Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

More simd/Neon functions #950

Merged
merged 8 commits into from
Nov 2, 2019
Merged

More simd/Neon functions #950

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
c35bc3d
Fix Arm 64bit detection
Sep 10, 2019
e508cc6
More simd/Neon functions
Sep 10, 2019
9b0a053
Fixing GLM_MODEL for Windows compilers targeting ARM
Sep 10, 2019
7b06a98
Let test return the number of errors
UniQP Sep 11, 2019
4ff530b
Merge pull request #949 from amaury-ml/neon
Groovounet Sep 12, 2019
7c07544
Merge pull request #952 from UniQP/fixReturnValue
Groovounet Sep 12, 2019
88a7aee
More simd/Neon functions
Sep 10, 2019
8a17335
More simd/Neon functions
Sep 18, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
66 changes: 66 additions & 0 deletions glm/detail/func_geometric_simd.inl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,4 +96,70 @@ namespace detail
}//namespace detail
}//namespace glm

#elif GLM_ARCH & GLM_ARCH_NEON_BIT
namespace glm{
namespace detail
{
template<qualifier Q>
struct compute_length<4, float, Q, true>
{
GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& v)
{
return compute_dot<vec<4, float, Q>, float, true>::call(v, v);
}
};

template<qualifier Q>
struct compute_distance<4, float, Q, true>
{
GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& p0, vec<4, float, Q> const& p1)
{
return compute_length<4, float, Q, true>::call(p1 - p0);
}
};


template<qualifier Q>
struct compute_dot<vec<4, float, Q>, float, true>
{
GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& x, vec<4, float, Q> const& y)
{
#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
float32x4_t v = vmulq_f32(x.data, y.data);
v = vpaddq_f32(v, v);
v = vpaddq_f32(v, v);
return vgetq_lane_f32(v, 0);
#else // Armv7a with Neon
float32x4_t p = vmulq_f32(x.data, y.data);
float32x2_t v = vpadd_f32(vget_low_f32(p), vget_high_f32(p));
v = vpadd_f32(v, v);
return vget_lane_f32(v, 0);
#endif
}
};

template<qualifier Q>
struct compute_normalize<4, float, Q, true>
{
GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
{
float32x4_t p = vmulq_f32(v.data, v.data);
#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
p = vpaddq_f32(p, p);
p = vpaddq_f32(p, p);
#else
float32x2_t t = vpadd_f32(vget_low_f32(p), vget_high_f32(p));
t = vpadd_f32(t, t);
p = vcombine_f32(t, t);
#endif

float32x4_t vd = vrsqrteq_f32(p);
vec<4, float, Q> Result;
Result.data = vmulq_f32(v, vd);
return Result;
}
};
}//namespace detail
}//namespace glm

#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
37 changes: 37 additions & 0 deletions glm/detail/func_matrix_simd.inl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -91,4 +91,41 @@ namespace detail
# endif
}//namespace glm

#elif GLM_ARCH & GLM_ARCH_NEON_BIT

namespace glm {
#if GLM_LANG & GLM_LANG_CXX11_FLAG
template <qualifier Q>
GLM_FUNC_QUALIFIER
typename std::enable_if<detail::is_aligned<Q>::value, mat<4, 4, float, Q>>::type
operator*(mat<4, 4, float, Q> const & m1, mat<4, 4, float, Q> const & m2)
{
auto MulRow = [&](int l) {
float32x4_t const SrcA = m2[l].data;

#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
float32x4_t r= vmulq_laneq_f32(m1[0].data, SrcA, 0);
r = vaddq_f32(r, vmulq_laneq_f32(m1[1].data, SrcA, 1));
r = vaddq_f32(r, vmulq_laneq_f32(m1[2].data, SrcA, 2));
r = vaddq_f32(r, vmulq_laneq_f32(m1[3].data, SrcA, 3));
#else
float32x4_t r= vmulq_f32(m1[0].data, vdupq_n_f32(vgetq_lane_f32(SrcA, 0)));
r = vaddq_f32(r, vmulq_f32(m1[1].data, vdupq_n_f32(vgetq_lane_f32(SrcA, 1))));
r = vaddq_f32(r, vmulq_f32(m1[2].data, vdupq_n_f32(vgetq_lane_f32(SrcA, 2))));
r = vaddq_f32(r, vmulq_f32(m1[3].data, vdupq_n_f32(vgetq_lane_f32(SrcA, 3))));
#endif

return r;
};

mat<4, 4, float, aligned_highp> Result;
Result[0].data = MulRow(0);
Result[1].data = MulRow(1);
Result[2].data = MulRow(2);
Result[3].data = MulRow(3);

return Result;
}
#endif // CXX11
}//namespace glm
#endif
4 changes: 2 additions & 2 deletions glm/detail/setup.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,9 +35,9 @@
///////////////////////////////////////////////////////////////////////////////////
// Build model

#if defined(__arch64__) || defined(__LP64__) || defined(_M_X64) || defined(__ppc64__) || defined(__x86_64__)
#if defined(_M_ARM64) || defined(__LP64__) || defined(_M_X64) || defined(__ppc64__) || defined(__x86_64__)
# define GLM_MODEL GLM_MODEL_64
#elif defined(__i386__) || defined(__ppc__)
#elif defined(__i386__) || defined(__ppc__) || defined(__ILP32__) || defined(_M_ARM)
# define GLM_MODEL GLM_MODEL_32
#else
# define GLM_MODEL GLM_MODEL_32
Expand Down
2 changes: 1 addition & 1 deletion test/gtx/gtx_fast_square_root.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ int test_fastInverseSqrt()
Error += glm::all(glm::epsilonEqual(glm::fastInverseSqrt(glm::dvec3(1.0)), glm::dvec3(1.0), 0.01)) ? 0 : 1;
Error += glm::all(glm::epsilonEqual(glm::fastInverseSqrt(glm::dvec4(1.0)), glm::dvec4(1.0), 0.01)) ? 0 : 1;

return 0;
return Error;
}

int test_fastDistance()
Expand Down