mirror/glm
1
0
Fork 0
mirror of https://github.com/g-truc/glm.git synced 2024-11-14 14:04:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixed SIMD smoothstep

Browse Source
This commit is contained in:
Christophe 2024-02-08 21:46:37 +01:00 committed by Christophe
parent b101e8f3de
commit 1ac95994c4
2 changed files with 21 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
glm/simd/geometric.h
View File

@ -102,17 +102,23 @@ GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_reflect(glm_vec4 I, glm_vec4 N)
GLM_FUNC_QUALIFIER __m128 glm_vec4_refract(glm_vec4 I, glm_vec4 N, glm_vec4 eta) GLM_FUNC_QUALIFIER __m128 glm_vec4_refract(glm_vec4 I, glm_vec4 N, glm_vec4 eta)
{ {
glm_vec4 const dot0 = glm_vec4_dot(N, I); // k = 1.0 - eta * eta * (1.0 - dot(N, I) * dot(N, I));
glm_vec4 const mul0 = _mm_mul_ps(eta, eta); // if (k < 0.0)
glm_vec4 const mul1 = _mm_mul_ps(dot0, dot0); // R = genType(0.0); // or genDType(0.0)
glm_vec4 const sub0 = _mm_sub_ps(_mm_set1_ps(1.0f), mul0); // else
glm_vec4 const sub1 = _mm_sub_ps(_mm_set1_ps(1.0f), mul1); // R = eta * I - (eta * dot(N, I) + sqrt(k)) * N;
glm_vec4 const mul2 = _mm_mul_ps(sub0, sub1);
if(_mm_movemask_ps(_mm_cmplt_ss(mul2, _mm_set1_ps(0.0f))) == 0) glm_vec4 const dot0 = glm_vec4_dot(N, I); // dot(N, I)
glm_vec4 const mul0 = _mm_mul_ps(eta, eta); // eta * eta
glm_vec4 const mul1 = _mm_mul_ps(dot0, dot0); // dot(N, I) * dot(N, I)
glm_vec4 const sub1 = _mm_sub_ps(_mm_set1_ps(1.0f), mul1); // (1.0 - dot(N, I) * dot(N, I))
glm_vec4 const mul2 = _mm_mul_ps(mul0, sub1); // eta * eta * (1.0 - dot(N, I) * dot(N, I))
glm_vec4 const sub0 = _mm_sub_ps(_mm_set1_ps(1.0f), mul2); // 1.0 - eta * eta * (1.0 - dot(N, I) * dot(N, I))
if(_mm_movemask_ps(_mm_cmplt_ss(sub0, _mm_set1_ps(0.0f))) == 0)
return _mm_set1_ps(0.0f); return _mm_set1_ps(0.0f);
glm_vec4 const sqt0 = _mm_sqrt_ps(mul2); glm_vec4 const sqt0 = _mm_sqrt_ps(sub0);
glm_vec4 const mad0 = glm_vec4_fma(eta, dot0, sqt0); glm_vec4 const mad0 = glm_vec4_fma(eta, dot0, sqt0);
glm_vec4 const mul4 = _mm_mul_ps(mad0, N); glm_vec4 const mul4 = _mm_mul_ps(mad0, N);
glm_vec4 const mul5 = _mm_mul_ps(eta, I); glm_vec4 const mul5 = _mm_mul_ps(eta, I);

7
test/core/core_func_geometric.cpp
View File

@ -181,6 +181,13 @@ namespace refract
Error += glm::all(glm::equal(C, glm::dvec2(0.0, -1.0), 0.0001)) ? 0 : 1; Error += glm::all(glm::equal(C, glm::dvec2(0.0, -1.0), 0.0001)) ? 0 : 1;
} }
{
glm::vec4 A(0.0f, -1.0f, 0.0f, 0.0f);
glm::vec4 B(0.0f, 1.0f, 0.0f, 0.0f);
glm::vec4 C = glm::refract(A, B, 0.5f);
Error += glm::all(glm::equal(C, glm::vec4(0.0, -1.0, 0.0f, 0.0f), 0.0001f)) ? 0 : 1;
}
return Error; return Error;
} }
}//namespace refract }//namespace refract