diff --git a/build_linux64/libamsculib2.linux64.a b/build_linux64/libamsculib2.linux64.a
index 00cf916..62524fa 100644
Binary files a/build_linux64/libamsculib2.linux64.a and b/build_linux64/libamsculib2.linux64.a differ
diff --git a/build_linux64/objstore/cuvect3f.o b/build_linux64/objstore/cuvect3f.o
index 04282a1..4f2708b 100644
Binary files a/build_linux64/objstore/cuvect3f.o and b/build_linux64/objstore/cuvect3f.o differ
diff --git a/build_linux64/test b/build_linux64/test
index 5cab227..6f13865 100644
Binary files a/build_linux64/test and b/build_linux64/test differ
diff --git a/src/amsculib2/cuvect3f.cu b/src/amsculib2/cuvect3f.cu
index 26996bb..3706b0b 100644
--- a/src/amsculib2/cuvect3f.cu
+++ b/src/amsculib2/cuvect3f.cu
@@ -348,15 +348,16 @@ __host__ __device__ cumat3f cumat3f::operator*(const float &rhs)
 __host__ __device__ cumat3f cumat3f::operator/(const float &rhs)
 {
     cumat3f ret;
-    ret.m00 = m00 / rhs;
-    ret.m10 = m10 / rhs;
-    ret.m20 = m20 / rhs;
-    ret.m01 = m01 / rhs;
-    ret.m11 = m11 / rhs;
-    ret.m21 = m21 / rhs;
-    ret.m02 = m02 / rhs;
-    ret.m12 = m12 / rhs;
-    ret.m22 = m22 / rhs;
+    float irhs = 1.0f/rhs;
+    ret.m00 = m00 * irhs;
+    ret.m10 = m10 * irhs;
+    ret.m20 = m20 * irhs;
+    ret.m01 = m01 * irhs;
+    ret.m11 = m11 * irhs;
+    ret.m21 = m21 * irhs;
+    ret.m02 = m02 * irhs;
+    ret.m12 = m12 * irhs;
+    ret.m22 = m22 * irhs;
     return ret;
 }
         
@@ -423,17 +424,19 @@ __host__ __device__ cumat3f cumat3f::inverse()
 {
     cumat3f q;
     float dt = det();
+    float idt;
     if(dt!=0.0)
     {
-        q(0,0) =  (at(1,1)*at(2,2)-at(1,2)*at(2,1))/dt;
-        q(1,0) = -(at(1,0)*at(2,2)-at(1,2)*at(2,0))/dt;
-        q(2,0) =  (at(1,0)*at(2,1)-at(1,1)*at(2,0))/dt;
-        q(0,1) = -(at(0,1)*at(2,2)-at(0,2)*at(2,1))/dt;
-        q(1,1) =  (at(0,0)*at(2,2)-at(0,2)*at(2,0))/dt;
-        q(2,1) = -(at(0,0)*at(2,1)-at(0,1)*at(2,0))/dt;
-        q(0,2) =  (at(0,1)*at(1,2)-at(0,2)*at(1,1))/dt;
-        q(1,2) = -(at(0,0)*at(1,2)-at(0,2)*at(1,0))/dt;
-        q(2,2) =  (at(0,0)*at(1,1)-at(0,1)*at(1,0))/dt;
+        idt = 1.0f/dt;
+        q(0,0) =  (at(1,1)*at(2,2)-at(1,2)*at(2,1))*idt;
+        q(1,0) = -(at(1,0)*at(2,2)-at(1,2)*at(2,0))*idt;
+        q(2,0) =  (at(1,0)*at(2,1)-at(1,1)*at(2,0))*idt;
+        q(0,1) = -(at(0,1)*at(2,2)-at(0,2)*at(2,1))*idt;
+        q(1,1) =  (at(0,0)*at(2,2)-at(0,2)*at(2,0))*idt;
+        q(2,1) = -(at(0,0)*at(2,1)-at(0,1)*at(2,0))*idt;
+        q(0,2) =  (at(0,1)*at(1,2)-at(0,2)*at(1,1))*idt;
+        q(1,2) = -(at(0,0)*at(1,2)-at(0,2)*at(1,0))*idt;
+        q(2,2) =  (at(0,0)*at(1,1)-at(0,1)*at(1,0))*idt;
         // q(0,0) =  (at(1,1)*at(2,2)-at(1,2)*at(2,1))/dt;
         // q(0,1) = -(at(1,0)*at(2,2)-at(1,2)*at(2,0))/dt;
         // q(0,2) =  (at(1,0)*at(2,1)-at(1,1)*at(2,0))/dt;
@@ -511,15 +514,16 @@ __host__ __device__ cumat3f& cumat3f::operator-=(const cumat3f &rhs)
 
 __host__ __device__ cumat3f& cumat3f::operator/=(const float &rhs)
 {
-    m00 /= rhs;
-    m10 /= rhs;
-    m20 /= rhs;
-    m01 /= rhs;
-    m11 /= rhs;
-    m21 /= rhs;
-    m02 /= rhs;
-    m12 /= rhs;
-    m22 /= rhs;
+    float irhs = 1.0f/rhs;
+    m00 *= irhs;
+    m10 *= irhs;
+    m20 *= irhs;
+    m01 *= irhs;
+    m11 *= irhs;
+    m21 *= irhs;
+    m02 *= irhs;
+    m12 *= irhs;
+    m22 *= irhs;
 
     return *this;
 }