vectorize bitfieldExtract

This commit is contained in:
Christophe Riccio 2014-10-21 03:08:32 +02:00
parent 2179695e02
commit 679c765a07
3 changed files with 232 additions and 151 deletions

View File

@ -55,11 +55,11 @@ namespace glm
///
/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/uaddCarry.xml">GLSL uaddCarry man page</a>
/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
template <typename genUType>
GLM_FUNC_DECL genUType uaddCarry(
genUType const & x,
genUType const & y,
genUType & carry);
template <precision P, template <typename, precision> class vecType>
GLM_FUNC_DECL vecType<uint, P> uaddCarry(
vecType<uint, P> const & x,
vecType<uint, P> const & y,
vecType<uint, P> & carry);
/// Subtracts the 32-bit unsigned integer y from x, returning
/// the difference if non-negative, or pow(2, 32) plus the difference
@ -98,12 +98,12 @@ namespace glm
///
/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/imulExtended.xml">GLSL imulExtended man page</a>
/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
template <typename genIType>
template <precision P, template <typename, precision> class vecType>
GLM_FUNC_DECL void imulExtended(
genIType const & x,
genIType const & y,
genIType & msb,
genIType & lsb);
vecType<int, P> const & x,
vecType<int, P> const & y,
vecType<int, P> & msb,
vecType<int, P> & lsb);
/// Extracts bits [offset, offset + bits - 1] from value,
/// returning them in the least significant bits of the result.

View File

@ -42,13 +42,7 @@
namespace glm
{
// uaddCarry
template <>
GLM_FUNC_QUALIFIER uint uaddCarry
(
uint const & x,
uint const & y,
uint & Carry
)
GLM_FUNC_QUALIFIER uint uaddCarry(uint const & x, uint const & y, uint & Carry)
{
uint64 Value64 = static_cast<uint64>(x) + static_cast<uint64>(y);
uint32 Result = static_cast<uint32>(Value64 % (static_cast<uint64>(1) << static_cast<uint64>(32)));
@ -56,46 +50,16 @@ namespace glm
return Result;
}
template <>
GLM_FUNC_QUALIFIER uvec2 uaddCarry
(
uvec2 const & x,
uvec2 const & y,
uvec2 & Carry
)
template <precision P, template <typename, precision> class vecType>
GLM_FUNC_QUALIFIER vecType<uint, P> uaddCarry(vecType<uint, P> const & x, vecType<uint, P> const & y, vecType<uint, P> & Carry)
{
return uvec2(
uaddCarry(x[0], y[0], Carry[0]),
uaddCarry(x[1], y[1], Carry[1]));
}
vecType<uint64, P> Value64(vecType<uint64, P>(x) + vecType<uint64, P>(y));
vecType<uint32, P> Result(Value64 % (static_cast<uint64>(1) << static_cast<uint64>(32)));
template <>
GLM_FUNC_QUALIFIER uvec3 uaddCarry
(
uvec3 const & x,
uvec3 const & y,
uvec3 & Carry
)
{
return uvec3(
uaddCarry(x[0], y[0], Carry[0]),
uaddCarry(x[1], y[1], Carry[1]),
uaddCarry(x[2], y[2], Carry[2]));
}
vecType<bool, P> DoCarry(greaterThan(Value64 % (static_cast<uint64>(1) << static_cast<uint64>(32)), vecType<uint64, P>(1)));
Carry = mix(vecType<uint32, P>(0), vecType<uint32, P>(1), DoCarry);
template <>
GLM_FUNC_QUALIFIER uvec4 uaddCarry
(
uvec4 const & x,
uvec4 const & y,
uvec4 & Carry
)
{
return uvec4(
uaddCarry(x[0], y[0], Carry[0]),
uaddCarry(x[1], y[1], Carry[1]),
uaddCarry(x[2], y[2], Carry[2]),
uaddCarry(x[3], y[3], Carry[3]));
return Result;
}
// usubBorrow
@ -142,14 +106,7 @@ namespace glm
}
// imulExtended
template <>
GLM_FUNC_QUALIFIER void imulExtended
(
int const & x,
int const & y,
int & msb,
int & lsb
)
GLM_FUNC_QUALIFIER void imulExtended(int x, int y, int & msb, int & lsb)
{
GLM_STATIC_ASSERT(sizeof(int) == sizeof(int32), "int and int32 size mismatch");
@ -160,109 +117,40 @@ namespace glm
lsb = *PointerLSB;
}
template <>
GLM_FUNC_QUALIFIER void imulExtended
(
ivec2 const & x,
ivec2 const & y,
ivec2 & msb,
ivec2 & lsb
)
template <precision P, template <typename, precision> class vecType>
GLM_FUNC_QUALIFIER void imulExtended(vecType<int, P> const & x, vecType<int, P> const & y, vecType<int, P> & msb, vecType<int, P> & lsb)
{
imulExtended(x[0], y[0], msb[0], lsb[0]),
imulExtended(x[1], y[1], msb[1], lsb[1]);
}
GLM_STATIC_ASSERT(sizeof(int) == sizeof(int32), "int and int32 size mismatch");
template <>
GLM_FUNC_QUALIFIER void imulExtended
(
ivec3 const & x,
ivec3 const & y,
ivec3 & msb,
ivec3 & lsb
)
{
imulExtended(x[0], y[0], msb[0], lsb[0]),
imulExtended(x[1], y[1], msb[1], lsb[1]);
imulExtended(x[2], y[2], msb[2], lsb[2]);
}
template <>
GLM_FUNC_QUALIFIER void imulExtended
(
ivec4 const & x,
ivec4 const & y,
ivec4 & msb,
ivec4 & lsb
)
{
imulExtended(x[0], y[0], msb[0], lsb[0]),
imulExtended(x[1], y[1], msb[1], lsb[1]);
imulExtended(x[2], y[2], msb[2], lsb[2]);
imulExtended(x[3], y[3], msb[3], lsb[3]);
vecType<int64, P> Value64(vecType<int64, P>(x) * vecType<int64, P>(y));
lsb = vecType<int32, P>(Value64 & static_cast<int64>(0xFFFFFFFF));
msb = vecType<int32, P>((Value64 >> static_cast<int64>(32)) & static_cast<int64>(0xFFFFFFFF));
}
// bitfieldExtract
template <typename genIUType>
GLM_FUNC_QUALIFIER genIUType bitfieldExtract
(
genIUType const & Value,
int const & Offset,
int const & Bits
)
GLM_FUNC_QUALIFIER genIUType bitfieldExtract(genIUType Value, int Offset, int Bits)
{
int GenSize = int(sizeof(genIUType)) << int(3);
return bitfieldExtract(tvec1<genIUType>(Value), Offset, Bits).x;
}
template <typename T, precision P, template <typename, precision> class vecType>
GLM_FUNC_QUALIFIER vecType<T, P> bitfieldExtract(vecType<T, P> const & Value, int const & Offset, int const & Bits)
{
GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldExtract' only accept integer inputs");
int GenSize = int(sizeof(T)) << int(3);
assert(Offset + Bits <= GenSize);
genIUType ShiftLeft = Bits ? Value << (GenSize - (Bits + Offset)) : genIUType(0);
genIUType ShiftBack = ShiftLeft >> genIUType(GenSize - Bits);
vecType<T, P> ShiftLeft(0);
if(Bits)
ShiftLeft = Value << static_cast<T>(GenSize - (Bits + Offset));
vecType<T, P> const ShiftBack = ShiftLeft >> static_cast<T>(GenSize - Bits);
return ShiftBack;
}
template <typename T, precision P>
GLM_FUNC_QUALIFIER tvec2<T, P> bitfieldExtract
(
tvec2<T, P> const & Value,
int const & Offset,
int const & Bits
)
{
return tvec2<T, P>(
bitfieldExtract(Value[0], Offset, Bits),
bitfieldExtract(Value[1], Offset, Bits));
}
template <typename T, precision P>
GLM_FUNC_QUALIFIER tvec3<T, P> bitfieldExtract
(
tvec3<T, P> const & Value,
int const & Offset,
int const & Bits
)
{
return tvec3<T, P>(
bitfieldExtract(Value[0], Offset, Bits),
bitfieldExtract(Value[1], Offset, Bits),
bitfieldExtract(Value[2], Offset, Bits));
}
template <typename T, precision P>
GLM_FUNC_QUALIFIER tvec4<T, P> bitfieldExtract
(
tvec4<T, P> const & Value,
int const & Offset,
int const & Bits
)
{
return tvec4<T, P>(
bitfieldExtract(Value[0], Offset, Bits),
bitfieldExtract(Value[1], Offset, Bits),
bitfieldExtract(Value[2], Offset, Bits),
bitfieldExtract(Value[3], Offset, Bits));
}
// bitfieldInsert
template <typename genIUType>
GLM_FUNC_QUALIFIER genIUType bitfieldInsert

View File

@ -220,6 +220,66 @@ namespace findLSB
}
}//findLSB
namespace uaddCarry
{
int test()
{
int Error(0);
{
glm::uint x = 16;
glm::uint y = 17;
glm::uint Carry = 0;
glm::uint Result = glm::uaddCarry(x, y, Carry);
Error += Carry == 1 ? 0 : 1;
Error += Result == 33 ? 0 : 1;
}
{
glm::uvec1 x(16);
glm::uvec1 y(17);
glm::uvec1 Carry(0);
glm::uvec1 Result(glm::uaddCarry(x, y, Carry));
Error += glm::all(glm::equal(Carry, glm::uvec1(1))) ? 0 : 1;
Error += glm::all(glm::equal(Result, glm::uvec1(33))) ? 0 : 1;
}
{
glm::uvec2 x(16);
glm::uvec2 y(17);
glm::uvec2 Carry(0);
glm::uvec2 Result(glm::uaddCarry(x, y, Carry));
Error += glm::all(glm::equal(Carry, glm::uvec2(1))) ? 0 : 1;
Error += glm::all(glm::equal(Result, glm::uvec2(33))) ? 0 : 1;
}
{
glm::uvec3 x(16);
glm::uvec3 y(17);
glm::uvec3 Carry(0);
glm::uvec3 Result(glm::uaddCarry(x, y, Carry));
Error += glm::all(glm::equal(Carry, glm::uvec3(1))) ? 0 : 1;
Error += glm::all(glm::equal(Result, glm::uvec3(33))) ? 0 : 1;
}
{
glm::uvec4 x(16);
glm::uvec4 y(17);
glm::uvec4 Carry(0);
glm::uvec4 Result(glm::uaddCarry(x, y, Carry));
Error += glm::all(glm::equal(Carry, glm::uvec4(1))) ? 0 : 1;
Error += glm::all(glm::equal(Result, glm::uvec4(33))) ? 0 : 1;
}
return Error;
}
}//namespace uaddCarry
namespace usubBorrow
{
int test()
@ -280,12 +340,145 @@ namespace usubBorrow
}
}//namespace usubBorrow
namespace umulExtended
{
int test()
{
int Error(0);
{
glm::uint x = 2;
glm::uint y = 3;
glm::uint msb = 0;
glm::uint lsb = 0;
glm::umulExtended(x, y, msb, lsb);
Error += msb == 0 ? 0 : 1;
Error += lsb == 6 ? 0 : 1;
}
{
glm::uvec1 x(2);
glm::uvec1 y(3);
glm::uvec1 msb(0);
glm::uvec1 lsb(0);
glm::umulExtended(x, y, msb, lsb);
Error += glm::all(glm::equal(msb, glm::uvec1(0))) ? 0 : 1;
Error += glm::all(glm::equal(lsb, glm::uvec1(6))) ? 0 : 1;
}
{
glm::uvec2 x(2);
glm::uvec2 y(3);
glm::uvec2 msb(0);
glm::uvec2 lsb(0);
glm::umulExtended(x, y, msb, lsb);
Error += glm::all(glm::equal(msb, glm::uvec2(0))) ? 0 : 1;
Error += glm::all(glm::equal(lsb, glm::uvec2(6))) ? 0 : 1;
}
{
glm::uvec3 x(2);
glm::uvec3 y(3);
glm::uvec3 msb(0);
glm::uvec3 lsb(0);
glm::umulExtended(x, y, msb, lsb);
Error += glm::all(glm::equal(msb, glm::uvec3(0))) ? 0 : 1;
Error += glm::all(glm::equal(lsb, glm::uvec3(6))) ? 0 : 1;
}
{
glm::uvec4 x(2);
glm::uvec4 y(3);
glm::uvec4 msb(0);
glm::uvec4 lsb(0);
glm::umulExtended(x, y, msb, lsb);
Error += glm::all(glm::equal(msb, glm::uvec4(0))) ? 0 : 1;
Error += glm::all(glm::equal(lsb, glm::uvec4(6))) ? 0 : 1;
}
return Error;
}
}//namespace umulExtended
namespace imulExtended
{
int test()
{
int Error(0);
{
int x = 2;
int y = 3;
int msb = 0;
int lsb = 0;
glm::imulExtended(x, y, msb, lsb);
Error += msb == 0 ? 0 : 1;
Error += lsb == 6 ? 0 : 1;
}
{
glm::ivec1 x(2);
glm::ivec1 y(3);
glm::ivec1 msb(0);
glm::ivec1 lsb(0);
glm::imulExtended(x, y, msb, lsb);
Error += glm::all(glm::equal(msb, glm::ivec1(0))) ? 0 : 1;
Error += glm::all(glm::equal(lsb, glm::ivec1(6))) ? 0 : 1;
}
{
glm::ivec2 x(2);
glm::ivec2 y(3);
glm::ivec2 msb(0);
glm::ivec2 lsb(0);
glm::imulExtended(x, y, msb, lsb);
Error += glm::all(glm::equal(msb, glm::ivec2(0))) ? 0 : 1;
Error += glm::all(glm::equal(lsb, glm::ivec2(6))) ? 0 : 1;
}
{
glm::ivec3 x(2);
glm::ivec3 y(3);
glm::ivec3 msb(0);
glm::ivec3 lsb(0);
glm::imulExtended(x, y, msb, lsb);
Error += glm::all(glm::equal(msb, glm::ivec3(0))) ? 0 : 1;
Error += glm::all(glm::equal(lsb, glm::ivec3(6))) ? 0 : 1;
}
{
glm::ivec4 x(2);
glm::ivec4 y(3);
glm::ivec4 msb(0);
glm::ivec4 lsb(0);
glm::imulExtended(x, y, msb, lsb);
Error += glm::all(glm::equal(msb, glm::ivec4(0))) ? 0 : 1;
Error += glm::all(glm::equal(lsb, glm::ivec4(6))) ? 0 : 1;
}
return Error;
}
}//namespace imulExtended
int main()
{
int Error = 0;
std::cout << "sizeof(glm::uint64): " << sizeof(glm::detail::uint64) << std::endl;
Error += ::umulExtended::test();
Error += ::imulExtended::test();
Error += ::uaddCarry::test();
Error += ::usubBorrow::test();
Error += ::bitfieldExtract::test();
Error += ::bitfieldReverse::test();