomx_isn_188:
 	; FSCAL_R f2
 	xorps xmm2, xmm15
@@ -780,7 +780,7 @@ randomx_isn_199:
 	subpd xmm3, xmm11
 randomx_isn_200:
 	; IADD_RS r2, r5, LSH 2
-	lea r10, [r10+r13*1]
+	lea r10, [r10+r13*4]
 randomx_isn_201:
 	; ISUB_M r6, L2[r3+376384700]
 	lea eax, [r11d+376384700]
@@ -810,7 +810,7 @@ randomx_isn_207:
 	xorps xmm1, xmm15
 randomx_isn_208:
 	; IADD_RS r6, r3, LSH 1
-	lea r14, [r14+r11*1]
+	lea r14, [r14+r11*2]
 randomx_isn_209:
 	; FSUB_M f0, L1[r4-557177119]
 	lea eax, [r12d-557177119]
@@ -873,7 +873,7 @@ randomx_isn_223:
 	xorps xmm2, xmm15
 randomx_isn_224:
 	; IADD_RS r5, r4, 312567979, LSH 1
-	lea r13, [r13+r12*4+312567979]
+	lea r13, [r13+r12*2+312567979]
 randomx_isn_225:
 	; ISTORE L2[r2+260885699], r1
 	lea eax, [r10d+260885699]
@@ -898,7 +898,7 @@ randomx_isn_229:
 	xchg r8, r14
 randomx_isn_230:
 	; IADD_RS r2, r7, LSH 2
-	lea r10, [r10+r15*1]
+	lea r10, [r10+r15*4]
 randomx_isn_231:
 	; FMUL_R e1, a0
 	mulpd xmm5, xmm8
@@ -924,7 +924,7 @@ randomx_isn_237:
 	subpd xmm1, xmm11
 randomx_isn_238:
 	; IADD_RS r4, r2, LSH 1
-	lea r12, [r12+r10*4]
+	lea r12, [r12+r10*2]
 randomx_isn_239:
 	; IMUL_RCP r7, 3065786637
 	mov rax, 12921343181238534701
@@ -978,13 +978,13 @@ randomx_isn_250:
 	addpd xmm3, xmm8
 randomx_isn_251:
 	; IADD_RS r0, r0, LSH 0
-	lea r8, [r8+r8*4]
+	lea r8, [r8+r8*1]
 randomx_isn_252:
 	; ISUB_R r4, r2
 	sub r12, r10
 randomx_isn_253:
 	; IADD_RS r5, r4, 256175395, LSH 0
-	lea r13, [r13+r12*4+256175395]
+	lea r13, [r13+r12*1+256175395]
 randomx_isn_254:
 	; IADD_RS r6, r7, LSH 2
 	lea r14, [r14+r15*4]

src/assembly_generator_x86.cpp

@@ -71,7 +71,7 @@ namespace randomx {
 				asmCode << "xor " << regR[instr.dst] << ", " << regR[instr.src] << std::endl;
 				break;
 			case SuperscalarInstructionType::IADD_RS:
-				asmCode << "lea " << regR[instr.dst] << ", [" << regR[instr.dst] << "+" << regR[instr.src] << "*" << (1 << (instr.getModShift2())) << "]" << std::endl;
+				asmCode << "lea " << regR[instr.dst] << ", [" << regR[instr.dst] << "+" << regR[instr.src] << "*" << (1 << (instr.getModMem())) << "]" << std::endl;
 				break;
 			case SuperscalarInstructionType::IMUL_R:
 				asmCode << "imul " << regR[instr.dst] << ", " << regR[instr.src] << std::endl;
@@ -180,7 +180,7 @@ namespace randomx {
 				asmCode << regR[instr.dst] << " ^= " << regR[instr.src] << ";" << std::endl;
 				break;
 			case SuperscalarInstructionType::IADD_RS:
-				asmCode << regR[instr.dst] << " += " << regR[instr.src] << "*" << (1 << (instr.getModShift2())) << ";" << std::endl;
+				asmCode << regR[instr.dst] << " += " << regR[instr.src] << "*" << (1 << (instr.getModMem())) << ";" << std::endl;
 				break;
 			case SuperscalarInstructionType::IMUL_R:
 				asmCode << regR[instr.dst] << " *= " << regR[instr.src] << ";" << std::endl;
@@ -275,9 +275,9 @@ namespace randomx {
 	void AssemblyGeneratorX86::h_IADD_RS(Instruction& instr, int i) {
 		registerUsage[instr.dst] = i;
 		if(instr.dst == RegisterNeedsDisplacement)
-			asmCode << "\tlea " << regR[instr.dst] << ", [" << regR[instr.dst] << "+" << regR[instr.src] << "*" << (1 << (instr.getModShift2())) << std::showpos << (int32_t)instr.getImm32() << std::noshowpos << "]" << std::endl;
+			asmCode << "\tlea " << regR[instr.dst] << ", [" << regR[instr.dst] << "+" << regR[instr.src] << "*" << (1 << (instr.getModMem())) << std::showpos << (int32_t)instr.getImm32() << std::noshowpos << "]" << std::endl;
 		else
-			asmCode << "\tlea " << regR[instr.dst] << ", [" << regR[instr.dst] << "+" << regR[instr.src] << "*" << (1 << (instr.getModShift2())) << "]" << std::endl;
+			asmCode << "\tlea " << regR[instr.dst] << ", [" << regR[instr.dst] << "+" << regR[instr.src] << "*" << (1 << (instr.getModMem())) << "]" << std::endl;
 		traceint(instr);
 	}
 
@@ -442,7 +442,6 @@ namespace randomx {
 	void AssemblyGeneratorX86::h_IMUL_RCP(Instruction& instr, int i) {
 		if (instr.getImm32() != 0) {
 			registerUsage[instr.dst] = i;
-			uint32_t divisor = instr.getImm32();
 			asmCode << "\tmov rax, " << randomx_reciprocal(instr.getImm32()) << std::endl;
 			asmCode << "\timul " << regR[instr.dst] << ", rax" << std::endl;
 			traceint(instr);
@@ -566,7 +565,7 @@ namespace randomx {
 	}
 
 	void AssemblyGeneratorX86::handleCondition(Instruction& instr, int i) {
-		const int shift = instr.getModShift3();
+		const int shift = instr.getModShift();
 		const int conditionMask = ((1 << RANDOMX_CONDITION_BITS) - 1) << shift;
 		int reg = getConditionRegister();
 		int target = registerUsage[reg] + 1;

src/configuration.h

@@ -31,7 +31,7 @@ along with RandomX.  If not, see<http://www.gnu.org/licenses/>.
 //Argon2d salt
 #define RANDOMX_ARGON_SALT         "RandomX\x03"
 
-//Number of random Cache accesses per Dataset block. Minimum is 2.
+//Number of random Cache accesses per Dataset item. Minimum is 2.
 #define RANDOMX_CACHE_ACCESSES     8
 
 #define RANDOMX_SUPERSCALAR_LATENCY   170

src/instruction.hpp

@@ -78,16 +78,13 @@ namespace randomx {
 			return os;
 		}
 		int getModMem() const {
-			return mod % 4;
+			return mod % 4; //bits 0-1
 		}
 		int getModCond() const {
-			return (mod >> 2) & 7;
+			return (mod >> 2) % 8; //bits 2-4
 		}
-		int getModShift3() const {
-			return mod >> 5;
-		}
-		int getModShift2() const {
-			return mod >> 6;
+		int getModShift() const {
+			return mod >> 5; //bits 5-7
 		}
 		void setMod(uint8_t val) {
 			mod = val;

src/jit_compiler_x86.cpp

@@ -357,7 +357,7 @@ namespace randomx {
 		case randomx::SuperscalarInstructionType::IADD_RS:
 			emit(REX_LEA);
 			emitByte(0x04 + 8 * instr.dst);
-			genSIB(instr.getModShift2(), instr.src, instr.dst);
+			genSIB(instr.getModMem(), instr.src, instr.dst);
 			break;
 		case randomx::SuperscalarInstructionType::IMUL_R:
 			emit(REX_IMUL_RR);
@@ -481,7 +481,7 @@ namespace randomx {
 			emitByte(0xac);
 		else
 			emitByte(0x04 + 8 * instr.dst);
-		genSIB(instr.getModShift2(), instr.src, instr.dst);
+		genSIB(instr.getModMem(), instr.src, instr.dst);
 		if (instr.dst == RegisterNeedsDisplacement)
 			emit32(instr.getImm32());
 	}
@@ -882,7 +882,7 @@ namespace randomx {
 	}
 
 	void JitCompilerX86::handleCondition(Instruction& instr, int i) {
-		const int shift = instr.getModShift3();
+		const int shift = instr.getModShift();
 		const int conditionMask = ((1 << RANDOMX_CONDITION_BITS) - 1) << shift;
 		int reg = getConditionRegister();
 		int target = registerUsage[reg] + 1;

src/superscalar.cpp

@@ -849,7 +849,7 @@ namespace randomx {
 				r[instr.dst] ^= r[instr.src];
 				break;
 			case randomx::SuperscalarInstructionType::IADD_RS:
-				r[instr.dst] += r[instr.src] << instr.getModShift2();
+				r[instr.dst] += r[instr.src] << instr.getModMem();
 				break;
 			case randomx::SuperscalarInstructionType::IMUL_R:
 				r[instr.dst] *= r[instr.src];

src/tests/benchmark.cpp

@@ -225,7 +225,7 @@ int main(int argc, char** argv) {
 		std::cout << "Calculated result: ";
 		result.print(std::cout);
 		if (noncesCount == 1000 && seedValue == 0)
-			std::cout << "Reference result:  918a8bc3ce0e537eec9d3c5e1a8bb3204ae3954f14c50c14810b38e49588a9e0" << std::endl;
+			std::cout << "Reference result:  89336a85bf6d1e83eb20fbc92170705ded9b42285b30178ed8e855d65c4c4b69" << std::endl;
 		if (!miningMode) {
 			std::cout << "Performance: " << 1000 * elapsed / noncesCount << " ms per hash" << std::endl;
 		}

src/vm_interpreted.cpp

@@ -434,12 +434,12 @@ namespace randomx {
 					ibc.idst = &r[dst];
 					if (dst != RegisterNeedsDisplacement) {
 						ibc.isrc = &r[src];
-						ibc.shift = instr.getModShift2();
+						ibc.shift = instr.getModMem();
 						ibc.imm = 0;
 					}
 					else {
 						ibc.isrc = &r[src];
-						ibc.shift = instr.getModShift2();
+						ibc.shift = instr.getModMem();
 						ibc.imm = signExtend2sCompl(instr.getImm32());
 					}
 					registerUsage[instr.dst] = i;
@@ -763,7 +763,7 @@ namespace randomx {
 					//jump condition
 					int reg = getConditionRegister(registerUsage);
 					ibc.target = registerUsage[reg];
-					ibc.shift = instr.getModShift3();
+					ibc.shift = instr.getModShift();
 					ibc.creg = &r[reg];
 					for (unsigned j = 0; j < 8; ++j) { //mark all registers as used
 						registerUsage[j] = i;