aboutsummaryrefslogtreecommitdiff
path: root/NorthstarDLL/core/math
diff options
context:
space:
mode:
authorEmma Miler <emma.pi@protonmail.com>2022-12-19 19:32:16 +0100
committerGitHub <noreply@github.com>2022-12-19 19:32:16 +0100
commite04f3b36accccb590a2d51b4829256b9964ac3fd (patch)
tree20ee30c82e6f53e6e772be2e1b9613eebca12bf3 /NorthstarDLL/core/math
parent33f18a735986dcd136bf8ba70ad8331306c28227 (diff)
downloadNorthstarLauncher-e04f3b36accccb590a2d51b4829256b9964ac3fd.tar.gz
NorthstarLauncher-e04f3b36accccb590a2d51b4829256b9964ac3fd.zip
Restructuring (#365)
* Remove launcher proxy * Restructuring * More restructuring * Fix include dirs * Fix merge * Remove clang thing * Filters * Oops
Diffstat (limited to 'NorthstarDLL/core/math')
-rw-r--r--NorthstarDLL/core/math/bitbuf.h1148
-rw-r--r--NorthstarDLL/core/math/bits.cpp46
-rw-r--r--NorthstarDLL/core/math/bits.h10
-rw-r--r--NorthstarDLL/core/math/color.cpp26
-rw-r--r--NorthstarDLL/core/math/color.h197
-rw-r--r--NorthstarDLL/core/math/vector.h52
6 files changed, 1479 insertions, 0 deletions
diff --git a/NorthstarDLL/core/math/bitbuf.h b/NorthstarDLL/core/math/bitbuf.h
new file mode 100644
index 00000000..8e8e216f
--- /dev/null
+++ b/NorthstarDLL/core/math/bitbuf.h
@@ -0,0 +1,1148 @@
+#pragma once
+
+#define INLINE inline
+
+#define BITS_PER_INT 32
+
+INLINE int GetBitForBitnum(int bitNum)
+{
+ static int bitsForBitnum[] = {
+ (1 << 0), (1 << 1), (1 << 2), (1 << 3), (1 << 4), (1 << 5), (1 << 6), (1 << 7), (1 << 8), (1 << 9), (1 << 10),
+ (1 << 11), (1 << 12), (1 << 13), (1 << 14), (1 << 15), (1 << 16), (1 << 17), (1 << 18), (1 << 19), (1 << 20), (1 << 21),
+ (1 << 22), (1 << 23), (1 << 24), (1 << 25), (1 << 26), (1 << 27), (1 << 28), (1 << 29), (1 << 30), (1 << 31),
+ };
+
+ return bitsForBitnum[(bitNum) & (BITS_PER_INT - 1)];
+}
+
+#undef BITS_PER_INT
+
+using u8 = uint8_t;
+using u16 = uint16_t;
+using u32 = uint32_t;
+using u64 = uint64_t;
+using uptr = uintptr_t;
+
+using i8 = int8_t;
+using i16 = int16_t;
+using i32 = int32_t;
+using i64 = int64_t;
+using iptr = intptr_t;
+
+// Endianess, don't use on PPC64 nor ARM64BE
+#define LittleDWord(val) (val)
+
+static INLINE void StoreLittleDWord(u32* base, size_t dwordIndex, u32 dword)
+{
+ base[dwordIndex] = LittleDWord(dword);
+}
+
+static INLINE u32 LoadLittleDWord(u32* base, size_t dwordIndex)
+{
+ return LittleDWord(base[dwordIndex]);
+}
+
+#include <algorithm>
+
+static inline const u32 s_nMaskTable[33] = {
+ 0,
+ (1 << 1) - 1,
+ (1 << 2) - 1,
+ (1 << 3) - 1,
+ (1 << 4) - 1,
+ (1 << 5) - 1,
+ (1 << 6) - 1,
+ (1 << 7) - 1,
+ (1 << 8) - 1,
+ (1 << 9) - 1,
+ (1 << 10) - 1,
+ (1 << 11) - 1,
+ (1 << 12) - 1,
+ (1 << 13) - 1,
+ (1 << 14) - 1,
+ (1 << 15) - 1,
+ (1 << 16) - 1,
+ (1 << 17) - 1,
+ (1 << 18) - 1,
+ (1 << 19) - 1,
+ (1 << 20) - 1,
+ (1 << 21) - 1,
+ (1 << 22) - 1,
+ (1 << 23) - 1,
+ (1 << 24) - 1,
+ (1 << 25) - 1,
+ (1 << 26) - 1,
+ (1 << 27) - 1,
+ (1 << 28) - 1,
+ (1 << 29) - 1,
+ (1 << 30) - 1,
+ 0x7fffffff,
+ 0xffffffff,
+};
+
+enum EBitCoordType
+{
+ kCW_None,
+ kCW_LowPrecision,
+ kCW_Integral
+};
+
+class BitBufferBase
+{
+ protected:
+ INLINE void SetName(const char* name)
+ {
+ m_BufferName = name;
+ }
+
+ public:
+ INLINE bool IsOverflowed()
+ {
+ return m_Overflow;
+ }
+ INLINE void SetOverflowed()
+ {
+ m_Overflow = true;
+ }
+
+ INLINE const char* GetName()
+ {
+ return m_BufferName;
+ }
+
+ private:
+ const char* m_BufferName = "";
+
+ protected:
+ u8 m_Overflow = false;
+};
+
+class BFRead : public BitBufferBase
+{
+ public:
+ BFRead() = default;
+
+ INLINE BFRead(uptr data, size_t byteLength, size_t startPos = 0, const char* bufferName = 0)
+ {
+ StartReading(data, byteLength, startPos);
+
+ if (bufferName)
+ SetName(bufferName);
+ }
+
+ public:
+ INLINE void StartReading(uptr data, size_t byteLength, size_t startPos = 0)
+ {
+ m_Data = reinterpret_cast<u32 const*>(data);
+ m_DataIn = m_Data;
+
+ m_DataBytes = byteLength;
+ m_DataBits = byteLength << 3;
+
+ m_DataEnd = reinterpret_cast<u32 const*>(reinterpret_cast<u8 const*>(m_Data) + m_DataBytes);
+
+ Seek(startPos);
+ }
+
+ INLINE void GrabNextDWord(bool overflow = false)
+ {
+ if (m_Data == m_DataEnd)
+ {
+ m_CachedBitsLeft = 1;
+ m_CachedBufWord = 0;
+
+ m_DataIn++;
+
+ if (overflow)
+ SetOverflowed();
+ }
+ else
+ {
+ if (m_DataIn > m_DataEnd)
+ {
+ SetOverflowed();
+ m_CachedBufWord = 0;
+ }
+ else
+ {
+ m_CachedBufWord = LittleDWord(*(m_DataIn++));
+ }
+ }
+ }
+
+ INLINE void FetchNext()
+ {
+ m_CachedBitsLeft = 32;
+ GrabNextDWord(false);
+ }
+
+ INLINE i32 ReadOneBit()
+ {
+ i32 ret = m_CachedBufWord & 1;
+
+ if (--m_CachedBitsLeft == 0)
+ FetchNext();
+ else
+ m_CachedBufWord >>= 1;
+
+ return ret;
+ }
+
+ INLINE u32 ReadUBitLong(i32 numBits)
+ {
+ if (m_CachedBitsLeft >= numBits)
+ {
+ u32 ret = m_CachedBufWord & s_nMaskTable[numBits];
+
+ m_CachedBitsLeft -= numBits;
+
+ if (m_CachedBitsLeft)
+ m_CachedBufWord >>= numBits;
+ else
+ FetchNext();
+
+ return ret;
+ }
+ else
+ {
+ // need to merge words
+ u32 ret = m_CachedBufWord;
+ numBits -= m_CachedBitsLeft;
+
+ GrabNextDWord(true);
+
+ if (IsOverflowed())
+ return 0;
+
+ ret |= ((m_CachedBufWord & s_nMaskTable[numBits]) << m_CachedBitsLeft);
+
+ m_CachedBitsLeft = 32 - numBits;
+ m_CachedBufWord >>= numBits;
+
+ return ret;
+ }
+ }
+
+ INLINE i32 ReadSBitLong(int numBits)
+ {
+ i32 ret = ReadUBitLong(numBits);
+ return (ret << (32 - numBits)) >> (32 - numBits);
+ }
+
+ INLINE u32 ReadUBitVar()
+ {
+ u32 ret = ReadUBitLong(6);
+
+ switch (ret & (16 | 32))
+ {
+ case 16:
+ ret = (ret & 15) | (ReadUBitLong(4) << 4);
+ // Assert(ret >= 16);
+ break;
+ case 32:
+ ret = (ret & 15) | (ReadUBitLong(8) << 4);
+ // Assert(ret >= 256);
+ break;
+ case 48:
+ ret = (ret & 15) | (ReadUBitLong(32 - 4) << 4);
+ // Assert(ret >= 4096);
+ break;
+ }
+
+ return ret;
+ }
+
+ INLINE u32 PeekUBitLong(i32 numBits)
+ {
+ i32 nSaveBA = m_CachedBitsLeft;
+ i32 nSaveW = m_CachedBufWord;
+ u32 const* pSaveP = m_DataIn;
+ u32 nRet = ReadUBitLong(numBits);
+
+ m_CachedBitsLeft = nSaveBA;
+ m_CachedBufWord = nSaveW;
+ m_DataIn = pSaveP;
+
+ return nRet;
+ }
+
+ INLINE float ReadBitFloat()
+ {
+ u32 value = ReadUBitLong(32);
+ return *reinterpret_cast<float*>(&value);
+ }
+
+ /*INLINE float ReadBitCoord() {
+ i32 intval = 0, fractval = 0, signbit = 0;
+ float value = 0.0;
+
+ // Read the required integer and fraction flags
+ intval = ReadOneBit();
+ fractval = ReadOneBit();
+
+ // If we got either parse them, otherwise it's a zero.
+ if (intval || fractval) {
+ // Read the sign bit
+ signbit = ReadOneBit();
+
+ // If there's an integer, read it in
+ if (intval) {
+ // Adjust the integers from [0..MAX_COORD_VALUE-1] to [1..MAX_COORD_VALUE]
+ intval = ReadUBitLong(COORD_INTEGER_BITS) + 1;
+ }
+
+ // If there's a fraction, read it in
+ if (fractval) {
+ fractval = ReadUBitLong(COORD_FRACTIONAL_BITS);
+ }
+
+ // Calculate the correct floating point value
+ value = intval + ((float)fractval * COORD_RESOLUTION);
+
+ // Fixup the sign if negative.
+ if (signbit)
+ value = -value;
+ }
+
+ return value;
+ }
+
+ INLINE float ReadBitCoordMP() {
+ i32 intval = 0, fractval = 0, signbit = 0;
+ float value = 0.0;
+
+ bool inBounds = ReadOneBit() ? true : false;
+
+ // Read the required integer and fraction flags
+ intval = ReadOneBit();
+
+ // If we got either parse them, otherwise it's a zero.
+ if (intval) {
+ // Read the sign bit
+ signbit = ReadOneBit();
+
+ // If there's an integer, read it in
+ // Adjust the integers from [0..MAX_COORD_VALUE-1] to [1..MAX_COORD_VALUE]
+ if (inBounds)
+ value = ReadUBitLong(COORD_INTEGER_BITS_MP) + 1;
+ else
+ value = ReadUBitLong(COORD_INTEGER_BITS) + 1;
+ }
+
+ // Fixup the sign if negative.
+ if (signbit)
+ value = -value;
+
+ return value;
+ }
+
+ INLINE float ReadBitCellCoord(int bits, EBitCoordType coordType) {
+ bool bIntegral = (coordType == kCW_Integral);
+ bool bLowPrecision = (coordType == kCW_LowPrecision);
+
+ int intval = 0, fractval = 0;
+ float value = 0.0;
+
+ if (bIntegral)
+ value = ReadUBitLong(bits);
+ else {
+ intval = ReadUBitLong(bits);
+
+ // If there's a fraction, read it in
+ fractval = ReadUBitLong(bLowPrecision ? COORD_FRACTIONAL_BITS_MP_LOWPRECISION : COORD_FRACTIONAL_BITS);
+
+ // Calculate the correct floating point value
+ value = intval + ((float)fractval * (bLowPrecision ? COORD_RESOLUTION_LOWPRECISION : COORD_RESOLUTION));
+ }
+
+ return value;
+ }
+
+ INLINE float ReadBitNormal() {
+ // Read the sign bit
+ i32 signbit = ReadOneBit();
+
+ // Read the fractional part
+ u32 fractval = ReadUBitLong(NORMAL_FRACTIONAL_BITS);
+
+ // Calculate the correct floating point value
+ float value = (float)fractval * NORMAL_RESOLUTION;
+
+ // Fixup the sign if negative.
+ if (signbit)
+ value = -value;
+
+ return value;
+ }
+
+ INLINE void ReadBitVec3Coord(Vector& fa) {
+ i32 xflag, yflag, zflag;
+
+ // This vector must be initialized! Otherwise, If any of the flags aren't set,
+ // the corresponding component will not be read and will be stack garbage.
+ fa.Init(0, 0, 0);
+
+ xflag = ReadOneBit();
+ yflag = ReadOneBit();
+ zflag = ReadOneBit();
+
+ if (xflag)
+ fa[0] = ReadBitCoord();
+ if (yflag)
+ fa[1] = ReadBitCoord();
+ if (zflag)
+ fa[2] = ReadBitCoord();
+ }
+
+ INLINE void ReadBitVec3Normal(Vector& fa) {
+ i32 xflag = ReadOneBit();
+ i32 yflag = ReadOneBit();
+
+ if (xflag)
+ fa[0] = ReadBitNormal();
+ else
+ fa[0] = 0.0f;
+
+ if (yflag)
+ fa[1] = ReadBitNormal();
+ else
+ fa[1] = 0.0f;
+
+ // The first two imply the third (but not its sign)
+ i32 znegative = ReadOneBit();
+
+ float fafafbfb = fa[0] * fa[0] + fa[1] * fa[1];
+ if (fafafbfb < 1.0f)
+ fa[2] = sqrt(1.0f - fafafbfb);
+ else
+ fa[2] = 0.0f;
+
+ if (znegative)
+ fa[2] = -fa[2];
+ }
+
+ INLINE void ReadBitAngles(QAngle& fa) {
+ Vector tmp;
+ ReadBitVec3Coord(tmp);
+ fa.Init(tmp.x, tmp.y, tmp.z);
+ }*/
+
+ INLINE float ReadBitAngle(int numBits)
+ {
+ float shift = (float)(GetBitForBitnum(numBits));
+
+ i32 i = ReadUBitLong(numBits);
+ float fReturn = (float)i * (360.0 / shift);
+
+ return fReturn;
+ }
+
+ INLINE i32 ReadChar()
+ {
+ return ReadSBitLong(sizeof(char) << 3);
+ }
+ INLINE u32 ReadByte()
+ {
+ return ReadUBitLong(sizeof(unsigned char) << 3);
+ }
+
+ INLINE i32 ReadShort()
+ {
+ return ReadSBitLong(sizeof(short) << 3);
+ }
+ INLINE u32 ReadWord()
+ {
+ return ReadUBitLong(sizeof(unsigned short) << 3);
+ }
+
+ INLINE i32 ReadLong()
+ {
+ return (i32)(ReadUBitLong(sizeof(i32) << 3));
+ }
+ INLINE float ReadFloat()
+ {
+ u32 temp = ReadUBitLong(sizeof(float) << 3);
+ return *reinterpret_cast<float*>(&temp);
+ }
+
+ INLINE u32 ReadVarInt32()
+ {
+ constexpr int kMaxVarint32Bytes = 5;
+
+ u32 result = 0;
+ int count = 0;
+ u32 b;
+
+ do
+ {
+ if (count == kMaxVarint32Bytes)
+ return result;
+
+ b = ReadUBitLong(8);
+ result |= (b & 0x7F) << (7 * count);
+ ++count;
+ } while (b & 0x80);
+
+ return result;
+ }
+
+ INLINE u64 ReadVarInt64()
+ {
+ constexpr int kMaxVarintBytes = 10;
+
+ u64 result = 0;
+ int count = 0;
+ u64 b;
+
+ do
+ {
+ if (count == kMaxVarintBytes)
+ return result;
+
+ b = ReadUBitLong(8);
+ result |= static_cast<u64>(b & 0x7F) << (7 * count);
+ ++count;
+ } while (b & 0x80);
+
+ return result;
+ }
+
+ INLINE void ReadBits(uptr outData, u32 bitLength)
+ {
+ u8* out = reinterpret_cast<u8*>(outData);
+ int bitsLeft = bitLength;
+
+ // align output to dword boundary
+ while (((uptr)out & 3) != 0 && bitsLeft >= 8)
+ {
+ *out = (unsigned char)ReadUBitLong(8);
+ ++out;
+ bitsLeft -= 8;
+ }
+
+ // read dwords
+ while (bitsLeft >= 32)
+ {
+ *((u32*)out) = ReadUBitLong(32);
+ out += sizeof(u32);
+ bitsLeft -= 32;
+ }
+
+ // read remaining bytes
+ while (bitsLeft >= 8)
+ {
+ *out = ReadUBitLong(8);
+ ++out;
+ bitsLeft -= 8;
+ }
+
+ // read remaining bits
+ if (bitsLeft)
+ *out = ReadUBitLong(bitsLeft);
+ }
+
+ INLINE bool ReadBytes(uptr outData, u32 byteLength)
+ {
+ ReadBits(outData, byteLength << 3);
+ return !IsOverflowed();
+ }
+
+ INLINE bool ReadString(char* str, i32 maxLength, bool stopAtLineTermination = false, i32* outNumChars = 0)
+ {
+ bool tooSmall = false;
+ int iChar = 0;
+
+ while (1)
+ {
+ char val = ReadChar();
+
+ if (val == 0)
+ break;
+ else if (stopAtLineTermination && val == '\n')
+ break;
+
+ if (iChar < (maxLength - 1))
+ {
+ str[iChar] = val;
+ ++iChar;
+ }
+ else
+ {
+ tooSmall = true;
+ }
+ }
+
+ // Make sure it's null-terminated.
+ // Assert(iChar < maxLength);
+ str[iChar] = 0;
+
+ if (outNumChars)
+ *outNumChars = iChar;
+
+ return !IsOverflowed() && !tooSmall;
+ }
+
+ INLINE char* ReadAndAllocateString(bool* hasOverflowed = 0)
+ {
+ char str[2048];
+
+ int chars = 0;
+ bool overflowed = !ReadString(str, sizeof(str), false, &chars);
+
+ if (hasOverflowed)
+ *hasOverflowed = overflowed;
+
+ // Now copy into the output and return it;
+ char* ret = new char[chars + 1];
+ for (u32 i = 0; i <= chars; i++)
+ ret[i] = str[i];
+
+ return ret;
+ }
+
+ INLINE i64 ReadLongLong()
+ {
+ i64 retval;
+ u32* longs = (u32*)&retval;
+
+ // Read the two DWORDs according to network endian
+ const short endianIndex = 0x0100;
+ u8* idx = (u8*)&endianIndex;
+
+ longs[*idx++] = ReadUBitLong(sizeof(i32) << 3);
+ longs[*idx] = ReadUBitLong(sizeof(i32) << 3);
+
+ return retval;
+ }
+
+ INLINE bool Seek(size_t startPos)
+ {
+ bool bSucc = true;
+
+ if (startPos < 0 || startPos > m_DataBits)
+ {
+ SetOverflowed();
+ bSucc = false;
+ startPos = m_DataBits;
+ }
+
+ // non-multiple-of-4 bytes at head of buffer. We put the "round off"
+ // at the head to make reading and detecting the end efficient.
+ int nHead = m_DataBytes & 3;
+
+ int posBytes = startPos / 8;
+ if ((m_DataBytes < 4) || (nHead && (posBytes < nHead)))
+ {
+ // partial first dword
+ u8 const* partial = (u8 const*)m_Data;
+
+ if (m_Data)
+ {
+ m_CachedBufWord = *(partial++);
+ if (nHead > 1)
+ m_CachedBufWord |= (*partial++) << 8;
+ if (nHead > 2)
+ m_CachedBufWord |= (*partial++) << 16;
+ }
+
+ m_DataIn = (u32 const*)partial;
+
+ m_CachedBufWord >>= (startPos & 31);
+ m_CachedBitsLeft = (nHead << 3) - (startPos & 31);
+ }
+ else
+ {
+ int adjustedPos = startPos - (nHead << 3);
+
+ m_DataIn = reinterpret_cast<u32 const*>(reinterpret_cast<u8 const*>(m_Data) + ((adjustedPos / 32) << 2) + nHead);
+
+ if (m_Data)
+ {
+ m_CachedBitsLeft = 32;
+ GrabNextDWord();
+ }
+ else
+ {
+ m_CachedBufWord = 0;
+ m_CachedBitsLeft = 1;
+ }
+
+ m_CachedBufWord >>= (adjustedPos & 31);
+ m_CachedBitsLeft = std::min(m_CachedBitsLeft, u32(32 - (adjustedPos & 31))); // in case grabnextdword overflowed
+ }
+
+ return bSucc;
+ }
+
+ INLINE size_t GetNumBitsRead()
+ {
+ if (!m_Data)
+ return 0;
+
+ size_t nCurOfs = size_t(((iptr(m_DataIn) - iptr(m_Data)) / 4) - 1);
+ nCurOfs *= 32;
+ nCurOfs += (32 - m_CachedBitsLeft);
+
+ size_t nAdjust = 8 * (m_DataBytes & 3);
+ return std::min(nCurOfs + nAdjust, m_DataBits);
+ }
+
+ INLINE bool SeekRelative(size_t offset)
+ {
+ return Seek(GetNumBitsRead() + offset);
+ }
+
+ INLINE size_t TotalBytesAvailable()
+ {
+ return m_DataBytes;
+ }
+
+ INLINE size_t GetNumBitsLeft()
+ {
+ return m_DataBits - GetNumBitsRead();
+ }
+ INLINE size_t GetNumBytesLeft()
+ {
+ return GetNumBitsLeft() >> 3;
+ }
+
+ private:
+ size_t m_DataBits; // 0x0010
+ size_t m_DataBytes; // 0x0018
+
+ u32 m_CachedBufWord; // 0x0020
+ u32 m_CachedBitsLeft; // 0x0024
+
+ const u32* m_DataIn; // 0x0028
+ const u32* m_DataEnd; // 0x0030
+ const u32* m_Data; // 0x0038
+};
+
+class BFWrite : public BitBufferBase
+{
+ public:
+ BFWrite() = default;
+
+ INLINE BFWrite(uptr data, size_t byteLength, const char* bufferName = 0)
+ {
+ StartWriting(data, byteLength);
+
+ if (bufferName)
+ SetName(bufferName);
+ }
+
+ public:
+ INLINE void StartWriting(uptr data, size_t byteLength)
+ {
+ m_Data = reinterpret_cast<u32*>(data);
+ m_DataOut = m_Data;
+
+ m_DataBytes = byteLength;
+ m_DataBits = byteLength << 3;
+
+ m_DataEnd = reinterpret_cast<u32*>(reinterpret_cast<u8*>(m_Data) + m_DataBytes);
+ }
+
+ INLINE int GetNumBitsLeft()
+ {
+ return m_OutBitsLeft + (32 * (m_DataEnd - m_DataOut - 1));
+ }
+
+ INLINE void Reset()
+ {
+ m_Overflow = false;
+ m_OutBufWord = 0;
+ m_OutBitsLeft = 32;
+ m_DataOut = m_Data;
+ }
+
+ INLINE void TempFlush()
+ {
+ if (m_OutBitsLeft != 32)
+ {
+ if (m_DataOut == m_DataEnd)
+ SetOverflowed();
+ else
+ StoreLittleDWord(m_DataOut, 0, LoadLittleDWord(m_DataOut, 0) & ~s_nMaskTable[32 - m_OutBitsLeft] | m_OutBufWord);
+ }
+
+ m_Flushed = true;
+ }
+
+ INLINE u8* GetBasePointer()
+ {
+ TempFlush();
+ return reinterpret_cast<u8*>(m_Data);
+ }
+
+ INLINE u8* GetData()
+ {
+ return GetBasePointer();
+ }
+
+ INLINE void Finish()
+ {
+ if (m_OutBitsLeft != 32)
+ {
+ if (m_DataOut == m_DataEnd)
+ SetOverflowed();
+
+ StoreLittleDWord(m_DataOut, 0, m_OutBufWord);
+ }
+ }
+
+ INLINE void FlushNoCheck()
+ {
+ StoreLittleDWord(m_DataOut++, 0, m_OutBufWord);
+
+ m_OutBitsLeft = 32;
+ m_OutBufWord = 0;
+ }
+
+ INLINE void Flush()
+ {
+ if (m_DataOut == m_DataEnd)
+ SetOverflowed();
+ else
+ StoreLittleDWord(m_DataOut++, 0, m_OutBufWord);
+
+ m_OutBitsLeft = 32;
+ m_OutBufWord = 0;
+ }
+
+ INLINE void WriteOneBitNoCheck(i32 value)
+ {
+ m_OutBufWord |= (value & 1) << (32 - m_OutBitsLeft);
+
+ if (--m_OutBitsLeft == 0)
+ FlushNoCheck();
+ }
+
+ INLINE void WriteOneBit(i32 value)
+ {
+ m_OutBufWord |= (value & 1) << (32 - m_OutBitsLeft);
+
+ if (--m_OutBitsLeft == 0)
+ Flush();
+ }
+
+ INLINE void WriteUBitLong(u32 data, i32 numBits, bool checkRange = true)
+ {
+ if (numBits <= m_OutBitsLeft)
+ {
+ if (checkRange)
+ m_OutBufWord |= (data) << (32 - m_OutBitsLeft);
+ else
+ m_OutBufWord |= (data & s_nMaskTable[numBits]) << (32 - m_OutBitsLeft);
+
+ m_OutBitsLeft -= numBits;
+
+ if (m_OutBitsLeft == 0)
+ Flush();
+ }
+ else
+ {
+ // split dwords case
+ i32 overflowBits = (numBits - m_OutBitsLeft);
+ m_OutBufWord |= (data & s_nMaskTable[m_OutBitsLeft]) << (32 - m_OutBitsLeft);
+ Flush();
+ m_OutBufWord = (data >> (numBits - overflowBits));
+ m_OutBitsLeft = 32 - overflowBits;
+ }
+ }
+
+ INLINE void WriteSBitLong(i32 data, i32 numBits)
+ {
+ WriteUBitLong((u32)data, numBits, false);
+ }
+
+ INLINE void WriteUBitVar(u32 n)
+ {
+ if (n < 16)
+ WriteUBitLong(n, 6);
+ else if (n < 256)
+ WriteUBitLong((n & 15) | 16 | ((n & (128 | 64 | 32 | 16)) << 2), 10);
+ else if (n < 4096)
+ WriteUBitLong((n & 15) | 32 | ((n & (2048 | 1024 | 512 | 256 | 128 | 64 | 32 | 16)) << 2), 14);
+ else
+ {
+ WriteUBitLong((n & 15) | 48, 6);
+ WriteUBitLong((n >> 4), 32 - 4);
+ }
+ }
+
+ INLINE void WriteBitFloat(float value)
+ {
+ auto temp = &value;
+ WriteUBitLong(*reinterpret_cast<u32*>(temp), 32);
+ }
+
+ INLINE void WriteFloat(float value)
+ {
+ auto temp = &value;
+ WriteUBitLong(*reinterpret_cast<u32*>(temp), 32);
+ }
+
+ INLINE bool WriteBits(const uptr data, i32 numBits)
+ {
+ u8* out = (u8*)data;
+ i32 numBitsLeft = numBits;
+
+ // Bounds checking..
+ if ((GetNumBitsWritten() + numBits) > m_DataBits)
+ {
+ SetOverflowed();
+ return false;
+ }
+
+ // !! speed!! need fast paths
+ // write remaining bytes
+ while (numBitsLeft >= 8)
+ {
+ WriteUBitLong(*out, 8, false);
+ ++out;
+ numBitsLeft -= 8;
+ }
+
+ // write remaining bits
+ if (numBitsLeft)
+ WriteUBitLong(*out, numBitsLeft, false);
+
+ return !IsOverflowed();
+ }
+
+ INLINE bool WriteBytes(const uptr data, i32 numBytes)
+ {
+ return WriteBits(data, numBytes << 3);
+ }
+
+ INLINE i32 GetNumBitsWritten()
+ {
+ return (32 - m_OutBitsLeft) + (32 * (m_DataOut - m_Data));
+ }
+
+ INLINE i32 GetNumBytesWritten()
+ {
+ return (GetNumBitsWritten() + 7) >> 3;
+ }
+
+ INLINE void WriteChar(i32 val)
+ {
+ WriteSBitLong(val, sizeof(char) << 3);
+ }
+
+ INLINE void WriteByte(i32 val)
+ {
+ WriteUBitLong(val, sizeof(unsigned char) << 3, false);
+ }
+
+ INLINE void WriteShort(i32 val)
+ {
+ WriteSBitLong(val, sizeof(short) << 3);
+ }
+
+ INLINE void WriteWord(i32 val)
+ {
+ WriteUBitLong(val, sizeof(unsigned short) << 3);
+ }
+
+ INLINE bool WriteString(const char* str)
+ {
+ if (str)
+ while (*str)
+ WriteChar(*(str++));
+
+ WriteChar(0);
+
+ return !IsOverflowed();
+ }
+
+ INLINE void WriteLongLong(i64 val)
+ {
+ u32* pLongs = (u32*)&val;
+
+ // Insert the two DWORDS according to network endian
+ const short endianIndex = 0x0100;
+ u8* idx = (u8*)&endianIndex;
+
+ WriteUBitLong(pLongs[*idx++], sizeof(i32) << 3);
+ WriteUBitLong(pLongs[*idx], sizeof(i32) << 3);
+ }
+
+ /*INLINE void WriteBitCoord(const float f) {
+ i32 signbit = (f <= -COORD_RESOLUTION);
+ i32 intval = (i32)abs(f);
+ i32 fractval = abs((i32)(f * COORD_DENOMINATOR)) & (COORD_DENOMINATOR - 1);
+
+ // Send the bit flags that indicate whether we have an integer part and/or a fraction part.
+ WriteOneBit(intval);
+ WriteOneBit(fractval);
+
+ if (intval || fractval) {
+ // Send the sign bit
+ WriteOneBit(signbit);
+
+ // Send the integer if we have one.
+ if (intval) {
+ // Adjust the integers from [1..MAX_COORD_VALUE] to [0..MAX_COORD_VALUE-1]
+ intval--;
+ WriteUBitLong((u32)intval, COORD_INTEGER_BITS);
+ }
+
+ // Send the fraction if we have one
+ if (fractval) {
+ WriteUBitLong((u32)fractval, COORD_FRACTIONAL_BITS);
+ }
+ }
+ }
+
+ INLINE void WriteBitCoordMP(const float f) {
+ i32 signbit = (f <= -COORD_RESOLUTION);
+ i32 intval = (i32)abs(f);
+ i32 fractval = (abs((i32)(f * COORD_DENOMINATOR)) & (COORD_DENOMINATOR - 1));
+
+ bool bInBounds = intval < (1 << COORD_INTEGER_BITS_MP);
+
+ WriteOneBit(bInBounds);
+
+ // Send the sign bit
+ WriteOneBit(intval);
+
+ if (intval) {
+ WriteOneBit(signbit);
+
+ // Send the integer if we have one.
+ // Adjust the integers from [1..MAX_COORD_VALUE] to [0..MAX_COORD_VALUE-1]
+ intval--;
+
+ if (bInBounds)
+ WriteUBitLong((u32)intval, COORD_INTEGER_BITS_MP);
+ else
+ WriteUBitLong((u32)intval, COORD_INTEGER_BITS);
+ }
+ }
+
+ INLINE void WriteBitCellCoord(const float f, int bits, EBitCoordType coordType) {
+ bool bIntegral = (coordType == kCW_Integral);
+ bool bLowPrecision = (coordType == kCW_LowPrecision);
+
+ i32 intval = (i32)abs(f);
+ i32 fractval = bLowPrecision ? (abs((i32)(f * COORD_DENOMINATOR_LOWPRECISION)) & (COORD_DENOMINATOR_LOWPRECISION - 1)) :
+ (abs((i32)(f * COORD_DENOMINATOR)) & (COORD_DENOMINATOR - 1));
+
+ if (bIntegral)
+ WriteUBitLong((u32)intval, bits);
+ else {
+ WriteUBitLong((u32)intval, bits);
+ WriteUBitLong((u32)fractval, bLowPrecision ? COORD_FRACTIONAL_BITS_MP_LOWPRECISION : COORD_FRACTIONAL_BITS);
+ }
+ }*/
+
+ INLINE void SeekToBit(int bit)
+ {
+ TempFlush();
+
+ m_DataOut = m_Data + (bit / 32);
+ m_OutBufWord = LoadLittleDWord(m_DataOut, 0);
+ m_OutBitsLeft = 32 - (bit & 31);
+ }
+
+ /*INLINE void WriteBitVec3Coord(const Vector& fa) {
+ i32 xflag, yflag, zflag;
+
+ xflag = (fa[0] >= COORD_RESOLUTION) || (fa[0] <= -COORD_RESOLUTION);
+ yflag = (fa[1] >= COORD_RESOLUTION) || (fa[1] <= -COORD_RESOLUTION);
+ zflag = (fa[2] >= COORD_RESOLUTION) || (fa[2] <= -COORD_RESOLUTION);
+
+ WriteOneBit(xflag);
+ WriteOneBit(yflag);
+ WriteOneBit(zflag);
+
+ if (xflag)
+ WriteBitCoord(fa[0]);
+ if (yflag)
+ WriteBitCoord(fa[1]);
+ if (zflag)
+ WriteBitCoord(fa[2]);
+ }
+
+ INLINE void WriteBitNormal(float f) {
+ i32 signbit = (f <= -NORMAL_RESOLUTION);
+
+ // NOTE: Since +/-1 are valid values for a normal, I'm going to encode that as all ones
+ u32 fractval = abs((i32)(f * NORMAL_DENOMINATOR));
+
+ // clamp..
+ if (fractval > NORMAL_DENOMINATOR)
+ fractval = NORMAL_DENOMINATOR;
+
+ // Send the sign bit
+ WriteOneBit(signbit);
+
+ // Send the fractional component
+ WriteUBitLong(fractval, NORMAL_FRACTIONAL_BITS);
+ }
+
+ INLINE void WriteBitVec3Normal(const Vector& fa) {
+ i32 xflag, yflag;
+
+ xflag = (fa[0] >= NORMAL_RESOLUTION) || (fa[0] <= -NORMAL_RESOLUTION);
+ yflag = (fa[1] >= NORMAL_RESOLUTION) || (fa[1] <= -NORMAL_RESOLUTION);
+
+ WriteOneBit(xflag);
+ WriteOneBit(yflag);
+
+ if (xflag)
+ WriteBitNormal(fa[0]);
+ if (yflag)
+ WriteBitNormal(fa[1]);
+
+ // Write z sign bit
+ i32 signbit = (fa[2] <= -NORMAL_RESOLUTION);
+ WriteOneBit(signbit);
+ }*/
+
+ INLINE void WriteBitAngle(float angle, int numBits)
+ {
+ u32 shift = GetBitForBitnum(numBits);
+ u32 mask = shift - 1;
+
+ i32 d = (i32)((angle / 360.0) * shift);
+ d &= mask;
+
+ WriteUBitLong((u32)d, numBits);
+ }
+
+ INLINE bool WriteBitsFromBuffer(BFRead* in, int numBits)
+ {
+ while (numBits > 32)
+ {
+ WriteUBitLong(in->ReadUBitLong(32), 32);
+ numBits -= 32;
+ }
+
+ WriteUBitLong(in->ReadUBitLong(numBits), numBits);
+ return !IsOverflowed() && !in->IsOverflowed();
+ }
+
+ /*INLINE void WriteBitAngles(const QAngle& fa) {
+ // FIXME:
+ Vector tmp(fa.x, fa.y, fa.z);
+ WriteBitVec3Coord(tmp);
+ }*/
+
+ private:
+ size_t m_DataBits = 0;
+ size_t m_DataBytes = 0;
+
+ u32 m_OutBufWord = 0;
+ u32 m_OutBitsLeft = 32;
+
+ u32* m_DataOut = nullptr;
+ u32* m_DataEnd = nullptr;
+ u32* m_Data = nullptr;
+
+ bool m_Flushed = false; // :flushed:
+};
+
+#undef INLINE
diff --git a/NorthstarDLL/core/math/bits.cpp b/NorthstarDLL/core/math/bits.cpp
new file mode 100644
index 00000000..014899f2
--- /dev/null
+++ b/NorthstarDLL/core/math/bits.cpp
@@ -0,0 +1,46 @@
+//=============================================================================//
+//
+// Purpose: look for NANs, infinities, and underflows.
+//
+//=============================================================================//
+
+#include "pch.h"
+#include "bits.h"
+
+//-----------------------------------------------------------------------------
+// This follows the ANSI/IEEE 754-1985 standard
+//-----------------------------------------------------------------------------
+unsigned long& FloatBits(float& f)
+{
+ return *reinterpret_cast<unsigned long*>(&f);
+}
+
+unsigned long const& FloatBits(float const& f)
+{
+ return *reinterpret_cast<unsigned long const*>(&f);
+}
+
+float BitsToFloat(unsigned long i)
+{
+ return *reinterpret_cast<float*>(&i);
+}
+
+bool IsFinite(float f)
+{
+ return ((FloatBits(f) & 0x7F800000) != 0x7F800000);
+}
+
+unsigned long FloatAbsBits(float f)
+{
+ return FloatBits(f) & 0x7FFFFFFF;
+}
+
+float FloatMakePositive(float f)
+{
+ return fabsf(f);
+}
+
+float FloatNegate(float f)
+{
+ return -f; // BitsToFloat( FloatBits(f) ^ 0x80000000 );
+}
diff --git a/NorthstarDLL/core/math/bits.h b/NorthstarDLL/core/math/bits.h
new file mode 100644
index 00000000..0532a9bd
--- /dev/null
+++ b/NorthstarDLL/core/math/bits.h
@@ -0,0 +1,10 @@
+#pragma once
+
+unsigned long& FloatBits(float& f);
+unsigned long const& FloatBits(float const& f);
+float BitsToFloat(unsigned long i);
+bool IsFinite(float f);
+unsigned long FloatAbsBits(float f);
+
+#define FLOAT32_NAN_BITS (std::uint32_t)0x7FC00000 // NaN!
+#define FLOAT32_NAN BitsToFloat(FLOAT32_NAN_BITS)
diff --git a/NorthstarDLL/core/math/color.cpp b/NorthstarDLL/core/math/color.cpp
new file mode 100644
index 00000000..05cf645f
--- /dev/null
+++ b/NorthstarDLL/core/math/color.cpp
@@ -0,0 +1,26 @@
+#include "pch.h"
+
+// clang-format off
+namespace NS::Colors
+{
+ Color SCRIPT_UI (100, 255, 255);
+ Color SCRIPT_CL (100, 255, 100);
+ Color SCRIPT_SV (255, 100, 255);
+ Color NATIVE_UI (50 , 150, 150);
+ Color NATIVE_CL (50 , 150, 50 );
+ Color NATIVE_SV (150, 50 , 150);
+ Color NATIVE_ENGINE (252, 133, 153);
+ Color FILESYSTEM (0 , 150, 150);
+ Color RPAK (255, 190, 0 );
+ Color NORTHSTAR (66 , 72 , 128);
+ Color ECHO (150, 150, 159);
+
+ Color TRACE (0 , 255, 255);
+ Color DEBUG (0 , 255, 255);
+ Color INFO (16 , 160, 16 );
+ Color WARN (255, 255, 0 );
+ Color ERR (255, 50 , 50 );
+ Color CRIT (255, 0 , 0 );
+ Color OFF (0 , 0 , 0 );
+};
+// clang-format on
diff --git a/NorthstarDLL/core/math/color.h b/NorthstarDLL/core/math/color.h
new file mode 100644
index 00000000..4dc9b36e
--- /dev/null
+++ b/NorthstarDLL/core/math/color.h
@@ -0,0 +1,197 @@
+#pragma once
+
+struct color24
+{
+ uint8_t r, g, b;
+};
+
+typedef struct color32_s
+{
+ bool operator!=(const struct color32_s& other) const
+ {
+ return r != other.r || g != other.g || b != other.b || a != other.a;
+ }
+ inline unsigned* asInt(void)
+ {
+ return reinterpret_cast<unsigned*>(this);
+ }
+ inline const unsigned* asInt(void) const
+ {
+ return reinterpret_cast<const unsigned*>(this);
+ }
+ inline void Copy(const color32_s& rhs)
+ {
+ *asInt() = *rhs.asInt();
+ }
+
+ uint8_t r, g, b, a;
+} color32;
+
+struct SourceColor
+{
+ unsigned char R;
+ unsigned char G;
+ unsigned char B;
+ unsigned char A;
+
+ SourceColor(unsigned char r, unsigned char g, unsigned char b, unsigned char a)
+ {
+ R = r;
+ G = g;
+ B = b;
+ A = a;
+ }
+
+ SourceColor()
+ {
+ R = 0;
+ G = 0;
+ B = 0;
+ A = 0;
+ }
+};
+
+//-----------------------------------------------------------------------------
+// Purpose: Basic handler for an rgb set of colors
+// This class is fully inline
+//-----------------------------------------------------------------------------
+class Color
+{
+ public:
+ Color(int r, int g, int b, int a = 255)
+ {
+ _color[0] = (unsigned char)r;
+ _color[1] = (unsigned char)g;
+ _color[2] = (unsigned char)b;
+ _color[3] = (unsigned char)a;
+ }
+ void SetColor(int _r, int _g, int _b, int _a = 0)
+ {
+ _color[0] = (unsigned char)_r;
+ _color[1] = (unsigned char)_g;
+ _color[2] = (unsigned char)_b;
+ _color[3] = (unsigned char)_a;
+ }
+ void GetColor(int& _r, int& _g, int& _b, int& _a) const
+ {
+ _r = _color[0];
+ _g = _color[1];
+ _b = _color[2];
+ _a = _color[3];
+ }
+ int GetValue(int index) const
+ {
+ return _color[index];
+ }
+ void SetRawColor(int color32)
+ {
+ *((int*)this) = color32;
+ }
+ int GetRawColor(void) const
+ {
+ return *((int*)this);
+ }
+
+ inline int r() const
+ {
+ return _color[0];
+ }
+ inline int g() const
+ {
+ return _color[1];
+ }
+ inline int b() const
+ {
+ return _color[2];
+ }
+ inline int a() const
+ {
+ return _color[3];
+ }
+
+ unsigned char& operator[](int index)
+ {
+ return _color[index];
+ }
+
+ const unsigned char& operator[](int index) const
+ {
+ return _color[index];
+ }
+
+ bool operator==(const Color& rhs) const
+ {
+ return (*((int*)this) == *((int*)&rhs));
+ }
+
+ bool operator!=(const Color& rhs) const
+ {
+ return !(operator==(rhs));
+ }
+
+ Color& operator=(const Color& rhs)
+ {
+ SetRawColor(rhs.GetRawColor());
+ return *this;
+ }
+
+ Color& operator=(const color32& rhs)
+ {
+ _color[0] = rhs.r;
+ _color[1] = rhs.g;
+ _color[2] = rhs.b;
+ _color[3] = rhs.a;
+ return *this;
+ }
+
+ color32 ToColor32(void) const
+ {
+ color32 newColor {};
+ newColor.r = _color[0];
+ newColor.g = _color[1];
+ newColor.b = _color[2];
+ newColor.a = _color[3];
+ return newColor;
+ }
+
+ std::string ToANSIColor()
+ {
+ std::string out = "\033[38;2;";
+ out += std::to_string(_color[0]) + ";";
+ out += std::to_string(_color[1]) + ";";
+ out += std::to_string(_color[2]) + ";";
+ out += "49m";
+ return out;
+ }
+
+ SourceColor ToSourceColor()
+ {
+ return SourceColor(_color[0], _color[1], _color[2], _color[3]);
+ }
+
+ private:
+ unsigned char _color[4];
+};
+
+namespace NS::Colors
+{
+ extern Color SCRIPT_UI;
+ extern Color SCRIPT_CL;
+ extern Color SCRIPT_SV;
+ extern Color NATIVE_UI;
+ extern Color NATIVE_CL;
+ extern Color NATIVE_SV;
+ extern Color NATIVE_ENGINE;
+ extern Color FILESYSTEM;
+ extern Color RPAK;
+ extern Color NORTHSTAR;
+ extern Color ECHO;
+
+ extern Color TRACE;
+ extern Color DEBUG;
+ extern Color INFO;
+ extern Color WARN;
+ extern Color ERR;
+ extern Color CRIT;
+ extern Color OFF;
+}; // namespace NS::Colors
diff --git a/NorthstarDLL/core/math/vector.h b/NorthstarDLL/core/math/vector.h
new file mode 100644
index 00000000..95eae7ca
--- /dev/null
+++ b/NorthstarDLL/core/math/vector.h
@@ -0,0 +1,52 @@
+#pragma once
+
+union Vector3
+{
+ struct
+ {
+ float x;
+ float y;
+ float z;
+ };
+
+ float raw[3];
+
+ Vector3() : x(0), y(0), z(0) {}
+ Vector3(float x, float y, float z) : x(x), y(y), z(z) {}
+ Vector3(float* pRawFloats) // assumes float[3] => vector
+ {
+ memcpy(raw, pRawFloats, sizeof(this));
+ }
+
+ void MakeValid()
+ {
+ for (auto& fl : raw)
+ if (isnan(fl))
+ fl = 0;
+ }
+
+ // todo: more operators maybe
+ bool operator==(const Vector3& other)
+ {
+ return x == other.x && y == other.y && z == other.z;
+ }
+};
+
+union QAngle
+{
+ struct
+ {
+ float x;
+ float y;
+ float z;
+ float w;
+ };
+
+ float raw[4];
+
+ // todo: more operators maybe
+ bool operator==(const QAngle& other)
+ {
+ return x == other.x && y == other.y && z == other.z && w == other.w;
+ }
+};