NOP instruction

register load/store from L3

src/AssemblyGeneratorX86.cpp

@@ -491,6 +491,10 @@ namespace RandomX {
 		asmCode << "\tmovapd xmmword ptr [rsi+rax], " << regFE[instr.src] << std::endl;
 	}
 
+	void AssemblyGeneratorX86::h_NOP(Instruction& instr, int i) {
+		asmCode << "\tnop" << std::endl;
+	}
+
 #include "instructionWeights.hpp"
 #define INST_HANDLE(x) REPN(&AssemblyGeneratorX86::h_##x, WT(x))
 
@@ -540,5 +544,7 @@ namespace RandomX {
 
 		INST_HANDLE(ISTORE)
 		INST_HANDLE(FSTORE)
+
+		INST_HANDLE(NOP)
 	};
 }

src/AssemblyGeneratorX86.hpp

@@ -79,5 +79,6 @@ namespace RandomX {
 		void  h_CFROUND(Instruction&, int);
 		void  h_ISTORE(Instruction&, int);
 		void  h_FSTORE(Instruction&, int);
+		void  h_NOP(Instruction&, int);
 	};
 }

src/Instruction.cpp

@@ -327,6 +327,10 @@ namespace RandomX {
 		os << ", " << reg << srcIndex << std::endl;
 	}
 
+	void  Instruction::h_NOP(std::ostream& os) const {
+		os << std::endl;
+	}
+
 #include "instructionWeights.hpp"
 #define INST_NAME(x) REPN(#x, WT(x))
 #define INST_HANDLE(x) REPN(&Instruction::h_##x, WT(x))
@@ -377,6 +381,8 @@ namespace RandomX {
 
 		INST_NAME(ISTORE)
 		INST_NAME(FSTORE)
+
+		INST_NAME(NOP)
 	};
 
 	InstructionVisualizer Instruction::engine[256] = {
@@ -425,6 +431,8 @@ namespace RandomX {
 
 		INST_HANDLE(ISTORE)
 		INST_HANDLE(FSTORE)
+
+		INST_HANDLE(NOP)
 	};
 
 }

src/Instruction.hpp

@@ -86,6 +86,7 @@ namespace RandomX {
 		void  h_CFROUND(std::ostream&) const;
 		void  h_ISTORE(std::ostream&) const;
 		void  h_FSTORE(std::ostream&) const;
+		void  h_NOP(std::ostream&) const;
 	};
 
 	static_assert(sizeof(Instruction) == 8, "Invalid alignment of struct Instruction");

src/JitCompilerX86.cpp

@@ -181,7 +181,7 @@ namespace RandomX {
 	static const uint8_t JMP = 0xe9;
 
 	size_t JitCompilerX86::getCodeSize() {
-		return codePos - prologueSize + readDatasetSize;
+		return codePos - prologueSize;
 	}
 
 	JitCompilerX86::JitCompilerX86() {
@@ -761,6 +761,10 @@ namespace RandomX {
 		emitByte(0x06);
 	}
 
+	void JitCompilerX86::h_NOP(Instruction& instr) {
+		emitByte(0x90);
+	}
+
 #include "instructionWeights.hpp"
 #define INST_HANDLE(x) REPN(&JitCompilerX86::h_##x, WT(x))
 
@@ -800,6 +804,7 @@ namespace RandomX {
 		INST_HANDLE(CFROUND)
 		INST_HANDLE(ISTORE)
 		INST_HANDLE(FSTORE)
+		INST_HANDLE(NOP)
 	};
 
 

src/JitCompilerX86.hpp

@@ -125,6 +125,7 @@ namespace RandomX {
 		void  h_CFROUND(Instruction&);
 		void  h_ISTORE(Instruction&);
 		void  h_FSTORE(Instruction&);
+		void  h_NOP(Instruction&);
 	};
 
 }

src/asm/program_load_flt.inc

@@ -1,4 +1,4 @@
-	and eax, 262080
+	and eax, 1048512
 	lea rcx, [rsi+rax]
 	cvtdq2pd xmm0, qword ptr [rcx+0]
 	cvtdq2pd xmm1, qword ptr [rcx+8]

src/asm/program_load_int.inc

@@ -1,4 +1,4 @@
-	and eax, 262080
+	and eax, 1048512
 	lea rcx, [rsi+rax]
 	xor r8,  qword ptr [rcx+0]
 	xor r9,  qword ptr [rcx+8]

src/asm/program_store_flt.inc

@@ -1,4 +1,4 @@
-	and eax, 262080
+	and eax, 1048512
 	lea rcx, [rsi+rax]
 	mulpd xmm0, xmm4
 	mulpd xmm1, xmm5

src/asm/program_store_int.inc

@@ -1,4 +1,4 @@
-	and eax, 262080
+	and eax, 1048512
 	lea rcx, [rsi+rax]
 	mov qword ptr [rcx+0], r8
 	mov qword ptr [rcx+8], r9

src/common.hpp

@@ -72,7 +72,7 @@ namespace RandomX {
 		convertible_t hi;
 	};
 
-	constexpr int ProgramLength = 256;
+	constexpr int ProgramLength = 128;
 	constexpr uint32_t InstructionCount = 1024;
 	constexpr uint32_t ScratchpadSize = 1024 * 1024;
 	constexpr uint32_t ScratchpadLength = ScratchpadSize / sizeof(convertible_t);

src/executeProgram-win64.asm

@@ -119,7 +119,7 @@ signMask:
 ALIGN 64
 program_begin:
 	xor eax, r8d                      ;# read address register 1
-	and eax, 262080
+	and eax, 1048512
 	lea rcx, [rsi+rax]
 	xor r8,  qword ptr [rcx+0]
 	xor r9,  qword ptr [rcx+8]
@@ -130,7 +130,7 @@ program_begin:
 	xor r14, qword ptr [rcx+48]
 	xor r15, qword ptr [rcx+56]
 	xor eax, r9d                      ;# read address register 2
-	and eax, 262080
+	and eax, 1048512
 	lea rcx, [rsi+rax]
 	cvtdq2pd xmm0, qword ptr [rcx+0]
 	cvtdq2pd xmm1, qword ptr [rcx+8]
@@ -166,7 +166,7 @@ program_begin:
 	xor r14, qword ptr [rcx+48]
 	xor r15, qword ptr [rcx+56]                 
 	mov eax, r12d                      ;# write address register 1
-	and eax, 262080
+	and eax, 1048512
 	lea rcx, [rsi+rax]
 	mov qword ptr [rcx+0], r8
 	mov qword ptr [rcx+8], r9
@@ -177,7 +177,7 @@ program_begin:
 	mov qword ptr [rcx+48], r14
 	mov qword ptr [rcx+56], r15
 	xor eax, r13d                      ;# write address register 2
-	and eax, 262080
+	and eax, 1048512
 	lea rcx, [rsi+rax]
 	mulpd xmm0, xmm4
 	mulpd xmm1, xmm5

src/instructionWeights.hpp

@@ -20,51 +20,51 @@ along with RandomX.  If not, see<http://www.gnu.org/licenses/>.
 #pragma once
 
 //Integer
-#define WT_IADD_R 10
+#define WT_IADD_R 12
 #define WT_IADD_M 3
-#define WT_IADD_RC 10
-#define WT_ISUB_R 10
+#define WT_IADD_RC 12
+#define WT_ISUB_R 12
 #define WT_ISUB_M 3
 #define WT_IMUL_9C 10
-#define WT_IMUL_R 20
-#define WT_IMUL_M 6
-#define WT_IMULH_R 6
-#define WT_IMULH_M 2
-#define WT_ISMULH_R 6
-#define WT_ISMULH_M 2
+#define WT_IMUL_R 16
+#define WT_IMUL_M 4
+#define WT_IMULH_R 4
+#define WT_IMULH_M 1
+#define WT_ISMULH_R 4
+#define WT_ISMULH_M 1
 #define WT_IDIV_C 4
 #define WT_ISDIV_C 4
 #define WT_INEG_R 2
 #define WT_IXOR_R 12
-#define WT_IXOR_M 4
-#define WT_IROR_R 10
-#define WT_IROL_R 10
+#define WT_IXOR_M 3
+#define WT_IROR_R 12
+#define WT_IROL_R 12
 
 //Common floating point
-#define WT_FPSWAP_R 6
+#define WT_FPSWAP_R 8
 
 //Floating point group F
-#define WT_FPADD_R 18
-#define WT_FPADD_M 3
-#define WT_FPSUB_R 18
-#define WT_FPSUB_M 3
-#define WT_FPNEG_R 5
+#define WT_FPADD_R 20
+#define WT_FPADD_M 5
+#define WT_FPSUB_R 20
+#define WT_FPSUB_M 5
+#define WT_FPNEG_R 6
 
 //Floating point group E
-#define WT_FPMUL_R 18
-#define WT_FPMUL_M 3
-#define WT_FPDIV_R 6
+#define WT_FPMUL_R 16
+#define WT_FPMUL_M 4
+#define WT_FPDIV_R 7
 #define WT_FPDIV_M 1
 #define WT_FPSQRT_R 6
 
 //Control
-#define WT_COND_R 12
-#define WT_COND_M 4
+#define WT_COND_R 7
+#define WT_COND_M 1
 #define WT_CFROUND 1
 
 //Store
-#define WT_ISTORE 12
-#define WT_FSTORE 6
+#define WT_ISTORE 18
+#define WT_FSTORE 0
 
 #define WT_NOP 0
 
@@ -115,6 +115,7 @@ static_assert(wtSum == 256,
 #define REP33(x) REP32(x) x,
 #define REP40(x) REP32(x) REP8(x)
 #define REP128(x) REP32(x) REP32(x) REP32(x) REP32(x)
+#define REP232(x) REP128(x) REP40(x) REP40(x) REP24(x)
 #define REP256(x) REP128(x) REP128(x)
 #define REPNX(x,N) REP##N(x)
 #define REPN(x,N) REPNX(x,N)

src/main.cpp

@@ -169,12 +169,10 @@ void mine(RandomX::VirtualMachine* vm, std::atomic<int>& atomicNonce, AtomicHash
 		blake2b(hash, sizeof(hash), blockTemplate, sizeof(blockTemplate), nullptr, 0);
 		int spIndex = ((uint8_t*)hash)[24] | ((((uint8_t*)hash)[25] & 15) << 8);
 		vm->initializeScratchpad(scratchpad, spIndex);
-		//vm->initializeProgram(hash);
+		vm->setScratchpad(scratchpad);
 		//dump((char*)((RandomX::CompiledVirtualMachine*)vm)->getProgram(), RandomX::CodeSize, "code-1337-jmp.txt");
 		for (int chain = 0; chain < 16; ++chain) {
 			vm->initializeProgram(hash);
-			int segment = hash[3] & 3;
-			vm->setScratchpad(scratchpad + segment * RandomX::ScratchpadSize / 4);
 			vm->execute();
 			vm->getResult(nullptr, 0, hash);
 		}

src/program.inc

@@ -10,54 +10,54 @@
 	mulpd xmm6, xmm10
 	; IMUL_R r6, r3
 	imul r14, r11
-	; FPMUL_R e1, a0
-	mulpd xmm5, xmm8
-	; IROR_R r5, r3
+	; FPSUB_M f1, L1[r4]
+	mov eax, r12d
+	and eax, 16376
+	cvtdq2pd xmm12, qword ptr [rsi+rax]
+	subpd xmm1, xmm12
+	; IROL_R r5, r3
 	mov ecx, r11d
-	ror r13, cl
+	rol r13, cl
 	; FPMUL_R e2, a0
 	mulpd xmm6, xmm8
-	; FPNEG_R f3
-	xorps xmm3, xmm15
+	; FPSUB_R f3, a0
+	subpd xmm3, xmm8
 	; IXOR_R r0, r4
 	xor r8, r12
-	; ISMULH_R r3, r7
+	; ISMULH_M r3, L1[r7]
+	mov ecx, r15d
+	and ecx, 16376
 	mov rax, r11
-	imul r15
+	imul qword ptr [rsi+rcx]
 	mov r11, rdx
 	; FPSWAP_R f2
 	shufpd xmm2, xmm2, 1
-	; ISMULH_R r6, r0
-	mov rax, r14
-	imul r8
-	mov r14, rdx
+	; IDIV_C r6, 1248528248
+	mov rax, 15864311168205210203
+	mul r14
+	shr rdx, 30
+	add r14, rdx
 	; FPMUL_R e0, a2
 	mulpd xmm4, xmm10
-	; ISUB_R r3, r4
-	sub r11, r12
+	; IADD_RC r3, r4, -52260428
+	lea r11, [r11+r12-52260428]
 	; IADD_R r7, -1138617760
 	add r15, -1138617760
-	; IROR_R r2, r6
+	; IROL_R r2, r6
 	mov ecx, r14d
-	ror r10, cl
-	; FPMUL_R e2, a1
-	mulpd xmm6, xmm9
+	rol r10, cl
+	; FPNEG_R f2
+	xorps xmm2, xmm15
 	; IROR_R r7, r1
 	mov ecx, r9d
 	ror r15, cl
-	; COND_M r2, lt(L1[r7], -41618808)
+	; COND_R r2, lt(r7, -41618808)
 	xor ecx, ecx
-	mov eax, r15d
-	and eax, 16376
-	cmp dword ptr [rsi+rax], -41618808
+	cmp r15d, -41618808
 	setl cl
 	add r10, rcx
-	; FPMUL_M e3, L1[r0]
-	mov eax, r8d
-	and eax, 16376
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	mulpd xmm7, xmm12
-	maxpd xmm7, xmm13
+	; FPMUL_R e3, a0
+	mulpd xmm7, xmm8
 	; CFROUND r1, 43
 	mov rax, r9
 	rol rax, 34
@@ -67,14 +67,17 @@
 	ldmxcsr dword ptr [rsp-8]
 	; FPADD_R f2, a1
 	addpd xmm2, xmm9
-	; FPNEG_R f0
-	xorps xmm0, xmm15
-	; FSTORE L1[r6], f2
+	; FPSUB_M f0, L1[r7]
+	mov eax, r15d
+	and eax, 16376
+	cvtdq2pd xmm12, qword ptr [rsi+rax]
+	subpd xmm0, xmm12
+	; ISTORE L1[r6], r2
 	mov eax, r14d
-	and eax, 16368
-	movapd xmmword ptr [rsi+rax], xmm2
-	; IMUL_9C r6, -45112665
-	lea r14, [r14+r14*8-45112665]
+	and eax, 16376
+	mov qword ptr [rsi+rax], r10
+	; ISUB_R r6, r5
+	sub r14, r13
 	; IADD_M r0, L1[r4]
 	mov eax, r12d
 	and eax, 16376
@@ -87,41 +90,30 @@
 	mov eax, r14d
 	and eax, 16376
 	mov qword ptr [rsi+rax], r14
-	; COND_R r4, sg(r1, -1189096105)
-	xor ecx, ecx
-	cmp r9d, -1189096105
-	sets cl
-	add r12, rcx
+	; FPSQRT_R e0
+	sqrtpd xmm4, xmm4
 	; IXOR_R r2, r5
 	xor r10, r13
-	; COND_R r1, be(r5, -965180434)
-	xor ecx, ecx
-	cmp r13d, -965180434
-	setbe cl
-	add r9, rcx
-	; FPMUL_M e1, L2[r3]
-	mov eax, r11d
-	and eax, 262136
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	mulpd xmm5, xmm12
-	maxpd xmm5, xmm13
+	; FPSQRT_R e1
+	sqrtpd xmm5, xmm5
+	; FPMUL_R e1, a3
+	mulpd xmm5, xmm11
 	; IMULH_R r7, r6
 	mov rax, r15
 	mul r14
 	mov r15, rdx
-	; ISMULH_M r0, L1[r4]
-	mov ecx, r12d
-	and ecx, 16376
-	mov rax, r8
-	imul qword ptr [rsi+rcx]
-	mov r8, rdx
+	; ISDIV_C r0, -1706892622
+	mov rax, -5802075764249827661
+	imul r8
+	xor eax, eax
+	sar rdx, 29
+	sets al
+	add rdx, rax
+	add r8, rdx
 	; IMUL_R r5, r3
 	imul r13, r11
-	; COND_R r2, of(r0, -1045938770)
-	xor ecx, ecx
-	cmp r8d, -1045938770
-	seto cl
-	add r10, rcx
+	; FPSQRT_R e2
+	sqrtpd xmm6, xmm6
 	; FPADD_M f3, L1[r4]
 	mov eax, r12d
 	and eax, 16376
@@ -131,18 +123,19 @@
 	add r11, r10
 	; FPADD_R f1, a0
 	addpd xmm1, xmm8
-	; FPSQRT_R e3
-	sqrtpd xmm7, xmm7
+	; FPDIV_R e3, a2
+	divpd xmm7, xmm10
+	maxpd xmm7, xmm13
 	; FPSUB_R f0, a1
 	subpd xmm0, xmm9
 	; IMUL_M r5, L1[r6]
 	mov eax, r14d
 	and eax, 16376
 	imul r13, qword ptr [rsi+rax]
-	; ISUB_R r1, r2
-	sub r9, r10
-	; IMUL_R r4, r6
-	imul r12, r14
+	; IADD_RC r1, r2, -1263285243
+	lea r9, [r9+r10-1263285243]
+	; IMUL_9C r4, 1994773931
+	lea r12, [r12+r12*8+1994773931]
 	; FPSWAP_R e3
 	shufpd xmm7, xmm7, 1
 	; IMUL_M r0, L1[r7]
@@ -152,69 +145,72 @@
 	; IROR_R r1, r6
 	mov ecx, r14d
 	ror r9, cl
-	; IROR_R r2, r4
+	; IROL_R r2, r4
 	mov ecx, r12d
-	ror r10, cl
+	rol r10, cl
 	; FPSUB_R f3, a1
 	subpd xmm3, xmm9
-	; FSTORE L1[r0], e1
+	; ISTORE L1[r0], r5
 	mov eax, r8d
-	and eax, 16368
-	movapd xmmword ptr [rsi+rax], xmm5
-	; COND_R r2, sg(r3, 1269153133)
-	xor ecx, ecx
-	cmp r11d, 1269153133
-	sets cl
-	add r10, rcx
+	and eax, 16376
+	mov qword ptr [rsi+rax], r13
+	; FPDIV_M e2, L2[r3]
+	mov eax, r11d
+	and eax, 262136
+	cvtdq2pd xmm12, qword ptr [rsi+rax]
+	divpd xmm6, xmm12
+	maxpd xmm6, xmm13
 	; FPSWAP_R f2
 	shufpd xmm2, xmm2, 1
 	; IADD_R r7, r5
 	add r15, r13
-	; COND_R r0, be(r4, -1486502150)
-	xor ecx, ecx
-	cmp r12d, -1486502150
-	setbe cl
-	add r8, rcx
-	; FPSUB_R f3, a1
-	subpd xmm3, xmm9
+	; FPDIV_M e0, L1[r4]
+	mov eax, r12d
+	and eax, 16376
+	cvtdq2pd xmm12, qword ptr [rsi+rax]
+	divpd xmm4, xmm12
+	maxpd xmm4, xmm13
+	; FPADD_M f3, L1[r5]
+	mov eax, r13d
+	and eax, 16376
+	cvtdq2pd xmm12, qword ptr [rsi+rax]
+	addpd xmm3, xmm12
 	; FPADD_R f0, a3
 	addpd xmm0, xmm11
 	; IADD_R r2, r0
 	add r10, r8
-	; FSTORE L1[r3], e2
+	; ISTORE L1[r3], r6
 	mov eax, r11d
-	and eax, 16368
-	movapd xmmword ptr [rsi+rax], xmm6
-	; IXOR_R r1, r7
-	xor r9, r15
-	; IMUL_R r5, r7
-	imul r13, r15
+	and eax, 16376
+	mov qword ptr [rsi+rax], r14
+	; IROR_R r1, r7
+	mov ecx, r15d
+	ror r9, cl
+	; IMUL_9C r5, 301671287
+	lea r13, [r13+r13*8+301671287]
 	; IXOR_R r7, 266992378
 	xor r15, 266992378
-	; COND_R r7, no(r4, 1983804692)
-	xor ecx, ecx
-	cmp r12d, 1983804692
-	setno cl
-	add r15, rcx
+	; FPSQRT_R e3
+	sqrtpd xmm7, xmm7
 	; IMUL_M r2, L2[r0]
 	mov eax, r8d
 	and eax, 262136
 	imul r10, qword ptr [rsi+rax]
-	; FPDIV_R e3, a2
-	divpd xmm7, xmm10
-	maxpd xmm7, xmm13
-	; IMUL_M r0, L2[r6]
-	mov eax, r14d
-	and eax, 262136
-	imul r8, qword ptr [rsi+rax]
+	; FPMUL_R e3, a2
+	mulpd xmm7, xmm10
+	; IMUL_R r0, r6
+	imul r8, r14
 	; ISTORE L1[r0], r7
 	mov eax, r8d
 	and eax, 16376
 	mov qword ptr [rsi+rax], r15
-	; FPMUL_R e0, a1
-	mulpd xmm4, xmm9
-	; FPSUB_R f3, a1
-	subpd xmm3, xmm9
+	; FPNEG_R f0
+	xorps xmm0, xmm15
+	; FPADD_M f3, L1[r5]
+	mov eax, r13d
+	and eax, 16376
+	cvtdq2pd xmm12, qword ptr [rsi+rax]
+	addpd xmm3, xmm12
 	; IROR_R r5, r4
 	mov ecx, r12d
 	ror r13, cl
@@ -222,17 +218,20 @@
 	mov eax, r15d
 	and eax, 262136
 	mov qword ptr [rsi+rax], r10
-	; FPSWAP_R e2
-	shufpd xmm6, xmm6, 1
+	; FPADD_R f2, a3
+	addpd xmm2, xmm11
 	; FPADD_M f3, L1[r2]
 	mov eax, r10d
 	and eax, 16376
 	cvtdq2pd xmm12, qword ptr [rsi+rax]
 	addpd xmm3, xmm12
-	; IDIV_C r5, 2218798981
-	mov rax, 17853839665672790751
-	mul r13
-	shr rdx, 31
+	; ISDIV_C r5, -2076168315
+	mov rax, -4770095103914078469
+	imul r13
+	xor eax, eax
+	sar rdx, 29
+	sets al
+	add rdx, rax
 	add r13, rdx
 	; IADD_RC r0, r4, -1321374359
 	lea r8, [r8+r12-1321374359]
@@ -250,28 +249,26 @@
 	rol r15, cl
 	; ISUB_R r2, r4
 	sub r10, r12
-	; IMULH_M r0, L1[12400]
-	mov rax, r8
-	mul qword ptr [rsi+12400]
-	mov r8, rdx
+	; ISMULH_R r0, -1500893068
+	mov rax, -1500893068
+	imul r8
+	add r8, rdx
 	; IADD_R r2, r3
 	add r10, r11
-	; COND_R r6, lt(r1, -1124202227)
-	xor ecx, ecx
-	cmp r9d, -1124202227
-	setl cl
-	add r14, rcx
-	; IROR_R r7, r4
+	; FPSQRT_R e2
+	sqrtpd xmm6, xmm6
+	; IROL_R r7, r4
 	mov ecx, r12d
-	ror r15, cl
+	rol r15, cl
 	; IMUL_R r4, r2
 	imul r12, r10
 	; ISUB_R r3, r7
 	sub r11, r15
 	; IADD_R r2, r7
 	add r10, r15
-	; FPSQRT_R e3
-	sqrtpd xmm7, xmm7
+	; FPDIV_R e3, a0
+	divpd xmm7, xmm8
+	maxpd xmm7, xmm13
 	; ISUB_R r6, 540663146
 	sub r14, 540663146
 	; IROL_R r5, 58
@@ -280,67 +277,65 @@
 	addpd xmm2, xmm9
 	; FPADD_R f2, a2
 	addpd xmm2, xmm10
-	; FPSQRT_R e1
-	sqrtpd xmm5, xmm5
+	; FPDIV_R e1, a2
+	divpd xmm5, xmm10
+	maxpd xmm5, xmm13
 	; FPADD_R f1, a2
 	addpd xmm1, xmm10
 	; IADD_R r5, r3
 	add r13, r11
-	; IADD_M r7, L1[880]
-	add r15, qword ptr [rsi+880]
+	; IADD_R r7, -1780268176
+	add r15, -1780268176
 	; ISUB_R r7, r0
 	sub r15, r8
 	; ISTORE L2[r0], r7
 	mov eax, r8d
 	and eax, 262136
 	mov qword ptr [rsi+rax], r15
-	; IDIV_C r2, 1014940364
-	mov rax, r10
-	shr rax, 2
-	mov rcx, 1219717022984988185
-	mul rcx
-	shr rdx, 24
-	add r10, rdx
-	; FPMUL_R e0, a2
-	mulpd xmm4, xmm10
-	; IDIV_C r2, 3059159304
-	mov rax, 12949335853590502915
-	mul r10
-	shr rdx, 31
-	add r10, rdx
+	; INEG_R r2
+	neg r10
+	; FPNEG_R f0
+	xorps xmm0, xmm15
+	; INEG_R r2
+	neg r10
 	; IADD_R r0, r3
 	add r8, r11
 	; IMUL_9C r7, -2124093035
 	lea r15, [r15+r15*8-2124093035]
-	; FPSUB_R f2, a0
-	subpd xmm2, xmm8
-	; FPDIV_R e0, a2
-	divpd xmm4, xmm10
+	; FPADD_M f2, L1[r0]
+	mov eax, r8d
+	and eax, 16376
+	cvtdq2pd xmm12, qword ptr [rsi+rax]
+	addpd xmm2, xmm12
+	; FPMUL_M e0, L1[r6]
+	mov eax, r14d
+	and eax, 16376
+	cvtdq2pd xmm12, qword ptr [rsi+rax]
+	mulpd xmm4, xmm12
 	maxpd xmm4, xmm13
 	; FPSUB_R f2, a3
 	subpd xmm2, xmm11
 	; IMUL_R r1, r2
 	imul r9, r10
-	; ISMULH_R r7, r5
-	mov rax, r15
-	imul r13
-	mov r15, rdx
+	; IDIV_C r7, 3214009572
+	mov rax, 12325439725582798855
+	mul r15
+	shr rdx, 31
+	add r15, rdx
 	; IMULH_R r3, r2
 	mov rax, r11
 	mul r10
 	mov r11, rdx
-	; IXOR_M r1, L2[r0]
-	mov eax, r8d
-	and eax, 262136
-	xor r9, qword ptr [rsi+rax]
+	; IROR_R r1, r0
+	mov ecx, r8d
+	ror r9, cl
 	; FPMUL_R e0, a1
 	mulpd xmm4, xmm9
-	; ISUB_R r4, 1456841848
-	sub r12, 1456841848
-	; IXOR_M r3, L2[r2]
-	mov eax, r10d
-	and eax, 262136
-	xor r11, qword ptr [rsi+rax]
+	; IADD_RC r4, r4, 1456841848
+	lea r12, [r12+r12+1456841848]
+	; IROR_R r3, r2
+	mov ecx, r10d
+	ror r11, cl
 	; COND_M r0, of(L1[r4], 1678513610)
 	xor ecx, ecx
 	mov eax, r12d
@@ -348,446 +343,39 @@
 	cmp dword ptr [rsi+rax], 1678513610
 	seto cl
 	add r8, rcx
-	; IDIV_C r4, 2674394209
-	mov rax, 925772300223658071
-	mul r12
-	shr rdx, 27
-	add r12, rdx
+	; INEG_R r4
+	neg r12
 	; IMUL_R r4, r1
 	imul r12, r9
 	; FPADD_R f1, a2
 	addpd xmm1, xmm10
 	; FPSUB_R f2, a0
 	subpd xmm2, xmm8
-	; FPMUL_M e1, L2[r6]
-	mov eax, r14d
-	and eax, 262136
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	mulpd xmm5, xmm12
-	maxpd xmm5, xmm13
-	; FPSUB_M f0, L2[r3]
-	mov eax, r11d
-	and eax, 262136
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	subpd xmm0, xmm12
+	; FPMUL_R e1, a2
+	mulpd xmm5, xmm10
+	; FPSUB_R f0, a3
+	subpd xmm0, xmm11
 	; IROR_R r0, r7
 	mov ecx, r15d
 	ror r8, cl
-	; FSTORE L2[r1], e0
+	; ISTORE L2[r1], r4
 	mov eax, r9d
-	and eax, 262128
-	movapd xmmword ptr [rsi+rax], xmm4
-	; IROR_R r7, r6
+	and eax, 262136
+	mov qword ptr [rsi+rax], r12
+	; IROL_R r7, r6
 	mov ecx, r14d
-	ror r15, cl
+	rol r15, cl
 	; IMUL_9C r2, 266593902
 	lea r10, [r10+r10*8+266593902]
 	; IMUL_R r4, r6
 	imul r12, r14
 	; FPSUB_R f2, a2
 	subpd xmm2, xmm10
-	; FPMUL_R e3, a0
-	mulpd xmm7, xmm8
-	; IXOR_M r7, L1[r2]
-	mov eax, r10d
-	and eax, 16376
-	xor r15, qword ptr [rsi+rax]
-	; IROR_R r0, r5
-	mov ecx, r13d
-	ror r8, cl
-	; FPADD_R f1, a2
-	addpd xmm1, xmm10
-	; FPSQRT_R e3
-	sqrtpd xmm7, xmm7
-	; FPADD_R f3, a1
-	addpd xmm3, xmm9
-	; FPADD_R f1, a0
-	addpd xmm1, xmm8
-	; COND_M r2, ge(L2[r2], -226330940)
-	xor ecx, ecx
-	mov eax, r10d
-	and eax, 262136
-	cmp dword ptr [rsi+rax], -226330940
-	setge cl
-	add r10, rcx
-	; FPDIV_R e2, a3
-	divpd xmm6, xmm11
-	maxpd xmm6, xmm13
-	; FPMUL_R e2, a1
-	mulpd xmm6, xmm9
-	; FPSUB_R f1, a0
-	subpd xmm1, xmm8
-	; IMUL_R r7, r5
-	imul r15, r13
-	; IMUL_R r0, r1
-	imul r8, r9
-	; FPSUB_R f3, a1
-	subpd xmm3, xmm9
-	; IROL_R r3, r5
-	mov ecx, r13d
-	rol r11, cl
-	; IADD_RC r5, r2, 795784298
-	lea r13, [r13+r10+795784298]
-	; ISUB_R r0, r4
-	sub r8, r12
-	; IMUL_R r5, r4
-	imul r13, r12
-	; FPSUB_R f0, a2
-	subpd xmm0, xmm10
-	; FPMUL_R e3, a1
-	mulpd xmm7, xmm9
-	; ISDIV_C r3, 1662492575
-	mov rax, 2978515652703905219
-	imul r11
-	xor eax, eax
-	sar rdx, 28
-	sets al
-	add rdx, rax
-	add r11, rdx
-	; ISMULH_R r5, r0
-	mov rax, r13
-	imul r8
-	mov r13, rdx
-	; ISDIV_C r4, 1963597892
-	mov rax, -8359627607928540073
-	imul r12
-	xor eax, eax
-	add rdx, r12
-	sar rdx, 30
-	sets al
-	add rdx, rax
-	add r12, rdx
-	; IMUL_R r7, r0
-	imul r15, r8
-	; IMULH_M r0, L1[r3]
-	mov ecx, r11d
-	and ecx, 16376
-	mov rax, r8
-	mul qword ptr [rsi+rcx]
-	mov r8, rdx
-	; IXOR_R r3, r7
-	xor r11, r15
-	; IDIV_C r4, 1146125335
-	mov rax, 8640870253760721727
-	mul r12
-	shr rdx, 29
-	add r12, rdx
-	; FPSWAP_R f3
-	shufpd xmm3, xmm3, 1
-	; IXOR_M r2, L1[r0]
-	mov eax, r8d
-	and eax, 16376
-	xor r10, qword ptr [rsi+rax]
-	; IROR_R r0, r1
-	mov ecx, r9d
-	ror r8, cl
-	; IXOR_R r7, r4
-	xor r15, r12
-	; ISMULH_R r6, r2
-	mov rax, r14
-	imul r10
-	mov r14, rdx
-	; FPMUL_R e3, a2
-	mulpd xmm7, xmm10
-	; IADD_RC r4, r2, 1704868083
-	lea r12, [r12+r10+1704868083]
-	; FPSUB_R f2, a0
-	subpd xmm2, xmm8
-	; ISTORE L1[r0], r0
-	mov eax, r8d
-	and eax, 16376
-	mov qword ptr [rsi+rax], r8
-	; FPSUB_R f0, a3
-	subpd xmm0, xmm11
-	; FPDIV_R e0, a3
-	divpd xmm4, xmm11
-	maxpd xmm4, xmm13
-	; FPMUL_R e3, a2
-	mulpd xmm7, xmm10
-	; ISUB_R r7, 1302457878
-	sub r15, 1302457878
-	; IMUL_9C r1, 1330165941
-	lea r9, [r9+r9*8+1330165941]
-	; FPMUL_R e1, a3
-	mulpd xmm5, xmm11
-	; IROL_R r0, r4
-	mov ecx, r12d
-	rol r8, cl
-	; FPSUB_M f1, L1[r0]
-	mov eax, r8d
-	and eax, 16376
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	subpd xmm1, xmm12
-	; IROL_R r5, r6
-	mov ecx, r14d
-	rol r13, cl
-	; COND_M r0, ab(L1[r1], -310933871)
-	xor ecx, ecx
-	mov eax, r9d
-	and eax, 16376
-	cmp dword ptr [rsi+rax], -310933871
-	seta cl
-	add r8, rcx
-	; CFROUND r7, 39
-	mov rax, r15
-	rol rax, 38
-	and eax, 24576
-	or eax, 40896
-	mov dword ptr [rsp-8], eax
-	ldmxcsr dword ptr [rsp-8]
-	; FPDIV_R e0, a1
-	divpd xmm4, xmm9
-	maxpd xmm4, xmm13
-	; IMUL_M r1, L1[r3]
-	mov eax, r11d
-	and eax, 16376
-	imul r9, qword ptr [rsi+rax]
-	; IMUL_9C r3, 1573236728
-	lea r11, [r11+r11*8+1573236728]
 	; FPNEG_R f3
 	xorps xmm3, xmm15
-	; COND_R r1, lt(r4, -1805702334)
-	xor ecx, ecx
-	cmp r12d, -1805702334
-	setl cl
-	add r9, rcx
-	; FPSWAP_R f1
-	shufpd xmm1, xmm1, 1
-	; IADD_R r7, -1421188024
-	add r15, -1421188024
-	; FPMUL_R e3, a2
-	mulpd xmm7, xmm10
-	; FPSUB_M f2, L2[r7]
-	mov eax, r15d
-	and eax, 262136
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	subpd xmm2, xmm12
-	; FPSUB_R f3, a1
-	subpd xmm3, xmm9
-	; FPSQRT_R e1
-	sqrtpd xmm5, xmm5
-	; ISUB_R r2, r4
-	sub r10, r12
-	; ISMULH_R r4, r5
-	mov rax, r12
-	imul r13
-	mov r12, rdx
-	; COND_R r1, of(r7, 1294727006)
-	xor ecx, ecx
-	cmp r15d, 1294727006
-	seto cl
-	add r9, rcx
-	; IADD_M r5, L2[r2]
-	mov eax, r10d
-	and eax, 262136
-	add r13, qword ptr [rsi+rax]
-	; IMUL_9C r4, 401020510
-	lea r12, [r12+r12*8+401020510]
-	; IROL_R r3, r0
-	mov ecx, r8d
-	rol r11, cl
-	; ISTORE L1[r7], r0
-	mov eax, r15d
-	and eax, 16376
-	mov qword ptr [rsi+rax], r8
-	; FPSUB_R f2, a1
-	subpd xmm2, xmm9
-	; FPSQRT_R e3
-	sqrtpd xmm7, xmm7
-	; IMUL_R r3, 720965215
-	imul r11, 720965215
-	; IMUL_R r6, r2
-	imul r14, r10
-	; ISTORE L1[r7], r3
-	mov eax, r15d
-	and eax, 16376
-	mov qword ptr [rsi+rax], r11
-	; IROR_R r2, r6
-	mov ecx, r14d
-	ror r10, cl
-	; FPSQRT_R e3
-	sqrtpd xmm7, xmm7
-	; IMUL_9C r4, 788211341
-	lea r12, [r12+r12*8+788211341]
-	; IMUL_9C r3, -67993446
-	lea r11, [r11+r11*8-67993446]
-	; FPSWAP_R e3
-	shufpd xmm7, xmm7, 1
-	; IMUL_M r2, L1[r6]
-	mov eax, r14d
-	and eax, 16376
-	imul r10, qword ptr [rsi+rax]
-	; COND_M r2, ge(L1[r2], -1892157506)
-	xor ecx, ecx
-	mov eax, r10d
-	and eax, 16376
-	cmp dword ptr [rsi+rax], -1892157506
-	setge cl
-	add r10, rcx
-	; FPADD_M f1, L1[r3]
-	mov eax, r11d
-	and eax, 16376
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	addpd xmm1, xmm12
-	; IADD_M r7, L1[r0]
-	mov eax, r8d
-	and eax, 16376
-	add r15, qword ptr [rsi+rax]
-	; ISDIV_C r1, 624867857
-	mov rax, 7924491717200811467
-	imul r9
-	xor eax, eax
-	sar rdx, 28
-	sets al
-	add rdx, rax
-	add r9, rdx
-	; FPADD_R f0, a1
-	addpd xmm0, xmm9
-	; ISUB_R r5, r7
-	sub r13, r15
-	; FPNEG_R f0
-	xorps xmm0, xmm15
-	; IMUL_R r6, r2
-	imul r14, r10
-	; FPMUL_M e3, L1[r1]
-	mov eax, r9d
-	and eax, 16376
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	mulpd xmm7, xmm12
-	maxpd xmm7, xmm13
-	; IADD_R r0, r4
-	add r8, r12
-	; FPSUB_M f3, L1[r1]
-	mov eax, r9d
-	and eax, 16376
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	subpd xmm3, xmm12
-	; FPMUL_R e2, a0
-	mulpd xmm6, xmm8
-	; INEG_R r2
-	neg r10
-	; FPMUL_R e2, a2
-	mulpd xmm6, xmm10
-	; FPSUB_M f3, L1[r6]
-	mov eax, r14d
-	and eax, 16376
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	subpd xmm3, xmm12
-	; FPADD_R f1, a3
-	addpd xmm1, xmm11
-	; IMULH_R r3, r2
-	mov rax, r11
-	mul r10
-	mov r11, rdx
-	; FPSUB_R f0, a3
-	subpd xmm0, xmm11
-	; IDIV_C r5, 2887845607
-	mov rax, 13717520480010955377
-	mul r13
-	shr rdx, 31
-	add r13, rdx
-	; ISMULH_M r6, L1[r2]
-	mov ecx, r10d
-	and ecx, 16376
-	mov rax, r14
-	imul qword ptr [rsi+rcx]
-	mov r14, rdx
-	; FPSUB_R f3, a3
-	subpd xmm3, xmm11
-	; IMUL_M r6, L1[r7]
-	mov eax, r15d
-	and eax, 16376
-	imul r14, qword ptr [rsi+rax]
-	; FPNEG_R f0
-	xorps xmm0, xmm15
-	; FPMUL_R e2, a0
-	mulpd xmm6, xmm8
-	; IMUL_9C r6, 295130073
-	lea r14, [r14+r14*8+295130073]
-	; FPADD_R f1, a1
-	addpd xmm1, xmm9
-	; IXOR_R r0, r5
-	xor r8, r13
-	; FPADD_R f2, a1
-	addpd xmm2, xmm9
-	; FPSWAP_R e3
-	shufpd xmm7, xmm7, 1
-	; FPSQRT_R e3
-	sqrtpd xmm7, xmm7
-	; IADD_RC r3, r6, -1317630728
-	lea r11, [r11+r14-1317630728]
-	; IMUL_M r2, L1[r3]
-	mov eax, r11d
-	and eax, 16376
-	imul r10, qword ptr [rsi+rax]
-	; IADD_RC r1, r4, 894105694
-	lea r9, [r9+r12+894105694]
-	; IMUL_R r7, r0
-	imul r15, r8
-	; FPSUB_R f1, a0
-	subpd xmm1, xmm8
-	; IMUL_M r7, L1[r1]
-	mov eax, r9d
-	and eax, 16376
-	imul r15, qword ptr [rsi+rax]
-	; IXOR_R r2, r4
-	xor r10, r12
-	; ISUB_M r0, L1[r1]
-	mov eax, r9d
-	and eax, 16376
-	sub r8, qword ptr [rsi+rax]
-	; INEG_R r4
-	neg r12
-	; IMUL_9C r4, -285272388
-	lea r12, [r12+r12*8-285272388]
-	; IMUL_R r7, r4
-	imul r15, r12
-	; IMULH_M r5, L1[r7]
-	mov ecx, r15d
-	and ecx, 16376
-	mov rax, r13
-	mul qword ptr [rsi+rcx]
-	mov r13, rdx
-	; IROL_R r1, r7
-	mov ecx, r15d
-	rol r9, cl
-	; IXOR_R r4, -757532727
-	xor r12, -757532727
-	; IMUL_R r3, 1863959234
-	imul r11, 1863959234
-	; IROL_R r4, 59
-	rol r12, 59
-	; ISMULH_R r1, 2122681086
-	mov rax, 2122681086
-	imul r9
-	add r9, rdx
-	; ISTORE L2[r6], r7
-	mov eax, r14d
-	and eax, 262136
-	mov qword ptr [rsi+rax], r15
-	; ISTORE L1[r1], r5
-	mov eax, r9d
-	and eax, 16376
-	mov qword ptr [rsi+rax], r13
-	; FPMUL_R e0, a1
-	mulpd xmm4, xmm9
-	; COND_R r2, ns(r1, 486049737)
-	xor ecx, ecx
-	cmp r9d, 486049737
-	setns cl
-	add r10, rcx
-	; FPMUL_M e0, L2[r7]
-	mov eax, r15d
-	and eax, 262136
-	cvtdq2pd xmm12, qword ptr [rsi+rax]
-	mulpd xmm4, xmm12
-	maxpd xmm4, xmm13
-	; FPMUL_R e3, a2
-	mulpd xmm7, xmm10
-	; IROL_R r5, r2
+	; IROR_R r7, r2
 	mov ecx, r10d
-	rol r13, cl
-	; IADD_M r0, L1[r4]
-	mov eax, r12d
-	and eax, 16376
-	add r8, qword ptr [rsi+rax]
+	ror r15, cl
+	; IROR_R r0, r5
+	mov ecx, r13d
+	ror r8, cl