update openal-soft

sync point: master-ac5d40e40a0155351fe1be4aab30017b6a13a859
This commit is contained in:
AzaezelX 2021-01-26 13:01:35 -06:00
parent 762a84550f
commit 3603188b7f
365 changed files with 76053 additions and 53126 deletions

View file

@ -0,0 +1,144 @@
#ifndef AL_BIT_H
#define AL_BIT_H
#include <limits>
#include <type_traits>
#if !defined(__GNUC__) && (defined(_WIN32) || defined(_WIN64))
#include <intrin.h>
#include "opthelpers.h"
#endif
namespace al {
#ifdef __BYTE_ORDER__
enum class endian {
little = __ORDER_LITTLE_ENDIAN__,
big = __ORDER_BIG_ENDIAN__,
native = __BYTE_ORDER__
};
#else
/* This doesn't support mixed-endian. */
namespace detail_ {
constexpr inline bool EndianTest() noexcept
{
static_assert(sizeof(char) < sizeof(int), "char is too big");
constexpr int test_val{1};
return static_cast<const char&>(test_val);
}
} // namespace detail_
enum class endian {
little = 0,
big = 1,
native = detail_::EndianTest() ? little : big
};
#endif
/* Define popcount (population count/count 1 bits) and countr_zero (count
* trailing zero bits, starting from the lsb) methods, for various integer
* types.
*/
#ifdef __GNUC__
namespace detail_ {
inline int popcount(unsigned long long val) noexcept { return __builtin_popcountll(val); }
inline int popcount(unsigned long val) noexcept { return __builtin_popcountl(val); }
inline int popcount(unsigned int val) noexcept { return __builtin_popcount(val); }
inline int countr_zero(unsigned long long val) noexcept { return __builtin_ctzll(val); }
inline int countr_zero(unsigned long val) noexcept { return __builtin_ctzl(val); }
inline int countr_zero(unsigned int val) noexcept { return __builtin_ctz(val); }
} // namespace detail_
template<typename T>
inline std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
int> popcount(T v) noexcept { return detail_::popcount(v); }
template<typename T>
inline std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
int> countr_zero(T val) noexcept
{ return val ? detail_::countr_zero(val) : std::numeric_limits<T>::digits; }
#else
/* There be black magics here. The popcount method is derived from
* https://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetParallel
* while the ctz-utilizing-popcount algorithm is shown here
* http://www.hackersdelight.org/hdcodetxt/ntz.c.txt
* as the ntz2 variant. These likely aren't the most efficient methods, but
* they're good enough if the GCC built-ins aren't available.
*/
namespace detail_ {
template<typename T>
constexpr T repbits(unsigned char bits) noexcept
{
T ret{bits};
for(size_t i{1};i < sizeof(T);++i)
ret = (ret<<8) | bits;
return ret;
}
} // namespace detail_
template<typename T>
constexpr std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
int> popcount(T v) noexcept
{
constexpr T m55{detail_::repbits<T>(0x55)};
constexpr T m33{detail_::repbits<T>(0x33)};
constexpr T m0f{detail_::repbits<T>(0x0f)};
constexpr T m01{detail_::repbits<T>(0x01)};
v = v - ((v >> 1) & m55);
v = (v & m33) + ((v >> 2) & m33);
v = (v + (v >> 4)) & m0f;
return static_cast<int>((v * m01) >> ((sizeof(T)-1)*8));
}
#if defined(_WIN64)
template<typename T>
inline std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
int> countr_zero(T v)
{
unsigned long idx{std::numeric_limits<T>::digits};
if_constexpr(std::numeric_limits<T>::digits <= 32)
_BitScanForward(&idx, static_cast<uint32_t>(v));
else // std::numeric_limits<T>::digits > 32
_BitScanForward64(&idx, v);
return static_cast<int>(idx);
}
#elif defined(_WIN32)
template<typename T>
inline std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
int> countr_zero(T v)
{
unsigned long idx{std::numeric_limits<T>::digits};
if_constexpr(std::numeric_limits<T>::digits <= 32)
_BitScanForward(&idx, static_cast<uint32_t>(v));
else if(!_BitScanForward(&idx, static_cast<uint32_t>(v)))
{
if(_BitScanForward(&idx, static_cast<uint32_t>(v>>32)))
idx += 32;
}
return static_cast<int>(idx);
}
#else
template<typename T>
constexpr std::enable_if_t<std::is_integral<T>::value && std::is_unsigned<T>::value,
int> countr_zero(T value)
{ return popcount(static_cast<T>(~value & (value - 1))); }
#endif
#endif
} // namespace al
#endif /* AL_BIT_H */

View file

@ -0,0 +1,67 @@
#ifndef AL_BYTE_H
#define AL_BYTE_H
#include <cstddef>
#include <cstdint>
#include <limits>
#include <type_traits>
using uint = unsigned int;
namespace al {
/* The "canonical" way to store raw byte data. Like C++17's std::byte, it's not
* treated as a character type and does not work with arithmatic ops. Only
* bitwise ops are allowed.
*/
enum class byte : unsigned char { };
template<typename T>
constexpr std::enable_if_t<std::is_integral<T>::value,T>
to_integer(al::byte b) noexcept { return T(b); }
template<typename T>
constexpr std::enable_if_t<std::is_integral<T>::value,al::byte>
operator<<(al::byte lhs, T rhs) noexcept { return al::byte(to_integer<uint>(lhs) << rhs); }
template<typename T>
constexpr std::enable_if_t<std::is_integral<T>::value,al::byte>
operator>>(al::byte lhs, T rhs) noexcept { return al::byte(to_integer<uint>(lhs) >> rhs); }
template<typename T>
constexpr std::enable_if_t<std::is_integral<T>::value,al::byte&>
operator<<=(al::byte &lhs, T rhs) noexcept { lhs = lhs << rhs; return lhs; }
template<typename T>
constexpr std::enable_if_t<std::is_integral<T>::value,al::byte&>
operator>>=(al::byte &lhs, T rhs) noexcept { lhs = lhs >> rhs; return lhs; }
#define AL_DECL_OP(op, opeq) \
template<typename T> \
constexpr std::enable_if_t<std::is_integral<T>::value,al::byte> \
operator op (al::byte lhs, T rhs) noexcept \
{ return al::byte(to_integer<uint>(lhs) op static_cast<uint>(rhs)); } \
\
template<typename T> \
constexpr std::enable_if_t<std::is_integral<T>::value,al::byte&> \
operator opeq (al::byte &lhs, T rhs) noexcept { lhs = lhs op rhs; return lhs; } \
\
constexpr al::byte operator op (al::byte lhs, al::byte rhs) noexcept \
{ return al::byte(lhs op to_integer<uint>(rhs)); } \
\
constexpr al::byte& operator opeq (al::byte &lhs, al::byte rhs) noexcept \
{ lhs = lhs op rhs; return lhs; }
AL_DECL_OP(|, |=)
AL_DECL_OP(&, &=)
AL_DECL_OP(^, ^=)
#undef AL_DECL_OP
constexpr al::byte operator~(al::byte b) noexcept
{ return al::byte(~to_integer<uint>(b)); }
} // namespace al
#endif /* AL_BYTE_H */

View file

@ -0,0 +1,80 @@
#include "config.h"
#include "alcomplex.h"
#include <algorithm>
#include <cmath>
#include <cstddef>
#include <utility>
#include "albit.h"
#include "alnumeric.h"
#include "math_defs.h"
void complex_fft(const al::span<std::complex<double>> buffer, const double sign)
{
const size_t fftsize{buffer.size()};
/* Get the number of bits used for indexing. Simplifies bit-reversal and
* the main loop count.
*/
const size_t log2_size{static_cast<size_t>(al::countr_zero(fftsize))};
/* Bit-reversal permutation applied to a sequence of fftsize items. */
for(size_t idx{1u};idx < fftsize-1;++idx)
{
size_t revidx{0u}, imask{idx};
for(size_t i{0};i < log2_size;++i)
{
revidx = (revidx<<1) | (imask&1);
imask >>= 1;
}
if(idx < revidx)
std::swap(buffer[idx], buffer[revidx]);
}
/* Iterative form of Danielson-Lanczos lemma */
size_t step2{1u};
for(size_t i{0};i < log2_size;++i)
{
const double arg{al::MathDefs<double>::Pi() / static_cast<double>(step2)};
const std::complex<double> w{std::cos(arg), std::sin(arg)*sign};
std::complex<double> u{1.0, 0.0};
const size_t step{step2 << 1};
for(size_t j{0};j < step2;j++)
{
for(size_t k{j};k < fftsize;k+=step)
{
std::complex<double> temp{buffer[k+step2] * u};
buffer[k+step2] = buffer[k] - temp;
buffer[k] += temp;
}
u *= w;
}
step2 <<= 1;
}
}
void complex_hilbert(const al::span<std::complex<double>> buffer)
{
inverse_fft(buffer);
const double inverse_size = 1.0/static_cast<double>(buffer.size());
auto bufiter = buffer.begin();
const auto halfiter = bufiter + (buffer.size()>>1);
*bufiter *= inverse_size; ++bufiter;
bufiter = std::transform(bufiter, halfiter, bufiter,
[inverse_size](const std::complex<double> &c) -> std::complex<double>
{ return c * (2.0*inverse_size); });
*bufiter *= inverse_size; ++bufiter;
std::fill(bufiter, buffer.end(), std::complex<double>{});
forward_fft(buffer);
}

View file

@ -0,0 +1,38 @@
#ifndef ALCOMPLEX_H
#define ALCOMPLEX_H
#include <complex>
#include "alspan.h"
/**
* Iterative implementation of 2-radix FFT (In-place algorithm). Sign = -1 is
* FFT and 1 is inverse FFT. Applies the Discrete Fourier Transform (DFT) to
* the data supplied in the buffer, which MUST BE power of two.
*/
void complex_fft(const al::span<std::complex<double>> buffer, const double sign);
/**
* Calculate the frequency-domain response of the time-domain signal in the
* provided buffer, which MUST BE power of two.
*/
inline void forward_fft(const al::span<std::complex<double>> buffer)
{ complex_fft(buffer, -1.0); }
/**
* Calculate the time-domain signal of the frequency-domain response in the
* provided buffer, which MUST BE power of two.
*/
inline void inverse_fft(const al::span<std::complex<double>> buffer)
{ complex_fft(buffer, 1.0); }
/**
* Calculate the complex helical sequence (discrete-time analytical signal) of
* the given input using the discrete Hilbert transform (In-place algorithm).
* Fills the buffer with the discrete-time analytical signal stored in the
* buffer. The buffer is an array of complex numbers and MUST BE power of two,
* and the imaginary components should be cleared to 0.
*/
void complex_hilbert(const al::span<std::complex<double>> buffer);
#endif /* ALCOMPLEX_H */

View file

@ -0,0 +1,16 @@
#ifndef ALDEQUE_H
#define ALDEQUE_H
#include <deque>
#include "almalloc.h"
namespace al {
template<typename T>
using deque = std::deque<T, al::allocator<T>>;
} // namespace al
#endif /* ALDEQUE_H */

View file

@ -0,0 +1,150 @@
#include "config.h"
#include "alfstream.h"
#include "strutils.h"
#ifdef _WIN32
namespace al {
auto filebuf::underflow() -> int_type
{
if(mFile != INVALID_HANDLE_VALUE && gptr() == egptr())
{
// Read in the next chunk of data, and set the pointers on success
DWORD got{};
if(ReadFile(mFile, mBuffer.data(), static_cast<DWORD>(mBuffer.size()), &got, nullptr))
setg(mBuffer.data(), mBuffer.data(), mBuffer.data()+got);
}
if(gptr() == egptr())
return traits_type::eof();
return traits_type::to_int_type(*gptr());
}
auto filebuf::seekoff(off_type offset, std::ios_base::seekdir whence, std::ios_base::openmode mode) -> pos_type
{
if(mFile == INVALID_HANDLE_VALUE || (mode&std::ios_base::out) || !(mode&std::ios_base::in))
return traits_type::eof();
LARGE_INTEGER fpos{};
switch(whence)
{
case std::ios_base::beg:
fpos.QuadPart = offset;
if(!SetFilePointerEx(mFile, fpos, &fpos, FILE_BEGIN))
return traits_type::eof();
break;
case std::ios_base::cur:
// If the offset remains in the current buffer range, just
// update the pointer.
if((offset >= 0 && offset < off_type(egptr()-gptr())) ||
(offset < 0 && -offset <= off_type(gptr()-eback())))
{
// Get the current file offset to report the correct read
// offset.
fpos.QuadPart = 0;
if(!SetFilePointerEx(mFile, fpos, &fpos, FILE_CURRENT))
return traits_type::eof();
setg(eback(), gptr()+offset, egptr());
return fpos.QuadPart - off_type(egptr()-gptr());
}
// Need to offset for the file offset being at egptr() while
// the requested offset is relative to gptr().
offset -= off_type(egptr()-gptr());
fpos.QuadPart = offset;
if(!SetFilePointerEx(mFile, fpos, &fpos, FILE_CURRENT))
return traits_type::eof();
break;
case std::ios_base::end:
fpos.QuadPart = offset;
if(!SetFilePointerEx(mFile, fpos, &fpos, FILE_END))
return traits_type::eof();
break;
default:
return traits_type::eof();
}
setg(nullptr, nullptr, nullptr);
return fpos.QuadPart;
}
auto filebuf::seekpos(pos_type pos, std::ios_base::openmode mode) -> pos_type
{
// Simplified version of seekoff
if(mFile == INVALID_HANDLE_VALUE || (mode&std::ios_base::out) || !(mode&std::ios_base::in))
return traits_type::eof();
LARGE_INTEGER fpos{};
fpos.QuadPart = pos;
if(!SetFilePointerEx(mFile, fpos, &fpos, FILE_BEGIN))
return traits_type::eof();
setg(nullptr, nullptr, nullptr);
return fpos.QuadPart;
}
filebuf::~filebuf()
{ close(); }
bool filebuf::open(const wchar_t *filename, std::ios_base::openmode mode)
{
if((mode&std::ios_base::out) || !(mode&std::ios_base::in))
return false;
HANDLE f{CreateFileW(filename, GENERIC_READ, FILE_SHARE_READ, nullptr, OPEN_EXISTING,
FILE_ATTRIBUTE_NORMAL, nullptr)};
if(f == INVALID_HANDLE_VALUE) return false;
if(mFile != INVALID_HANDLE_VALUE)
CloseHandle(mFile);
mFile = f;
setg(nullptr, nullptr, nullptr);
return true;
}
bool filebuf::open(const char *filename, std::ios_base::openmode mode)
{
std::wstring wname{utf8_to_wstr(filename)};
return open(wname.c_str(), mode);
}
void filebuf::close()
{
if(mFile != INVALID_HANDLE_VALUE)
CloseHandle(mFile);
mFile = INVALID_HANDLE_VALUE;
}
ifstream::ifstream(const wchar_t *filename, std::ios_base::openmode mode)
: std::istream{nullptr}
{
init(&mStreamBuf);
// Set the failbit if the file failed to open.
if((mode&std::ios_base::out) || !mStreamBuf.open(filename, mode|std::ios_base::in))
clear(failbit);
}
ifstream::ifstream(const char *filename, std::ios_base::openmode mode)
: std::istream{nullptr}
{
init(&mStreamBuf);
// Set the failbit if the file failed to open.
if((mode&std::ios_base::out) || !mStreamBuf.open(filename, mode|std::ios_base::in))
clear(failbit);
}
/* This is only here to ensure the compiler doesn't define an implicit
* destructor, which it tries to automatically inline and subsequently complain
* it can't inline without excessive code growth.
*/
ifstream::~ifstream() { }
} // namespace al
#endif

View file

@ -0,0 +1,74 @@
#ifndef AL_FSTREAM_H
#define AL_FSTREAM_H
#ifdef _WIN32
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#include <array>
#include <string>
#include <fstream>
namespace al {
// Windows' std::ifstream fails with non-ANSI paths since the standard only
// specifies names using const char* (or std::string). MSVC has a non-standard
// extension using const wchar_t* (or std::wstring?) to handle Unicode paths,
// but not all Windows compilers support it. So we have to make our own istream
// that accepts UTF-8 paths and forwards to Unicode-aware I/O functions.
class filebuf final : public std::streambuf {
std::array<char_type,4096> mBuffer;
HANDLE mFile{INVALID_HANDLE_VALUE};
int_type underflow() override;
pos_type seekoff(off_type offset, std::ios_base::seekdir whence, std::ios_base::openmode mode) override;
pos_type seekpos(pos_type pos, std::ios_base::openmode mode) override;
public:
filebuf() = default;
~filebuf() override;
bool open(const wchar_t *filename, std::ios_base::openmode mode);
bool open(const char *filename, std::ios_base::openmode mode);
bool is_open() const noexcept { return mFile != INVALID_HANDLE_VALUE; }
void close();
};
// Inherit from std::istream to use our custom streambuf
class ifstream final : public std::istream {
filebuf mStreamBuf;
public:
ifstream(const wchar_t *filename, std::ios_base::openmode mode = std::ios_base::in);
ifstream(const std::wstring &filename, std::ios_base::openmode mode = std::ios_base::in)
: ifstream(filename.c_str(), mode) { }
ifstream(const char *filename, std::ios_base::openmode mode = std::ios_base::in);
ifstream(const std::string &filename, std::ios_base::openmode mode = std::ios_base::in)
: ifstream(filename.c_str(), mode) { }
~ifstream() override;
bool is_open() const noexcept { return mStreamBuf.is_open(); }
void close() { mStreamBuf.close(); }
};
} // namespace al
#else /* _WIN32 */
#include <fstream>
namespace al {
using filebuf = std::filebuf;
using ifstream = std::ifstream;
} // namespace al
#endif /* _WIN32 */
#endif /* AL_FSTREAM_H */

View file

@ -1,21 +0,0 @@
#ifndef AL_ALIGN_H
#define AL_ALIGN_H
#if defined(HAVE_STDALIGN_H) && defined(HAVE_C11_ALIGNAS)
#include <stdalign.h>
#endif
#ifndef alignas
#if defined(IN_IDE_PARSER)
/* KDevelop has problems with our align macro, so just use nothing for parsing. */
#define alignas(x)
#elif defined(HAVE_C11_ALIGNAS)
#define alignas _Alignas
#else
/* NOTE: Our custom ALIGN macro can't take a type name like alignas can. For
* maximum compatibility, only provide constant integer values to alignas. */
#define alignas(_x) ALIGN(_x)
#endif
#endif
#endif /* AL_ALIGN_H */

View file

@ -1,110 +0,0 @@
#include "config.h"
#include "almalloc.h"
#include <stdlib.h>
#include <string.h>
#ifdef HAVE_MALLOC_H
#include <malloc.h>
#endif
#ifdef HAVE_WINDOWS_H
#include <windows.h>
#else
#include <unistd.h>
#endif
#ifdef __GNUC__
#define LIKELY(x) __builtin_expect(!!(x), !0)
#define UNLIKELY(x) __builtin_expect(!!(x), 0)
#else
#define LIKELY(x) (!!(x))
#define UNLIKELY(x) (!!(x))
#endif
void *al_malloc(size_t alignment, size_t size)
{
#if defined(HAVE_ALIGNED_ALLOC)
size = (size+(alignment-1))&~(alignment-1);
return aligned_alloc(alignment, size);
#elif defined(HAVE_POSIX_MEMALIGN)
void *ret;
if(posix_memalign(&ret, alignment, size) == 0)
return ret;
return NULL;
#elif defined(HAVE__ALIGNED_MALLOC)
return _aligned_malloc(size, alignment);
#else
char *ret = malloc(size+alignment);
if(ret != NULL)
{
*(ret++) = 0x00;
while(((ptrdiff_t)ret&(alignment-1)) != 0)
*(ret++) = 0x55;
}
return ret;
#endif
}
void *al_calloc(size_t alignment, size_t size)
{
void *ret = al_malloc(alignment, size);
if(ret) memset(ret, 0, size);
return ret;
}
void al_free(void *ptr)
{
#if defined(HAVE_ALIGNED_ALLOC) || defined(HAVE_POSIX_MEMALIGN)
free(ptr);
#elif defined(HAVE__ALIGNED_MALLOC)
_aligned_free(ptr);
#else
if(ptr != NULL)
{
char *finder = ptr;
do {
--finder;
} while(*finder == 0x55);
free(finder);
}
#endif
}
size_t al_get_page_size(void)
{
static size_t psize = 0;
if(UNLIKELY(!psize))
{
#ifdef HAVE_SYSCONF
#if defined(_SC_PAGESIZE)
if(!psize) psize = sysconf(_SC_PAGESIZE);
#elif defined(_SC_PAGE_SIZE)
if(!psize) psize = sysconf(_SC_PAGE_SIZE);
#endif
#endif /* HAVE_SYSCONF */
#ifdef _WIN32
if(!psize)
{
SYSTEM_INFO sysinfo;
memset(&sysinfo, 0, sizeof(sysinfo));
GetSystemInfo(&sysinfo);
psize = sysinfo.dwPageSize;
}
#endif
if(!psize) psize = DEF_ALIGN;
}
return psize;
}
int al_is_sane_alignment_allocator(void)
{
#if defined(HAVE_ALIGNED_ALLOC) || defined(HAVE_POSIX_MEMALIGN) || defined(HAVE__ALIGNED_MALLOC)
return 1;
#else
return 0;
#endif
}

View file

@ -0,0 +1,61 @@
#include "config.h"
#include "almalloc.h"
#include <cassert>
#include <cstddef>
#include <cstdlib>
#include <cstring>
#include <memory>
#ifdef HAVE_MALLOC_H
#include <malloc.h>
#endif
void *al_malloc(size_t alignment, size_t size)
{
assert((alignment & (alignment-1)) == 0);
alignment = std::max(alignment, alignof(std::max_align_t));
#if defined(HAVE_POSIX_MEMALIGN)
void *ret{};
if(posix_memalign(&ret, alignment, size) == 0)
return ret;
return nullptr;
#elif defined(HAVE__ALIGNED_MALLOC)
return _aligned_malloc(size, alignment);
#else
size_t total_size{size + alignment-1 + sizeof(void*)};
void *base{std::malloc(total_size)};
if(base != nullptr)
{
void *aligned_ptr{static_cast<char*>(base) + sizeof(void*)};
total_size -= sizeof(void*);
std::align(alignment, size, aligned_ptr, total_size);
*(static_cast<void**>(aligned_ptr)-1) = base;
base = aligned_ptr;
}
return base;
#endif
}
void *al_calloc(size_t alignment, size_t size)
{
void *ret{al_malloc(alignment, size)};
if(ret) std::memset(ret, 0, size);
return ret;
}
void al_free(void *ptr) noexcept
{
#if defined(HAVE_POSIX_MEMALIGN)
std::free(ptr);
#elif defined(HAVE__ALIGNED_MALLOC)
_aligned_free(ptr);
#else
if(ptr != nullptr)
std::free(*(static_cast<void**>(ptr) - 1));
#endif
}

View file

@ -1,30 +1,291 @@
#ifndef AL_MALLOC_H
#define AL_MALLOC_H
#include <stddef.h>
#include <algorithm>
#include <cstddef>
#include <iterator>
#include <limits>
#include <memory>
#include <new>
#include <type_traits>
#include <utility>
#ifdef __cplusplus
extern "C" {
#endif
#include "pragmadefs.h"
/* Minimum alignment required by posix_memalign. */
#define DEF_ALIGN sizeof(void*)
void *al_malloc(size_t alignment, size_t size);
void *al_calloc(size_t alignment, size_t size);
void al_free(void *ptr);
[[gnu::alloc_align(1), gnu::alloc_size(2)]] void *al_malloc(size_t alignment, size_t size);
[[gnu::alloc_align(1), gnu::alloc_size(2)]] void *al_calloc(size_t alignment, size_t size);
void al_free(void *ptr) noexcept;
size_t al_get_page_size(void);
/**
* Returns non-0 if the allocation function has direct alignment handling.
* Otherwise, the standard malloc is used with an over-allocation and pointer
* offset strategy.
#define DISABLE_ALLOC() \
void *operator new(size_t) = delete; \
void *operator new[](size_t) = delete; \
void operator delete(void*) noexcept = delete; \
void operator delete[](void*) noexcept = delete;
#define DEF_NEWDEL(T) \
void *operator new(size_t size) \
{ \
void *ret = al_malloc(alignof(T), size); \
if(!ret) throw std::bad_alloc(); \
return ret; \
} \
void *operator new[](size_t size) { return operator new(size); } \
void operator delete(void *block) noexcept { al_free(block); } \
void operator delete[](void *block) noexcept { operator delete(block); }
#define DEF_PLACE_NEWDEL() \
void *operator new(size_t /*size*/, void *ptr) noexcept { return ptr; } \
void *operator new[](size_t /*size*/, void *ptr) noexcept { return ptr; } \
void operator delete(void *block, void*) noexcept { al_free(block); } \
void operator delete(void *block) noexcept { al_free(block); } \
void operator delete[](void *block, void*) noexcept { al_free(block); } \
void operator delete[](void *block) noexcept { al_free(block); }
enum FamCount : size_t { };
#define DEF_FAM_NEWDEL(T, FamMem) \
static constexpr size_t Sizeof(size_t count) noexcept \
{ \
return std::max<size_t>(sizeof(T), \
decltype(FamMem)::Sizeof(count, offsetof(T, FamMem))); \
} \
\
void *operator new(size_t /*size*/, FamCount count) \
{ \
if(void *ret{al_malloc(alignof(T), T::Sizeof(count))}) \
return ret; \
throw std::bad_alloc(); \
} \
void *operator new[](size_t /*size*/) = delete; \
void operator delete(void *block, FamCount) { al_free(block); } \
void operator delete(void *block) noexcept { al_free(block); } \
void operator delete[](void* /*block*/) = delete;
namespace al {
template<typename T, std::size_t alignment=alignof(T)>
struct allocator {
using value_type = T;
using reference = T&;
using const_reference = const T&;
using pointer = T*;
using const_pointer = const T*;
using size_type = std::size_t;
using difference_type = std::ptrdiff_t;
using is_always_equal = std::true_type;
template<typename U>
struct rebind {
using other = allocator<U, (alignment<alignof(U))?alignof(U):alignment>;
};
constexpr explicit allocator() noexcept = default;
template<typename U, std::size_t N>
constexpr explicit allocator(const allocator<U,N>&) noexcept { }
T *allocate(std::size_t n)
{
if(n > std::numeric_limits<std::size_t>::max()/sizeof(T)) throw std::bad_alloc();
if(auto p = al_malloc(alignment, n*sizeof(T))) return static_cast<T*>(p);
throw std::bad_alloc();
}
void deallocate(T *p, std::size_t) noexcept { al_free(p); }
};
template<typename T, std::size_t N, typename U, std::size_t M>
bool operator==(const allocator<T,N>&, const allocator<U,M>&) noexcept { return true; }
template<typename T, std::size_t N, typename U, std::size_t M>
bool operator!=(const allocator<T,N>&, const allocator<U,M>&) noexcept { return false; }
template<size_t alignment, typename T>
[[gnu::assume_aligned(alignment)]] inline T* assume_aligned(T *ptr) noexcept { return ptr; }
/* At least VS 2015 complains that 'ptr' is unused when the given type's
* destructor is trivial (a no-op). So disable that warning for this call.
*/
int al_is_sane_alignment_allocator(void);
#ifdef __cplusplus
DIAGNOSTIC_PUSH
msc_pragma(warning(disable : 4100))
template<typename T>
constexpr std::enable_if_t<!std::is_array<T>::value>
destroy_at(T *ptr) noexcept(std::is_nothrow_destructible<T>::value)
{ ptr->~T(); }
DIAGNOSTIC_POP
template<typename T>
constexpr std::enable_if_t<std::is_array<T>::value>
destroy_at(T *ptr) noexcept(std::is_nothrow_destructible<T>::value)
{
for(auto &elem : *ptr)
al::destroy_at(std::addressof(elem));
}
#endif
template<typename T>
constexpr void destroy(T first, T end)
{
while(first != end)
{
al::destroy_at(std::addressof(*first));
++first;
}
}
template<typename T, typename N>
constexpr std::enable_if_t<std::is_integral<N>::value,T>
destroy_n(T first, N count)
{
if(count != 0)
{
do {
al::destroy_at(std::addressof(*first));
++first;
} while(--count);
}
return first;
}
template<typename T, typename N>
inline std::enable_if_t<std::is_integral<N>::value,T>
uninitialized_default_construct_n(T first, N count)
{
using ValueT = typename std::iterator_traits<T>::value_type;
T current{first};
if(count != 0)
{
try {
do {
::new(static_cast<void*>(std::addressof(*current))) ValueT;
++current;
} while(--count);
}
catch(...) {
al::destroy(first, current);
throw;
}
}
return current;
}
/* Storage for flexible array data. This is trivially destructible if type T is
* trivially destructible.
*/
template<typename T, size_t alignment, bool = std::is_trivially_destructible<T>::value>
struct FlexArrayStorage;
template<typename T, size_t alignment>
struct FlexArrayStorage<T,alignment,true> {
const size_t mSize;
union {
char mDummy;
alignas(alignment) T mArray[1];
};
static constexpr size_t Sizeof(size_t count, size_t base=0u) noexcept
{
return std::max<size_t>(offsetof(FlexArrayStorage, mArray) + sizeof(T)*count,
sizeof(FlexArrayStorage)) + base;
}
FlexArrayStorage(size_t size) : mSize{size}
{ al::uninitialized_default_construct_n(mArray, mSize); }
~FlexArrayStorage() = default;
FlexArrayStorage(const FlexArrayStorage&) = delete;
FlexArrayStorage& operator=(const FlexArrayStorage&) = delete;
};
template<typename T, size_t alignment>
struct FlexArrayStorage<T,alignment,false> {
const size_t mSize;
union {
char mDummy;
alignas(alignment) T mArray[1];
};
static constexpr size_t Sizeof(size_t count, size_t base) noexcept
{
return std::max<size_t>(offsetof(FlexArrayStorage, mArray) + sizeof(T)*count,
sizeof(FlexArrayStorage)) + base;
}
FlexArrayStorage(size_t size) : mSize{size}
{ al::uninitialized_default_construct_n(mArray, mSize); }
~FlexArrayStorage() { al::destroy_n(mArray, mSize); }
FlexArrayStorage(const FlexArrayStorage&) = delete;
FlexArrayStorage& operator=(const FlexArrayStorage&) = delete;
};
/* A flexible array type. Used either standalone or at the end of a parent
* struct, with placement new, to have a run-time-sized array that's embedded
* with its size.
*/
template<typename T, size_t alignment=alignof(T)>
struct FlexArray {
using element_type = T;
using value_type = std::remove_cv_t<T>;
using index_type = size_t;
using difference_type = ptrdiff_t;
using pointer = T*;
using const_pointer = const T*;
using reference = T&;
using const_reference = const T&;
using iterator = pointer;
using const_iterator = const_pointer;
using reverse_iterator = std::reverse_iterator<iterator>;
using const_reverse_iterator = std::reverse_iterator<const_iterator>;
using Storage_t_ = FlexArrayStorage<element_type,alignment>;
Storage_t_ mStore;
static constexpr index_type Sizeof(index_type count, index_type base=0u) noexcept
{ return Storage_t_::Sizeof(count, base); }
static std::unique_ptr<FlexArray> Create(index_type count)
{
void *ptr{al_calloc(alignof(FlexArray), Sizeof(count))};
return std::unique_ptr<FlexArray>{new(ptr) FlexArray{count}};
}
FlexArray(index_type size) : mStore{size} { }
~FlexArray() = default;
index_type size() const noexcept { return mStore.mSize; }
bool empty() const noexcept { return mStore.mSize == 0; }
pointer data() noexcept { return mStore.mArray; }
const_pointer data() const noexcept { return mStore.mArray; }
reference operator[](index_type i) noexcept { return mStore.mArray[i]; }
const_reference operator[](index_type i) const noexcept { return mStore.mArray[i]; }
reference front() noexcept { return mStore.mArray[0]; }
const_reference front() const noexcept { return mStore.mArray[0]; }
reference back() noexcept { return mStore.mArray[mStore.mSize-1]; }
const_reference back() const noexcept { return mStore.mArray[mStore.mSize-1]; }
iterator begin() noexcept { return mStore.mArray; }
const_iterator begin() const noexcept { return mStore.mArray; }
const_iterator cbegin() const noexcept { return mStore.mArray; }
iterator end() noexcept { return mStore.mArray + mStore.mSize; }
const_iterator end() const noexcept { return mStore.mArray + mStore.mSize; }
const_iterator cend() const noexcept { return mStore.mArray + mStore.mSize; }
reverse_iterator rbegin() noexcept { return end(); }
const_reverse_iterator rbegin() const noexcept { return end(); }
const_reverse_iterator crbegin() const noexcept { return cend(); }
reverse_iterator rend() noexcept { return begin(); }
const_reverse_iterator rend() const noexcept { return begin(); }
const_reverse_iterator crend() const noexcept { return cbegin(); }
DEF_PLACE_NEWDEL()
};
} // namespace al
#endif /* AL_MALLOC_H */

View file

@ -0,0 +1,274 @@
#ifndef AL_NUMERIC_H
#define AL_NUMERIC_H
#include <cstddef>
#include <cstdint>
#ifdef HAVE_INTRIN_H
#include <intrin.h>
#endif
#ifdef HAVE_SSE_INTRINSICS
#include <xmmintrin.h>
#endif
#include "opthelpers.h"
inline constexpr int64_t operator "" _i64(unsigned long long int n) noexcept { return static_cast<int64_t>(n); }
inline constexpr uint64_t operator "" _u64(unsigned long long int n) noexcept { return static_cast<uint64_t>(n); }
constexpr inline float minf(float a, float b) noexcept
{ return ((a > b) ? b : a); }
constexpr inline float maxf(float a, float b) noexcept
{ return ((a > b) ? a : b); }
constexpr inline float clampf(float val, float min, float max) noexcept
{ return minf(max, maxf(min, val)); }
constexpr inline double mind(double a, double b) noexcept
{ return ((a > b) ? b : a); }
constexpr inline double maxd(double a, double b) noexcept
{ return ((a > b) ? a : b); }
constexpr inline double clampd(double val, double min, double max) noexcept
{ return mind(max, maxd(min, val)); }
constexpr inline unsigned int minu(unsigned int a, unsigned int b) noexcept
{ return ((a > b) ? b : a); }
constexpr inline unsigned int maxu(unsigned int a, unsigned int b) noexcept
{ return ((a > b) ? a : b); }
constexpr inline unsigned int clampu(unsigned int val, unsigned int min, unsigned int max) noexcept
{ return minu(max, maxu(min, val)); }
constexpr inline int mini(int a, int b) noexcept
{ return ((a > b) ? b : a); }
constexpr inline int maxi(int a, int b) noexcept
{ return ((a > b) ? a : b); }
constexpr inline int clampi(int val, int min, int max) noexcept
{ return mini(max, maxi(min, val)); }
constexpr inline int64_t mini64(int64_t a, int64_t b) noexcept
{ return ((a > b) ? b : a); }
constexpr inline int64_t maxi64(int64_t a, int64_t b) noexcept
{ return ((a > b) ? a : b); }
constexpr inline int64_t clampi64(int64_t val, int64_t min, int64_t max) noexcept
{ return mini64(max, maxi64(min, val)); }
constexpr inline uint64_t minu64(uint64_t a, uint64_t b) noexcept
{ return ((a > b) ? b : a); }
constexpr inline uint64_t maxu64(uint64_t a, uint64_t b) noexcept
{ return ((a > b) ? a : b); }
constexpr inline uint64_t clampu64(uint64_t val, uint64_t min, uint64_t max) noexcept
{ return minu64(max, maxu64(min, val)); }
constexpr inline size_t minz(size_t a, size_t b) noexcept
{ return ((a > b) ? b : a); }
constexpr inline size_t maxz(size_t a, size_t b) noexcept
{ return ((a > b) ? a : b); }
constexpr inline size_t clampz(size_t val, size_t min, size_t max) noexcept
{ return minz(max, maxz(min, val)); }
constexpr inline float lerp(float val1, float val2, float mu) noexcept
{ return val1 + (val2-val1)*mu; }
constexpr inline float cubic(float val1, float val2, float val3, float val4, float mu) noexcept
{
const float mu2{mu*mu}, mu3{mu2*mu};
const float a0{-0.5f*mu3 + mu2 + -0.5f*mu};
const float a1{ 1.5f*mu3 + -2.5f*mu2 + 1.0f};
const float a2{-1.5f*mu3 + 2.0f*mu2 + 0.5f*mu};
const float a3{ 0.5f*mu3 + -0.5f*mu2};
return val1*a0 + val2*a1 + val3*a2 + val4*a3;
}
/** Find the next power-of-2 for non-power-of-2 numbers. */
inline uint32_t NextPowerOf2(uint32_t value) noexcept
{
if(value > 0)
{
value--;
value |= value>>1;
value |= value>>2;
value |= value>>4;
value |= value>>8;
value |= value>>16;
}
return value+1;
}
/** Round up a value to the next multiple. */
inline size_t RoundUp(size_t value, size_t r) noexcept
{
value += r-1;
return value - (value%r);
}
/**
* Fast float-to-int conversion. No particular rounding mode is assumed; the
* IEEE-754 default is round-to-nearest with ties-to-even, though an app could
* change it on its own threads. On some systems, a truncating conversion may
* always be the fastest method.
*/
inline int fastf2i(float f) noexcept
{
#if defined(HAVE_SSE_INTRINSICS)
return _mm_cvt_ss2si(_mm_set_ss(f));
#elif defined(_MSC_VER) && defined(_M_IX86_FP)
int i;
__asm fld f
__asm fistp i
return i;
#elif (defined(__GNUC__) || defined(__clang__)) && (defined(__i386__) || defined(__x86_64__))
int i;
#ifdef __SSE_MATH__
__asm__("cvtss2si %1, %0" : "=r"(i) : "x"(f));
#else
__asm__ __volatile__("fistpl %0" : "=m"(i) : "t"(f) : "st");
#endif
return i;
#else
return static_cast<int>(f);
#endif
}
inline unsigned int fastf2u(float f) noexcept
{ return static_cast<unsigned int>(fastf2i(f)); }
/** Converts float-to-int using standard behavior (truncation). */
inline int float2int(float f) noexcept
{
#if defined(HAVE_SSE_INTRINSICS)
return _mm_cvtt_ss2si(_mm_set_ss(f));
#elif (defined(_MSC_VER) && defined(_M_IX86_FP) && _M_IX86_FP == 0) \
|| ((defined(__GNUC__) || defined(__clang__)) && (defined(__i386__) || defined(__x86_64__)) \
&& !defined(__SSE_MATH__))
int sign, shift, mant;
union {
float f;
int i;
} conv;
conv.f = f;
sign = (conv.i>>31) | 1;
shift = ((conv.i>>23)&0xff) - (127+23);
/* Over/underflow */
if UNLIKELY(shift >= 31 || shift < -23)
return 0;
mant = (conv.i&0x7fffff) | 0x800000;
if LIKELY(shift < 0)
return (mant >> -shift) * sign;
return (mant << shift) * sign;
#else
return static_cast<int>(f);
#endif
}
inline unsigned int float2uint(float f) noexcept
{ return static_cast<unsigned int>(float2int(f)); }
/** Converts double-to-int using standard behavior (truncation). */
inline int double2int(double d) noexcept
{
#if defined(HAVE_SSE_INTRINSICS)
return _mm_cvttsd_si32(_mm_set_sd(d));
#elif (defined(_MSC_VER) && defined(_M_IX86_FP) && _M_IX86_FP < 2) \
|| ((defined(__GNUC__) || defined(__clang__)) && (defined(__i386__) || defined(__x86_64__)) \
&& !defined(__SSE2_MATH__))
int sign, shift;
int64_t mant;
union {
double d;
int64_t i64;
} conv;
conv.d = d;
sign = (conv.i64 >> 63) | 1;
shift = ((conv.i64 >> 52) & 0x7ff) - (1023 + 52);
/* Over/underflow */
if UNLIKELY(shift >= 63 || shift < -52)
return 0;
mant = (conv.i64 & 0xfffffffffffff_i64) | 0x10000000000000_i64;
if LIKELY(shift < 0)
return (int)(mant >> -shift) * sign;
return (int)(mant << shift) * sign;
#else
return static_cast<int>(d);
#endif
}
/**
* Rounds a float to the nearest integral value, according to the current
* rounding mode. This is essentially an inlined version of rintf, although
* makes fewer promises (e.g. -0 or -0.25 rounded to 0 may result in +0).
*/
inline float fast_roundf(float f) noexcept
{
#if (defined(__GNUC__) || defined(__clang__)) && (defined(__i386__) || defined(__x86_64__)) \
&& !defined(__SSE_MATH__)
float out;
__asm__ __volatile__("frndint" : "=t"(out) : "0"(f));
return out;
#elif (defined(__GNUC__) || defined(__clang__)) && defined(__aarch64__)
float out;
__asm__ volatile("frintx %s0, %s1" : "=w"(out) : "w"(f));
return out;
#else
/* Integral limit, where sub-integral precision is not available for
* floats.
*/
static const float ilim[2]{
8388608.0f /* 0x1.0p+23 */,
-8388608.0f /* -0x1.0p+23 */
};
unsigned int sign, expo;
union {
float f;
unsigned int i;
} conv;
conv.f = f;
sign = (conv.i>>31)&0x01;
expo = (conv.i>>23)&0xff;
if UNLIKELY(expo >= 150/*+23*/)
{
/* An exponent (base-2) of 23 or higher is incapable of sub-integral
* precision, so it's already an integral value. We don't need to worry
* about infinity or NaN here.
*/
return f;
}
/* Adding the integral limit to the value (with a matching sign) forces a
* result that has no sub-integral precision, and is consequently forced to
* round to an integral value. Removing the integral limit then restores
* the initial value rounded to the integral. The compiler should not
* optimize this out because of non-associative rules on floating-point
* math (as long as you don't use -fassociative-math,
* -funsafe-math-optimizations, -ffast-math, or -Ofast, in which case this
* may break).
*/
f += ilim[sign];
return f - ilim[sign];
#endif
}
#endif /* AL_NUMERIC_H */

View file

@ -0,0 +1,161 @@
#ifndef AL_OPTIONAL_H
#define AL_OPTIONAL_H
#include <initializer_list>
#include <type_traits>
#include <utility>
#include "almalloc.h"
namespace al {
struct nullopt_t { };
struct in_place_t { };
constexpr nullopt_t nullopt{};
constexpr in_place_t in_place{};
template<typename T, bool = std::is_trivially_destructible<T>::value>
struct optional_storage;
template<typename T>
struct optional_storage<T, true> {
bool mHasValue{false};
union {
char mDummy;
T mValue;
};
optional_storage() { }
template<typename ...Args>
explicit optional_storage(in_place_t, Args&& ...args)
: mHasValue{true}, mValue{std::forward<Args>(args)...}
{ }
~optional_storage() = default;
};
template<typename T>
struct optional_storage<T, false> {
bool mHasValue{false};
union {
char mDummy;
T mValue;
};
optional_storage() { }
template<typename ...Args>
explicit optional_storage(in_place_t, Args&& ...args)
: mHasValue{true}, mValue{std::forward<Args>(args)...}
{ }
~optional_storage() { if(mHasValue) al::destroy_at(std::addressof(mValue)); }
};
template<typename T>
class optional {
using storage_t = optional_storage<T>;
storage_t mStore;
template<typename... Args>
void doConstruct(Args&& ...args)
{
::new(std::addressof(mStore.mValue)) T{std::forward<Args>(args)...};
mStore.mHasValue = true;
}
public:
using value_type = T;
optional() = default;
optional(nullopt_t) noexcept { }
optional(const optional &rhs) { if(rhs) doConstruct(*rhs); }
optional(optional&& rhs) { if(rhs) doConstruct(std::move(*rhs)); }
template<typename ...Args>
explicit optional(in_place_t, Args&& ...args)
: mStore{al::in_place, std::forward<Args>(args)...}
{ }
~optional() = default;
optional& operator=(nullopt_t) noexcept { reset(); return *this; }
std::enable_if_t<std::is_copy_constructible<T>::value && std::is_copy_assignable<T>::value,
optional&> operator=(const optional &rhs)
{
if(!rhs)
reset();
else if(*this)
mStore.mValue = *rhs;
else
doConstruct(*rhs);
return *this;
}
std::enable_if_t<std::is_move_constructible<T>::value && std::is_move_assignable<T>::value,
optional&> operator=(optional&& rhs)
{
if(!rhs)
reset();
else if(*this)
mStore.mValue = std::move(*rhs);
else
doConstruct(std::move(*rhs));
return *this;
}
template<typename U=T>
std::enable_if_t<std::is_constructible<T, U>::value
&& std::is_assignable<T&, U>::value
&& !std::is_same<std::decay_t<U>, optional<T>>::value
&& (!std::is_same<std::decay_t<U>, T>::value || !std::is_scalar<U>::value),
optional&> operator=(U&& rhs)
{
if(*this)
mStore.mValue = std::forward<U>(rhs);
else
doConstruct(std::forward<U>(rhs));
return *this;
}
const T* operator->() const { return std::addressof(mStore.mValue); }
T* operator->() { return std::addressof(mStore.mValue); }
const T& operator*() const& { return this->mValue; }
T& operator*() & { return mStore.mValue; }
const T&& operator*() const&& { return std::move(mStore.mValue); }
T&& operator*() && { return std::move(mStore.mValue); }
operator bool() const noexcept { return mStore.mHasValue; }
bool has_value() const noexcept { return mStore.mHasValue; }
T& value() & { return mStore.mValue; }
const T& value() const& { return mStore.mValue; }
T&& value() && { return std::move(mStore.mValue); }
const T&& value() const&& { return std::move(mStore.mValue); }
template<typename U>
T value_or(U&& defval) const&
{ return bool{*this} ? **this : static_cast<T>(std::forward<U>(defval)); }
template<typename U>
T value_or(U&& defval) &&
{ return bool{*this} ? std::move(**this) : static_cast<T>(std::forward<U>(defval)); }
void reset() noexcept
{
if(mStore.mHasValue)
al::destroy_at(std::addressof(mStore.mValue));
mStore.mHasValue = false;
}
};
template<typename T>
inline optional<std::decay_t<T>> make_optional(T&& arg)
{ return optional<std::decay_t<T>>{in_place, std::forward<T>(arg)}; }
template<typename T, typename... Args>
inline optional<T> make_optional(Args&& ...args)
{ return optional<T>{in_place, std::forward<Args>(args)...}; }
template<typename T, typename U, typename... Args>
inline optional<T> make_optional(std::initializer_list<U> il, Args&& ...args)
{ return optional<T>{in_place, il, std::forward<Args>(args)...}; }
} // namespace al
#endif /* AL_OPTIONAL_H */

View file

@ -0,0 +1,315 @@
#ifndef AL_SPAN_H
#define AL_SPAN_H
#include <array>
#include <cstddef>
#include <initializer_list>
#include <iterator>
#include <type_traits>
namespace al {
template<typename T>
constexpr auto size(T &cont) noexcept(noexcept(cont.size())) -> decltype(cont.size())
{ return cont.size(); }
template<typename T>
constexpr auto size(const T &cont) noexcept(noexcept(cont.size())) -> decltype(cont.size())
{ return cont.size(); }
template<typename T, size_t N>
constexpr size_t size(T (&)[N]) noexcept
{ return N; }
template<typename T>
constexpr size_t size(std::initializer_list<T> list) noexcept
{ return list.size(); }
template<typename T>
constexpr auto data(T &cont) noexcept(noexcept(cont.data())) -> decltype(cont.data())
{ return cont.data(); }
template<typename T>
constexpr auto data(const T &cont) noexcept(noexcept(cont.data())) -> decltype(cont.data())
{ return cont.data(); }
template<typename T, size_t N>
constexpr T* data(T (&arr)[N]) noexcept
{ return arr; }
template<typename T>
constexpr const T* data(std::initializer_list<T> list) noexcept
{ return list.begin(); }
constexpr size_t dynamic_extent{static_cast<size_t>(-1)};
template<typename T, size_t E=dynamic_extent>
class span;
namespace detail_ {
template<typename... Ts>
struct make_void { using type = void; };
template<typename... Ts>
using void_t = typename make_void<Ts...>::type;
template<typename T>
struct is_span_ : std::false_type { };
template<typename T, size_t E>
struct is_span_<span<T,E>> : std::true_type { };
template<typename T>
using is_span = is_span_<std::remove_cv_t<T>>;
template<typename T>
struct is_std_array_ : std::false_type { };
template<typename T, size_t N>
struct is_std_array_<std::array<T,N>> : std::true_type { };
template<typename T>
using is_std_array = is_std_array_<std::remove_cv_t<T>>;
template<typename T, typename = void>
struct has_size_and_data : std::false_type { };
template<typename T>
struct has_size_and_data<T,
void_t<decltype(al::size(std::declval<T>())), decltype(al::data(std::declval<T>()))>>
: std::true_type { };
} // namespace detail_
#define REQUIRES(...) bool rt_=true, std::enable_if_t<rt_ && (__VA_ARGS__),bool> = true
#define IS_VALID_CONTAINER(C) \
!detail_::is_span<C>::value && !detail_::is_std_array<C>::value && \
!std::is_array<C>::value && detail_::has_size_and_data<C>::value && \
std::is_convertible<std::remove_pointer_t<decltype(al::data(std::declval<C&>()))>(*)[],element_type(*)[]>::value
template<typename T, size_t E>
class span {
public:
using element_type = T;
using value_type = std::remove_cv_t<T>;
using index_type = size_t;
using difference_type = ptrdiff_t;
using pointer = T*;
using const_pointer = const T*;
using reference = T&;
using const_reference = const T&;
using iterator = pointer;
using const_iterator = const_pointer;
using reverse_iterator = std::reverse_iterator<iterator>;
using const_reverse_iterator = std::reverse_iterator<const_iterator>;
static constexpr size_t extent{E};
template<REQUIRES(extent==0)>
constexpr span() noexcept { }
constexpr span(pointer ptr, index_type /*count*/) : mData{ptr} { }
constexpr span(pointer first, pointer /*last*/) : mData{first} { }
constexpr span(element_type (&arr)[E]) noexcept : span{al::data(arr), al::size(arr)} { }
constexpr span(std::array<value_type,E> &arr) noexcept : span{al::data(arr), al::size(arr)} { }
template<REQUIRES(std::is_const<element_type>::value)>
constexpr span(const std::array<value_type,E> &arr) noexcept
: span{al::data(arr), al::size(arr)}
{ }
template<typename U, REQUIRES(IS_VALID_CONTAINER(U))>
constexpr span(U &cont) : span{al::data(cont), al::size(cont)} { }
template<typename U, REQUIRES(IS_VALID_CONTAINER(const U))>
constexpr span(const U &cont) : span{al::data(cont), al::size(cont)} { }
template<typename U, REQUIRES(!std::is_same<element_type,U>::value
&& std::is_convertible<U(*)[],element_type(*)[]>::value)>
constexpr span(const span<U,E> &span_) noexcept : span{al::data(span_), al::size(span_)} { }
constexpr span(const span&) noexcept = default;
constexpr span& operator=(const span &rhs) noexcept = default;
constexpr reference front() const { return *mData; }
constexpr reference back() const { return *(mData+E-1); }
constexpr reference operator[](index_type idx) const { return mData[idx]; }
constexpr pointer data() const noexcept { return mData; }
constexpr index_type size() const noexcept { return E; }
constexpr index_type size_bytes() const noexcept { return E * sizeof(value_type); }
constexpr bool empty() const noexcept { return E == 0; }
constexpr iterator begin() const noexcept { return mData; }
constexpr iterator end() const noexcept { return mData+E; }
constexpr const_iterator cbegin() const noexcept { return mData; }
constexpr const_iterator cend() const noexcept { return mData+E; }
constexpr reverse_iterator rbegin() const noexcept { return reverse_iterator{end()}; }
constexpr reverse_iterator rend() const noexcept { return reverse_iterator{begin()}; }
constexpr const_reverse_iterator crbegin() const noexcept
{ return const_reverse_iterator{cend()}; }
constexpr const_reverse_iterator crend() const noexcept
{ return const_reverse_iterator{cbegin()}; }
template<size_t C>
constexpr span<element_type,C> first() const
{
static_assert(E >= C, "New size exceeds original capacity");
return span<element_type,C>{mData, C};
}
template<size_t C>
constexpr span<element_type,C> last() const
{
static_assert(E >= C, "New size exceeds original capacity");
return span<element_type,C>{mData+(E-C), C};
}
template<size_t O, size_t C>
constexpr auto subspan() const -> std::enable_if_t<C!=dynamic_extent,span<element_type,C>>
{
static_assert(E >= O, "Offset exceeds extent");
static_assert(E-O >= C, "New size exceeds original capacity");
return span<element_type,C>{mData+O, C};
}
template<size_t O, size_t C=dynamic_extent>
constexpr auto subspan() const -> std::enable_if_t<C==dynamic_extent,span<element_type,E-O>>
{
static_assert(E >= O, "Offset exceeds extent");
return span<element_type,E-O>{mData+O, E-O};
}
/* NOTE: Can't declare objects of a specialized template class prior to
* defining the specialization. As a result, these methods need to be
* defined later.
*/
constexpr span<element_type,dynamic_extent> first(size_t count) const;
constexpr span<element_type,dynamic_extent> last(size_t count) const;
constexpr span<element_type,dynamic_extent> subspan(size_t offset,
size_t count=dynamic_extent) const;
private:
pointer mData{nullptr};
};
template<typename T>
class span<T,dynamic_extent> {
public:
using element_type = T;
using value_type = std::remove_cv_t<T>;
using index_type = size_t;
using difference_type = ptrdiff_t;
using pointer = T*;
using const_pointer = const T*;
using reference = T&;
using const_reference = const T&;
using iterator = pointer;
using const_iterator = const_pointer;
using reverse_iterator = std::reverse_iterator<iterator>;
using const_reverse_iterator = std::reverse_iterator<const_iterator>;
static constexpr size_t extent{dynamic_extent};
constexpr span() noexcept = default;
constexpr span(pointer ptr, index_type count) : mData{ptr}, mDataEnd{ptr+count} { }
constexpr span(pointer first, pointer last) : mData{first}, mDataEnd{last} { }
template<size_t N>
constexpr span(element_type (&arr)[N]) noexcept : span{al::data(arr), al::size(arr)} { }
template<size_t N>
constexpr span(std::array<value_type,N> &arr) noexcept : span{al::data(arr), al::size(arr)} { }
template<size_t N, REQUIRES(std::is_const<element_type>::value)>
constexpr span(const std::array<value_type,N> &arr) noexcept
: span{al::data(arr), al::size(arr)}
{ }
template<typename U, REQUIRES(IS_VALID_CONTAINER(U))>
constexpr span(U &cont) : span{al::data(cont), al::size(cont)} { }
template<typename U, REQUIRES(IS_VALID_CONTAINER(const U))>
constexpr span(const U &cont) : span{al::data(cont), al::size(cont)} { }
template<typename U, size_t N, REQUIRES((!std::is_same<element_type,U>::value || extent != N)
&& std::is_convertible<U(*)[],element_type(*)[]>::value)>
constexpr span(const span<U,N> &span_) noexcept : span{al::data(span_), al::size(span_)} { }
constexpr span(const span&) noexcept = default;
constexpr span& operator=(const span &rhs) noexcept = default;
constexpr reference front() const { return *mData; }
constexpr reference back() const { return *(mDataEnd-1); }
constexpr reference operator[](index_type idx) const { return mData[idx]; }
constexpr pointer data() const noexcept { return mData; }
constexpr index_type size() const noexcept { return static_cast<index_type>(mDataEnd-mData); }
constexpr index_type size_bytes() const noexcept
{ return static_cast<index_type>(mDataEnd-mData) * sizeof(value_type); }
constexpr bool empty() const noexcept { return mData == mDataEnd; }
constexpr iterator begin() const noexcept { return mData; }
constexpr iterator end() const noexcept { return mDataEnd; }
constexpr const_iterator cbegin() const noexcept { return mData; }
constexpr const_iterator cend() const noexcept { return mDataEnd; }
constexpr reverse_iterator rbegin() const noexcept { return reverse_iterator{end()}; }
constexpr reverse_iterator rend() const noexcept { return reverse_iterator{begin()}; }
constexpr const_reverse_iterator crbegin() const noexcept
{ return const_reverse_iterator{cend()}; }
constexpr const_reverse_iterator crend() const noexcept
{ return const_reverse_iterator{cbegin()}; }
template<size_t C>
constexpr span<element_type,C> first() const
{ return span<element_type,C>{mData, C}; }
constexpr span first(size_t count) const
{ return (count >= size()) ? *this : span{mData, mData+count}; }
template<size_t C>
constexpr span<element_type,C> last() const
{ return span<element_type,C>{mDataEnd-C, C}; }
constexpr span last(size_t count) const
{ return (count >= size()) ? *this : span{mDataEnd-count, mDataEnd}; }
template<size_t O, size_t C>
constexpr auto subspan() const -> std::enable_if_t<C!=dynamic_extent,span<element_type,C>>
{ return span<element_type,C>{mData+O, C}; }
template<size_t O, size_t C=dynamic_extent>
constexpr auto subspan() const -> std::enable_if_t<C==dynamic_extent,span<element_type,C>>
{ return span<element_type,C>{mData+O, mDataEnd}; }
constexpr span subspan(size_t offset, size_t count=dynamic_extent) const
{
return (offset > size()) ? span{} :
(count >= size()-offset) ? span{mData+offset, mDataEnd} :
span{mData+offset, mData+offset+count};
}
private:
pointer mData{nullptr};
pointer mDataEnd{nullptr};
};
template<typename T, size_t E>
constexpr inline auto span<T,E>::first(size_t count) const -> span<element_type,dynamic_extent>
{
return (count >= size()) ? span<element_type>{mData, extent} :
span<element_type>{mData, count};
}
template<typename T, size_t E>
constexpr inline auto span<T,E>::last(size_t count) const -> span<element_type,dynamic_extent>
{
return (count >= size()) ? span<element_type>{mData, extent} :
span<element_type>{mData+extent-count, count};
}
template<typename T, size_t E>
constexpr inline auto span<T,E>::subspan(size_t offset, size_t count) const
-> span<element_type,dynamic_extent>
{
return (offset > size()) ? span<element_type>{} :
(count >= size()-offset) ? span<element_type>{mData+offset, mData+extent} :
span<element_type>{mData+offset, mData+offset+count};
}
#undef IS_VALID_CONTAINER
#undef REQUIRES
} // namespace al
#endif /* AL_SPAN_H */

View file

@ -0,0 +1,45 @@
#include "config.h"
#include "alstring.h"
#include <cctype>
#include <string>
namespace {
int to_upper(const char ch)
{
using char8_traits = std::char_traits<char>;
return std::toupper(char8_traits::to_int_type(ch));
}
} // namespace
namespace al {
int strcasecmp(const char *str0, const char *str1) noexcept
{
do {
const int diff{to_upper(*str0) - to_upper(*str1)};
if(diff < 0) return -1;
if(diff > 0) return 1;
} while(*(str0++) && *(str1++));
return 0;
}
int strncasecmp(const char *str0, const char *str1, std::size_t len) noexcept
{
if(len > 0)
{
do {
const int diff{to_upper(*str0) - to_upper(*str1)};
if(diff < 0) return -1;
if(diff > 0) return 1;
} while(--len && *(str0++) && *(str1++));
}
return 0;
}
} // namespace al

View file

@ -0,0 +1,30 @@
#ifndef AL_STRING_H
#define AL_STRING_H
#include <cstddef>
#include <cstring>
#include <string>
#include "almalloc.h"
namespace al {
template<typename T, typename Tr=std::char_traits<T>>
using basic_string = std::basic_string<T, Tr, al::allocator<T>>;
using string = basic_string<char>;
using wstring = basic_string<wchar_t>;
using u16string = basic_string<char16_t>;
using u32string = basic_string<char32_t>;
/* These would be better served by using a string_view-like span/view with
* case-insensitive char traits.
*/
int strcasecmp(const char *str0, const char *str1) noexcept;
int strncasecmp(const char *str0, const char *str1, std::size_t len) noexcept;
} // namespace al
#endif /* AL_STRING_H */

View file

@ -1,10 +0,0 @@
#include "config.h"
#include "atomic.h"
extern inline void InitRef(RefCount *ptr, uint value);
extern inline uint ReadRef(RefCount *ptr);
extern inline uint IncrementRef(RefCount *ptr);
extern inline uint DecrementRef(RefCount *ptr);

View file

@ -1,439 +1,33 @@
#ifndef AL_ATOMIC_H
#define AL_ATOMIC_H
#include "static_assert.h"
#include "bool.h"
#include <atomic>
#ifdef __GNUC__
/* This helps cast away the const-ness of a pointer without accidentally
* changing the pointer type. This is necessary due to Clang's inability to use
* atomic_load on a const _Atomic variable.
*/
#define CONST_CAST(T, V) __extension__({ \
const T _tmp = (V); \
(T)_tmp; \
})
#else
#define CONST_CAST(T, V) ((T)(V))
#endif
#ifdef __cplusplus
extern "C" {
#endif
using RefCount = std::atomic<unsigned int>;
/* Atomics using C11 */
#ifdef HAVE_C11_ATOMIC
#include <stdatomic.h>
#define almemory_order memory_order
#define almemory_order_relaxed memory_order_relaxed
#define almemory_order_consume memory_order_consume
#define almemory_order_acquire memory_order_acquire
#define almemory_order_release memory_order_release
#define almemory_order_acq_rel memory_order_acq_rel
#define almemory_order_seq_cst memory_order_seq_cst
#define ATOMIC(T) T _Atomic
#define ATOMIC_FLAG atomic_flag
#define ATOMIC_INIT atomic_init
#define ATOMIC_INIT_STATIC ATOMIC_VAR_INIT
/*#define ATOMIC_FLAG_INIT ATOMIC_FLAG_INIT*/
#define ATOMIC_LOAD atomic_load_explicit
#define ATOMIC_STORE atomic_store_explicit
#define ATOMIC_ADD atomic_fetch_add_explicit
#define ATOMIC_SUB atomic_fetch_sub_explicit
#define ATOMIC_EXCHANGE atomic_exchange_explicit
#define ATOMIC_COMPARE_EXCHANGE_STRONG atomic_compare_exchange_strong_explicit
#define ATOMIC_COMPARE_EXCHANGE_WEAK atomic_compare_exchange_weak_explicit
#define ATOMIC_FLAG_TEST_AND_SET atomic_flag_test_and_set_explicit
#define ATOMIC_FLAG_CLEAR atomic_flag_clear_explicit
#define ATOMIC_THREAD_FENCE atomic_thread_fence
/* Atomics using GCC intrinsics */
#elif defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 1)) && !defined(__QNXNTO__)
enum almemory_order {
almemory_order_relaxed,
almemory_order_consume,
almemory_order_acquire,
almemory_order_release,
almemory_order_acq_rel,
almemory_order_seq_cst
};
#define ATOMIC(T) struct { T volatile value; }
#define ATOMIC_FLAG ATOMIC(int)
#define ATOMIC_INIT(_val, _newval) do { (_val)->value = (_newval); } while(0)
#define ATOMIC_INIT_STATIC(_newval) {(_newval)}
#define ATOMIC_FLAG_INIT ATOMIC_INIT_STATIC(0)
#define ATOMIC_LOAD(_val, _MO) __extension__({ \
__typeof((_val)->value) _r = (_val)->value; \
__asm__ __volatile__("" ::: "memory"); \
_r; \
})
#define ATOMIC_STORE(_val, _newval, _MO) do { \
__asm__ __volatile__("" ::: "memory"); \
(_val)->value = (_newval); \
} while(0)
#define ATOMIC_ADD(_val, _incr, _MO) __sync_fetch_and_add(&(_val)->value, (_incr))
#define ATOMIC_SUB(_val, _decr, _MO) __sync_fetch_and_sub(&(_val)->value, (_decr))
#define ATOMIC_EXCHANGE(_val, _newval, _MO) __extension__({ \
__asm__ __volatile__("" ::: "memory"); \
__sync_lock_test_and_set(&(_val)->value, (_newval)); \
})
#define ATOMIC_COMPARE_EXCHANGE_STRONG(_val, _oldval, _newval, _MO1, _MO2) __extension__({ \
__typeof(*(_oldval)) _o = *(_oldval); \
*(_oldval) = __sync_val_compare_and_swap(&(_val)->value, _o, (_newval)); \
*(_oldval) == _o; \
})
#define ATOMIC_FLAG_TEST_AND_SET(_val, _MO) __extension__({ \
__asm__ __volatile__("" ::: "memory"); \
__sync_lock_test_and_set(&(_val)->value, 1); \
})
#define ATOMIC_FLAG_CLEAR(_val, _MO) __extension__({ \
__sync_lock_release(&(_val)->value); \
__asm__ __volatile__("" ::: "memory"); \
})
#define ATOMIC_THREAD_FENCE(order) do { \
enum { must_be_constant = (order) }; \
const int _o = must_be_constant; \
if(_o > almemory_order_relaxed) \
__asm__ __volatile__("" ::: "memory"); \
} while(0)
/* Atomics using x86/x86-64 GCC inline assembly */
#elif defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
#define WRAP_ADD(S, ret, dest, incr) __asm__ __volatile__( \
"lock; xadd"S" %0,(%1)" \
: "=r" (ret) \
: "r" (dest), "0" (incr) \
: "memory" \
)
#define WRAP_SUB(S, ret, dest, decr) __asm__ __volatile__( \
"lock; xadd"S" %0,(%1)" \
: "=r" (ret) \
: "r" (dest), "0" (-(decr)) \
: "memory" \
)
#define WRAP_XCHG(S, ret, dest, newval) __asm__ __volatile__( \
"lock; xchg"S" %0,(%1)" \
: "=r" (ret) \
: "r" (dest), "0" (newval) \
: "memory" \
)
#define WRAP_CMPXCHG(S, ret, dest, oldval, newval) __asm__ __volatile__( \
"lock; cmpxchg"S" %2,(%1)" \
: "=a" (ret) \
: "r" (dest), "r" (newval), "0" (oldval) \
: "memory" \
)
enum almemory_order {
almemory_order_relaxed,
almemory_order_consume,
almemory_order_acquire,
almemory_order_release,
almemory_order_acq_rel,
almemory_order_seq_cst
};
#define ATOMIC(T) struct { T volatile value; }
#define ATOMIC_INIT(_val, _newval) do { (_val)->value = (_newval); } while(0)
#define ATOMIC_INIT_STATIC(_newval) {(_newval)}
#define ATOMIC_LOAD(_val, _MO) __extension__({ \
__typeof((_val)->value) _r = (_val)->value; \
__asm__ __volatile__("" ::: "memory"); \
_r; \
})
#define ATOMIC_STORE(_val, _newval, _MO) do { \
__asm__ __volatile__("" ::: "memory"); \
(_val)->value = (_newval); \
} while(0)
#define ATOMIC_ADD(_val, _incr, _MO) __extension__({ \
static_assert(sizeof((_val)->value)==4 || sizeof((_val)->value)==8, "Unsupported size!"); \
__typeof((_val)->value) _r; \
if(sizeof((_val)->value) == 4) WRAP_ADD("l", _r, &(_val)->value, _incr); \
else if(sizeof((_val)->value) == 8) WRAP_ADD("q", _r, &(_val)->value, _incr); \
_r; \
})
#define ATOMIC_SUB(_val, _decr, _MO) __extension__({ \
static_assert(sizeof((_val)->value)==4 || sizeof((_val)->value)==8, "Unsupported size!"); \
__typeof((_val)->value) _r; \
if(sizeof((_val)->value) == 4) WRAP_SUB("l", _r, &(_val)->value, _decr); \
else if(sizeof((_val)->value) == 8) WRAP_SUB("q", _r, &(_val)->value, _decr); \
_r; \
})
#define ATOMIC_EXCHANGE(_val, _newval, _MO) __extension__({ \
__typeof((_val)->value) _r; \
if(sizeof((_val)->value) == 4) WRAP_XCHG("l", _r, &(_val)->value, (_newval)); \
else if(sizeof((_val)->value) == 8) WRAP_XCHG("q", _r, &(_val)->value, (_newval)); \
_r; \
})
#define ATOMIC_COMPARE_EXCHANGE_STRONG(_val, _oldval, _newval, _MO1, _MO2) __extension__({ \
__typeof(*(_oldval)) _old = *(_oldval); \
if(sizeof((_val)->value) == 4) WRAP_CMPXCHG("l", *(_oldval), &(_val)->value, _old, (_newval)); \
else if(sizeof((_val)->value) == 8) WRAP_CMPXCHG("q", *(_oldval), &(_val)->value, _old, (_newval)); \
*(_oldval) == _old; \
})
#define ATOMIC_EXCHANGE_PTR(_val, _newval, _MO) __extension__({ \
void *_r; \
if(sizeof(void*) == 4) WRAP_XCHG("l", _r, &(_val)->value, (_newval)); \
else if(sizeof(void*) == 8) WRAP_XCHG("q", _r, &(_val)->value, (_newval));\
_r; \
})
#define ATOMIC_COMPARE_EXCHANGE_PTR_STRONG(_val, _oldval, _newval, _MO1, _MO2) __extension__({ \
void *_old = *(_oldval); \
if(sizeof(void*) == 4) WRAP_CMPXCHG("l", *(_oldval), &(_val)->value, _old, (_newval)); \
else if(sizeof(void*) == 8) WRAP_CMPXCHG("q", *(_oldval), &(_val)->value, _old, (_newval)); \
*(_oldval) == _old; \
})
#define ATOMIC_THREAD_FENCE(order) do { \
enum { must_be_constant = (order) }; \
const int _o = must_be_constant; \
if(_o > almemory_order_relaxed) \
__asm__ __volatile__("" ::: "memory"); \
} while(0)
/* Atomics using Windows methods */
#elif defined(_WIN32)
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
/* NOTE: This mess is *extremely* touchy. It lacks quite a bit of safety
* checking due to the lack of multi-statement expressions, typeof(), and C99
* compound literals. It is incapable of properly exchanging floats, which get
* casted to LONG/int, and could cast away potential warnings.
*
* Unfortunately, it's the only semi-safe way that doesn't rely on C99 (because
* MSVC).
*/
inline LONG AtomicAdd32(volatile LONG *dest, LONG incr)
{
return InterlockedExchangeAdd(dest, incr);
}
inline LONGLONG AtomicAdd64(volatile LONGLONG *dest, LONGLONG incr)
{
return InterlockedExchangeAdd64(dest, incr);
}
inline LONG AtomicSub32(volatile LONG *dest, LONG decr)
{
return InterlockedExchangeAdd(dest, -decr);
}
inline LONGLONG AtomicSub64(volatile LONGLONG *dest, LONGLONG decr)
{
return InterlockedExchangeAdd64(dest, -decr);
}
inline LONG AtomicSwap32(volatile LONG *dest, LONG newval)
{
return InterlockedExchange(dest, newval);
}
inline LONGLONG AtomicSwap64(volatile LONGLONG *dest, LONGLONG newval)
{
return InterlockedExchange64(dest, newval);
}
inline void *AtomicSwapPtr(void *volatile *dest, void *newval)
{
return InterlockedExchangePointer(dest, newval);
}
inline bool CompareAndSwap32(volatile LONG *dest, LONG newval, LONG *oldval)
{
LONG old = *oldval;
*oldval = InterlockedCompareExchange(dest, newval, *oldval);
return old == *oldval;
}
inline bool CompareAndSwap64(volatile LONGLONG *dest, LONGLONG newval, LONGLONG *oldval)
{
LONGLONG old = *oldval;
*oldval = InterlockedCompareExchange64(dest, newval, *oldval);
return old == *oldval;
}
inline bool CompareAndSwapPtr(void *volatile *dest, void *newval, void **oldval)
{
void *old = *oldval;
*oldval = InterlockedCompareExchangePointer(dest, newval, *oldval);
return old == *oldval;
}
#define WRAP_ADDSUB(T, _func, _ptr, _amnt) _func((T volatile*)(_ptr), (_amnt))
#define WRAP_XCHG(T, _func, _ptr, _newval) _func((T volatile*)(_ptr), (_newval))
#define WRAP_CMPXCHG(T, _func, _ptr, _newval, _oldval) _func((T volatile*)(_ptr), (_newval), (T*)(_oldval))
enum almemory_order {
almemory_order_relaxed,
almemory_order_consume,
almemory_order_acquire,
almemory_order_release,
almemory_order_acq_rel,
almemory_order_seq_cst
};
#define ATOMIC(T) struct { T volatile value; }
#define ATOMIC_INIT(_val, _newval) do { (_val)->value = (_newval); } while(0)
#define ATOMIC_INIT_STATIC(_newval) {(_newval)}
#define ATOMIC_LOAD(_val, _MO) ((_val)->value)
#define ATOMIC_STORE(_val, _newval, _MO) do { \
(_val)->value = (_newval); \
} while(0)
int _al_invalid_atomic_size(); /* not defined */
void *_al_invalid_atomic_ptr_size(); /* not defined */
#define ATOMIC_ADD(_val, _incr, _MO) \
((sizeof((_val)->value)==4) ? WRAP_ADDSUB(LONG, AtomicAdd32, &(_val)->value, (_incr)) : \
(sizeof((_val)->value)==8) ? WRAP_ADDSUB(LONGLONG, AtomicAdd64, &(_val)->value, (_incr)) : \
_al_invalid_atomic_size())
#define ATOMIC_SUB(_val, _decr, _MO) \
((sizeof((_val)->value)==4) ? WRAP_ADDSUB(LONG, AtomicSub32, &(_val)->value, (_decr)) : \
(sizeof((_val)->value)==8) ? WRAP_ADDSUB(LONGLONG, AtomicSub64, &(_val)->value, (_decr)) : \
_al_invalid_atomic_size())
#define ATOMIC_EXCHANGE(_val, _newval, _MO) \
((sizeof((_val)->value)==4) ? WRAP_XCHG(LONG, AtomicSwap32, &(_val)->value, (_newval)) : \
(sizeof((_val)->value)==8) ? WRAP_XCHG(LONGLONG, AtomicSwap64, &(_val)->value, (_newval)) : \
(LONG)_al_invalid_atomic_size())
#define ATOMIC_COMPARE_EXCHANGE_STRONG(_val, _oldval, _newval, _MO1, _MO2) \
((sizeof((_val)->value)==4) ? WRAP_CMPXCHG(LONG, CompareAndSwap32, &(_val)->value, (_newval), (_oldval)) : \
(sizeof((_val)->value)==8) ? WRAP_CMPXCHG(LONGLONG, CompareAndSwap64, &(_val)->value, (_newval), (_oldval)) : \
(bool)_al_invalid_atomic_size())
#define ATOMIC_EXCHANGE_PTR(_val, _newval, _MO) \
((sizeof((_val)->value)==sizeof(void*)) ? AtomicSwapPtr((void*volatile*)&(_val)->value, (_newval)) : \
_al_invalid_atomic_ptr_size())
#define ATOMIC_COMPARE_EXCHANGE_PTR_STRONG(_val, _oldval, _newval, _MO1, _MO2)\
((sizeof((_val)->value)==sizeof(void*)) ? CompareAndSwapPtr((void*volatile*)&(_val)->value, (_newval), (void**)(_oldval)) : \
(bool)_al_invalid_atomic_size())
#define ATOMIC_THREAD_FENCE(order) do { \
enum { must_be_constant = (order) }; \
const int _o = must_be_constant; \
if(_o > almemory_order_relaxed) \
_ReadWriteBarrier(); \
} while(0)
#else
#error "No atomic functions available on this platform!"
#define ATOMIC(T) T
#define ATOMIC_INIT(_val, _newval) ((void)0)
#define ATOMIC_INIT_STATIC(_newval) (0)
#define ATOMIC_LOAD(...) (0)
#define ATOMIC_STORE(...) ((void)0)
#define ATOMIC_ADD(...) (0)
#define ATOMIC_SUB(...) (0)
#define ATOMIC_EXCHANGE(...) (0)
#define ATOMIC_COMPARE_EXCHANGE_STRONG(...) (0)
#define ATOMIC_THREAD_FENCE(...) ((void)0)
#endif
/* If no PTR xchg variants are provided, the normal ones can handle it. */
#ifndef ATOMIC_EXCHANGE_PTR
#define ATOMIC_EXCHANGE_PTR ATOMIC_EXCHANGE
#define ATOMIC_COMPARE_EXCHANGE_PTR_STRONG ATOMIC_COMPARE_EXCHANGE_STRONG
#define ATOMIC_COMPARE_EXCHANGE_PTR_WEAK ATOMIC_COMPARE_EXCHANGE_WEAK
#endif
/* If no weak cmpxchg is provided (not all systems will have one), substitute a
* strong cmpxchg. */
#ifndef ATOMIC_COMPARE_EXCHANGE_WEAK
#define ATOMIC_COMPARE_EXCHANGE_WEAK ATOMIC_COMPARE_EXCHANGE_STRONG
#endif
#ifndef ATOMIC_COMPARE_EXCHANGE_PTR_WEAK
#define ATOMIC_COMPARE_EXCHANGE_PTR_WEAK ATOMIC_COMPARE_EXCHANGE_PTR_STRONG
#endif
/* If no ATOMIC_FLAG is defined, simulate one with an atomic int using exchange
* and store ops.
*/
#ifndef ATOMIC_FLAG
#define ATOMIC_FLAG ATOMIC(int)
#define ATOMIC_FLAG_INIT ATOMIC_INIT_STATIC(0)
#define ATOMIC_FLAG_TEST_AND_SET(_val, _MO) ATOMIC_EXCHANGE(_val, 1, _MO)
#define ATOMIC_FLAG_CLEAR(_val, _MO) ATOMIC_STORE(_val, 0, _MO)
#endif
#define ATOMIC_LOAD_SEQ(_val) ATOMIC_LOAD(_val, almemory_order_seq_cst)
#define ATOMIC_STORE_SEQ(_val, _newval) ATOMIC_STORE(_val, _newval, almemory_order_seq_cst)
#define ATOMIC_ADD_SEQ(_val, _incr) ATOMIC_ADD(_val, _incr, almemory_order_seq_cst)
#define ATOMIC_SUB_SEQ(_val, _decr) ATOMIC_SUB(_val, _decr, almemory_order_seq_cst)
#define ATOMIC_EXCHANGE_SEQ(_val, _newval) ATOMIC_EXCHANGE(_val, _newval, almemory_order_seq_cst)
#define ATOMIC_COMPARE_EXCHANGE_STRONG_SEQ(_val, _oldval, _newval) \
ATOMIC_COMPARE_EXCHANGE_STRONG(_val, _oldval, _newval, almemory_order_seq_cst, almemory_order_seq_cst)
#define ATOMIC_COMPARE_EXCHANGE_WEAK_SEQ(_val, _oldval, _newval) \
ATOMIC_COMPARE_EXCHANGE_WEAK(_val, _oldval, _newval, almemory_order_seq_cst, almemory_order_seq_cst)
#define ATOMIC_EXCHANGE_PTR_SEQ(_val, _newval) ATOMIC_EXCHANGE_PTR(_val, _newval, almemory_order_seq_cst)
#define ATOMIC_COMPARE_EXCHANGE_PTR_STRONG_SEQ(_val, _oldval, _newval) \
ATOMIC_COMPARE_EXCHANGE_PTR_STRONG(_val, _oldval, _newval, almemory_order_seq_cst, almemory_order_seq_cst)
#define ATOMIC_COMPARE_EXCHANGE_PTR_WEAK_SEQ(_val, _oldval, _newval) \
ATOMIC_COMPARE_EXCHANGE_PTR_WEAK(_val, _oldval, _newval, almemory_order_seq_cst, almemory_order_seq_cst)
typedef unsigned int uint;
typedef ATOMIC(uint) RefCount;
inline void InitRef(RefCount *ptr, uint value)
{ ATOMIC_INIT(ptr, value); }
inline uint ReadRef(RefCount *ptr)
{ return ATOMIC_LOAD(ptr, almemory_order_acquire); }
inline uint IncrementRef(RefCount *ptr)
{ return ATOMIC_ADD(ptr, 1, almemory_order_acq_rel)+1; }
inline uint DecrementRef(RefCount *ptr)
{ return ATOMIC_SUB(ptr, 1, almemory_order_acq_rel)-1; }
inline void InitRef(RefCount &ref, unsigned int value)
{ ref.store(value, std::memory_order_relaxed); }
inline unsigned int ReadRef(RefCount &ref)
{ return ref.load(std::memory_order_acquire); }
inline unsigned int IncrementRef(RefCount &ref)
{ return ref.fetch_add(1u, std::memory_order_acq_rel)+1u; }
inline unsigned int DecrementRef(RefCount &ref)
{ return ref.fetch_sub(1u, std::memory_order_acq_rel)-1u; }
/* WARNING: A livelock is theoretically possible if another thread keeps
* changing the head without giving this a chance to actually swap in the new
* one (practically impossible with this little code, but...).
*/
#define ATOMIC_REPLACE_HEAD(T, _head, _entry) do { \
T _first = ATOMIC_LOAD(_head, almemory_order_acquire); \
do { \
ATOMIC_STORE(&(_entry)->next, _first, almemory_order_relaxed); \
} while(ATOMIC_COMPARE_EXCHANGE_PTR_WEAK(_head, &_first, _entry, \
almemory_order_acq_rel, almemory_order_acquire) == 0); \
} while(0)
#ifdef __cplusplus
template<typename T>
inline void AtomicReplaceHead(std::atomic<T> &head, T newhead)
{
T first_ = head.load(std::memory_order_acquire);
do {
newhead->next.store(first_, std::memory_order_relaxed);
} while(!head.compare_exchange_weak(first_, newhead,
std::memory_order_acq_rel, std::memory_order_acquire));
}
#endif
#endif /* AL_ATOMIC_H */

View file

@ -1,18 +0,0 @@
#ifndef AL_BOOL_H
#define AL_BOOL_H
#ifdef HAVE_STDBOOL_H
#include <stdbool.h>
#endif
#ifndef bool
#ifdef HAVE_C99_BOOL
#define bool _Bool
#else
#define bool int
#endif
#define false 0
#define true 1
#endif
#endif /* AL_BOOL_H */

View file

@ -0,0 +1,44 @@
#include "config.h"
#include "dynload.h"
#include "strutils.h"
#ifdef _WIN32
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
void *LoadLib(const char *name)
{
std::wstring wname{utf8_to_wstr(name)};
return LoadLibraryW(wname.c_str());
}
void CloseLib(void *handle)
{ FreeLibrary(static_cast<HMODULE>(handle)); }
void *GetSymbol(void *handle, const char *name)
{ return reinterpret_cast<void*>(GetProcAddress(static_cast<HMODULE>(handle), name)); }
#elif defined(HAVE_DLFCN_H)
#include <dlfcn.h>
void *LoadLib(const char *name)
{
dlerror();
void *handle{dlopen(name, RTLD_NOW)};
const char *err{dlerror()};
if(err) handle = nullptr;
return handle;
}
void CloseLib(void *handle)
{ dlclose(handle); }
void *GetSymbol(void *handle, const char *name)
{
dlerror();
void *sym{dlsym(handle, name)};
const char *err{dlerror()};
if(err) sym = nullptr;
return sym;
}
#endif

View file

@ -0,0 +1,14 @@
#ifndef AL_DYNLOAD_H
#define AL_DYNLOAD_H
#if defined(_WIN32) || defined(HAVE_DLFCN_H)
#define HAVE_DYNLOAD
void *LoadLib(const char *name);
void CloseLib(void *handle);
void *GetSymbol(void *handle, const char *name);
#endif
#endif /* AL_DYNLOAD_H */

View file

@ -0,0 +1,120 @@
#ifndef INTRUSIVE_PTR_H
#define INTRUSIVE_PTR_H
#include "atomic.h"
#include "opthelpers.h"
namespace al {
template<typename T>
class intrusive_ref {
RefCount mRef{1u};
public:
unsigned int add_ref() noexcept { return IncrementRef(mRef); }
unsigned int release() noexcept
{
auto ref = DecrementRef(mRef);
if UNLIKELY(ref == 0)
delete static_cast<T*>(this);
return ref;
}
/**
* Release only if doing so would not bring the object to 0 references and
* delete it. Returns false if the object could not be released.
*
* NOTE: The caller is responsible for handling a failed release, as it
* means the object has no other references and needs to be be deleted
* somehow.
*/
bool releaseIfNoDelete() noexcept
{
auto val = mRef.load(std::memory_order_acquire);
while(val > 1 && !mRef.compare_exchange_strong(val, val-1, std::memory_order_acq_rel))
{
/* val was updated with the current value on failure, so just try
* again.
*/
}
return val >= 2;
}
};
template<typename T>
class intrusive_ptr {
T *mPtr{nullptr};
public:
intrusive_ptr() noexcept = default;
intrusive_ptr(const intrusive_ptr &rhs) noexcept : mPtr{rhs.mPtr}
{ if(mPtr) mPtr->add_ref(); }
intrusive_ptr(intrusive_ptr&& rhs) noexcept : mPtr{rhs.mPtr}
{ rhs.mPtr = nullptr; }
intrusive_ptr(std::nullptr_t) noexcept { }
explicit intrusive_ptr(T *ptr) noexcept : mPtr{ptr} { }
~intrusive_ptr() { if(mPtr) mPtr->release(); }
intrusive_ptr& operator=(const intrusive_ptr &rhs) noexcept
{
if(rhs.mPtr) rhs.mPtr->add_ref();
if(mPtr) mPtr->release();
mPtr = rhs.mPtr;
return *this;
}
intrusive_ptr& operator=(intrusive_ptr&& rhs) noexcept
{
if(mPtr)
mPtr->release();
mPtr = rhs.mPtr;
rhs.mPtr = nullptr;
return *this;
}
operator bool() const noexcept { return mPtr != nullptr; }
T& operator*() const noexcept { return *mPtr; }
T* operator->() const noexcept { return mPtr; }
T* get() const noexcept { return mPtr; }
void reset(T *ptr=nullptr) noexcept
{
if(mPtr)
mPtr->release();
mPtr = ptr;
}
T* release() noexcept
{
T *ret{mPtr};
mPtr = nullptr;
return ret;
}
void swap(intrusive_ptr &rhs) noexcept { std::swap(mPtr, rhs.mPtr); }
void swap(intrusive_ptr&& rhs) noexcept { std::swap(mPtr, rhs.mPtr); }
};
#define AL_DECL_OP(op) \
template<typename T> \
inline bool operator op(const intrusive_ptr<T> &lhs, const T *rhs) noexcept \
{ return lhs.get() op rhs; } \
template<typename T> \
inline bool operator op(const T *lhs, const intrusive_ptr<T> &rhs) noexcept \
{ return lhs op rhs.get(); }
AL_DECL_OP(==)
AL_DECL_OP(!=)
AL_DECL_OP(<=)
AL_DECL_OP(>=)
AL_DECL_OP(<)
AL_DECL_OP(>)
#undef AL_DECL_OP
} // namespace al
#endif /* INTRUSIVE_PTR_H */

View file

@ -1,46 +1,26 @@
#ifndef AL_MATH_DEFS_H
#define AL_MATH_DEFS_H
#include <math.h>
#ifdef HAVE_FLOAT_H
#include <float.h>
#endif
constexpr float Deg2Rad(float x) noexcept { return x * 1.74532925199432955e-02f/*pi/180*/; }
constexpr float Rad2Deg(float x) noexcept { return x * 5.72957795130823229e+01f/*180/pi*/; }
#ifndef M_PI
#define M_PI (3.14159265358979323846)
#endif
namespace al {
#define F_PI (3.14159265358979323846f)
#define F_PI_2 (1.57079632679489661923f)
#define F_TAU (6.28318530717958647692f)
template<typename Real>
struct MathDefs { };
#ifndef FLT_EPSILON
#define FLT_EPSILON (1.19209290e-07f)
#endif
template<>
struct MathDefs<float> {
static constexpr float Pi() noexcept { return 3.14159265358979323846e+00f; }
static constexpr float Tau() noexcept { return 6.28318530717958647692e+00f; }
};
#ifndef HUGE_VALF
static const union msvc_inf_hack {
unsigned char b[4];
float f;
} msvc_inf_union = {{ 0x00, 0x00, 0x80, 0x7F }};
#define HUGE_VALF (msvc_inf_union.f)
#endif
template<>
struct MathDefs<double> {
static constexpr double Pi() noexcept { return 3.14159265358979323846e+00; }
static constexpr double Tau() noexcept { return 6.28318530717958647692e+00; }
};
#ifndef HAVE_LOG2F
static inline float log2f(float f)
{
return logf(f) / logf(2.0f);
}
#endif
#ifndef HAVE_CBRTF
static inline float cbrtf(float f)
{
return powf(f, 1.0f/3.0f);
}
#endif
#define DEG2RAD(x) ((float)(x) * (F_PI/180.0f))
#define RAD2DEG(x) ((float)(x) * (180.0f/F_PI))
} // namespace al
#endif /* AL_MATH_DEFS_H */

View file

@ -0,0 +1,45 @@
#ifndef OPTHELPERS_H
#define OPTHELPERS_H
#ifdef __has_builtin
#define HAS_BUILTIN __has_builtin
#else
#define HAS_BUILTIN(x) (0)
#endif
#if defined(__GNUC__) || HAS_BUILTIN(__builtin_expect)
/* LIKELY optimizes the case where the condition is true. The condition is not
* required to be true, but it can result in more optimal code for the true
* path at the expense of a less optimal false path.
*/
#define LIKELY(x) (__builtin_expect(!!(x), !false))
/* The opposite of LIKELY, optimizing the case where the condition is false. */
#define UNLIKELY(x) (__builtin_expect(!!(x), false))
#else
#define LIKELY(x) (!!(x))
#define UNLIKELY(x) (!!(x))
#endif
#if HAS_BUILTIN(__builtin_assume)
/* Unlike LIKELY, ASSUME requires the condition to be true or else it invokes
* undefined behavior. It's essentially an assert without actually checking the
* condition at run-time, allowing for stronger optimizations than LIKELY.
*/
#define ASSUME __builtin_assume
#elif defined(_MSC_VER)
#define ASSUME __assume
#elif defined(__GNUC__)
#define ASSUME(x) do { if(!(x)) __builtin_unreachable(); } while(0)
#else
#define ASSUME(x) ((void)0)
#endif
#if __cplusplus >= 201703L || defined(__cpp_if_constexpr)
#define if_constexpr if constexpr
#else
#define if_constexpr if
#endif
#endif /* OPTHELPERS_H */

View file

@ -0,0 +1,222 @@
#include "polyphase_resampler.h"
#include <algorithm>
#include <cmath>
#include "math_defs.h"
#include "opthelpers.h"
namespace {
constexpr double Epsilon{1e-9};
using uint = unsigned int;
/* This is the normalized cardinal sine (sinc) function.
*
* sinc(x) = { 1, x = 0
* { sin(pi x) / (pi x), otherwise.
*/
double Sinc(const double x)
{
if UNLIKELY(std::abs(x) < Epsilon)
return 1.0;
return std::sin(al::MathDefs<double>::Pi()*x) / (al::MathDefs<double>::Pi()*x);
}
/* The zero-order modified Bessel function of the first kind, used for the
* Kaiser window.
*
* I_0(x) = sum_{k=0}^inf (1 / k!)^2 (x / 2)^(2 k)
* = sum_{k=0}^inf ((x / 2)^k / k!)^2
*/
constexpr double BesselI_0(const double x)
{
// Start at k=1 since k=0 is trivial.
const double x2{x/2.0};
double term{1.0};
double sum{1.0};
int k{1};
// Let the integration converge until the term of the sum is no longer
// significant.
double last_sum{};
do {
const double y{x2 / k};
++k;
last_sum = sum;
term *= y * y;
sum += term;
} while(sum != last_sum);
return sum;
}
/* Calculate a Kaiser window from the given beta value and a normalized k
* [-1, 1].
*
* w(k) = { I_0(B sqrt(1 - k^2)) / I_0(B), -1 <= k <= 1
* { 0, elsewhere.
*
* Where k can be calculated as:
*
* k = i / l, where -l <= i <= l.
*
* or:
*
* k = 2 i / M - 1, where 0 <= i <= M.
*/
double Kaiser(const double b, const double k)
{
if(!(k >= -1.0 && k <= 1.0))
return 0.0;
return BesselI_0(b * std::sqrt(1.0 - k*k)) / BesselI_0(b);
}
// Calculates the greatest common divisor of a and b.
constexpr uint Gcd(uint x, uint y)
{
while(y > 0)
{
const uint z{y};
y = x % y;
x = z;
}
return x;
}
/* Calculates the size (order) of the Kaiser window. Rejection is in dB and
* the transition width is normalized frequency (0.5 is nyquist).
*
* M = { ceil((r - 7.95) / (2.285 2 pi f_t)), r > 21
* { ceil(5.79 / 2 pi f_t), r <= 21.
*
*/
constexpr uint CalcKaiserOrder(const double rejection, const double transition)
{
const double w_t{2.0 * al::MathDefs<double>::Pi() * transition};
if LIKELY(rejection > 21.0)
return static_cast<uint>(std::ceil((rejection - 7.95) / (2.285 * w_t)));
return static_cast<uint>(std::ceil(5.79 / w_t));
}
// Calculates the beta value of the Kaiser window. Rejection is in dB.
constexpr double CalcKaiserBeta(const double rejection)
{
if LIKELY(rejection > 50.0)
return 0.1102 * (rejection - 8.7);
if(rejection >= 21.0)
return (0.5842 * std::pow(rejection - 21.0, 0.4)) +
(0.07886 * (rejection - 21.0));
return 0.0;
}
/* Calculates a point on the Kaiser-windowed sinc filter for the given half-
* width, beta, gain, and cutoff. The point is specified in non-normalized
* samples, from 0 to M, where M = (2 l + 1).
*
* w(k) 2 p f_t sinc(2 f_t x)
*
* x -- centered sample index (i - l)
* k -- normalized and centered window index (x / l)
* w(k) -- window function (Kaiser)
* p -- gain compensation factor when sampling
* f_t -- normalized center frequency (or cutoff; 0.5 is nyquist)
*/
double SincFilter(const uint l, const double b, const double gain, const double cutoff,
const uint i)
{
const double x{static_cast<double>(i) - l};
return Kaiser(b, x / l) * 2.0 * gain * cutoff * Sinc(2.0 * cutoff * x);
}
} // namespace
// Calculate the resampling metrics and build the Kaiser-windowed sinc filter
// that's used to cut frequencies above the destination nyquist.
void PPhaseResampler::init(const uint srcRate, const uint dstRate)
{
const uint gcd{Gcd(srcRate, dstRate)};
mP = dstRate / gcd;
mQ = srcRate / gcd;
/* The cutoff is adjusted by half the transition width, so the transition
* ends before the nyquist (0.5). Both are scaled by the downsampling
* factor.
*/
double cutoff, width;
if(mP > mQ)
{
cutoff = 0.475 / mP;
width = 0.05 / mP;
}
else
{
cutoff = 0.475 / mQ;
width = 0.05 / mQ;
}
// A rejection of -180 dB is used for the stop band. Round up when
// calculating the left offset to avoid increasing the transition width.
const uint l{(CalcKaiserOrder(180.0, width)+1) / 2};
const double beta{CalcKaiserBeta(180.0)};
mM = l*2 + 1;
mL = l;
mF.resize(mM);
for(uint i{0};i < mM;i++)
mF[i] = SincFilter(l, beta, mP, cutoff, i);
}
// Perform the upsample-filter-downsample resampling operation using a
// polyphase filter implementation.
void PPhaseResampler::process(const uint inN, const double *in, const uint outN, double *out)
{
if UNLIKELY(outN == 0)
return;
// Handle in-place operation.
std::vector<double> workspace;
double *work{out};
if UNLIKELY(work == in)
{
workspace.resize(outN);
work = workspace.data();
}
// Resample the input.
const uint p{mP}, q{mQ}, m{mM}, l{mL};
const double *f{mF.data()};
for(uint i{0};i < outN;i++)
{
// Input starts at l to compensate for the filter delay. This will
// drop any build-up from the first half of the filter.
size_t j_f{(l + q*i) % p};
size_t j_s{(l + q*i) / p};
// Only take input when 0 <= j_s < inN.
double r{0.0};
if LIKELY(j_f < m)
{
size_t filt_len{(m-j_f+p-1) / p};
if LIKELY(j_s+1 > inN)
{
size_t skip{std::min<size_t>(j_s+1 - inN, filt_len)};
j_f += p*skip;
j_s -= skip;
filt_len -= skip;
}
if(size_t todo{std::min<size_t>(j_s+1, filt_len)})
{
do {
r += f[j_f] * in[j_s];
j_f += p;
--j_s;
} while(--todo);
}
}
work[i] = r;
}
// Clean up after in-place operation.
if(work != out)
std::copy_n(work, outN, out);
}

View file

@ -0,0 +1,45 @@
#ifndef POLYPHASE_RESAMPLER_H
#define POLYPHASE_RESAMPLER_H
#include <vector>
/* This is a polyphase sinc-filtered resampler. It is built for very high
* quality results, rather than real-time performance.
*
* Upsample Downsample
*
* p/q = 3/2 p/q = 3/5
*
* M-+-+-+-> M-+-+-+->
* -------------------+ ---------------------+
* p s * f f f f|f| | p s * f f f f f |
* | 0 * 0 0 0|0|0 | | 0 * 0 0 0 0|0| |
* v 0 * 0 0|0|0 0 | v 0 * 0 0 0|0|0 |
* s * f|f|f f f | s * f f|f|f f |
* 0 * |0|0 0 0 0 | 0 * 0|0|0 0 0 |
* --------+=+--------+ 0 * |0|0 0 0 0 |
* d . d .|d|. d . d ----------+=+--------+
* d . . . .|d|. . . .
* q->
* q-+-+-+->
*
* P_f(i,j) = q i mod p + pj
* P_s(i,j) = floor(q i / p) - j
* d[i=0..N-1] = sum_{j=0}^{floor((M - 1) / p)} {
* { f[P_f(i,j)] s[P_s(i,j)], P_f(i,j) < M
* { 0, P_f(i,j) >= M. }
*/
struct PPhaseResampler {
using uint = unsigned int;
void init(const uint srcRate, const uint dstRate);
void process(const uint inN, const double *in, const uint outN, double *out);
private:
uint mP, mQ, mM, mL;
std::vector<double> mF;
};
#endif /* POLYPHASE_RESAMPLER_H */

View file

@ -0,0 +1,21 @@
#ifndef PRAGMADEFS_H
#define PRAGMADEFS_H
#if defined(_MSC_VER)
#define DIAGNOSTIC_PUSH __pragma(warning(push))
#define DIAGNOSTIC_POP __pragma(warning(pop))
#define std_pragma(...)
#define msc_pragma __pragma
#else
#if defined(__GNUC__) || defined(__clang__)
#define DIAGNOSTIC_PUSH _Pragma("GCC diagnostic push")
#define DIAGNOSTIC_POP _Pragma("GCC diagnostic pop")
#else
#define DIAGNOSTIC_PUSH
#define DIAGNOSTIC_POP
#endif
#define std_pragma _Pragma
#define msc_pragma(...)
#endif
#endif /* PRAGMADEFS_H */

View file

@ -0,0 +1,224 @@
/**
* OpenAL cross platform audio library
* Copyright (C) 1999-2007 by authors.
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
* Or go to http://www.gnu.org/copyleft/lgpl.html
*/
#include "config.h"
#include "ringbuffer.h"
#include <algorithm>
#include <climits>
#include <stdexcept>
#include "almalloc.h"
RingBufferPtr RingBuffer::Create(size_t sz, size_t elem_sz, int limit_writes)
{
size_t power_of_two{0u};
if(sz > 0)
{
power_of_two = sz;
power_of_two |= power_of_two>>1;
power_of_two |= power_of_two>>2;
power_of_two |= power_of_two>>4;
power_of_two |= power_of_two>>8;
power_of_two |= power_of_two>>16;
#if SIZE_MAX > UINT_MAX
power_of_two |= power_of_two>>32;
#endif
}
++power_of_two;
if(power_of_two <= sz || power_of_two > std::numeric_limits<size_t>::max()/elem_sz)
throw std::overflow_error{"Ring buffer size overflow"};
const size_t bufbytes{power_of_two * elem_sz};
RingBufferPtr rb{new(FamCount(bufbytes)) RingBuffer{bufbytes}};
rb->mWriteSize = limit_writes ? sz : (power_of_two-1);
rb->mSizeMask = power_of_two - 1;
rb->mElemSize = elem_sz;
return rb;
}
void RingBuffer::reset() noexcept
{
mWritePtr.store(0, std::memory_order_relaxed);
mReadPtr.store(0, std::memory_order_relaxed);
std::fill_n(mBuffer.begin(), (mSizeMask+1)*mElemSize, al::byte{});
}
size_t RingBuffer::read(void *dest, size_t cnt) noexcept
{
const size_t free_cnt{readSpace()};
if(free_cnt == 0) return 0;
const size_t to_read{std::min(cnt, free_cnt)};
size_t read_ptr{mReadPtr.load(std::memory_order_relaxed) & mSizeMask};
size_t n1, n2;
const size_t cnt2{read_ptr + to_read};
if(cnt2 > mSizeMask+1)
{
n1 = mSizeMask+1 - read_ptr;
n2 = cnt2 & mSizeMask;
}
else
{
n1 = to_read;
n2 = 0;
}
auto outiter = std::copy_n(mBuffer.begin() + read_ptr*mElemSize, n1*mElemSize,
static_cast<al::byte*>(dest));
read_ptr += n1;
if(n2 > 0)
{
std::copy_n(mBuffer.begin(), n2*mElemSize, outiter);
read_ptr += n2;
}
mReadPtr.store(read_ptr, std::memory_order_release);
return to_read;
}
size_t RingBuffer::peek(void *dest, size_t cnt) const noexcept
{
const size_t free_cnt{readSpace()};
if(free_cnt == 0) return 0;
const size_t to_read{std::min(cnt, free_cnt)};
size_t read_ptr{mReadPtr.load(std::memory_order_relaxed) & mSizeMask};
size_t n1, n2;
const size_t cnt2{read_ptr + to_read};
if(cnt2 > mSizeMask+1)
{
n1 = mSizeMask+1 - read_ptr;
n2 = cnt2 & mSizeMask;
}
else
{
n1 = to_read;
n2 = 0;
}
auto outiter = std::copy_n(mBuffer.begin() + read_ptr*mElemSize, n1*mElemSize,
static_cast<al::byte*>(dest));
if(n2 > 0)
std::copy_n(mBuffer.begin(), n2*mElemSize, outiter);
return to_read;
}
size_t RingBuffer::write(const void *src, size_t cnt) noexcept
{
const size_t free_cnt{writeSpace()};
if(free_cnt == 0) return 0;
const size_t to_write{std::min(cnt, free_cnt)};
size_t write_ptr{mWritePtr.load(std::memory_order_relaxed) & mSizeMask};
size_t n1, n2;
const size_t cnt2{write_ptr + to_write};
if(cnt2 > mSizeMask+1)
{
n1 = mSizeMask+1 - write_ptr;
n2 = cnt2 & mSizeMask;
}
else
{
n1 = to_write;
n2 = 0;
}
auto srcbytes = static_cast<const al::byte*>(src);
std::copy_n(srcbytes, n1*mElemSize, mBuffer.begin() + write_ptr*mElemSize);
write_ptr += n1;
if(n2 > 0)
{
std::copy_n(srcbytes + n1*mElemSize, n2*mElemSize, mBuffer.begin());
write_ptr += n2;
}
mWritePtr.store(write_ptr, std::memory_order_release);
return to_write;
}
ll_ringbuffer_data_pair RingBuffer::getReadVector() const noexcept
{
ll_ringbuffer_data_pair ret;
size_t w{mWritePtr.load(std::memory_order_acquire)};
size_t r{mReadPtr.load(std::memory_order_acquire)};
w &= mSizeMask;
r &= mSizeMask;
const size_t free_cnt{(w-r) & mSizeMask};
const size_t cnt2{r + free_cnt};
if(cnt2 > mSizeMask+1)
{
/* Two part vector: the rest of the buffer after the current read ptr,
* plus some from the start of the buffer. */
ret.first.buf = const_cast<al::byte*>(mBuffer.data() + r*mElemSize);
ret.first.len = mSizeMask+1 - r;
ret.second.buf = const_cast<al::byte*>(mBuffer.data());
ret.second.len = cnt2 & mSizeMask;
}
else
{
/* Single part vector: just the rest of the buffer */
ret.first.buf = const_cast<al::byte*>(mBuffer.data() + r*mElemSize);
ret.first.len = free_cnt;
ret.second.buf = nullptr;
ret.second.len = 0;
}
return ret;
}
ll_ringbuffer_data_pair RingBuffer::getWriteVector() const noexcept
{
ll_ringbuffer_data_pair ret;
size_t w{mWritePtr.load(std::memory_order_acquire)};
size_t r{mReadPtr.load(std::memory_order_acquire) + mWriteSize - mSizeMask};
w &= mSizeMask;
r &= mSizeMask;
const size_t free_cnt{(r-w-1) & mSizeMask};
const size_t cnt2{w + free_cnt};
if(cnt2 > mSizeMask+1)
{
/* Two part vector: the rest of the buffer after the current write ptr,
* plus some from the start of the buffer. */
ret.first.buf = const_cast<al::byte*>(mBuffer.data() + w*mElemSize);
ret.first.len = mSizeMask+1 - w;
ret.second.buf = const_cast<al::byte*>(mBuffer.data());
ret.second.len = cnt2 & mSizeMask;
}
else
{
ret.first.buf = const_cast<al::byte*>(mBuffer.data() + w*mElemSize);
ret.first.len = free_cnt;
ret.second.buf = nullptr;
ret.second.len = 0;
}
return ret;
}

View file

@ -0,0 +1,113 @@
#ifndef RINGBUFFER_H
#define RINGBUFFER_H
#include <atomic>
#include <memory>
#include <stddef.h>
#include <utility>
#include "albyte.h"
#include "almalloc.h"
/* NOTE: This lockless ringbuffer implementation is copied from JACK, extended
* to include an element size. Consequently, parameters and return values for a
* size or count is in 'elements', not bytes. Additionally, it only supports
* single-consumer/single-provider operation.
*/
struct ll_ringbuffer_data {
al::byte *buf;
size_t len;
};
using ll_ringbuffer_data_pair = std::pair<ll_ringbuffer_data,ll_ringbuffer_data>;
struct RingBuffer {
private:
std::atomic<size_t> mWritePtr{0u};
std::atomic<size_t> mReadPtr{0u};
size_t mWriteSize{0u};
size_t mSizeMask{0u};
size_t mElemSize{0u};
al::FlexArray<al::byte, 16> mBuffer;
public:
RingBuffer(const size_t count) : mBuffer{count} { }
/** Reset the read and write pointers to zero. This is not thread safe. */
void reset() noexcept;
/**
* The non-copying data reader. Returns two ringbuffer data pointers that
* hold the current readable data. If the readable data is in one segment
* the second segment has zero length.
*/
ll_ringbuffer_data_pair getReadVector() const noexcept;
/**
* The non-copying data writer. Returns two ringbuffer data pointers that
* hold the current writeable data. If the writeable data is in one segment
* the second segment has zero length.
*/
ll_ringbuffer_data_pair getWriteVector() const noexcept;
/**
* Return the number of elements available for reading. This is the number
* of elements in front of the read pointer and behind the write pointer.
*/
size_t readSpace() const noexcept
{
const size_t w{mWritePtr.load(std::memory_order_acquire)};
const size_t r{mReadPtr.load(std::memory_order_acquire)};
return (w-r) & mSizeMask;
}
/**
* The copying data reader. Copy at most `cnt' elements into `dest'.
* Returns the actual number of elements copied.
*/
size_t read(void *dest, size_t cnt) noexcept;
/**
* The copying data reader w/o read pointer advance. Copy at most `cnt'
* elements into `dest'. Returns the actual number of elements copied.
*/
size_t peek(void *dest, size_t cnt) const noexcept;
/** Advance the read pointer `cnt' places. */
void readAdvance(size_t cnt) noexcept
{ mReadPtr.fetch_add(cnt, std::memory_order_acq_rel); }
/**
* Return the number of elements available for writing. This is the number
* of elements in front of the write pointer and behind the read pointer.
*/
size_t writeSpace() const noexcept
{
const size_t w{mWritePtr.load(std::memory_order_acquire)};
const size_t r{mReadPtr.load(std::memory_order_acquire) + mWriteSize - mSizeMask};
return (r-w-1) & mSizeMask;
}
/**
* The copying data writer. Copy at most `cnt' elements from `src'. Returns
* the actual number of elements copied.
*/
size_t write(const void *src, size_t cnt) noexcept;
/** Advance the write pointer `cnt' places. */
void writeAdvance(size_t cnt) noexcept
{ mWritePtr.fetch_add(cnt, std::memory_order_acq_rel); }
/**
* Create a new ringbuffer to hold at least `sz' elements of `elem_sz'
* bytes. The number of elements is rounded up to the next power of two
* (even if it is already a power of two, to ensure the requested amount
* can be written).
*/
static std::unique_ptr<RingBuffer> Create(size_t sz, size_t elem_sz, int limit_writes);
DEF_FAM_NEWDEL(RingBuffer, mBuffer)
};
using RingBufferPtr = std::unique_ptr<RingBuffer>;
#endif /* RINGBUFFER_H */

View file

@ -1,59 +0,0 @@
#include "config.h"
#include "rwlock.h"
#include "bool.h"
#include "atomic.h"
#include "threads.h"
/* A simple spinlock. Yield the thread while the given integer is set by
* another. Could probably be improved... */
#define LOCK(l) do { \
while(ATOMIC_FLAG_TEST_AND_SET(&(l), almemory_order_acq_rel) == true) \
althrd_yield(); \
} while(0)
#define UNLOCK(l) ATOMIC_FLAG_CLEAR(&(l), almemory_order_release)
void RWLockInit(RWLock *lock)
{
InitRef(&lock->read_count, 0);
InitRef(&lock->write_count, 0);
ATOMIC_FLAG_CLEAR(&lock->read_lock, almemory_order_relaxed);
ATOMIC_FLAG_CLEAR(&lock->read_entry_lock, almemory_order_relaxed);
ATOMIC_FLAG_CLEAR(&lock->write_lock, almemory_order_relaxed);
}
void ReadLock(RWLock *lock)
{
LOCK(lock->read_entry_lock);
LOCK(lock->read_lock);
/* NOTE: ATOMIC_ADD returns the *old* value! */
if(ATOMIC_ADD(&lock->read_count, 1, almemory_order_acq_rel) == 0)
LOCK(lock->write_lock);
UNLOCK(lock->read_lock);
UNLOCK(lock->read_entry_lock);
}
void ReadUnlock(RWLock *lock)
{
/* NOTE: ATOMIC_SUB returns the *old* value! */
if(ATOMIC_SUB(&lock->read_count, 1, almemory_order_acq_rel) == 1)
UNLOCK(lock->write_lock);
}
void WriteLock(RWLock *lock)
{
if(ATOMIC_ADD(&lock->write_count, 1, almemory_order_acq_rel) == 0)
LOCK(lock->read_lock);
LOCK(lock->write_lock);
}
void WriteUnlock(RWLock *lock)
{
UNLOCK(lock->write_lock);
if(ATOMIC_SUB(&lock->write_count, 1, almemory_order_acq_rel) == 1)
UNLOCK(lock->read_lock);
}

View file

@ -1,31 +0,0 @@
#ifndef AL_RWLOCK_H
#define AL_RWLOCK_H
#include "bool.h"
#include "atomic.h"
#ifdef __cplusplus
extern "C" {
#endif
typedef struct {
RefCount read_count;
RefCount write_count;
ATOMIC_FLAG read_lock;
ATOMIC_FLAG read_entry_lock;
ATOMIC_FLAG write_lock;
} RWLock;
#define RWLOCK_STATIC_INITIALIZE { ATOMIC_INIT_STATIC(0), ATOMIC_INIT_STATIC(0), \
ATOMIC_FLAG_INIT, ATOMIC_FLAG_INIT, ATOMIC_FLAG_INIT }
void RWLockInit(RWLock *lock);
void ReadLock(RWLock *lock);
void ReadUnlock(RWLock *lock);
void WriteLock(RWLock *lock);
void WriteUnlock(RWLock *lock);
#ifdef __cplusplus
}
#endif
#endif /* AL_RWLOCK_H */

View file

@ -1,21 +0,0 @@
#ifndef AL_STATIC_ASSERT_H
#define AL_STATIC_ASSERT_H
#include <assert.h>
#ifndef static_assert
#ifdef HAVE_C11_STATIC_ASSERT
#define static_assert _Static_assert
#else
#define CTASTR2(_pre,_post) _pre##_post
#define CTASTR(_pre,_post) CTASTR2(_pre,_post)
#if defined(__COUNTER__)
#define static_assert(_cond, _msg) typedef struct { int CTASTR(static_assert_failed_at_line_,__LINE__) : !!(_cond); } CTASTR(static_assertion_,__COUNTER__)
#else
#define static_assert(_cond, _msg) struct { int CTASTR(static_assert_failed_at_line_,__LINE__) : !!(_cond); }
#endif
#endif
#endif
#endif /* AL_STATIC_ASSERT_H */

View file

@ -0,0 +1,64 @@
#include "config.h"
#include "strutils.h"
#include <cstdlib>
#ifdef _WIN32
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
std::string wstr_to_utf8(const WCHAR *wstr)
{
std::string ret;
int len = WideCharToMultiByte(CP_UTF8, 0, wstr, -1, nullptr, 0, nullptr, nullptr);
if(len > 0)
{
ret.resize(len);
WideCharToMultiByte(CP_UTF8, 0, wstr, -1, &ret[0], len, nullptr, nullptr);
ret.pop_back();
}
return ret;
}
std::wstring utf8_to_wstr(const char *str)
{
std::wstring ret;
int len = MultiByteToWideChar(CP_UTF8, 0, str, -1, nullptr, 0);
if(len > 0)
{
ret.resize(len);
MultiByteToWideChar(CP_UTF8, 0, str, -1, &ret[0], len);
ret.pop_back();
}
return ret;
}
#endif
namespace al {
al::optional<std::string> getenv(const char *envname)
{
const char *str{std::getenv(envname)};
if(str && str[0] != '\0')
return al::make_optional<std::string>(str);
return al::nullopt;
}
#ifdef _WIN32
al::optional<std::wstring> getenv(const WCHAR *envname)
{
const WCHAR *str{_wgetenv(envname)};
if(str && str[0] != L'\0')
return al::make_optional<std::wstring>(str);
return al::nullopt;
}
#endif
} // namespace al

View file

@ -0,0 +1,24 @@
#ifndef AL_STRUTILS_H
#define AL_STRUTILS_H
#include <string>
#include "aloptional.h"
#ifdef _WIN32
#include <wchar.h>
std::string wstr_to_utf8(const wchar_t *wstr);
std::wstring utf8_to_wstr(const char *str);
#endif
namespace al {
al::optional<std::string> getenv(const char *envname);
#ifdef _WIN32
al::optional<std::wstring> getenv(const wchar_t *envname);
#endif
} // namespace al
#endif /* AL_STRUTILS_H */

View file

@ -1,712 +0,0 @@
/**
* OpenAL cross platform audio library
* Copyright (C) 1999-2007 by authors.
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
* Or go to http://www.gnu.org/copyleft/lgpl.html
*/
#include "config.h"
#include "threads.h"
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include "uintmap.h"
extern inline althrd_t althrd_current(void);
extern inline int althrd_equal(althrd_t thr0, althrd_t thr1);
extern inline void althrd_exit(int res);
extern inline void althrd_yield(void);
extern inline int almtx_lock(almtx_t *mtx);
extern inline int almtx_unlock(almtx_t *mtx);
extern inline int almtx_trylock(almtx_t *mtx);
extern inline void *altss_get(altss_t tss_id);
extern inline int altss_set(altss_t tss_id, void *val);
#ifndef UNUSED
#if defined(__cplusplus)
#define UNUSED(x)
#elif defined(__GNUC__)
#define UNUSED(x) UNUSED_##x __attribute__((unused))
#elif defined(__LCLINT__)
#define UNUSED(x) /*@unused@*/ x
#else
#define UNUSED(x) x
#endif
#endif
#define THREAD_STACK_SIZE (2*1024*1024) /* 2MB */
#ifdef _WIN32
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#include <mmsystem.h>
/* An associative map of uint:void* pairs. The key is the unique Thread ID and
* the value is the thread HANDLE. The thread ID is passed around as the
* althrd_t since there is only one ID per thread, whereas a thread may be
* referenced by multiple different HANDLEs. This map allows retrieving the
* original handle which is needed to join the thread and get its return value.
*/
static UIntMap ThrdIdHandle = UINTMAP_STATIC_INITIALIZE;
/* An associative map of uint:void* pairs. The key is the TLS index (given by
* TlsAlloc), and the value is the altss_dtor_t callback. When a thread exits,
* we iterate over the TLS indices for their thread-local value and call the
* destructor function with it if they're both not NULL.
*/
static UIntMap TlsDestructors = UINTMAP_STATIC_INITIALIZE;
void althrd_setname(althrd_t thr, const char *name)
{
#if defined(_MSC_VER)
#define MS_VC_EXCEPTION 0x406D1388
#pragma pack(push,8)
struct {
DWORD dwType; // Must be 0x1000.
LPCSTR szName; // Pointer to name (in user addr space).
DWORD dwThreadID; // Thread ID (-1=caller thread).
DWORD dwFlags; // Reserved for future use, must be zero.
} info;
#pragma pack(pop)
info.dwType = 0x1000;
info.szName = name;
info.dwThreadID = thr;
info.dwFlags = 0;
__try {
RaiseException(MS_VC_EXCEPTION, 0, sizeof(info)/sizeof(ULONG_PTR), (ULONG_PTR*)&info);
}
__except(EXCEPTION_CONTINUE_EXECUTION) {
}
#undef MS_VC_EXCEPTION
#else
(void)thr;
(void)name;
#endif
}
typedef struct thread_cntr {
althrd_start_t func;
void *arg;
} thread_cntr;
static DWORD WINAPI althrd_starter(void *arg)
{
thread_cntr cntr;
memcpy(&cntr, arg, sizeof(cntr));
free(arg);
return (DWORD)((*cntr.func)(cntr.arg));
}
int althrd_create(althrd_t *thr, althrd_start_t func, void *arg)
{
thread_cntr *cntr;
DWORD thrid;
HANDLE hdl;
cntr = malloc(sizeof(*cntr));
if(!cntr) return althrd_nomem;
cntr->func = func;
cntr->arg = arg;
hdl = CreateThread(NULL, THREAD_STACK_SIZE, althrd_starter, cntr, 0, &thrid);
if(!hdl)
{
free(cntr);
return althrd_error;
}
InsertUIntMapEntry(&ThrdIdHandle, thrid, hdl);
*thr = thrid;
return althrd_success;
}
int althrd_detach(althrd_t thr)
{
HANDLE hdl = RemoveUIntMapKey(&ThrdIdHandle, thr);
if(!hdl) return althrd_error;
CloseHandle(hdl);
return althrd_success;
}
int althrd_join(althrd_t thr, int *res)
{
DWORD code;
HANDLE hdl = RemoveUIntMapKey(&ThrdIdHandle, thr);
if(!hdl) return althrd_error;
WaitForSingleObject(hdl, INFINITE);
GetExitCodeThread(hdl, &code);
CloseHandle(hdl);
if(res != NULL)
*res = (int)code;
return althrd_success;
}
int althrd_sleep(const struct timespec *ts, struct timespec* UNUSED(rem))
{
DWORD msec;
if(ts->tv_sec < 0 || ts->tv_sec >= (0x7fffffff / 1000) ||
ts->tv_nsec < 0 || ts->tv_nsec >= 1000000000)
return -2;
msec = (DWORD)(ts->tv_sec * 1000);
msec += (DWORD)((ts->tv_nsec+999999) / 1000000);
Sleep(msec);
return 0;
}
int almtx_init(almtx_t *mtx, int type)
{
if(!mtx) return althrd_error;
type &= ~almtx_recursive;
if(type != almtx_plain)
return althrd_error;
InitializeCriticalSection(mtx);
return althrd_success;
}
void almtx_destroy(almtx_t *mtx)
{
DeleteCriticalSection(mtx);
}
#if defined(_WIN32_WINNT) && _WIN32_WINNT >= 0x0600
int alcnd_init(alcnd_t *cond)
{
InitializeConditionVariable(cond);
return althrd_success;
}
int alcnd_signal(alcnd_t *cond)
{
WakeConditionVariable(cond);
return althrd_success;
}
int alcnd_broadcast(alcnd_t *cond)
{
WakeAllConditionVariable(cond);
return althrd_success;
}
int alcnd_wait(alcnd_t *cond, almtx_t *mtx)
{
if(SleepConditionVariableCS(cond, mtx, INFINITE) != 0)
return althrd_success;
return althrd_error;
}
void alcnd_destroy(alcnd_t* UNUSED(cond))
{
/* Nothing to delete? */
}
#else
/* WARNING: This is a rather poor implementation of condition variables, with
* known problems. However, it's simple, efficient, and good enough for now to
* not require Vista. Based on "Strategies for Implementing POSIX Condition
* Variables" by Douglas C. Schmidt and Irfan Pyarali:
* http://www.cs.wustl.edu/~schmidt/win32-cv-1.html
*/
/* A better solution may be using Wine's implementation. It requires internals
* (NtCreateKeyedEvent, NtReleaseKeyedEvent, and NtWaitForKeyedEvent) from
* ntdll, and implemention of exchange and compare-exchange for RefCounts.
*/
typedef struct {
RefCount wait_count;
HANDLE events[2];
} _int_alcnd_t;
enum {
SIGNAL = 0,
BROADCAST = 1
};
int alcnd_init(alcnd_t *cond)
{
_int_alcnd_t *icond = calloc(1, sizeof(*icond));
if(!icond) return althrd_nomem;
InitRef(&icond->wait_count, 0);
icond->events[SIGNAL] = CreateEventW(NULL, FALSE, FALSE, NULL);
icond->events[BROADCAST] = CreateEventW(NULL, TRUE, FALSE, NULL);
if(!icond->events[SIGNAL] || !icond->events[BROADCAST])
{
if(icond->events[SIGNAL])
CloseHandle(icond->events[SIGNAL]);
if(icond->events[BROADCAST])
CloseHandle(icond->events[BROADCAST]);
free(icond);
return althrd_error;
}
cond->Ptr = icond;
return althrd_success;
}
int alcnd_signal(alcnd_t *cond)
{
_int_alcnd_t *icond = cond->Ptr;
if(ReadRef(&icond->wait_count) > 0)
SetEvent(icond->events[SIGNAL]);
return althrd_success;
}
int alcnd_broadcast(alcnd_t *cond)
{
_int_alcnd_t *icond = cond->Ptr;
if(ReadRef(&icond->wait_count) > 0)
SetEvent(icond->events[BROADCAST]);
return althrd_success;
}
int alcnd_wait(alcnd_t *cond, almtx_t *mtx)
{
_int_alcnd_t *icond = cond->Ptr;
int res;
IncrementRef(&icond->wait_count);
LeaveCriticalSection(mtx);
res = WaitForMultipleObjects(2, icond->events, FALSE, INFINITE);
if(DecrementRef(&icond->wait_count) == 0 && res == WAIT_OBJECT_0+BROADCAST)
ResetEvent(icond->events[BROADCAST]);
EnterCriticalSection(mtx);
return althrd_success;
}
void alcnd_destroy(alcnd_t *cond)
{
_int_alcnd_t *icond = cond->Ptr;
CloseHandle(icond->events[SIGNAL]);
CloseHandle(icond->events[BROADCAST]);
free(icond);
}
#endif /* defined(_WIN32_WINNT) && _WIN32_WINNT >= 0x0600 */
int alsem_init(alsem_t *sem, unsigned int initial)
{
*sem = CreateSemaphore(NULL, initial, INT_MAX, NULL);
if(*sem != NULL) return althrd_success;
return althrd_error;
}
void alsem_destroy(alsem_t *sem)
{
CloseHandle(*sem);
}
int alsem_post(alsem_t *sem)
{
DWORD ret = ReleaseSemaphore(*sem, 1, NULL);
if(ret) return althrd_success;
return althrd_error;
}
int alsem_wait(alsem_t *sem)
{
DWORD ret = WaitForSingleObject(*sem, INFINITE);
if(ret == WAIT_OBJECT_0) return althrd_success;
return althrd_error;
}
int alsem_trywait(alsem_t *sem)
{
DWORD ret = WaitForSingleObject(*sem, 0);
if(ret == WAIT_OBJECT_0) return althrd_success;
if(ret == WAIT_TIMEOUT) return althrd_busy;
return althrd_error;
}
int altss_create(altss_t *tss_id, altss_dtor_t callback)
{
DWORD key = TlsAlloc();
if(key == TLS_OUT_OF_INDEXES)
return althrd_error;
*tss_id = key;
if(callback != NULL)
InsertUIntMapEntry(&TlsDestructors, key, callback);
return althrd_success;
}
void altss_delete(altss_t tss_id)
{
RemoveUIntMapKey(&TlsDestructors, tss_id);
TlsFree(tss_id);
}
int altimespec_get(struct timespec *ts, int base)
{
static_assert(sizeof(FILETIME) == sizeof(ULARGE_INTEGER),
"Size of FILETIME does not match ULARGE_INTEGER");
if(base == AL_TIME_UTC)
{
union {
FILETIME ftime;
ULARGE_INTEGER ulint;
} systime;
GetSystemTimeAsFileTime(&systime.ftime);
/* FILETIME is in 100-nanosecond units, or 1/10th of a microsecond. */
ts->tv_sec = systime.ulint.QuadPart/10000000;
ts->tv_nsec = (systime.ulint.QuadPart%10000000) * 100;
return base;
}
return 0;
}
void alcall_once(alonce_flag *once, void (*callback)(void))
{
LONG ret;
while((ret=InterlockedExchange(once, 1)) == 1)
althrd_yield();
if(ret == 0)
(*callback)();
InterlockedExchange(once, 2);
}
void althrd_deinit(void)
{
ResetUIntMap(&ThrdIdHandle);
ResetUIntMap(&TlsDestructors);
}
void althrd_thread_detach(void)
{
ALsizei i;
LockUIntMapRead(&TlsDestructors);
for(i = 0;i < TlsDestructors.size;i++)
{
void *ptr = altss_get(TlsDestructors.keys[i]);
altss_dtor_t callback = (altss_dtor_t)TlsDestructors.values[i];
if(ptr && callback) callback(ptr);
}
UnlockUIntMapRead(&TlsDestructors);
}
#else
#include <sys/time.h>
#include <unistd.h>
#include <pthread.h>
#ifdef HAVE_PTHREAD_NP_H
#include <pthread_np.h>
#endif
extern inline int althrd_sleep(const struct timespec *ts, struct timespec *rem);
extern inline void alcall_once(alonce_flag *once, void (*callback)(void));
extern inline void althrd_deinit(void);
extern inline void althrd_thread_detach(void);
void althrd_setname(althrd_t thr, const char *name)
{
#if defined(HAVE_PTHREAD_SETNAME_NP)
#if defined(PTHREAD_SETNAME_NP_ONE_PARAM)
if(althrd_equal(thr, althrd_current()))
pthread_setname_np(name);
#elif defined(PTHREAD_SETNAME_NP_THREE_PARAMS)
pthread_setname_np(thr, "%s", (void*)name);
#else
pthread_setname_np(thr, name);
#endif
#elif defined(HAVE_PTHREAD_SET_NAME_NP)
pthread_set_name_np(thr, name);
#else
(void)thr;
(void)name;
#endif
}
typedef struct thread_cntr {
althrd_start_t func;
void *arg;
} thread_cntr;
static void *althrd_starter(void *arg)
{
thread_cntr cntr;
memcpy(&cntr, arg, sizeof(cntr));
free(arg);
return (void*)(intptr_t)((*cntr.func)(cntr.arg));
}
int althrd_create(althrd_t *thr, althrd_start_t func, void *arg)
{
thread_cntr *cntr;
pthread_attr_t attr;
size_t stackmult = 1;
int err;
cntr = malloc(sizeof(*cntr));
if(!cntr) return althrd_nomem;
if(pthread_attr_init(&attr) != 0)
{
free(cntr);
return althrd_error;
}
retry_stacksize:
if(pthread_attr_setstacksize(&attr, THREAD_STACK_SIZE*stackmult) != 0)
{
pthread_attr_destroy(&attr);
free(cntr);
return althrd_error;
}
cntr->func = func;
cntr->arg = arg;
if((err=pthread_create(thr, &attr, althrd_starter, cntr)) == 0)
{
pthread_attr_destroy(&attr);
return althrd_success;
}
if(err == EINVAL)
{
/* If an invalid stack size, try increasing it (limit x4, 8MB). */
if(stackmult < 4)
{
stackmult *= 2;
goto retry_stacksize;
}
/* If still nothing, try defaults and hope they're good enough. */
if(pthread_create(thr, NULL, althrd_starter, cntr) == 0)
{
pthread_attr_destroy(&attr);
return althrd_success;
}
}
pthread_attr_destroy(&attr);
free(cntr);
return althrd_error;
}
int althrd_detach(althrd_t thr)
{
if(pthread_detach(thr) != 0)
return althrd_error;
return althrd_success;
}
int althrd_join(althrd_t thr, int *res)
{
void *code;
if(pthread_join(thr, &code) != 0)
return althrd_error;
if(res != NULL)
*res = (int)(intptr_t)code;
return althrd_success;
}
int almtx_init(almtx_t *mtx, int type)
{
int ret;
if(!mtx) return althrd_error;
if((type&~almtx_recursive) != 0)
return althrd_error;
if(type == almtx_plain)
ret = pthread_mutex_init(mtx, NULL);
else
{
pthread_mutexattr_t attr;
ret = pthread_mutexattr_init(&attr);
if(ret) return althrd_error;
if(type == almtx_recursive)
{
ret = pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE);
#ifdef HAVE_PTHREAD_MUTEXATTR_SETKIND_NP
if(ret != 0)
ret = pthread_mutexattr_setkind_np(&attr, PTHREAD_MUTEX_RECURSIVE);
#endif
}
else
ret = 1;
if(ret == 0)
ret = pthread_mutex_init(mtx, &attr);
pthread_mutexattr_destroy(&attr);
}
return ret ? althrd_error : althrd_success;
}
void almtx_destroy(almtx_t *mtx)
{
pthread_mutex_destroy(mtx);
}
int alcnd_init(alcnd_t *cond)
{
if(pthread_cond_init(cond, NULL) == 0)
return althrd_success;
return althrd_error;
}
int alcnd_signal(alcnd_t *cond)
{
if(pthread_cond_signal(cond) == 0)
return althrd_success;
return althrd_error;
}
int alcnd_broadcast(alcnd_t *cond)
{
if(pthread_cond_broadcast(cond) == 0)
return althrd_success;
return althrd_error;
}
int alcnd_wait(alcnd_t *cond, almtx_t *mtx)
{
if(pthread_cond_wait(cond, mtx) == 0)
return althrd_success;
return althrd_error;
}
void alcnd_destroy(alcnd_t *cond)
{
pthread_cond_destroy(cond);
}
int alsem_init(alsem_t *sem, unsigned int initial)
{
if(sem_init(sem, 0, initial) == 0)
return althrd_success;
return althrd_error;
}
void alsem_destroy(alsem_t *sem)
{
sem_destroy(sem);
}
int alsem_post(alsem_t *sem)
{
if(sem_post(sem) == 0)
return althrd_success;
return althrd_error;
}
int alsem_wait(alsem_t *sem)
{
if(sem_wait(sem) == 0) return althrd_success;
if(errno == EINTR) return -2;
return althrd_error;
}
int alsem_trywait(alsem_t *sem)
{
if(sem_trywait(sem) == 0) return althrd_success;
if(errno == EWOULDBLOCK) return althrd_busy;
if(errno == EINTR) return -2;
return althrd_error;
}
int altss_create(altss_t *tss_id, altss_dtor_t callback)
{
if(pthread_key_create(tss_id, callback) != 0)
return althrd_error;
return althrd_success;
}
void altss_delete(altss_t tss_id)
{
pthread_key_delete(tss_id);
}
int altimespec_get(struct timespec *ts, int base)
{
if(base == AL_TIME_UTC)
{
int ret;
#if _POSIX_TIMERS > 0
ret = clock_gettime(CLOCK_REALTIME, ts);
if(ret == 0) return base;
#else /* _POSIX_TIMERS > 0 */
struct timeval tv;
ret = gettimeofday(&tv, NULL);
if(ret == 0)
{
ts->tv_sec = tv.tv_sec;
ts->tv_nsec = tv.tv_usec * 1000;
return base;
}
#endif
}
return 0;
}
#endif
void al_nssleep(unsigned long nsec)
{
struct timespec ts, rem;
ts.tv_sec = nsec / 1000000000ul;
ts.tv_nsec = nsec % 1000000000ul;
while(althrd_sleep(&ts, &rem) == -1)
ts = rem;
}

View file

@ -0,0 +1,176 @@
/**
* OpenAL cross platform audio library
* Copyright (C) 1999-2007 by authors.
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
* Or go to http://www.gnu.org/copyleft/lgpl.html
*/
#include "config.h"
#include "opthelpers.h"
#include "threads.h"
#include <system_error>
#ifdef _WIN32
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#include <limits>
void althrd_setname(const char *name)
{
#if defined(_MSC_VER)
#define MS_VC_EXCEPTION 0x406D1388
#pragma pack(push,8)
struct {
DWORD dwType; // Must be 0x1000.
LPCSTR szName; // Pointer to name (in user addr space).
DWORD dwThreadID; // Thread ID (-1=caller thread).
DWORD dwFlags; // Reserved for future use, must be zero.
} info;
#pragma pack(pop)
info.dwType = 0x1000;
info.szName = name;
info.dwThreadID = ~DWORD{0};
info.dwFlags = 0;
__try {
RaiseException(MS_VC_EXCEPTION, 0, sizeof(info)/sizeof(ULONG_PTR), (ULONG_PTR*)&info);
}
__except(EXCEPTION_CONTINUE_EXECUTION) {
}
#undef MS_VC_EXCEPTION
#else
(void)name;
#endif
}
namespace al {
semaphore::semaphore(unsigned int initial)
{
if(initial > static_cast<unsigned int>(std::numeric_limits<int>::max()))
throw std::system_error(std::make_error_code(std::errc::value_too_large));
mSem = CreateSemaphore(nullptr, initial, std::numeric_limits<int>::max(), nullptr);
if(mSem == nullptr)
throw std::system_error(std::make_error_code(std::errc::resource_unavailable_try_again));
}
semaphore::~semaphore()
{ CloseHandle(mSem); }
void semaphore::post()
{
if UNLIKELY(!ReleaseSemaphore(static_cast<HANDLE>(mSem), 1, nullptr))
throw std::system_error(std::make_error_code(std::errc::value_too_large));
}
void semaphore::wait() noexcept
{ WaitForSingleObject(static_cast<HANDLE>(mSem), INFINITE); }
bool semaphore::try_wait() noexcept
{ return WaitForSingleObject(static_cast<HANDLE>(mSem), 0) == WAIT_OBJECT_0; }
} // namespace al
#else
#if defined(HAVE_PTHREAD_SETNAME_NP) || defined(HAVE_PTHREAD_SET_NAME_NP)
#include <pthread.h>
#ifdef HAVE_PTHREAD_NP_H
#include <pthread_np.h>
#endif
void althrd_setname(const char *name)
{
#if defined(HAVE_PTHREAD_SET_NAME_NP)
pthread_set_name_np(pthread_self(), name);
#elif defined(PTHREAD_SETNAME_NP_ONE_PARAM)
pthread_setname_np(name);
#elif defined(PTHREAD_SETNAME_NP_THREE_PARAMS)
pthread_setname_np(pthread_self(), "%s", (void*)name);
#else
pthread_setname_np(pthread_self(), name);
#endif
}
#else
void althrd_setname(const char*) { }
#endif
#ifdef __APPLE__
namespace al {
semaphore::semaphore(unsigned int initial)
{
mSem = dispatch_semaphore_create(initial);
if(!mSem)
throw std::system_error(std::make_error_code(std::errc::resource_unavailable_try_again));
}
semaphore::~semaphore()
{ dispatch_release(mSem); }
void semaphore::post()
{ dispatch_semaphore_signal(mSem); }
void semaphore::wait() noexcept
{ dispatch_semaphore_wait(mSem, DISPATCH_TIME_FOREVER); }
bool semaphore::try_wait() noexcept
{ return dispatch_semaphore_wait(mSem, DISPATCH_TIME_NOW) == 0; }
} // namespace al
#else /* !__APPLE__ */
#include <cerrno>
namespace al {
semaphore::semaphore(unsigned int initial)
{
if(sem_init(&mSem, 0, initial) != 0)
throw std::system_error(std::make_error_code(std::errc::resource_unavailable_try_again));
}
semaphore::~semaphore()
{ sem_destroy(&mSem); }
void semaphore::post()
{
if(sem_post(&mSem) != 0)
throw std::system_error(std::make_error_code(std::errc::value_too_large));
}
void semaphore::wait() noexcept
{
while(sem_wait(&mSem) == -1 && errno == EINTR) {
}
}
bool semaphore::try_wait() noexcept
{ return sem_trywait(&mSem) == 0; }
} // namespace al
#endif /* __APPLE__ */
#endif /* _WIN32 */

View file

@ -1,8 +1,6 @@
#ifndef AL_THREADS_H
#define AL_THREADS_H
#include <time.h>
#if defined(__GNUC__) && defined(__i386__)
/* force_align_arg_pointer is required for proper function arguments aligning
* when SSE code is used. Some systems (Windows, QNX) do not guarantee our
@ -13,248 +11,38 @@
#define FORCE_ALIGN
#endif
#ifdef __cplusplus
extern "C" {
#endif
enum {
althrd_success = 0,
althrd_error,
althrd_nomem,
althrd_timedout,
althrd_busy
};
enum {
almtx_plain = 0,
almtx_recursive = 1,
};
typedef int (*althrd_start_t)(void*);
typedef void (*altss_dtor_t)(void*);
#define AL_TIME_UTC 1
#ifdef _WIN32
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#ifndef HAVE_STRUCT_TIMESPEC
struct timespec {
time_t tv_sec;
long tv_nsec;
};
#endif
typedef DWORD althrd_t;
typedef CRITICAL_SECTION almtx_t;
#if defined(_WIN32_WINNT) && _WIN32_WINNT >= 0x0600
typedef CONDITION_VARIABLE alcnd_t;
#else
typedef struct { void *Ptr; } alcnd_t;
#endif
typedef HANDLE alsem_t;
typedef DWORD altss_t;
typedef LONG alonce_flag;
#define AL_ONCE_FLAG_INIT 0
int althrd_sleep(const struct timespec *ts, struct timespec *rem);
void alcall_once(alonce_flag *once, void (*callback)(void));
void althrd_deinit(void);
void althrd_thread_detach(void);
inline althrd_t althrd_current(void)
{
return GetCurrentThreadId();
}
inline int althrd_equal(althrd_t thr0, althrd_t thr1)
{
return thr0 == thr1;
}
inline void althrd_exit(int res)
{
ExitThread(res);
}
inline void althrd_yield(void)
{
SwitchToThread();
}
inline int almtx_lock(almtx_t *mtx)
{
if(!mtx) return althrd_error;
EnterCriticalSection(mtx);
return althrd_success;
}
inline int almtx_unlock(almtx_t *mtx)
{
if(!mtx) return althrd_error;
LeaveCriticalSection(mtx);
return althrd_success;
}
inline int almtx_trylock(almtx_t *mtx)
{
if(!mtx) return althrd_error;
if(!TryEnterCriticalSection(mtx))
return althrd_busy;
return althrd_success;
}
inline void *altss_get(altss_t tss_id)
{
return TlsGetValue(tss_id);
}
inline int altss_set(altss_t tss_id, void *val)
{
if(TlsSetValue(tss_id, val) == 0)
return althrd_error;
return althrd_success;
}
#else
#include <stdint.h>
#include <errno.h>
#include <pthread.h>
#if defined(__APPLE__)
#include <dispatch/dispatch.h>
#elif !defined(_WIN32)
#include <semaphore.h>
typedef pthread_t althrd_t;
typedef pthread_mutex_t almtx_t;
typedef pthread_cond_t alcnd_t;
typedef sem_t alsem_t;
typedef pthread_key_t altss_t;
typedef pthread_once_t alonce_flag;
#define AL_ONCE_FLAG_INIT PTHREAD_ONCE_INIT
inline althrd_t althrd_current(void)
{
return pthread_self();
}
inline int althrd_equal(althrd_t thr0, althrd_t thr1)
{
return pthread_equal(thr0, thr1);
}
inline void althrd_exit(int res)
{
pthread_exit((void*)(intptr_t)res);
}
inline void althrd_yield(void)
{
sched_yield();
}
inline int althrd_sleep(const struct timespec *ts, struct timespec *rem)
{
int ret = nanosleep(ts, rem);
if(ret != 0)
{
ret = ((errno==EINTR) ? -1 : -2);
errno = 0;
}
return ret;
}
inline int almtx_lock(almtx_t *mtx)
{
if(pthread_mutex_lock(mtx) != 0)
return althrd_error;
return althrd_success;
}
inline int almtx_unlock(almtx_t *mtx)
{
if(pthread_mutex_unlock(mtx) != 0)
return althrd_error;
return althrd_success;
}
inline int almtx_trylock(almtx_t *mtx)
{
int ret = pthread_mutex_trylock(mtx);
switch(ret)
{
case 0: return althrd_success;
case EBUSY: return althrd_busy;
}
return althrd_error;
}
inline void *altss_get(altss_t tss_id)
{
return pthread_getspecific(tss_id);
}
inline int altss_set(altss_t tss_id, void *val)
{
if(pthread_setspecific(tss_id, val) != 0)
return althrd_error;
return althrd_success;
}
inline void alcall_once(alonce_flag *once, void (*callback)(void))
{
pthread_once(once, callback);
}
inline void althrd_deinit(void) { }
inline void althrd_thread_detach(void) { }
#endif
void althrd_setname(const char *name);
int althrd_create(althrd_t *thr, althrd_start_t func, void *arg);
int althrd_detach(althrd_t thr);
int althrd_join(althrd_t thr, int *res);
void althrd_setname(althrd_t thr, const char *name);
namespace al {
int almtx_init(almtx_t *mtx, int type);
void almtx_destroy(almtx_t *mtx);
int alcnd_init(alcnd_t *cond);
int alcnd_signal(alcnd_t *cond);
int alcnd_broadcast(alcnd_t *cond);
int alcnd_wait(alcnd_t *cond, almtx_t *mtx);
void alcnd_destroy(alcnd_t *cond);
int alsem_init(alsem_t *sem, unsigned int initial);
void alsem_destroy(alsem_t *sem);
int alsem_post(alsem_t *sem);
int alsem_wait(alsem_t *sem);
int alsem_trywait(alsem_t *sem);
int altss_create(altss_t *tss_id, altss_dtor_t callback);
void altss_delete(altss_t tss_id);
int altimespec_get(struct timespec *ts, int base);
void al_nssleep(unsigned long nsec);
#ifdef __cplusplus
}
class semaphore {
#ifdef _WIN32
using native_type = void*;
#elif defined(__APPLE__)
using native_type = dispatch_semaphore_t;
#else
using native_type = sem_t;
#endif
native_type mSem;
public:
semaphore(unsigned int initial=0);
semaphore(const semaphore&) = delete;
~semaphore();
semaphore& operator=(const semaphore&) = delete;
void post();
void wait() noexcept;
bool try_wait() noexcept;
};
} // namespace al
#endif /* AL_THREADS_H */

View file

@ -1,182 +0,0 @@
#include "config.h"
#include "uintmap.h"
#include <stdlib.h>
#include <string.h>
#include "almalloc.h"
extern inline void LockUIntMapRead(UIntMap *map);
extern inline void UnlockUIntMapRead(UIntMap *map);
extern inline void LockUIntMapWrite(UIntMap *map);
extern inline void UnlockUIntMapWrite(UIntMap *map);
void InitUIntMap(UIntMap *map, ALsizei limit)
{
map->keys = NULL;
map->values = NULL;
map->size = 0;
map->capacity = 0;
map->limit = limit;
RWLockInit(&map->lock);
}
void ResetUIntMap(UIntMap *map)
{
WriteLock(&map->lock);
al_free(map->keys);
map->keys = NULL;
map->values = NULL;
map->size = 0;
map->capacity = 0;
WriteUnlock(&map->lock);
}
ALenum InsertUIntMapEntry(UIntMap *map, ALuint key, ALvoid *value)
{
ALsizei pos = 0;
WriteLock(&map->lock);
if(map->size > 0)
{
ALsizei count = map->size;
do {
ALsizei step = count>>1;
ALsizei i = pos+step;
if(!(map->keys[i] < key))
count = step;
else
{
pos = i+1;
count -= step+1;
}
} while(count > 0);
}
if(pos == map->size || map->keys[pos] != key)
{
if(map->size >= map->limit)
{
WriteUnlock(&map->lock);
return AL_OUT_OF_MEMORY;
}
if(map->size == map->capacity)
{
ALuint *keys = NULL;
ALvoid **values;
ALsizei newcap, keylen;
newcap = (map->capacity ? (map->capacity<<1) : 4);
if(map->limit > 0 && newcap > map->limit)
newcap = map->limit;
if(newcap > map->capacity)
{
/* Round the memory size for keys up to a multiple of the
* pointer size.
*/
keylen = newcap * sizeof(map->keys[0]);
keylen += sizeof(map->values[0]) - 1;
keylen -= keylen%sizeof(map->values[0]);
keys = al_malloc(16, keylen + newcap*sizeof(map->values[0]));
}
if(!keys)
{
WriteUnlock(&map->lock);
return AL_OUT_OF_MEMORY;
}
values = (ALvoid**)((ALbyte*)keys + keylen);
if(map->keys)
{
memcpy(keys, map->keys, map->size*sizeof(map->keys[0]));
memcpy(values, map->values, map->size*sizeof(map->values[0]));
}
al_free(map->keys);
map->keys = keys;
map->values = values;
map->capacity = newcap;
}
if(pos < map->size)
{
memmove(&map->keys[pos+1], &map->keys[pos],
(map->size-pos)*sizeof(map->keys[0]));
memmove(&map->values[pos+1], &map->values[pos],
(map->size-pos)*sizeof(map->values[0]));
}
map->size++;
}
map->keys[pos] = key;
map->values[pos] = value;
WriteUnlock(&map->lock);
return AL_NO_ERROR;
}
ALvoid *RemoveUIntMapKey(UIntMap *map, ALuint key)
{
ALvoid *ptr = NULL;
WriteLock(&map->lock);
if(map->size > 0)
{
ALsizei pos = 0;
ALsizei count = map->size;
do {
ALsizei step = count>>1;
ALsizei i = pos+step;
if(!(map->keys[i] < key))
count = step;
else
{
pos = i+1;
count -= step+1;
}
} while(count > 0);
if(pos < map->size && map->keys[pos] == key)
{
ptr = map->values[pos];
if(pos < map->size-1)
{
memmove(&map->keys[pos], &map->keys[pos+1],
(map->size-1-pos)*sizeof(map->keys[0]));
memmove(&map->values[pos], &map->values[pos+1],
(map->size-1-pos)*sizeof(map->values[0]));
}
map->size--;
}
}
WriteUnlock(&map->lock);
return ptr;
}
ALvoid *LookupUIntMapKey(UIntMap *map, ALuint key)
{
ALvoid *ptr = NULL;
ReadLock(&map->lock);
if(map->size > 0)
{
ALsizei pos = 0;
ALsizei count = map->size;
do {
ALsizei step = count>>1;
ALsizei i = pos+step;
if(!(map->keys[i] < key))
count = step;
else
{
pos = i+1;
count -= step+1;
}
} while(count > 0);
if(pos < map->size && map->keys[pos] == key)
ptr = map->values[pos];
}
ReadUnlock(&map->lock);
return ptr;
}

View file

@ -1,41 +0,0 @@
#ifndef AL_UINTMAP_H
#define AL_UINTMAP_H
#include <limits.h>
#include "AL/al.h"
#include "rwlock.h"
#ifdef __cplusplus
extern "C" {
#endif
typedef struct UIntMap {
ALuint *keys;
/* Shares memory with keys. */
ALvoid **values;
ALsizei size;
ALsizei capacity;
ALsizei limit;
RWLock lock;
} UIntMap;
#define UINTMAP_STATIC_INITIALIZE_N(_n) { NULL, NULL, 0, 0, (_n), RWLOCK_STATIC_INITIALIZE }
#define UINTMAP_STATIC_INITIALIZE UINTMAP_STATIC_INITIALIZE_N(INT_MAX)
void InitUIntMap(UIntMap *map, ALsizei limit);
void ResetUIntMap(UIntMap *map);
ALenum InsertUIntMapEntry(UIntMap *map, ALuint key, ALvoid *value);
ALvoid *RemoveUIntMapKey(UIntMap *map, ALuint key);
ALvoid *LookupUIntMapKey(UIntMap *map, ALuint key);
inline void LockUIntMapRead(UIntMap *map) { ReadLock(&map->lock); }
inline void UnlockUIntMapRead(UIntMap *map) { ReadUnlock(&map->lock); }
inline void LockUIntMapWrite(UIntMap *map) { WriteLock(&map->lock); }
inline void UnlockUIntMapWrite(UIntMap *map) { WriteUnlock(&map->lock); }
#ifdef __cplusplus
}
#endif
#endif /* AL_UINTMAP_H */

View file

@ -0,0 +1,118 @@
#ifndef COMMON_VECMAT_H
#define COMMON_VECMAT_H
#include <array>
#include <cmath>
#include <cstddef>
#include <limits>
#include "alspan.h"
namespace alu {
template<typename T>
class VectorR {
static_assert(std::is_floating_point<T>::value, "Must use floating-point types");
alignas(16) std::array<T,4> mVals;
public:
constexpr VectorR() noexcept = default;
constexpr VectorR(const VectorR&) noexcept = default;
constexpr VectorR(T a, T b, T c, T d) noexcept : mVals{{a, b, c, d}} { }
constexpr VectorR& operator=(const VectorR&) noexcept = default;
T& operator[](size_t idx) noexcept { return mVals[idx]; }
constexpr const T& operator[](size_t idx) const noexcept { return mVals[idx]; }
VectorR& operator+=(const VectorR &rhs) noexcept
{
mVals[0] += rhs.mVals[0];
mVals[1] += rhs.mVals[1];
mVals[2] += rhs.mVals[2];
mVals[3] += rhs.mVals[3];
return *this;
}
T normalize()
{
const T length{std::sqrt(mVals[0]*mVals[0] + mVals[1]*mVals[1] + mVals[2]*mVals[2])};
if(length > std::numeric_limits<T>::epsilon())
{
T inv_length{T{1}/length};
mVals[0] *= inv_length;
mVals[1] *= inv_length;
mVals[2] *= inv_length;
return length;
}
mVals[0] = mVals[1] = mVals[2] = T{0};
return T{0};
}
constexpr VectorR cross_product(const alu::VectorR<T> &rhs) const
{
return VectorR{
(*this)[1]*rhs[2] - (*this)[2]*rhs[1],
(*this)[2]*rhs[0] - (*this)[0]*rhs[2],
(*this)[0]*rhs[1] - (*this)[1]*rhs[0],
T{0}};
}
constexpr T dot_product(const alu::VectorR<T> &rhs) const
{ return (*this)[0]*rhs[0] + (*this)[1]*rhs[1] + (*this)[2]*rhs[2]; }
};
using Vector = VectorR<float>;
template<typename T>
class MatrixR {
static_assert(std::is_floating_point<T>::value, "Must use floating-point types");
alignas(16) std::array<T,16> mVals;
public:
constexpr MatrixR() noexcept = default;
constexpr MatrixR(const MatrixR&) noexcept = default;
constexpr MatrixR(T aa, T ab, T ac, T ad,
T ba, T bb, T bc, T bd,
T ca, T cb, T cc, T cd,
T da, T db, T dc, T dd) noexcept
: mVals{{aa,ab,ac,ad, ba,bb,bc,bd, ca,cb,cc,cd, da,db,dc,dd}}
{ }
constexpr MatrixR& operator=(const MatrixR&) noexcept = default;
auto operator[](size_t idx) noexcept { return al::span<T,4>{&mVals[idx*4], 4}; }
constexpr auto operator[](size_t idx) const noexcept
{ return al::span<const T,4>{&mVals[idx*4], 4}; }
static constexpr MatrixR Identity() noexcept
{
return MatrixR{
T{1}, T{0}, T{0}, T{0},
T{0}, T{1}, T{0}, T{0},
T{0}, T{0}, T{1}, T{0},
T{0}, T{0}, T{0}, T{1}};
}
};
using Matrix = MatrixR<float>;
template<typename T>
inline VectorR<T> operator*(const MatrixR<T> &mtx, const VectorR<T> &vec) noexcept
{
return VectorR<T>{
vec[0]*mtx[0][0] + vec[1]*mtx[1][0] + vec[2]*mtx[2][0] + vec[3]*mtx[3][0],
vec[0]*mtx[0][1] + vec[1]*mtx[1][1] + vec[2]*mtx[2][1] + vec[3]*mtx[3][1],
vec[0]*mtx[0][2] + vec[1]*mtx[1][2] + vec[2]*mtx[2][2] + vec[3]*mtx[3][2],
vec[0]*mtx[0][3] + vec[1]*mtx[1][3] + vec[2]*mtx[2][3] + vec[3]*mtx[3][3]};
}
template<typename U, typename T>
inline VectorR<U> cast_to(const VectorR<T> &vec) noexcept
{
return VectorR<U>{static_cast<U>(vec[0]), static_cast<U>(vec[1]),
static_cast<U>(vec[2]), static_cast<U>(vec[3])};
}
} // namespace alu
#endif /* COMMON_VECMAT_H */

View file

@ -0,0 +1,15 @@
#ifndef AL_VECTOR_H
#define AL_VECTOR_H
#include <vector>
#include "almalloc.h"
namespace al {
template<typename T, size_t alignment=alignof(T)>
using vector = std::vector<T, al::allocator<T, alignment>>;
} // namespace al
#endif /* AL_VECTOR_H */

View file

@ -13,10 +13,23 @@
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#include <shellapi.h>
#include <wchar.h>
#ifdef __cplusplus
#include <memory>
#define STATIC_CAST(...) static_cast<__VA_ARGS__>
#define REINTERPRET_CAST(...) reinterpret_cast<__VA_ARGS__>
#else
#define STATIC_CAST(...) (__VA_ARGS__)
#define REINTERPRET_CAST(...) (__VA_ARGS__)
#endif
static FILE *my_fopen(const char *fname, const char *mode)
{
WCHAR *wname=NULL, *wmode=NULL;
wchar_t *wname=NULL, *wmode=NULL;
int namelen, modelen;
FILE *file = NULL;
errno_t err;
@ -30,7 +43,12 @@ static FILE *my_fopen(const char *fname, const char *mode)
return NULL;
}
wname = calloc(sizeof(WCHAR), namelen+modelen);
#ifdef __cplusplus
auto strbuf = std::make_unique<wchar_t[]>(static_cast<size_t>(namelen+modelen));
wname = strbuf.get();
#else
wname = (wchar_t*)calloc(sizeof(wchar_t), (size_t)(namelen+modelen));
#endif
wmode = wname + namelen;
MultiByteToWideChar(CP_UTF8, 0, fname, -1, wname, namelen);
MultiByteToWideChar(CP_UTF8, 0, mode, -1, wmode, modelen);
@ -42,56 +60,58 @@ static FILE *my_fopen(const char *fname, const char *mode)
file = NULL;
}
#ifndef __cplusplus
free(wname);
#endif
return file;
}
#define fopen my_fopen
static char **arglist;
static void cleanup_arglist(void)
{
free(arglist);
}
static void GetUnicodeArgs(int *argc, char ***argv)
{
size_t total;
wchar_t **args;
int nargs, i;
args = CommandLineToArgvW(GetCommandLineW(), &nargs);
if(!args)
{
fprintf(stderr, "Failed to get command line args: %ld\n", GetLastError());
exit(EXIT_FAILURE);
}
total = sizeof(**argv) * nargs;
for(i = 0;i < nargs;i++)
total += WideCharToMultiByte(CP_UTF8, 0, args[i], -1, NULL, 0, NULL, NULL);
atexit(cleanup_arglist);
arglist = *argv = calloc(1, total);
(*argv)[0] = (char*)(*argv + nargs);
for(i = 0;i < nargs-1;i++)
{
int len = WideCharToMultiByte(CP_UTF8, 0, args[i], -1, (*argv)[i], 65535, NULL, NULL);
(*argv)[i+1] = (*argv)[i] + len;
}
WideCharToMultiByte(CP_UTF8, 0, args[i], -1, (*argv)[i], 65535, NULL, NULL);
*argc = nargs;
LocalFree(args);
}
#define GET_UNICODE_ARGS(argc, argv) GetUnicodeArgs(argc, argv)
#else
/* Do nothing. */
#define GET_UNICODE_ARGS(argc, argv)
/* SDL overrides main and provides UTF-8 args for us. */
#if !defined(SDL_MAIN_NEEDED) && !defined(SDL_MAIN_AVAILABLE)
int my_main(int, char**);
#define main my_main
#ifdef __cplusplus
extern "C"
#endif
int wmain(int argc, wchar_t **wargv)
{
char **argv;
size_t total;
int i;
total = sizeof(*argv) * STATIC_CAST(size_t)(argc);
for(i = 0;i < argc;i++)
total += STATIC_CAST(size_t)(WideCharToMultiByte(CP_UTF8, 0, wargv[i], -1, NULL, 0, NULL,
NULL));
#ifdef __cplusplus
auto argbuf = std::make_unique<char[]>(total);
argv = reinterpret_cast<char**>(argbuf.get());
#else
argv = (char**)calloc(1, total);
#endif
argv[0] = REINTERPRET_CAST(char*)(argv + argc);
for(i = 0;i < argc-1;i++)
{
int len = WideCharToMultiByte(CP_UTF8, 0, wargv[i], -1, argv[i], 65535, NULL, NULL);
argv[i+1] = argv[i] + len;
}
WideCharToMultiByte(CP_UTF8, 0, wargv[i], -1, argv[i], 65535, NULL, NULL);
#ifdef __cplusplus
return main(argc, argv);
#else
i = main(argc, argv);
free(argv);
return i;
#endif
}
#endif /* !defined(SDL_MAIN_NEEDED) && !defined(SDL_MAIN_AVAILABLE) */
#endif /* _WIN32 */
#endif /* WIN_MAIN_UTF8_H */