diff --git a/doc/program.asm b/doc/program.asm
index cced195..a1909c4 100644
--- a/doc/program.asm
+++ b/doc/program.asm
@@ -80,7 +80,7 @@ randomx_isn_20:
add r8, 2
test r8, 254
jz randomx_isn_0
- xor rcx, rcx
+ xor ecx, ecx
cmp r11d, 1593588996
seto cl
add r14, rcx
@@ -102,7 +102,7 @@ randomx_isn_24:
add r8, 64
test r8, 8128
jz randomx_isn_21
- xor rcx, rcx
+ xor ecx, ecx
cmp r8d, 149087159
setno cl
add r14, rcx
@@ -197,7 +197,7 @@ randomx_isn_51:
add r12, 128
test r12, 16256
jz randomx_isn_25
- xor rcx, rcx
+ xor ecx, ecx
cmp r11d, -1975981803
setbe cl
add r10, rcx
@@ -212,7 +212,7 @@ randomx_isn_54:
add r8, 64
test r8, 8128
jz randomx_isn_52
- xor rcx, rcx
+ xor ecx, ecx
cmp r9d, 1917049931
setns cl
add r13, rcx
@@ -290,7 +290,7 @@ randomx_isn_74:
add r9, 4
test r9, 508
jz randomx_isn_55
- xor rcx, rcx
+ xor ecx, ecx
cmp r11d, -1200328848
setns cl
add r14, rcx
@@ -352,7 +352,7 @@ randomx_isn_89:
add r8, 64
test r8, 8128
jz randomx_isn_75
- xor rcx, rcx
+ xor ecx, ecx
cmp r8d, -122257389
setno cl
add r10, rcx
@@ -556,7 +556,7 @@ randomx_isn_143:
add r14, 4
test r14, 508
jz randomx_isn_110
- xor rcx, rcx
+ xor ecx, ecx
cmp r9d, 880467599
setge cl
add r13, rcx
@@ -960,7 +960,7 @@ randomx_isn_247:
add r9, 4
test r9, 508
jz randomx_isn_223
- xor rcx, rcx
+ xor ecx, ecx
cmp r13d, -8545330
setbe cl
add r10, rcx
diff --git a/makefile b/makefile
index e295359..5585b2b 100644
--- a/makefile
+++ b/makefile
@@ -4,6 +4,7 @@ AR=gcc-ar
PLATFORM=$(shell uname -m)
CXXFLAGS=-std=c++11
CCFLAGS=
+ARFLAGS=rcs
BINDIR=bin
SRCDIR=src
TESTDIR=src/tests
@@ -21,13 +22,17 @@ ifeq ($(PLATFORM),x86_64)
CXXFLAGS += -maes
endif
-release: CXXFLAGS += -march=native -O3 -flto
-release: CCFLAGS += -march=native -O3 -flto
+release: CXXFLAGS += -O3 -flto
+release: CCFLAGS += -O3 -flto
release: LDFLAGS += -flto
release: $(BINARIES)
-nolto: CXXFLAGS += -march=native -O3
-nolto: CCFLAGS += -march=native -O3
+native: CXXFLAGS += -march=native -O3 -flto
+native: CCFLAGS += -march=native -O3 -flto
+native: $(BINARIES)
+
+nolto: CXXFLAGS += -O3
+nolto: CCFLAGS += -O3
nolto: $(BINARIES)
debug: CXXFLAGS += -g
@@ -42,10 +47,8 @@ profile: $(BINDIR)/benchmark
test: CXXFLAGS += -O0
-$(RXA): $(RXOBJS)
- $(AR) rcs $@ $(RXOBJS)
-$(OBJDIR)/%.o: | $(OBJDIR)
-$(BINDIR)/%: | $(BINDIR)
+$(RXA): $(RXOBJS) | $(BINDIR)
+ $(AR) $(ARFLAGS) $@ $(RXOBJS)
$(OBJDIR):
mkdir $(OBJDIR)
$(BINDIR):
@@ -65,7 +68,7 @@ $(OBJDIR)/code-generator.o: $(TESTDIR)/code-generator.cpp $(TESTDIR)/utility.hpp
$(CXX) $(CXXFLAGS) -c $< -o $@
$(BINDIR)/code-generator: $(OBJDIR)/code-generator.o $(RXA)
$(CXX) $(LDFLAGS) $< $(RXA) -o $@
-$(OBJDIR)/aes_hash.o: $(SRCDIR)/aes_hash.cpp $(SRCDIR)/soft_aes.h $(SRCDIR)/intrin_portable.h
+$(OBJDIR)/aes_hash.o: $(SRCDIR)/aes_hash.cpp $(SRCDIR)/soft_aes.h $(SRCDIR)/intrin_portable.h | $(OBJDIR)
$(OBJDIR)/argon2_ref.o: $(SRCDIR)/argon2_ref.c $(SRCDIR)/argon2.h $(SRCDIR)/argon2_core.h \
$(SRCDIR)/blake2/blamka-round-ref.h $(SRCDIR)/blake2/blake2.h \
$(SRCDIR)/blake2/blake2-impl.h $(SRCDIR)/blake2/endian.h $(SRCDIR)/blake2/blake2-impl.h \
diff --git a/src/assembly_generator_x86.cpp b/src/assembly_generator_x86.cpp
index 18fa18e..0d93752 100644
--- a/src/assembly_generator_x86.cpp
+++ b/src/assembly_generator_x86.cpp
@@ -34,15 +34,13 @@ namespace randomx {
static const char* regE[4] = { "xmm4", "xmm5", "xmm6", "xmm7" };
static const char* regA[4] = { "xmm8", "xmm9", "xmm10", "xmm11" };
- static const char* regA4 = "xmm12";
- static const char* dblMin = "xmm13";
- static const char* absMask = "xmm14";
- static const char* signMask = "xmm15";
- static const char* regMx = "rbp";
+ static const char* tempRegx = "xmm12";
+ static const char* mantissaMask = "xmm13";
+ static const char* exponentMask = "xmm14";
+ static const char* scaleMask = "xmm15";
static const char* regIc = "rbx";
static const char* regIc32 = "ebx";
static const char* regIc8 = "bl";
- static const char* regDatasetAddr = "rdi";
static const char* regScratchpadAddr = "rsi";
void AssemblyGeneratorX86::generateProgram(Program& prog) {
@@ -274,7 +272,6 @@ namespace randomx {
return (int32_t)instr.getImm32() & ScratchpadL3Mask;
}
- //1 uOP
void AssemblyGeneratorX86::h_IADD_RS(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if(instr.dst == RegisterNeedsDisplacement)
@@ -284,27 +281,18 @@ namespace randomx {
traceint(instr);
}
- //2.75 uOP
void AssemblyGeneratorX86::h_IADD_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr);
- asmCode << "\tadd " << regR[instr.dst] << ", qword ptr [rsi+rax]" << std::endl;
+ asmCode << "\tadd " << regR[instr.dst] << ", qword ptr [" << regScratchpadAddr << "+rax]" << std::endl;
}
else {
- asmCode << "\tadd " << regR[instr.dst] << ", qword ptr [rsi+" << genAddressImm(instr) << "]" << std::endl;
+ asmCode << "\tadd " << regR[instr.dst] << ", qword ptr [" << regScratchpadAddr << "+" << genAddressImm(instr) << "]" << std::endl;
}
traceint(instr);
}
- //1 uOP
- void AssemblyGeneratorX86::h_IADD_RC(Instruction& instr, int i) {
- registerUsage[instr.dst] = i;
- asmCode << "\tlea " << regR[instr.dst] << ", [" << regR[instr.dst] << "+" << regR[instr.src] << std::showpos << (int32_t)instr.getImm32() << std::noshowpos << "]" << std::endl;
- traceint(instr);
- }
-
- //1 uOP
void AssemblyGeneratorX86::h_ISUB_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
@@ -316,27 +304,18 @@ namespace randomx {
traceint(instr);
}
- //2.75 uOP
void AssemblyGeneratorX86::h_ISUB_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr);
- asmCode << "\tsub " << regR[instr.dst] << ", qword ptr [rsi+rax]" << std::endl;
+ asmCode << "\tsub " << regR[instr.dst] << ", qword ptr [" << regScratchpadAddr << "+rax]" << std::endl;
}
else {
- asmCode << "\tsub " << regR[instr.dst] << ", qword ptr [rsi+" << genAddressImm(instr) << "]" << std::endl;
+ asmCode << "\tsub " << regR[instr.dst] << ", qword ptr [" << regScratchpadAddr << "+" << genAddressImm(instr) << "]" << std::endl;
}
traceint(instr);
}
- //1 uOP
- void AssemblyGeneratorX86::h_IMUL_9C(Instruction& instr, int i) {
- registerUsage[instr.dst] = i;
- asmCode << "\tlea " << regR[instr.dst] << ", [" << regR[instr.dst] << "+" << regR[instr.dst] << "*8" << std::showpos << (int32_t)instr.getImm32() << std::noshowpos << "]" << std::endl;
- traceint(instr);
- }
-
- //1 uOP
void AssemblyGeneratorX86::h_IMUL_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
@@ -348,15 +327,14 @@ namespace randomx {
traceint(instr);
}
- //2.75 uOP
void AssemblyGeneratorX86::h_IMUL_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr);
- asmCode << "\timul " << regR[instr.dst] << ", qword ptr [rsi+rax]" << std::endl;
+ asmCode << "\timul " << regR[instr.dst] << ", qword ptr [" << regScratchpadAddr << "+rax]" << std::endl;
}
else {
- asmCode << "\timul " << regR[instr.dst] << ", qword ptr [rsi+" << genAddressImm(instr) << "]" << std::endl;
+ asmCode << "\timul " << regR[instr.dst] << ", qword ptr [" << regScratchpadAddr << "+" << genAddressImm(instr) << "]" << std::endl;
}
traceint(instr);
}
@@ -370,23 +348,21 @@ namespace randomx {
traceint(instr);
}
- //5.75 uOPs
void AssemblyGeneratorX86::h_IMULH_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr, "ecx");
asmCode << "\tmov rax, " << regR[instr.dst] << std::endl;
- asmCode << "\tmul qword ptr [rsi+rcx]" << std::endl;
+ asmCode << "\tmul qword ptr [" << regScratchpadAddr << "+rcx]" << std::endl;
}
else {
asmCode << "\tmov rax, " << regR[instr.dst] << std::endl;
- asmCode << "\tmul qword ptr [rsi+" << genAddressImm(instr) << "]" << std::endl;
+ asmCode << "\tmul qword ptr [" << regScratchpadAddr << "+" << genAddressImm(instr) << "]" << std::endl;
}
asmCode << "\tmov " << regR[instr.dst] << ", rdx" << std::endl;
traceint(instr);
}
- //4 uOPs
void AssemblyGeneratorX86::h_ISMULH_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
asmCode << "\tmov rax, " << regR[instr.dst] << std::endl;
@@ -395,30 +371,27 @@ namespace randomx {
traceint(instr);
}
- //5.75 uOPs
void AssemblyGeneratorX86::h_ISMULH_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr, "ecx");
asmCode << "\tmov rax, " << regR[instr.dst] << std::endl;
- asmCode << "\timul qword ptr [rsi+rcx]" << std::endl;
+ asmCode << "\timul qword ptr [" << regScratchpadAddr << "+rcx]" << std::endl;
}
else {
asmCode << "\tmov rax, " << regR[instr.dst] << std::endl;
- asmCode << "\timul qword ptr [rsi+" << genAddressImm(instr) << "]" << std::endl;
+ asmCode << "\timul qword ptr [" << regScratchpadAddr << "+" << genAddressImm(instr) << "]" << std::endl;
}
asmCode << "\tmov " << regR[instr.dst] << ", rdx" << std::endl;
traceint(instr);
}
- //1 uOP
void AssemblyGeneratorX86::h_INEG_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
asmCode << "\tneg " << regR[instr.dst] << std::endl;
traceint(instr);
}
- //1 uOP
void AssemblyGeneratorX86::h_IXOR_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
@@ -430,20 +403,18 @@ namespace randomx {
traceint(instr);
}
- //2.75 uOP
void AssemblyGeneratorX86::h_IXOR_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr);
- asmCode << "\txor " << regR[instr.dst] << ", qword ptr [rsi+rax]" << std::endl;
+ asmCode << "\txor " << regR[instr.dst] << ", qword ptr [" << regScratchpadAddr << "+rax]" << std::endl;
}
else {
- asmCode << "\txor " << regR[instr.dst] << ", qword ptr [rsi+" << genAddressImm(instr) << "]" << std::endl;
+ asmCode << "\txor " << regR[instr.dst] << ", qword ptr [" << regScratchpadAddr << "+" << genAddressImm(instr) << "]" << std::endl;
}
traceint(instr);
}
- //1.75 uOPs
void AssemblyGeneratorX86::h_IROR_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
@@ -456,7 +427,6 @@ namespace randomx {
traceint(instr);
}
- //1.75 uOPs
void AssemblyGeneratorX86::h_IROL_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
@@ -469,7 +439,6 @@ namespace randomx {
traceint(instr);
}
- //2 uOPs
void AssemblyGeneratorX86::h_IMUL_RCP(Instruction& instr, int i) {
if (instr.getImm32() != 0) {
registerUsage[instr.dst] = i;
@@ -483,12 +452,6 @@ namespace randomx {
}
}
- //~8.5 uOPs
- void AssemblyGeneratorX86::h_ISDIV_C(Instruction& instr, int i) {
- tracenop(instr);
- }
-
- //2 uOPs
void AssemblyGeneratorX86::h_ISWAP_R(Instruction& instr, int i) {
if (instr.src != instr.dst) {
registerUsage[instr.dst] = i;
@@ -501,13 +464,11 @@ namespace randomx {
}
}
- //1 uOPs
void AssemblyGeneratorX86::h_FSWAP_R(Instruction& instr, int i) {
asmCode << "\tshufpd " << regFE[instr.dst] << ", " << regFE[instr.dst] << ", 1" << std::endl;
traceflt(instr);
}
- //1 uOP
void AssemblyGeneratorX86::h_FADD_R(Instruction& instr, int i) {
instr.dst %= 4;
instr.src %= 4;
@@ -515,16 +476,14 @@ namespace randomx {
traceflt(instr);
}
- //5 uOPs
void AssemblyGeneratorX86::h_FADD_M(Instruction& instr, int i) {
instr.dst %= 4;
genAddressReg(instr);
- asmCode << "\tcvtdq2pd xmm12, qword ptr [rsi+rax]" << std::endl;
- asmCode << "\taddpd " << regF[instr.dst] << ", xmm12" << std::endl;
+ asmCode << "\tcvtdq2pd " << tempRegx << ", qword ptr [" << regScratchpadAddr << "+rax]" << std::endl;
+ asmCode << "\taddpd " << regF[instr.dst] << ", " << tempRegx << std::endl;
traceflt(instr);
}
- //1 uOP
void AssemblyGeneratorX86::h_FSUB_R(Instruction& instr, int i) {
instr.dst %= 4;
instr.src %= 4;
@@ -532,23 +491,20 @@ namespace randomx {
traceflt(instr);
}
- //5 uOPs
void AssemblyGeneratorX86::h_FSUB_M(Instruction& instr, int i) {
instr.dst %= 4;
genAddressReg(instr);
- asmCode << "\tcvtdq2pd xmm12, qword ptr [rsi+rax]" << std::endl;
- asmCode << "\tsubpd " << regF[instr.dst] << ", xmm12" << std::endl;
+ asmCode << "\tcvtdq2pd " << tempRegx << ", qword ptr [" << regScratchpadAddr << "+rax]" << std::endl;
+ asmCode << "\tsubpd " << regF[instr.dst] << ", " << tempRegx << std::endl;
traceflt(instr);
}
- //1 uOP
void AssemblyGeneratorX86::h_FSCAL_R(Instruction& instr, int i) {
instr.dst %= 4;
- asmCode << "\txorps " << regF[instr.dst] << ", " << signMask << std::endl;
+ asmCode << "\txorps " << regF[instr.dst] << ", " << scaleMask << std::endl;
traceflt(instr);
}
- //1 uOPs
void AssemblyGeneratorX86::h_FMUL_R(Instruction& instr, int i) {
instr.dst %= 4;
instr.src %= 4;
@@ -556,45 +512,22 @@ namespace randomx {
traceflt(instr);
}
- //7 uOPs
- void AssemblyGeneratorX86::h_FMUL_M(Instruction& instr, int i) {
- instr.dst %= 4;
- genAddressReg(instr);
- asmCode << "\tcvtdq2pd xmm12, qword ptr [rsi+rax]" << std::endl;
- asmCode << "\tandps xmm12, xmm14" << std::endl;
- asmCode << "\tmulpd " << regE[instr.dst] << ", xmm12" << std::endl;
- asmCode << "\tmaxpd " << regE[instr.dst] << ", " << dblMin << std::endl;
- traceflt(instr);
- }
-
- //2 uOPs
- void AssemblyGeneratorX86::h_FDIV_R(Instruction& instr, int i) {
- instr.dst %= 4;
- instr.src %= 4;
- asmCode << "\tdivpd " << regE[instr.dst] << ", " << regA[instr.src] << std::endl;
- asmCode << "\tmaxpd " << regE[instr.dst] << ", " << dblMin << std::endl;
- traceflt(instr);
- }
-
- //7 uOPs
void AssemblyGeneratorX86::h_FDIV_M(Instruction& instr, int i) {
instr.dst %= 4;
genAddressReg(instr);
- asmCode << "\tcvtdq2pd xmm12, qword ptr [rsi+rax]" << std::endl;
- asmCode << "\tandps xmm12, xmm13" << std::endl;
- asmCode << "\torps xmm12, xmm14" << std::endl;
- asmCode << "\tdivpd " << regE[instr.dst] << ", xmm12" << std::endl;
+ asmCode << "\tcvtdq2pd " << tempRegx << ", qword ptr [" << regScratchpadAddr << "+rax]" << std::endl;
+ asmCode << "\tandps " << tempRegx << ", " << mantissaMask << std::endl;
+ asmCode << "\torps " << tempRegx << ", " << exponentMask << std::endl;
+ asmCode << "\tdivpd " << regE[instr.dst] << ", " << tempRegx << std::endl;
traceflt(instr);
}
- //1 uOP
void AssemblyGeneratorX86::h_FSQRT_R(Instruction& instr, int i) {
instr.dst %= 4;
asmCode << "\tsqrtpd " << regE[instr.dst] << ", " << regE[instr.dst] << std::endl;
traceflt(instr);
}
- //6 uOPs
void AssemblyGeneratorX86::h_CFROUND(Instruction& instr, int i) {
asmCode << "\tmov rax, " << regR[instr.src] << std::endl;
int rotate = (13 - (instr.getImm32() & 63)) & 63;
@@ -645,7 +578,6 @@ namespace randomx {
}
}
- //4 uOPs
void AssemblyGeneratorX86::h_COND_R(Instruction& instr, int i) {
handleCondition(instr, i);
asmCode << "\txor ecx, ecx" << std::endl;
@@ -655,28 +587,9 @@ namespace randomx {
traceint(instr);
}
- //6 uOPs
- void AssemblyGeneratorX86::h_COND_M(Instruction& instr, int i) {
- handleCondition(instr, i);
- asmCode << "\txor ecx, ecx" << std::endl;
- genAddressReg(instr);
- asmCode << "\tcmp dword ptr [rsi+rax], " << (int32_t)instr.getImm32() << std::endl;
- asmCode << "\tset" << condition(instr) << " cl" << std::endl;
- asmCode << "\tadd " << regR[instr.dst] << ", rcx" << std::endl;
- traceint(instr);
- }
-
- //3 uOPs
void AssemblyGeneratorX86::h_ISTORE(Instruction& instr, int i) {
genAddressRegDst(instr);
- asmCode << "\tmov qword ptr [rsi+rax], " << regR[instr.src] << std::endl;
- tracenop(instr);
- }
-
- //3 uOPs
- void AssemblyGeneratorX86::h_FSTORE(Instruction& instr, int i) {
- genAddressRegDst(instr, 16);
- asmCode << "\tmovapd xmmword ptr [rsi+rax], " << regFE[instr.src] << std::endl;
+ asmCode << "\tmov qword ptr [" << regScratchpadAddr << "+rax], " << regR[instr.src] << std::endl;
tracenop(instr);
}
@@ -692,10 +605,8 @@ namespace randomx {
//Integer
INST_HANDLE(IADD_RS)
INST_HANDLE(IADD_M)
- INST_HANDLE(IADD_RC)
INST_HANDLE(ISUB_R)
INST_HANDLE(ISUB_M)
- INST_HANDLE(IMUL_9C)
INST_HANDLE(IMUL_R)
INST_HANDLE(IMUL_M)
INST_HANDLE(IMULH_R)
@@ -727,7 +638,6 @@ namespace randomx {
//Control
INST_HANDLE(COND_R)
- INST_HANDLE(COND_M)
INST_HANDLE(CFROUND)
INST_HANDLE(ISTORE)
diff --git a/src/assembly_generator_x86.hpp b/src/assembly_generator_x86.hpp
index f5b523f..60ea7ab 100644
--- a/src/assembly_generator_x86.hpp
+++ b/src/assembly_generator_x86.hpp
@@ -50,10 +50,8 @@ namespace randomx {
void tracenop(Instruction&);
void h_IADD_RS(Instruction&, int);
void h_IADD_M(Instruction&, int);
- void h_IADD_RC(Instruction&, int);
void h_ISUB_R(Instruction&, int);
void h_ISUB_M(Instruction&, int);
- void h_IMUL_9C(Instruction&, int);
void h_IMUL_R(Instruction&, int);
void h_IMUL_M(Instruction&, int);
void h_IMULH_R(Instruction&, int);
@@ -75,15 +73,11 @@ namespace randomx {
void h_FSUB_M(Instruction&, int);
void h_FSCAL_R(Instruction&, int);
void h_FMUL_R(Instruction&, int);
- void h_FMUL_M(Instruction&, int);
- void h_FDIV_R(Instruction&, int);
void h_FDIV_M(Instruction&, int);
void h_FSQRT_R(Instruction&, int);
void h_COND_R(Instruction&, int);
- void h_COND_M(Instruction&, int);
void h_CFROUND(Instruction&, int);
void h_ISTORE(Instruction&, int);
- void h_FSTORE(Instruction&, int);
void h_NOP(Instruction&, int);
static InstructionGenerator engine[256];
diff --git a/src/common.hpp b/src/common.hpp
index a35f8ec..84d0e26 100644
--- a/src/common.hpp
+++ b/src/common.hpp
@@ -40,13 +40,13 @@ namespace randomx {
static_assert((RANDOMX_SCRATCHPAD_L1 & (RANDOMX_SCRATCHPAD_L1 - 1)) == 0, "RANDOMX_SCRATCHPAD_L1 must be a power of 2.");
static_assert(RANDOMX_CACHE_ACCESSES > 1, "RANDOMX_CACHE_ACCESSES must be greater than 1");
- constexpr int wtSum = RANDOMX_FREQ_IADD_RS + RANDOMX_FREQ_IADD_M + RANDOMX_FREQ_IADD_RC + RANDOMX_FREQ_ISUB_R + \
- RANDOMX_FREQ_ISUB_M + RANDOMX_FREQ_IMUL_9C + RANDOMX_FREQ_IMUL_R + RANDOMX_FREQ_IMUL_M + RANDOMX_FREQ_IMULH_R + \
+ constexpr int wtSum = RANDOMX_FREQ_IADD_RS + RANDOMX_FREQ_IADD_M + RANDOMX_FREQ_ISUB_R + \
+ RANDOMX_FREQ_ISUB_M + RANDOMX_FREQ_IMUL_R + RANDOMX_FREQ_IMUL_M + RANDOMX_FREQ_IMULH_R + \
RANDOMX_FREQ_IMULH_M + RANDOMX_FREQ_ISMULH_R + RANDOMX_FREQ_ISMULH_M + RANDOMX_FREQ_IMUL_RCP + \
RANDOMX_FREQ_INEG_R + RANDOMX_FREQ_IXOR_R + RANDOMX_FREQ_IXOR_M + RANDOMX_FREQ_IROR_R + RANDOMX_FREQ_ISWAP_R + \
RANDOMX_FREQ_FSWAP_R + RANDOMX_FREQ_FADD_R + RANDOMX_FREQ_FADD_M + RANDOMX_FREQ_FSUB_R + RANDOMX_FREQ_FSUB_M + \
RANDOMX_FREQ_FSCAL_R + RANDOMX_FREQ_FMUL_R + RANDOMX_FREQ_FDIV_M + RANDOMX_FREQ_FSQRT_R + RANDOMX_FREQ_COND_R + \
- RANDOMX_FREQ_COND_M + RANDOMX_FREQ_CFROUND + RANDOMX_FREQ_ISTORE + RANDOMX_FREQ_NOP;
+ RANDOMX_FREQ_CFROUND + RANDOMX_FREQ_ISTORE + RANDOMX_FREQ_NOP;
static_assert(wtSum == 256, "Sum of instruction frequencies must be 256.");
diff --git a/src/configuration.h b/src/configuration.h
index ef57775..bf10f51 100644
--- a/src/configuration.h
+++ b/src/configuration.h
@@ -68,10 +68,8 @@ Total sum of frequencies must be 256
#define RANDOMX_FREQ_IADD_RS 32
#define RANDOMX_FREQ_IADD_M 7
-#define RANDOMX_FREQ_IADD_RC 0
#define RANDOMX_FREQ_ISUB_R 17
#define RANDOMX_FREQ_ISUB_M 7
-#define RANDOMX_FREQ_IMUL_9C 0
#define RANDOMX_FREQ_IMUL_R 16
#define RANDOMX_FREQ_IMUL_M 4
#define RANDOMX_FREQ_IMULH_R 4
@@ -97,7 +95,6 @@ Total sum of frequencies must be 256
#define RANDOMX_FREQ_FSQRT_R 6
#define RANDOMX_FREQ_COND_R 8
-#define RANDOMX_FREQ_COND_M 0
#define RANDOMX_FREQ_CFROUND 1
#define RANDOMX_FREQ_ISTORE 16
diff --git a/src/instruction.cpp b/src/instruction.cpp
index 24a44b7..9f1b681 100644
--- a/src/instruction.cpp
+++ b/src/instruction.cpp
@@ -65,10 +65,6 @@ namespace randomx {
}
}
- void Instruction::h_IADD_RC(std::ostream& os) const {
- os << "r" << (int)dst << ", r" << (int)src << ", " << (int32_t)getImm32() << std::endl;
- }
-
//1 uOP
void Instruction::h_ISUB_R(std::ostream& os) const {
if (src != dst) {
@@ -92,10 +88,6 @@ namespace randomx {
}
}
- void Instruction::h_IMUL_9C(std::ostream& os) const {
- os << "r" << (int)dst << ", " << (int32_t)getImm32() << std::endl;
- }
-
void Instruction::h_IMUL_R(std::ostream& os) const {
if (src != dst) {
os << "r" << (int)dst << ", r" << (int)src << std::endl;
@@ -200,10 +192,6 @@ namespace randomx {
os << "r" << (int)dst << ", " << getImm32() << std::endl;
}
- void Instruction::h_ISDIV_C(std::ostream& os) const {
- os << "r" << (int)dst << ", " << (int32_t)getImm32() << std::endl;
- }
-
void Instruction::h_ISWAP_R(std::ostream& os) const {
os << "r" << (int)dst << ", r" << (int)src << std::endl;
}
@@ -251,19 +239,6 @@ namespace randomx {
os << "e" << dstIndex << ", a" << srcIndex << std::endl;
}
- void Instruction::h_FMUL_M(std::ostream& os) const {
- auto dstIndex = dst % 4;
- os << "e" << dstIndex << ", ";
- genAddressReg(os);
- os << std::endl;
- }
-
- void Instruction::h_FDIV_R(std::ostream& os) const {
- auto dstIndex = dst % 4;
- auto srcIndex = src % 4;
- os << "e" << dstIndex << ", a" << srcIndex << std::endl;
- }
-
void Instruction::h_FDIV_M(std::ostream& os) const {
auto dstIndex = dst % 4;
os << "e" << dstIndex << ", ";
@@ -308,24 +283,11 @@ namespace randomx {
os << "r" << (int)dst << ", " << condition((mod >> 2) & 7) << "(r" << (int)src << ", " << (int32_t)getImm32() << "), LSH " << (int)(mod >> 5) << std::endl;
}
- void Instruction::h_COND_M(std::ostream& os) const {
- os << "r" << (int)dst << ", " << condition((mod >> 2) & 7) << "(";
- genAddressReg(os);
- os << ", " << (int32_t)getImm32() << "), LSH " << (int)(mod >> 5) << std::endl;
- }
-
void Instruction::h_ISTORE(std::ostream& os) const {
genAddressRegDst(os);
os << ", r" << (int)src << std::endl;
}
- void Instruction::h_FSTORE(std::ostream& os) const {
- const char reg = (src >= 4) ? 'e' : 'f';
- genAddressRegDst(os);
- auto srcIndex = src % 4;
- os << ", " << reg << srcIndex << std::endl;
- }
-
void Instruction::h_NOP(std::ostream& os) const {
os << std::endl;
}
@@ -338,10 +300,8 @@ namespace randomx {
//Integer
INST_NAME(IADD_RS)
INST_NAME(IADD_M)
- INST_NAME(IADD_RC)
INST_NAME(ISUB_R)
INST_NAME(ISUB_M)
- INST_NAME(IMUL_9C)
INST_NAME(IMUL_R)
INST_NAME(IMUL_M)
INST_NAME(IMULH_R)
@@ -372,7 +332,6 @@ namespace randomx {
//Control
INST_NAME(COND_R)
- INST_NAME(COND_M)
INST_NAME(CFROUND)
INST_NAME(ISTORE)
@@ -384,10 +343,8 @@ namespace randomx {
//Integer
INST_HANDLE(IADD_RS)
INST_HANDLE(IADD_M)
- INST_HANDLE(IADD_RC)
INST_HANDLE(ISUB_R)
INST_HANDLE(ISUB_M)
- INST_HANDLE(IMUL_9C)
INST_HANDLE(IMUL_R)
INST_HANDLE(IMUL_M)
INST_HANDLE(IMULH_R)
@@ -417,13 +374,9 @@ namespace randomx {
INST_HANDLE(FDIV_M)
INST_HANDLE(FSQRT_R)
- //Control
INST_HANDLE(COND_R)
- INST_HANDLE(COND_M)
INST_HANDLE(CFROUND)
-
INST_HANDLE(ISTORE)
-
INST_HANDLE(NOP)
};
diff --git a/src/instruction.hpp b/src/instruction.hpp
index 323c1f5..3420cda 100644
--- a/src/instruction.hpp
+++ b/src/instruction.hpp
@@ -32,38 +32,34 @@ namespace randomx {
namespace InstructionType {
constexpr int IADD_RS = 0;
constexpr int IADD_M = 1;
- constexpr int IADD_RC = 2;
- constexpr int ISUB_R = 3;
- constexpr int ISUB_M = 4;
- constexpr int IMUL_9C = 5;
- constexpr int IMUL_R = 6;
- constexpr int IMUL_M = 7;
- constexpr int IMULH_R = 8;
- constexpr int IMULH_M = 9;
- constexpr int ISMULH_R = 10;
- constexpr int ISMULH_M = 11;
- constexpr int IMUL_RCP = 12;
- constexpr int INEG_R = 13;
- constexpr int IXOR_R = 14;
- constexpr int IXOR_M = 15;
- constexpr int IROR_R = 16;
- constexpr int IROL_R = 17;
- constexpr int ISWAP_R = 18;
- constexpr int FSWAP_R = 19;
- constexpr int FADD_R = 20;
- constexpr int FADD_M = 21;
- constexpr int FSUB_R = 22;
- constexpr int FSUB_M = 23;
- constexpr int FSCAL_R = 24;
- constexpr int FMUL_R = 25;
- constexpr int FDIV_M = 26;
- constexpr int FSQRT_R = 27;
- constexpr int COND_R = 28;
- constexpr int COND_M = 29;
- constexpr int CFROUND = 30;
- constexpr int ISTORE = 31;
- constexpr int FSTORE = 32;
- constexpr int NOP = 33;
+ constexpr int ISUB_R = 2;
+ constexpr int ISUB_M = 3;
+ constexpr int IMUL_R = 4;
+ constexpr int IMUL_M = 5;
+ constexpr int IMULH_R = 6;
+ constexpr int IMULH_M = 7;
+ constexpr int ISMULH_R = 8;
+ constexpr int ISMULH_M = 9;
+ constexpr int IMUL_RCP = 10;
+ constexpr int INEG_R = 11;
+ constexpr int IXOR_R = 12;
+ constexpr int IXOR_M = 13;
+ constexpr int IROR_R = 14;
+ constexpr int IROL_R = 15;
+ constexpr int ISWAP_R = 16;
+ constexpr int FSWAP_R = 17;
+ constexpr int FADD_R = 18;
+ constexpr int FADD_M = 19;
+ constexpr int FSUB_R = 20;
+ constexpr int FSUB_M = 21;
+ constexpr int FSCAL_R = 22;
+ constexpr int FMUL_R = 23;
+ constexpr int FDIV_M = 24;
+ constexpr int FSQRT_R = 25;
+ constexpr int COND_R = 26;
+ constexpr int CFROUND = 27;
+ constexpr int ISTORE = 28;
+ constexpr int NOP = 29;
}
class Instruction {
@@ -112,10 +108,8 @@ namespace randomx {
void genAddressRegDst(std::ostream&) const;
void h_IADD_RS(std::ostream&) const;
void h_IADD_M(std::ostream&) const;
- void h_IADD_RC(std::ostream&) const;
void h_ISUB_R(std::ostream&) const;
void h_ISUB_M(std::ostream&) const;
- void h_IMUL_9C(std::ostream&) const;
void h_IMUL_R(std::ostream&) const;
void h_IMUL_M(std::ostream&) const;
void h_IMULH_R(std::ostream&) const;
@@ -123,7 +117,6 @@ namespace randomx {
void h_ISMULH_R(std::ostream&) const;
void h_ISMULH_M(std::ostream&) const;
void h_IMUL_RCP(std::ostream&) const;
- void h_ISDIV_C(std::ostream&) const;
void h_INEG_R(std::ostream&) const;
void h_IXOR_R(std::ostream&) const;
void h_IXOR_M(std::ostream&) const;
@@ -137,15 +130,11 @@ namespace randomx {
void h_FSUB_M(std::ostream&) const;
void h_FSCAL_R(std::ostream&) const;
void h_FMUL_R(std::ostream&) const;
- void h_FMUL_M(std::ostream&) const;
- void h_FDIV_R(std::ostream&) const;
void h_FDIV_M(std::ostream&) const;
void h_FSQRT_R(std::ostream&) const;
void h_COND_R(std::ostream&) const;
- void h_COND_M(std::ostream&) const;
void h_CFROUND(std::ostream&) const;
void h_ISTORE(std::ostream&) const;
- void h_FSTORE(std::ostream&) const;
void h_NOP(std::ostream&) const;
};
diff --git a/src/instructions_portable.cpp b/src/instructions_portable.cpp
index cfa20ab..b77e93b 100644
--- a/src/instructions_portable.cpp
+++ b/src/instructions_portable.cpp
@@ -19,7 +19,7 @@ along with RandomX. If not, see.
//#define DEBUG
-#pragma STDC FENV_ACCESS on
+#pragma STDC FENV_ACCESS ON
#include
#include
#ifdef DEBUG
diff --git a/src/jit_compiler_x86.cpp b/src/jit_compiler_x86.cpp
index 1d1f80b..b24eba3 100644
--- a/src/jit_compiler_x86.cpp
+++ b/src/jit_compiler_x86.cpp
@@ -953,10 +953,8 @@ namespace randomx {
InstructionGeneratorX86 JitCompilerX86::engine[256] = {
INST_HANDLE(IADD_RS)
INST_HANDLE(IADD_M)
- INST_HANDLE(IADD_RC)
INST_HANDLE(ISUB_R)
INST_HANDLE(ISUB_M)
- INST_HANDLE(IMUL_9C)
INST_HANDLE(IMUL_R)
INST_HANDLE(IMUL_M)
INST_HANDLE(IMULH_R)
@@ -980,7 +978,6 @@ namespace randomx {
INST_HANDLE(FDIV_M)
INST_HANDLE(FSQRT_R)
INST_HANDLE(COND_R)
- INST_HANDLE(COND_M)
INST_HANDLE(CFROUND)
INST_HANDLE(ISTORE)
INST_HANDLE(NOP)
diff --git a/src/vm_interpreted.cpp b/src/vm_interpreted.cpp
index 22109d3..3d82b18 100644
--- a/src/vm_interpreted.cpp
+++ b/src/vm_interpreted.cpp
@@ -125,10 +125,6 @@ namespace randomx {
*ibc.idst += load64(getScratchpadAddress(ibc));
} break;
- case InstructionType::IADD_RC: {
- *ibc.idst += *ibc.isrc + ibc.imm;
- } break;
-
case InstructionType::ISUB_R: {
*ibc.idst -= *ibc.isrc;
} break;
@@ -137,10 +133,6 @@ namespace randomx {
*ibc.idst -= load64(getScratchpadAddress(ibc));
} break;
- case InstructionType::IMUL_9C: {
- *ibc.idst += 8 * *ibc.idst + ibc.imm;
- } break;
-
case InstructionType::IMUL_R: { //also handles IMUL_RCP
*ibc.idst *= *ibc.isrc;
} break;
@@ -243,18 +235,6 @@ namespace randomx {
*ibc.idst += condition(ibc.condition, *ibc.isrc, ibc.imm) ? 1 : 0;
} break;
- case InstructionType::COND_M: {
-#ifdef RANDOMX_JUMP
- *ibc.creg += (1uLL << ibc.shift);
- const uint64_t conditionMask = ((1ULL << RANDOMX_CONDITION_BITS) - 1) << ibc.shift;
- if ((*ibc.creg & conditionMask) == 0) {
- ic = ibc.target;
- break;
- }
-#endif
- *ibc.idst += condition(ibc.condition, load64(getScratchpadAddress(ibc)), ibc.imm) ? 1 : 0;
- } break;
-
case InstructionType::CFROUND: {
setRoundMode(rotr(*ibc.isrc, ibc.imm) % 4);
} break;
@@ -482,16 +462,6 @@ namespace randomx {
registerUsage[instr.dst] = i;
} break;
- CASE_REP(IADD_RC) {
- auto dst = instr.dst % RegistersCount;
- auto src = instr.src % RegistersCount;
- ibc.type = InstructionType::IADD_RC;
- ibc.idst = &r[dst];
- ibc.isrc = &r[src];
- ibc.imm = signExtend2sCompl(instr.getImm32());
- registerUsage[instr.dst] = i;
- } break;
-
CASE_REP(ISUB_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
@@ -524,14 +494,6 @@ namespace randomx {
registerUsage[instr.dst] = i;
} break;
- CASE_REP(IMUL_9C) {
- auto dst = instr.dst % RegistersCount;
- ibc.type = InstructionType::IMUL_9C;
- ibc.idst = &r[dst];
- ibc.imm = signExtend2sCompl(instr.getImm32());
- registerUsage[instr.dst] = i;
- } break;
-
CASE_REP(IMUL_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
@@ -808,25 +770,6 @@ namespace randomx {
}
} break;
- CASE_REP(COND_M) {
- auto dst = instr.dst % RegistersCount;
- auto src = instr.src % RegistersCount;
- ibc.type = InstructionType::COND_M;
- ibc.idst = &r[dst];
- ibc.isrc = &r[src];
- ibc.condition = instr.getModCond();
- ibc.imm = instr.getImm32();
- ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
- //jump condition
- int reg = getConditionRegister(registerUsage);
- ibc.target = registerUsage[reg];
- ibc.shift = instr.getModShift3();
- ibc.creg = &r[reg];
- for (unsigned j = 0; j < 8; ++j) { //mark all registers as used
- registerUsage[j] = i;
- }
- } break;
-
CASE_REP(CFROUND) {
auto src = instr.src % 8;
ibc.isrc = &r[src];
diff --git a/src/vm_interpreted_light.hpp b/src/vm_interpreted_light.hpp
index 67bf49a..9e68fee 100644
--- a/src/vm_interpreted_light.hpp
+++ b/src/vm_interpreted_light.hpp
@@ -40,7 +40,7 @@ namespace randomx {
void setDataset(randomx_dataset* dataset) override { }
void setCache(randomx_cache* cache) override;
protected:
- virtual void datasetRead(uint32_t address, int_reg_t(&r)[8]);
+ void datasetRead(uint32_t address, int_reg_t(&r)[8]) override;
private:
randomx_cache* cachePtr;
};