Skip to content

R
rapidjson
Commit c4ce48cd authored by Milo Yip's avatar Milo Yip
Browse files
Options

Merge pull request #80 from miloyip/issue31optimizeitoa 

Issue31optimizeitoa
parents a7dca0d4 6f164613
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 378 additions and 31 deletions
include/rapidjson/internal/itoa.h 0 → 100644
View file @ c4ce48cd
#ifndef RAPIDJSON_ITOA_
#define RAPIDJSON_ITOA_
namespace rapidjson {
namespace internal {
// Modified from https://github.com/miloyip/itoa-benchmark/blob/master/src/branchlut.cpp
// API is changed to return the character passed the end of string, without writing '\0'
inline const char* GetDigitsLut() {
static const char cDigitsLut[200] = {
'0','0','0','1','0','2','0','3','0','4','0','5','0','6','0','7','0','8','0','9',
'1','0','1','1','1','2','1','3','1','4','1','5','1','6','1','7','1','8','1','9',
'2','0','2','1','2','2','2','3','2','4','2','5','2','6','2','7','2','8','2','9',
'3','0','3','1','3','2','3','3','3','4','3','5','3','6','3','7','3','8','3','9',
'4','0','4','1','4','2','4','3','4','4','4','5','4','6','4','7','4','8','4','9',
'5','0','5','1','5','2','5','3','5','4','5','5','5','6','5','7','5','8','5','9',
'6','0','6','1','6','2','6','3','6','4','6','5','6','6','6','7','6','8','6','9',
'7','0','7','1','7','2','7','3','7','4','7','5','7','6','7','7','7','8','7','9',
'8','0','8','1','8','2','8','3','8','4','8','5','8','6','8','7','8','8','8','9',
'9','0','9','1','9','2','9','3','9','4','9','5','9','6','9','7','9','8','9','9'
};
return cDigitsLut;
}
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++ = '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;
}
inline char* i32toa(int32_t value, char* buffer) {
if (value < 0) {
*buffer++ = '-';
value = -value;
}
return u32toa(static_cast<uint32_t>(value), buffer);
}
inline char* u64toa(uint64_t value, char* buffer) {
const char* cDigitsLut = GetDigitsLut();
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 >= 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 < 10000000000000000) {
const uint32_t v0 = static_cast<uint32_t>(value / 100000000);
const uint32_t v1 = static_cast<uint32_t>(value % 100000000);
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 >= 1000000000000000)
*buffer++ = cDigitsLut[d1];
if (value >= 100000000000000)
*buffer++ = cDigitsLut[d1 + 1];
if (value >= 10000000000000)
*buffer++ = cDigitsLut[d2];
if (value >= 1000000000000)
*buffer++ = cDigitsLut[d2 + 1];
if (value >= 100000000000)
*buffer++ = cDigitsLut[d3];
if (value >= 10000000000)
*buffer++ = cDigitsLut[d3 + 1];
if (value >= 1000000000)
*buffer++ = cDigitsLut[d4];
if (value >= 100000000)
*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 / 10000000000000000); // 1 to 1844
value %= 10000000000000000;
if (a < 10)
*buffer++ = '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++ = '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 / 100000000);
const uint32_t v1 = static_cast<uint32_t>(value % 100000000);
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;
}
inline char* i64toa(int64_t value, char* buffer) {
if (value < 0) {
*buffer++ = '-';
value = -value;
}
return u64toa(static_cast<uint64_t>(value), buffer);
}
} // namespace internal
} // namespace rapidjson
#endif // RAPIDJSON_ITOA_
include/rapidjson/msinttypes/stdint.h
View file @ c4ce48cd
...@@ -42,8 +42,44 @@ ...@@ -42,8 +42,44 @@
#endif #endif
// miloyip: Originally Visual Studio 2010 uses its own stdint.h. However it generates warning with INT64_C(), so change to use this file for vs2010. // miloyip: Originally Visual Studio 2010 uses its own stdint.h. However it generates warning with INT64_C(), so change to use this file for vs2010.
#if _MSC_VER >= 1700 // [ #if _MSC_VER >= 1600 // [
#include <stdint.h> #include <stdint.h>
#if !defined(__cplusplus) || defined(__STDC_CONSTANT_MACROS) // [ See footnote 224 at page 260
#undef INT8_C
#undef INT16_C
#undef INT32_C
#undef INT64_C
#undef UINT8_C
#undef UINT16_C
#undef UINT32_C
#undef UINT64_C
// 7.18.4.1 Macros for minimum-width integer constants
#define INT8_C(val) val##i8
#define INT16_C(val) val##i16
#define INT32_C(val) val##i32
#define INT64_C(val) val##i64
#define UINT8_C(val) val##ui8
#define UINT16_C(val) val##ui16
#define UINT32_C(val) val##ui32
#define UINT64_C(val) val##ui64
// 7.18.4.2 Macros for greatest-width integer constants
// These #ifndef's are needed to prevent collisions with <boost/cstdint.hpp>.
// Check out Issue 9 for the details.
#ifndef INTMAX_C // [
# define INTMAX_C INT64_C
#endif // INTMAX_C ]
#ifndef UINTMAX_C // [
# define UINTMAX_C UINT64_C
#endif // UINTMAX_C ]
#endif // __STDC_CONSTANT_MACROS ]
#else // ] _MSC_VER >= 1700 [ #else // ] _MSC_VER >= 1700 [
#include <limits.h> #include <limits.h>
......
include/rapidjson/rapidjson.h
View file @ c4ce48cd
...@@ -360,10 +360,14 @@ struct GenericInsituStringStream { ...@@ -360,10 +360,14 @@ struct GenericInsituStringStream {
size_t Tell() { return static_cast<size_t>(src_ - head_); } size_t Tell() { return static_cast<size_t>(src_ - head_); }
// Write // Write
Ch* PutBegin() { return dst_ = src_; }
void Put(Ch c) { RAPIDJSON_ASSERT(dst_ != 0); *dst_++ = c; } void Put(Ch c) { RAPIDJSON_ASSERT(dst_ != 0); *dst_++ = c; }
void Flush() {}
Ch* PutBegin() { return dst_ = src_; }
size_t PutEnd(Ch* begin) { return static_cast<size_t>(dst_ - begin); } size_t PutEnd(Ch* begin) { return static_cast<size_t>(dst_ - begin); }
void Flush() {}
Ch* Push(size_t count) { Ch* begin = dst_; dst_ += count; return begin; }
void Pop(size_t count) { dst_ -= count; }
Ch* src_; Ch* src_;
Ch* dst_; Ch* dst_;
......
include/rapidjson/stringbuffer.h
View file @ c4ce48cd
...@@ -22,6 +22,8 @@ struct GenericStringBuffer { ...@@ -22,6 +22,8 @@ struct GenericStringBuffer {
void Flush() {} void Flush() {}
void Clear() { stack_.Clear(); } void Clear() { stack_.Clear(); }
Ch* Push(size_t count) { return stack_.template Push<Ch>(count); }
void Pop(size_t count) { stack_.template Pop<Ch>(count); }
const Ch* GetString() const { const Ch* GetString() const {
// Push and pop a null terminator. This is safe. // Push and pop a null terminator. This is safe.
......
include/rapidjson/writer.h
View file @ c4ce48cd
...@@ -4,6 +4,8 @@ ...@@ -4,6 +4,8 @@
#include "rapidjson.h" #include "rapidjson.h"
#include "internal/stack.h" #include "internal/stack.h"
#include "internal/strfunc.h" #include "internal/strfunc.h"
#include "internal/itoa.h"
#include "stringbuffer.h"
#include <cstdio> // snprintf() or _sprintf_s() #include <cstdio> // snprintf() or _sprintf_s()
#include <new> // placement new #include <new> // placement new
...@@ -44,6 +46,10 @@ public: ...@@ -44,6 +46,10 @@ public:
os_(&os), level_stack_(allocator, levelDepth * sizeof(Level)), os_(&os), level_stack_(allocator, levelDepth * sizeof(Level)),
doublePrecision_(kDefaultDoublePrecision), hasRoot_(false) {} doublePrecision_(kDefaultDoublePrecision), hasRoot_(false) {}
Writer(Allocator* allocator = 0, size_t levelDepth = kDefaultLevelDepth) :
os_(0), level_stack_(allocator, levelDepth * sizeof(Level)),
doublePrecision_(kDefaultDoublePrecision), hasRoot_(false) {}
//! Reset the writer with a new stream. //! Reset the writer with a new stream.
/*! /*!
This function reset the writer with a new stream and default settings, This function reset the writer with a new stream and default settings,
...@@ -208,49 +214,34 @@ protected: ...@@ -208,49 +214,34 @@ protected:
} }
bool WriteInt(int i) { bool WriteInt(int i) {
if (i < 0) { char buffer[11];
os_->Put('-'); const char* end = internal::i32toa(i, buffer);
i = -i; for (const char* p = buffer; p != end; ++p)
} os_->Put(*p);
return WriteUint((unsigned)i); return true;
} }
bool WriteUint(unsigned u) { bool WriteUint(unsigned u) {
char buffer[10]; char buffer[10];
char *p = buffer; const char* end = internal::u32toa(u, buffer);
do { for (const char* p = buffer; p != end; ++p)
*p++ = char(u % 10) + '0';
u /= 10;
} while (u > 0);
do {
--p;
os_->Put(*p); os_->Put(*p);
} while (p != buffer);
return true; return true;
} }
bool WriteInt64(int64_t i64) { bool WriteInt64(int64_t i64) {
if (i64 < 0) { char buffer[21];
os_->Put('-'); const char* end = internal::i64toa(i64, buffer);
i64 = -i64; for (const char* p = buffer; p != end; ++p)
} os_->Put(*p);
WriteUint64((uint64_t)i64);
return true; return true;
} }
bool WriteUint64(uint64_t u64) { bool WriteUint64(uint64_t u64) {
char buffer[20]; char buffer[20];
char *p = buffer; const char* end = internal::u64toa(u64, buffer);
do { for (const char* p = buffer; p != end; ++p)
*p++ = char(u64 % 10) + '0';
u64 /= 10;
} while (u64 > 0);
do {
--p;
os_->Put(*p); os_->Put(*p);
} while (p != buffer);
return true; return true;
} }
...@@ -378,6 +369,40 @@ private: ...@@ -378,6 +369,40 @@ private:
Writer& operator=(const Writer&); Writer& operator=(const Writer&);
}; };
// Full specialization for StringStream to prevent memory copying
template<>
inline bool Writer<StringBuffer>::WriteInt(int i) {
char *buffer = os_->Push(11);
const char* end = internal::i32toa(i, buffer);
os_->Pop(11 - (end - buffer));
return true;
}
template<>
inline bool Writer<StringBuffer>::WriteUint(unsigned u) {
char *buffer = os_->Push(10);
const char* end = internal::u32toa(u, buffer);
os_->Pop(10 - (end - buffer));
return true;
}
template<>
inline bool Writer<StringBuffer>::WriteInt64(int64_t i64) {
char *buffer = os_->Push(21);
const char* end = internal::i64toa(i64, buffer);
os_->Pop(21 - (end - buffer));
return true;
}
template<>
inline bool Writer<StringBuffer>::WriteUint64(uint64_t u) {
char *buffer = os_->Push(20);
const char* end = internal::u64toa(u, buffer);
os_->Pop(20 - (end - buffer));
return true;
}
} // namespace rapidjson } // namespace rapidjson
#ifdef _MSC_VER #ifdef _MSC_VER
......
test/perftest/misctest.cpp
View file @ c4ce48cd
This diff is collapsed.
thirdparty/cppformat/format.cc 0 → 100644
View file @ c4ce48cd
This diff is collapsed.
thirdparty/cppformat/format.h 0 → 100644
View file @ c4ce48cd
This diff is collapsed.
thirdparty/strtk/strtk.hpp 0 → 100644
View file @ c4ce48cd
This diff is collapsed.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment