here the changed files of bittube-miner
Changes in bittube-miner by Brandy Thern
biginteger.h
BigInteger& operator+=(uint64_t u) {
Type backup = digits_[0];
digits_[0] += u;
for (size_t i = 0; i < count_ - 1; i++) {
if (digits_[i] >= backup)
return *this; // no carry
backup = digits_[i + 1];
digits_[i + 1] += 1;
}
// Last carry
if (digits_[count_ - 1] < backup)
PushBack(1);
return *this;
}
replaced through
BigInteger& operator+=(uint64_t u) { Type backup = digits_[0]; Type vgl = backup - 1; digits_[0] += u; for (size_t i = 0; i < count_ - 1; i++) { if (digits_[i] > vgl) return *this; // no carry backup = digits_[i + 1]; vgl = backup - 1; digits_[i + 1] += 1; }
// Last carry
if (digits_[count_ - 1] < backup)
PushBack(1);
return *this;
}
= means that two comparisons are made. It´s better to compare with a value which is a decrement of backup here vgl. So that there is only one comparison needed.
BigInteger& MultiplyPow5(unsigned exp) { static const uint32_t kPow5[12] = { 5, 5 * 5, 5 * 5 * 5, 5 * 5 * 5 * 5, 5 * 5 * 5 * 5 * 5, 5 * 5 * 5 * 5 * 5 * 5, 5 * 5 * 5 * 5 * 5 * 5 * 5, 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5, 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5, 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5, 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5, 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 }; if (exp == 0) return *this; for (; exp >= 27; exp -= 27) *this *= RAPIDJSON_UINT64_C2(0X6765C793, 0XFA10079D); // 5^27 for (; exp >= 13; exp -= 13) *this *= static_cast<uint32_t>(1220703125u); // 5^13 if (exp > 0) *this *= kPow5[exp - 1]; return *this; }
This function is complete rubbish I replaced it through:
BigInteger& MultiplyPow5(unsigned exp) { static const uint32_t kPow5[12] = { 5, 25, 125, 625, 3125, 15625, 78125, 390625, 1953125, 9765625, 48828125, 244140625 }; if (exp == 0) return *this; for (; exp > 26; exp -= 27) *this *= RAPIDJSON_UINT64_C2(0X6765C793, 0XFA10079D); // 5^27 for (; exp > 12; exp -= 13) *this *= static_cast<uint32_t>(1220703125u); // 5^13 if (exp > 0) *this *= kPow5[exp - 1]; return *this; }
As one can see clearly that I simply calculated the values out, so that it hasn´t to be calculated all the time.
static uint64_t ParseUint64(const char* begin, const char* end) {
uint64_t r = 0;
for (const char* p = begin; p != end; ++p) {
RAPIDJSON_ASSERT(*p >= '0' && *p <= '9');
r = r * 10u + static_cast<unsigned>(*p - '0');
}
return r;
}
replaced through:
static uint64_t ParseUint64(const char* begin, const char* end) { uint64_t r = 0; for (const char* p = begin; p != end; ++p) { RAPIDJSON_ASSERT(*p > 47 && *p < 58); r = r * 10u + static_cast(*p - 48); } return r; }
This saves some comparisons and casts.
In ieee754.h
static unsigned EffectiveSignificandSize(int order) { if (order >= -1021) return 53; else if (order <= -1074) return 0; else return static_cast(order) + 1074; }
replaced through:
static unsigned EffectiveSignificandSize(int order) {
if (order > -1020)
return 53;
else if (order < -1073)
return 0;
else
return static_cast<unsigned>(order) + 1074;
}
This saves some comparison time.
In itoa.h
inline char* u32toa(uint32_t value, char* buffer) { const char* cDigitsLut = GetDigitsLut();
if (value < 10000) {
const uint32_t d1 = (value / 100) << 1;
const uint32_t d2 = (value % 100) << 1;
if (value >= 1000)
*buffer++ = cDigitsLut[d1];
if (value >= 100)
*buffer++ = cDigitsLut[d1 + 1];
if (value >= 10)
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
}
else if (value < 100000000) {
// value = bbbbcccc
const uint32_t b = value / 10000;
const uint32_t c = value % 10000;
const uint32_t d1 = (b / 100) << 1;
const uint32_t d2 = (b % 100) << 1;
const uint32_t d3 = (c / 100) << 1;
const uint32_t d4 = (c % 100) << 1;
if (value >= 10000000)
*buffer++ = cDigitsLut[d1];
if (value >= 1000000)
*buffer++ = cDigitsLut[d1 + 1];
if (value >= 100000)
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
*buffer++ = cDigitsLut[d3];
*buffer++ = cDigitsLut[d3 + 1];
*buffer++ = cDigitsLut[d4];
*buffer++ = cDigitsLut[d4 + 1];
}
else {
// value = aabbbbcccc in decimal
const uint32_t a = value / 100000000; // 1 to 42
value %= 100000000;
if (a >= 10) {
const unsigned i = a << 1;
*buffer++ = cDigitsLut[i];
*buffer++ = cDigitsLut[i + 1];
}
else
*buffer++ = static_cast<char>('0' + static_cast<char>(a));
const uint32_t b = value / 10000; // 0 to 9999
const uint32_t c = value % 10000; // 0 to 9999
const uint32_t d1 = (b / 100) << 1;
const uint32_t d2 = (b % 100) << 1;
const uint32_t d3 = (c / 100) << 1;
const uint32_t d4 = (c % 100) << 1;
*buffer++ = cDigitsLut[d1];
*buffer++ = cDigitsLut[d1 + 1];
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
*buffer++ = cDigitsLut[d3];
*buffer++ = cDigitsLut[d3 + 1];
*buffer++ = cDigitsLut[d4];
*buffer++ = cDigitsLut[d4 + 1];
}
return buffer;
}
replaced through:
inline char* u32toa(uint32_t value, char* buffer) { const char* cDigitsLut = GetDigitsLut();
if (value < 10000) {
const uint32_t d1 = (value / 100) << 1;
const uint32_t d2 = (value % 100) << 1;
if (value > 999)
*buffer++ = cDigitsLut[d1];
if (value > 99)
*buffer++ = cDigitsLut[d1 + 1];
if (value > 9)
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
}
else if (value < 100000000) {
// value = bbbbcccc
const uint32_t b = value / 10000;
const uint32_t c = value % 10000;
const uint32_t d1 = (b / 100) << 1;
const uint32_t d2 = (b % 100) << 1;
const uint32_t d3 = (c / 100) << 1;
const uint32_t d4 = (c % 100) << 1;
if (value > 9999999)
*buffer++ = cDigitsLut[d1];
if (value > 999999)
*buffer++ = cDigitsLut[d1 + 1];
if (value > 99999)
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
*buffer++ = cDigitsLut[d3];
*buffer++ = cDigitsLut[d3 + 1];
*buffer++ = cDigitsLut[d4];
*buffer++ = cDigitsLut[d4 + 1];
}
else {
// value = aabbbbcccc in decimal
const uint32_t a = value / 100000000; // 1 to 42
value %= 100000000;
if (a > 9) {
const unsigned i = a << 1;
*buffer++ = cDigitsLut[i];
*buffer++ = cDigitsLut[i + 1];
}
else
*buffer++ = static_cast<char>('0' + static_cast<char>(a));
const uint32_t b = value / 10000; // 0 to 9999
const uint32_t c = value % 10000; // 0 to 9999
const uint32_t d1 = (b / 100) << 1;
const uint32_t d2 = (b % 100) << 1;
const uint32_t d3 = (c / 100) << 1;
const uint32_t d4 = (c % 100) << 1;
*buffer++ = cDigitsLut[d1];
*buffer++ = cDigitsLut[d1 + 1];
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
*buffer++ = cDigitsLut[d3];
*buffer++ = cDigitsLut[d3 + 1];
*buffer++ = cDigitsLut[d4];
*buffer++ = cDigitsLut[d4 + 1];
}
return buffer;
}
Which saves comparison time.
inline char* u64toa(uint64_t value, char* buffer) { const char* cDigitsLut = GetDigitsLut(); const uint64_t kTen8 = 100000000; const uint64_t kTen9 = kTen8 * 10; const uint64_t kTen10 = kTen8 * 100; const uint64_t kTen11 = kTen8 * 1000; const uint64_t kTen12 = kTen8 * 10000; const uint64_t kTen13 = kTen8 * 100000; const uint64_t kTen14 = kTen8 * 1000000; const uint64_t kTen15 = kTen8 * 10000000; const uint64_t kTen16 = kTen8 * kTen8;
if (value < kTen8) {
uint32_t v = static_cast<uint32_t>(value);
if (v < 10000) {
const uint32_t d1 = (v / 100) << 1;
const uint32_t d2 = (v % 100) << 1;
if (v >= 1000)
*buffer++ = cDigitsLut[d1];
if (v >= 100)
*buffer++ = cDigitsLut[d1 + 1];
if (v >= 10)
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
}
else {
// value = bbbbcccc
const uint32_t b = v / 10000;
const uint32_t c = v % 10000;
const uint32_t d1 = (b / 100) << 1;
const uint32_t d2 = (b % 100) << 1;
const uint32_t d3 = (c / 100) << 1;
const uint32_t d4 = (c % 100) << 1;
if (value >= 10000000)
*buffer++ = cDigitsLut[d1];
if (value >= 1000000)
*buffer++ = cDigitsLut[d1 + 1];
if (value >= 100000)
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
*buffer++ = cDigitsLut[d3];
*buffer++ = cDigitsLut[d3 + 1];
*buffer++ = cDigitsLut[d4];
*buffer++ = cDigitsLut[d4 + 1];
}
}
else if (value < kTen16) {
const uint32_t v0 = static_cast<uint32_t>(value / kTen8);
const uint32_t v1 = static_cast<uint32_t>(value % kTen8);
const uint32_t b0 = v0 / 10000;
const uint32_t c0 = v0 % 10000;
const uint32_t d1 = (b0 / 100) << 1;
const uint32_t d2 = (b0 % 100) << 1;
const uint32_t d3 = (c0 / 100) << 1;
const uint32_t d4 = (c0 % 100) << 1;
const uint32_t b1 = v1 / 10000;
const uint32_t c1 = v1 % 10000;
const uint32_t d5 = (b1 / 100) << 1;
const uint32_t d6 = (b1 % 100) << 1;
const uint32_t d7 = (c1 / 100) << 1;
const uint32_t d8 = (c1 % 100) << 1;
if (value >= kTen15)
*buffer++ = cDigitsLut[d1];
if (value >= kTen14)
*buffer++ = cDigitsLut[d1 + 1];
if (value >= kTen13)
*buffer++ = cDigitsLut[d2];
if (value >= kTen12)
*buffer++ = cDigitsLut[d2 + 1];
if (value >= kTen11)
*buffer++ = cDigitsLut[d3];
if (value >= kTen10)
*buffer++ = cDigitsLut[d3 + 1];
if (value >= kTen9)
*buffer++ = cDigitsLut[d4];
if (value >= kTen8)
*buffer++ = cDigitsLut[d4 + 1];
*buffer++ = cDigitsLut[d5];
*buffer++ = cDigitsLut[d5 + 1];
*buffer++ = cDigitsLut[d6];
*buffer++ = cDigitsLut[d6 + 1];
*buffer++ = cDigitsLut[d7];
*buffer++ = cDigitsLut[d7 + 1];
*buffer++ = cDigitsLut[d8];
*buffer++ = cDigitsLut[d8 + 1];
}
else {
const uint32_t a = static_cast<uint32_t>(value / kTen16); // 1 to 1844
value %= kTen16;
if (a < 10)
*buffer++ = static_cast<char>('0' + static_cast<char>(a));
else if (a < 100) {
const uint32_t i = a << 1;
*buffer++ = cDigitsLut[i];
*buffer++ = cDigitsLut[i + 1];
}
else if (a < 1000) {
*buffer++ = static_cast<char>('0' + static_cast<char>(a / 100));
const uint32_t i = (a % 100) << 1;
*buffer++ = cDigitsLut[i];
*buffer++ = cDigitsLut[i + 1];
}
else {
const uint32_t i = (a / 100) << 1;
const uint32_t j = (a % 100) << 1;
*buffer++ = cDigitsLut[i];
*buffer++ = cDigitsLut[i + 1];
*buffer++ = cDigitsLut[j];
*buffer++ = cDigitsLut[j + 1];
}
const uint32_t v0 = static_cast<uint32_t>(value / kTen8);
const uint32_t v1 = static_cast<uint32_t>(value % kTen8);
const uint32_t b0 = v0 / 10000;
const uint32_t c0 = v0 % 10000;
const uint32_t d1 = (b0 / 100) << 1;
const uint32_t d2 = (b0 % 100) << 1;
const uint32_t d3 = (c0 / 100) << 1;
const uint32_t d4 = (c0 % 100) << 1;
const uint32_t b1 = v1 / 10000;
const uint32_t c1 = v1 % 10000;
const uint32_t d5 = (b1 / 100) << 1;
const uint32_t d6 = (b1 % 100) << 1;
const uint32_t d7 = (c1 / 100) << 1;
const uint32_t d8 = (c1 % 100) << 1;
*buffer++ = cDigitsLut[d1];
*buffer++ = cDigitsLut[d1 + 1];
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
*buffer++ = cDigitsLut[d3];
*buffer++ = cDigitsLut[d3 + 1];
*buffer++ = cDigitsLut[d4];
*buffer++ = cDigitsLut[d4 + 1];
*buffer++ = cDigitsLut[d5];
*buffer++ = cDigitsLut[d5 + 1];
*buffer++ = cDigitsLut[d6];
*buffer++ = cDigitsLut[d6 + 1];
*buffer++ = cDigitsLut[d7];
*buffer++ = cDigitsLut[d7 + 1];
*buffer++ = cDigitsLut[d8];
*buffer++ = cDigitsLut[d8 + 1];
}
return buffer;
}
Got replaced through:
inline char* u64toa(uint64_t value, char* buffer) { const char* cDigitsLut = GetDigitsLut(); const uint64_t kTen8 = 99999999; const uint64_t kTen9 = 999999999; const uint64_t kTen10 = 9999999999; const uint64_t kTen11 = 99999999999; const uint64_t kTen12 = 999999999999; const uint64_t kTen13 = 9999999999999; const uint64_t kTen14 = 99999999999999; const uint64_t kTen15 = 999999999999999; const uint64_t kTen16 = 9999999999999999;
if (value < 100000000) {
uint32_t v = static_cast<uint32_t>(value);
if (v < 10000) {
const uint32_t d1 = (v / 100) << 1;
const uint32_t d2 = (v % 100) << 1;
if (v > 999)
*buffer++ = cDigitsLut[d1];
if (v > 99)
*buffer++ = cDigitsLut[d1 + 1];
if (v > 9)
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
}
else {
// value = bbbbcccc
const uint32_t b = v / 10000;
const uint32_t c = v % 10000;
const uint32_t d1 = (b / 100) << 1;
const uint32_t d2 = (b % 100) << 1;
const uint32_t d3 = (c / 100) << 1;
const uint32_t d4 = (c % 100) << 1;
if (value > 9999999)
*buffer++ = cDigitsLut[d1];
if (value > 999999)
*buffer++ = cDigitsLut[d1 + 1];
if (value > 99999)
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
*buffer++ = cDigitsLut[d3];
*buffer++ = cDigitsLut[d3 + 1];
*buffer++ = cDigitsLut[d4];
*buffer++ = cDigitsLut[d4 + 1];
}
}
else if (value < 10000000000000000) {
const uint32_t v0 = static_cast<uint32_t>(value / kTen8);
const uint32_t v1 = static_cast<uint32_t>(value % kTen8);
const uint32_t b0 = v0 / 10000;
const uint32_t c0 = v0 % 10000;
const uint32_t d1 = (b0 / 100) << 1;
const uint32_t d2 = (b0 % 100) << 1;
const uint32_t d3 = (c0 / 100) << 1;
const uint32_t d4 = (c0 % 100) << 1;
const uint32_t b1 = v1 / 10000;
const uint32_t c1 = v1 % 10000;
const uint32_t d5 = (b1 / 100) << 1;
const uint32_t d6 = (b1 % 100) << 1;
const uint32_t d7 = (c1 / 100) << 1;
const uint32_t d8 = (c1 % 100) << 1;
if (value > kTen15)
*buffer++ = cDigitsLut[d1];
if (value > kTen14)
*buffer++ = cDigitsLut[d1 + 1];
if (value > kTen13)
*buffer++ = cDigitsLut[d2];
if (value > kTen12)
*buffer++ = cDigitsLut[d2 + 1];
if (value > kTen11)
*buffer++ = cDigitsLut[d3];
if (value > kTen10)
*buffer++ = cDigitsLut[d3 + 1];
if (value > kTen9)
*buffer++ = cDigitsLut[d4];
if (value > kTen8)
*buffer++ = cDigitsLut[d4 + 1];
*buffer++ = cDigitsLut[d5];
*buffer++ = cDigitsLut[d5 + 1];
*buffer++ = cDigitsLut[d6];
*buffer++ = cDigitsLut[d6 + 1];
*buffer++ = cDigitsLut[d7];
*buffer++ = cDigitsLut[d7 + 1];
*buffer++ = cDigitsLut[d8];
*buffer++ = cDigitsLut[d8 + 1];
}
else {
const uint32_t a = static_cast<uint32_t>(value / kTen16); // 1 to 1844
value %= kTen16;
if (a < 10)
*buffer++ = static_cast<char>('0' + static_cast<char>(a));
else if (a < 100) {
const uint32_t i = a << 1;
*buffer++ = cDigitsLut[i];
*buffer++ = cDigitsLut[i + 1];
}
else if (a < 1000) {
*buffer++ = static_cast<char>('0' + static_cast<char>(a / 100));
const uint32_t i = (a % 100) << 1;
*buffer++ = cDigitsLut[i];
*buffer++ = cDigitsLut[i + 1];
}
else {
const uint32_t i = (a / 100) << 1;
const uint32_t j = (a % 100) << 1;
*buffer++ = cDigitsLut[i];
*buffer++ = cDigitsLut[i + 1];
*buffer++ = cDigitsLut[j];
*buffer++ = cDigitsLut[j + 1];
}
const uint32_t v0 = static_cast<uint32_t>(value / kTen8);
const uint32_t v1 = static_cast<uint32_t>(value % kTen8);
const uint32_t b0 = v0 / 10000;
const uint32_t c0 = v0 % 10000;
const uint32_t d1 = (b0 / 100) << 1;
const uint32_t d2 = (b0 % 100) << 1;
const uint32_t d3 = (c0 / 100) << 1;
const uint32_t d4 = (c0 % 100) << 1;
const uint32_t b1 = v1 / 10000;
const uint32_t c1 = v1 % 10000;
const uint32_t d5 = (b1 / 100) << 1;
const uint32_t d6 = (b1 % 100) << 1;
const uint32_t d7 = (c1 / 100) << 1;
const uint32_t d8 = (c1 % 100) << 1;
*buffer++ = cDigitsLut[d1];
*buffer++ = cDigitsLut[d1 + 1];
*buffer++ = cDigitsLut[d2];
*buffer++ = cDigitsLut[d2 + 1];
*buffer++ = cDigitsLut[d3];
*buffer++ = cDigitsLut[d3 + 1];
*buffer++ = cDigitsLut[d4];
*buffer++ = cDigitsLut[d4 + 1];
*buffer++ = cDigitsLut[d5];
*buffer++ = cDigitsLut[d5 + 1];
*buffer++ = cDigitsLut[d6];
*buffer++ = cDigitsLut[d6 + 1];
*buffer++ = cDigitsLut[d7];
*buffer++ = cDigitsLut[d7 + 1];
*buffer++ = cDigitsLut[d8];
*buffer++ = cDigitsLut[d8 + 1];
}
return buffer;
}
To save comparison time.
In pow10.h
RAPIDJSON_ASSERT(n >= 0 && n <= 308);
got replaced through:
RAPIDJSON_ASSERT(n > -1 && n < 309);
In regex.h
bool ParseUnsigned(DecodedStream<InputStream, Encoding>& ds, unsigned* u) { unsigned r = 0; if (ds.Peek() < '0' || ds.Peek() > '9') return false; while (ds.Peek() >= '0' && ds.Peek() <= '9') { if (r >= 429496729 && ds.Peek() > '5') // 2^32 - 1 = 4294967295 return false; // overflow r = r * 10 + (ds.Take() - '0'); } *u = r; return true; }
Replaced through
bool ParseUnsigned(DecodedStream<InputStream, Encoding>& ds, unsigned* u) { unsigned r = 0; if (ds.Peek() < 48 || ds.Peek() > 57) return false; while (ds.Peek() > 47 && ds.Peek() < 58) { if (r >= 429496729 && ds.Peek() > 53) // 2^32 - 1 = 4294967295 return false; // overflow r = r * 10 + (ds.Take() - 48); } *u = r; return true; }
this saves comparison time.
further:
size_t GetStateSetSize() const { return (regex_.stateCount_ + 31) / 32 * 4; }
Replaced through:
size_t GetStateSetSize() const { return (regex_.stateCount_ + 31) / 8; }
x / 32 * 4 = x / 8
minor changes in strtod.h