diff --git a/readme.txt b/readme.txt
index aabb7f7e..15978bfc 100644
--- a/readme.txt
+++ b/readme.txt
@@ -47,6 +47,7 @@ GLM 0.9.4.4: 2013-05-29
 - Fixed CUDA warnings #86
 - Fixed GLM natvis for VC11 #82
 - Fixed GLM_GTX_multiple with negative values #79
+- Fixed glm::perspective when zNear is zero #71
 
 ================================================================================
 GLM 0.9.4.3: 2013-03-20
diff --git a/test/gtx/gtx_simd_mat4.cpp b/test/gtx/gtx_simd_mat4.cpp
index 3634c8c0..ed9b5af9 100644
--- a/test/gtx/gtx_simd_mat4.cpp
+++ b/test/gtx/gtx_simd_mat4.cpp
@@ -203,21 +203,21 @@ int test_compute_gtx()
 
 	for(std::size_t k = 0; k < Output.size(); ++k)
 	{
-		float i = float(k) / 1000.f;
+		float i = float(k) / 1000.f + 0.001f;
 		glm::vec3 A = glm::normalize(glm::vec3(i));
-		glm::vec3 B = glm::cross(A, glm::vec3(0, 0, 1));
+		glm::vec3 B = glm::cross(A, glm::normalize(glm::vec3(1, 1, 2)));
 		glm::mat4 C = glm::rotate(glm::mat4(1.0f), i, B);
 		glm::mat4 D = glm::scale(C, glm::vec3(0.8f, 1.0f, 1.2f));
 		glm::mat4 E = glm::translate(D, glm::vec3(1.4f, 1.2f, 1.1f));
 		glm::mat4 F = glm::perspective(i, 1.5f, 0.1f, 1000.f);
 		glm::mat4 G = glm::inverse(F * E);
 		glm::vec3 H = glm::unProject(glm::vec3(i), G, F, E[3]);
-		glm::vec3 I = glm::project(H, G, F, E[3]);
-		glm::mat4 J = glm::lookAt(glm::normalize(B), H, I);
+		glm::vec3 I = glm::any(glm::isnan(glm::project(H, G, F, E[3]))) ? glm::vec3(2) : glm::vec3(1);
+		glm::mat4 J = glm::lookAt(glm::normalize(glm::max(B, glm::vec3(0.001f))), H, I);
 		glm::mat4 K = glm::transpose(J);
 		glm::quat L = glm::normalize(glm::quat_cast(K));
 		glm::vec4 M = L * glm::smoothstep(K[3], J[3], glm::vec4(i));
-		glm::mat4 N = glm::mat4(glm::normalize(M), K[3], J[3], glm::vec4(i));
+		glm::mat4 N = glm::mat4(glm::normalize(glm::max(M, glm::vec4(0.001f))), K[3], J[3], glm::vec4(i));
 		glm::mat4 O = N * glm::inverse(N);
 		glm::vec4 P = O * glm::reflect(N[3], glm::vec4(A, 1.0f));
 		glm::vec4 Q = glm::vec4(glm::dot(M, P));