AuroraMiddleware/glm
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Fixed sign with signed integer function on non-x86 architecture

Browse Source
This commit is contained in:
Christophe Riccio 2016-05-24 13:29:18 +02:00
parent 1208eb63f7
commit 72c741d8ea
5 changed files with 117 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
glm/detail/func_common.inl
View File

@ -115,7 +115,7 @@ namespace detail
}
};
template <typename T, precision P, template <class, precision> class vecType, bool isFloat = true, bool isSigned = true>
template <typename T, precision P, template <class, precision> class vecType, bool isFloat = true>
struct compute_sign
{
GLM_FUNC_QUALIFIER static vecType<T, P> call(vecType<T, P> const & x)
@ -124,17 +124,9 @@ namespace detail
}
};
# if GLM_ARCH == GLM_ARCH_X86
template <typename T, precision P, template <class, precision> class vecType>
struct compute_sign<T, P, vecType, false, false>
{
GLM_FUNC_QUALIFIER static vecType<T, P> call(vecType<T, P> const & x)
{
return vecType<T, P>(glm::greaterThan(x , vecType<T, P>(0)));
}
};
template <typename T, precision P, template <class, precision> class vecType>
struct compute_sign<T, P, vecType, false, true>
struct compute_sign<T, P, vecType, false>
{
GLM_FUNC_QUALIFIER static vecType<T, P> call(vecType<T, P> const & x)
{
@ -144,6 +136,7 @@ namespace detail
return (x >> Shift) | y;
}
};
# endif
template <typename T, precision P, template <class, precision> class vecType, typename genType, bool isFloat = true>
struct compute_mod

88
glm/detail/type_vec4_simd.inl
View File

@ -62,60 +62,114 @@ namespace detail
};
template <precision P>
struct compute_vec4_and<int, P>
struct compute_vec4_and<int32, P>
{
static tvec4<int, P> call(tvec4<int, P> const& a, tvec4<int, P> const& b)
static tvec4<int32, P> call(tvec4<int32, P> const& a, tvec4<int32, P> const& b)
{
tvec4<int, P> Result(uninitialize);
tvec4<int32, P> Result(uninitialize);
Result.data = _mm_and_si128(a.data, b.data);
return Result;
}
};
template <precision P>
struct compute_vec4_or<int, P>
struct compute_vec4_and<uint32, P>
{
static tvec4<int, P> call(tvec4<int, P> const& a, tvec4<int, P> const& b)
static tvec4<uint32, P> call(tvec4<uint32, P> const& a, tvec4<uint32, P> const& b)
{
tvec4<int, P> Result(uninitialize);
tvec4<uint32, P> Result(uninitialize);
Result.data = _mm_and_si128(a.data, b.data);
return Result;
}
};
template <precision P>
struct compute_vec4_or<int32, P>
{
static tvec4<int32, P> call(tvec4<int32, P> const& a, tvec4<int32, P> const& b)
{
tvec4<int32, P> Result(uninitialize);
Result.data = _mm_or_si128(a.data, b.data);
return Result;
}
};
template <precision P>
struct compute_vec4_xor<int, P>
struct compute_vec4_or<uint32, P>
{
static tvec4<int, P> call(tvec4<int, P> const& a, tvec4<int, P> const& b)
static tvec4<uint32, P> call(tvec4<uint32, P> const& a, tvec4<uint32, P> const& b)
{
tvec4<int, P> Result(uninitialize);
tvec4<uint32, P> Result(uninitialize);
Result.data = _mm_or_si128(a.data, b.data);
return Result;
}
};
template <precision P>
struct compute_vec4_xor<int32, P>
{
static tvec4<int32, P> call(tvec4<int32, P> const& a, tvec4<int32, P> const& b)
{
tvec4<int32, P> Result(uninitialize);
Result.data = _mm_xor_si128(a.data, b.data);
return Result;
}
};
/*
template <precision P>
struct compute_vec4_shift_left<int, P>
struct compute_vec4_xor<uint32, P>
{
static tvec4<int, P> call(tvec4<int, P> const& a, tvec4<int, P> const& b)
static tvec4<uint32, P> call(tvec4<uint32, P> const& a, tvec4<uint32, P> const& b)
{
tvec4<int, P> Result(uninitialize);
tvec4<uint32, P> Result(uninitialize);
Result.data = _mm_xor_si128(a.data, b.data);
return Result;
}
};
template <precision P>
struct compute_vec4_shift_left<int32, P>
{
static tvec4<int32, P> call(tvec4<int32, P> const& a, tvec4<int32, P> const& b)
{
tvec4<int32, P> Result(uninitialize);
Result.data = _mm_sll_epi32(a.data, b.data);
return Result;
}
};
template <precision P>
struct compute_vec4_shift_right<int, P>
struct compute_vec4_shift_left<uint32, P>
{
static tvec4<int, P> call(tvec4<int, P> const& a, tvec4<int, P> const& b)
static tvec4<uint32, P> call(tvec4<uint32, P> const& a, tvec4<uint32, P> const& b)
{
tvec4<int, P> Result(uninitialize);
tvec4<uint32, P> Result(uninitialize);
Result.data = _mm_sll_epi32(a.data, b.data);
return Result;
}
};
template <precision P>
struct compute_vec4_shift_right<int32, P>
{
static tvec4<int32, P> call(tvec4<int32, P> const& a, tvec4<int32, P> const& b)
{
tvec4<int32, P> Result(uninitialize);
Result.data = _mm_srl_epi32(a.data, b.data);
return Result;
}
};
template <precision P>
struct compute_vec4_shift_right<uint32, P>
{
static tvec4<uint32, P> call(tvec4<uint32, P> const& a, tvec4<uint32, P> const& b)
{
tvec4<uint32, P> Result(uninitialize);
Result.data = _mm_srl_epi32(a.data, b.data);
return Result;
}
};
*/
}//namespace detail
# if !GLM_HAS_DEFAULTED_FUNCTIONS

1
readme.md
View File

@ -80,6 +80,7 @@ glm::mat4 camera(float Translate, glm::vec2 const & Rotate)
- Fixed GTX_extended_min_max filename typo #386
- Fixed intersectRayTriangle to not do any unintentional backface culling
- Fixed long long warnings when using C++98 on GCC and Clang #482
- Fixed sign with signed integer function on non-x86 architecture
##### Deprecation:
- Removed GLM_FORCE_SIZE_FUNC define

2
test/core/core_func_common.cpp
View File

@ -1235,7 +1235,7 @@ namespace ldexp_
int main()
{
int Error(0);
int Error = 0;
Error += sign::test();
Error += floor_::test();

40
test/core/core_type_vec4.cpp
View File

@ -502,7 +502,47 @@ int test_vec4_simd()
int main()
{
int Error(0);
/*
{
glm::ivec4 const a1(2);
glm::ivec4 const b1 = a1 >> 1;
__m128i const e1 = _mm_set1_epi32(2);
__m128i const f1 = _mm_srli_epi32(e1, 1);
glm::ivec4 const g1 = *reinterpret_cast<glm::ivec4 const* const>(&f1);
glm::ivec4 const a2(-2);
glm::ivec4 const b2 = a2 >> 1;
__m128i const e2 = _mm_set1_epi32(-1);
__m128i const f2 = _mm_srli_epi32(e2, 1);
glm::ivec4 const g2 = *reinterpret_cast<glm::ivec4 const* const>(&f2);
printf("GNI\n");
}
{
glm::uvec4 const a1(2);
glm::uvec4 const b1 = a1 >> 1u;
__m128i const e1 = _mm_set1_epi32(2);
__m128i const f1 = _mm_srli_epi32(e1, 1);
glm::uvec4 const g1 = *reinterpret_cast<glm::uvec4 const* const>(&f1);
glm::uvec4 const a2(-1);
glm::uvec4 const b2 = a2 >> 1u;
__m128i const e2 = _mm_set1_epi32(-1);
__m128i const f2 = _mm_srli_epi32(e2, 1);
glm::uvec4 const g2 = *reinterpret_cast<glm::uvec4 const* const>(&f2);
printf("GNI\n");
}
*/
glm::vec4 v;
assert(v.length() == 4);