Merge branch 'vectorize' of https://github.com/g-truc/glm into vectorize
This commit is contained in:
commit
378074ee66
@ -44,7 +44,7 @@ namespace detail
|
||||
{
|
||||
GLM_FUNC_QUALIFIER int mask(int Bits)
|
||||
{
|
||||
return Bits >= 32 ? 0xffffffff : (static_cast<int>(1) << Bits) - static_cast<int>(1);
|
||||
return ~((~0) << Bits);
|
||||
}
|
||||
}//namespace detail
|
||||
|
||||
|
@ -57,13 +57,14 @@ namespace glm
|
||||
/// Build a mask of 'count' bits
|
||||
///
|
||||
/// @see gtc_bitfield
|
||||
GLM_FUNC_DECL int mask(int Bits);
|
||||
template <typename genType>
|
||||
GLM_FUNC_DECL genType mask(genType Bits);
|
||||
|
||||
/// Build a mask of 'count' bits
|
||||
///
|
||||
/// @see gtc_bitfield
|
||||
template <precision P, template <typename, precision> class vecType>
|
||||
GLM_FUNC_DECL vecType<int, P> mask(vecType<int, P> const & v);
|
||||
template <typename T, precision P, template <typename, precision> class vecType>
|
||||
GLM_FUNC_DECL vecType<T, P> mask(vecType<T, P> const & v);
|
||||
|
||||
/// Rotate all bits to the right. All the bits dropped in the right side are inserted back on the left side.
|
||||
///
|
||||
|
@ -107,23 +107,23 @@ namespace detail
|
||||
glm::uint32 REG1(x);
|
||||
glm::uint32 REG2(y);
|
||||
glm::uint32 REG3(z);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 16) | REG1) & glm::uint32(0x00FF0000FF0000FF);
|
||||
REG2 = ((REG2 << 16) | REG2) & glm::uint32(0x00FF0000FF0000FF);
|
||||
REG3 = ((REG3 << 16) | REG3) & glm::uint32(0x00FF0000FF0000FF);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 8) | REG1) & glm::uint32(0xF00F00F00F00F00F);
|
||||
REG2 = ((REG2 << 8) | REG2) & glm::uint32(0xF00F00F00F00F00F);
|
||||
REG3 = ((REG3 << 8) | REG3) & glm::uint32(0xF00F00F00F00F00F);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 4) | REG1) & glm::uint32(0x30C30C30C30C30C3);
|
||||
REG2 = ((REG2 << 4) | REG2) & glm::uint32(0x30C30C30C30C30C3);
|
||||
REG3 = ((REG3 << 4) | REG3) & glm::uint32(0x30C30C30C30C30C3);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 2) | REG1) & glm::uint32(0x9249249249249249);
|
||||
REG2 = ((REG2 << 2) | REG2) & glm::uint32(0x9249249249249249);
|
||||
REG3 = ((REG3 << 2) | REG3) & glm::uint32(0x9249249249249249);
|
||||
|
||||
|
||||
return REG1 | (REG2 << 1) | (REG3 << 2);
|
||||
}
|
||||
|
||||
@ -133,30 +133,30 @@ namespace detail
|
||||
glm::uint64 REG1(x);
|
||||
glm::uint64 REG2(y);
|
||||
glm::uint64 REG3(z);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 32) | REG1) & glm::uint64(0xFFFF00000000FFFF);
|
||||
REG2 = ((REG2 << 32) | REG2) & glm::uint64(0xFFFF00000000FFFF);
|
||||
REG3 = ((REG3 << 32) | REG3) & glm::uint64(0xFFFF00000000FFFF);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 16) | REG1) & glm::uint64(0x00FF0000FF0000FF);
|
||||
REG2 = ((REG2 << 16) | REG2) & glm::uint64(0x00FF0000FF0000FF);
|
||||
REG3 = ((REG3 << 16) | REG3) & glm::uint64(0x00FF0000FF0000FF);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 8) | REG1) & glm::uint64(0xF00F00F00F00F00F);
|
||||
REG2 = ((REG2 << 8) | REG2) & glm::uint64(0xF00F00F00F00F00F);
|
||||
REG3 = ((REG3 << 8) | REG3) & glm::uint64(0xF00F00F00F00F00F);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 4) | REG1) & glm::uint64(0x30C30C30C30C30C3);
|
||||
REG2 = ((REG2 << 4) | REG2) & glm::uint64(0x30C30C30C30C30C3);
|
||||
REG3 = ((REG3 << 4) | REG3) & glm::uint64(0x30C30C30C30C30C3);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 2) | REG1) & glm::uint64(0x9249249249249249);
|
||||
REG2 = ((REG2 << 2) | REG2) & glm::uint64(0x9249249249249249);
|
||||
REG3 = ((REG3 << 2) | REG3) & glm::uint64(0x9249249249249249);
|
||||
|
||||
|
||||
return REG1 | (REG2 << 1) | (REG3 << 2);
|
||||
}
|
||||
|
||||
|
||||
template <>
|
||||
GLM_FUNC_QUALIFIER glm::uint64 bitfieldInterleave(glm::uint32 x, glm::uint32 y, glm::uint32 z)
|
||||
{
|
||||
@ -194,25 +194,25 @@ namespace detail
|
||||
glm::uint32 REG2(y);
|
||||
glm::uint32 REG3(z);
|
||||
glm::uint32 REG4(w);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 12) | REG1) & glm::uint32(0x000F000F000F000F);
|
||||
REG2 = ((REG2 << 12) | REG2) & glm::uint32(0x000F000F000F000F);
|
||||
REG3 = ((REG3 << 12) | REG3) & glm::uint32(0x000F000F000F000F);
|
||||
REG4 = ((REG4 << 12) | REG4) & glm::uint32(0x000F000F000F000F);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 6) | REG1) & glm::uint32(0x0303030303030303);
|
||||
REG2 = ((REG2 << 6) | REG2) & glm::uint32(0x0303030303030303);
|
||||
REG3 = ((REG3 << 6) | REG3) & glm::uint32(0x0303030303030303);
|
||||
REG4 = ((REG4 << 6) | REG4) & glm::uint32(0x0303030303030303);
|
||||
|
||||
|
||||
REG1 = ((REG1 << 3) | REG1) & glm::uint32(0x1111111111111111);
|
||||
REG2 = ((REG2 << 3) | REG2) & glm::uint32(0x1111111111111111);
|
||||
REG3 = ((REG3 << 3) | REG3) & glm::uint32(0x1111111111111111);
|
||||
REG4 = ((REG4 << 3) | REG4) & glm::uint32(0x1111111111111111);
|
||||
|
||||
|
||||
return REG1 | (REG2 << 1) | (REG3 << 2) | (REG4 << 3);
|
||||
}
|
||||
|
||||
|
||||
template <>
|
||||
GLM_FUNC_QUALIFIER glm::uint64 bitfieldInterleave(glm::uint16 x, glm::uint16 y, glm::uint16 z, glm::uint16 w)
|
||||
{
|
||||
@ -245,21 +245,26 @@ namespace detail
|
||||
}
|
||||
}//namespace detail
|
||||
|
||||
GLM_FUNC_QUALIFIER int mask(int Bits)
|
||||
template <typename genType>
|
||||
GLM_FUNC_QUALIFIER genType mask(genType Bits)
|
||||
{
|
||||
return Bits >= sizeof(Bits) * 8 ? ~static_cast<int>(0) : (static_cast<int>(1) << Bits) - static_cast<int>(1);
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<genIType>::is_integer, "'mask' accepts only integer values");
|
||||
|
||||
return ~((~static_cast<genType>(0)) << Bits);
|
||||
}
|
||||
|
||||
template <precision P, template <typename, precision> class vecType>
|
||||
GLM_FUNC_QUALIFIER vecType<int, P> mask(vecType<int, P> const & v)
|
||||
template <typename T, precision P, template <typename, precision> class vecType>
|
||||
GLM_FUNC_QUALIFIER vecType<T, P> mask(vecType<T, P> const & v)
|
||||
{
|
||||
return detail::functor1<int, int, P, vecType>::call(mask, v);
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'mask' accepts only integer values");
|
||||
|
||||
return ~((~static_cast<T>(0)) << v);
|
||||
}
|
||||
|
||||
template <typename genIType>
|
||||
GLM_FUNC_QUALIFIER genIType bitfieldRotateRight(genIType In, int Shift)
|
||||
{
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<genIType>::is_integer, "'bitfieldRotateRight' only accept integer values");
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<genIType>::is_integer, "'bitfieldRotateRight' accepts only integer values");
|
||||
|
||||
int const BitSize = static_cast<genIType>(sizeof(genIType) * 8);
|
||||
return (In << static_cast<genIType>(Shift)) | (In >> static_cast<genIType>(BitSize - Shift));
|
||||
@ -268,7 +273,7 @@ namespace detail
|
||||
template <typename T, precision P, template <typename, precision> class vecType>
|
||||
GLM_FUNC_QUALIFIER vecType<T, P> bitfieldRotateRight(vecType<T, P> const & In, int Shift)
|
||||
{
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldRotateRight' only accept integer values");
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldRotateRight' accepts only integer values");
|
||||
|
||||
int const BitSize = static_cast<int>(sizeof(T) * 8);
|
||||
return (In << static_cast<T>(Shift)) | (In >> static_cast<T>(BitSize - Shift));
|
||||
@ -277,7 +282,7 @@ namespace detail
|
||||
template <typename genIType>
|
||||
GLM_FUNC_QUALIFIER genIType bitfieldRotateLeft(genIType In, int Shift)
|
||||
{
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<genIType>::is_integer, "'bitfieldRotateLeft' only accept integer values");
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<genIType>::is_integer, "'bitfieldRotateLeft' accepts only integer values");
|
||||
|
||||
int const BitSize = static_cast<genIType>(sizeof(genIType) * 8);
|
||||
return (In >> static_cast<genIType>(Shift)) | (In << static_cast<genIType>(BitSize - Shift));
|
||||
@ -286,7 +291,7 @@ namespace detail
|
||||
template <typename T, precision P, template <typename, precision> class vecType>
|
||||
GLM_FUNC_QUALIFIER vecType<T, P> bitfieldRotateLeft(vecType<T, P> const & In, int Shift)
|
||||
{
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldRotateLeft' only accept integer values");
|
||||
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldRotateLeft' accepts only integer values");
|
||||
|
||||
int const BitSize = static_cast<int>(sizeof(T) * 8);
|
||||
return (In >> static_cast<T>(Shift)) | (In << static_cast<T>(BitSize - Shift));
|
||||
|
@ -9,6 +9,7 @@
|
||||
|
||||
#include <glm/gtc/bitfield.hpp>
|
||||
#include <glm/gtc/type_precision.hpp>
|
||||
#include <glm/vector_relational.hpp>
|
||||
//#include <glm/vec2.hpp>
|
||||
#include <ctime>
|
||||
#include <cstdio>
|
||||
@ -16,6 +17,18 @@
|
||||
|
||||
namespace mask
|
||||
{
|
||||
template <typename genType>
|
||||
struct type
|
||||
{
|
||||
genType Value;
|
||||
genType Return;
|
||||
};
|
||||
|
||||
inline int mask_zero(int Bits)
|
||||
{
|
||||
return ~((~0) << Bits);
|
||||
}
|
||||
|
||||
inline int mask_mix(int Bits)
|
||||
{
|
||||
return Bits >= 32 ? 0xffffffff : (static_cast<int>(1) << Bits) - static_cast<int>(1);
|
||||
@ -62,16 +75,97 @@ namespace mask
|
||||
|
||||
std::clock_t Timestamp4 = std::clock();
|
||||
|
||||
{
|
||||
std::vector<int> Mask;
|
||||
Mask.resize(Count);
|
||||
for(int i = 0; i < Count; ++i)
|
||||
Mask[i] = mask_zero(i % 32);
|
||||
}
|
||||
|
||||
std::clock_t Timestamp5 = std::clock();
|
||||
|
||||
std::clock_t TimeMix = Timestamp2 - Timestamp1;
|
||||
std::clock_t TimeLoop = Timestamp3 - Timestamp2;
|
||||
std::clock_t TimeDefault = Timestamp4 - Timestamp3;
|
||||
std::clock_t TimeZero = Timestamp5 - Timestamp4;
|
||||
|
||||
printf("mask[mix]: %d\n", TimeMix);
|
||||
printf("mask[loop]: %d\n", TimeLoop);
|
||||
printf("mask[default]: %d\n", TimeDefault);
|
||||
printf("mask[zero]: %d\n", TimeZero);
|
||||
|
||||
return TimeDefault < TimeLoop ? 0 : 1;
|
||||
}
|
||||
|
||||
int test_uint()
|
||||
{
|
||||
type<glm::uint> const Data[] =
|
||||
{
|
||||
{0, 0x00000000},
|
||||
{1, 0x00000001},
|
||||
{2, 0x00000003},
|
||||
{3, 0x00000007}
|
||||
};
|
||||
|
||||
int Error(0);
|
||||
|
||||
for(std::size_t i = 0; i < sizeof(Data) / sizeof(type<int>); ++i)
|
||||
{
|
||||
int Result = mask_zero(Data[i].Value);
|
||||
Error += Data[i].Return == Result ? 0 : 1;
|
||||
}
|
||||
|
||||
for(std::size_t i = 0; i < sizeof(Data) / sizeof(type<int>); ++i)
|
||||
{
|
||||
int Result = mask_mix(Data[i].Value);
|
||||
Error += Data[i].Return == Result ? 0 : 1;
|
||||
}
|
||||
|
||||
for(std::size_t i = 0; i < sizeof(Data) / sizeof(type<int>); ++i)
|
||||
{
|
||||
int Result = mask_loop(Data[i].Value);
|
||||
Error += Data[i].Return == Result ? 0 : 1;
|
||||
}
|
||||
|
||||
for(std::size_t i = 0; i < sizeof(Data) / sizeof(type<int>); ++i)
|
||||
{
|
||||
int Result = glm::mask(Data[i].Value);
|
||||
Error += Data[i].Return == Result ? 0 : 1;
|
||||
}
|
||||
|
||||
return Error;
|
||||
}
|
||||
|
||||
int test_uvec4()
|
||||
{
|
||||
type<glm::ivec4> const Data[] =
|
||||
{
|
||||
{glm::ivec4(0), glm::ivec4(0x00000000)},
|
||||
{glm::ivec4(1), glm::ivec4(0x00000001)},
|
||||
{glm::ivec4(2), glm::ivec4(0x00000003)},
|
||||
{glm::ivec4(3), glm::ivec4(0x00000007)}
|
||||
};
|
||||
|
||||
int Error(0);
|
||||
|
||||
for(std::size_t i = 0, n = sizeof(Data) / sizeof(type<glm::ivec4>); i < n; ++i)
|
||||
{
|
||||
glm::ivec4 Result = glm::mask(Data[i].Value);
|
||||
Error += glm::all(glm::equal(Data[i].Return, Result)) ? 0 : 1;
|
||||
}
|
||||
|
||||
return Error;
|
||||
}
|
||||
|
||||
int test()
|
||||
{
|
||||
int Error(0);
|
||||
|
||||
Error += test_uint();
|
||||
Error += test_uvec4();
|
||||
|
||||
return Error;
|
||||
}
|
||||
}//namespace mask
|
||||
|
||||
|
||||
@ -502,11 +596,13 @@ int main()
|
||||
{
|
||||
int Error(0);
|
||||
|
||||
Error += ::mask::perf();
|
||||
Error += ::mask::test();
|
||||
Error += ::bitfieldInterleave3::test();
|
||||
Error += ::bitfieldInterleave4::test();
|
||||
Error += ::bitfieldInterleave::test();
|
||||
//Error += ::bitRevert::test();
|
||||
|
||||
Error += ::mask::perf();
|
||||
|
||||
return Error;
|
||||
}
|
||||
|
Loading…
Reference in New Issue
Block a user