Bug fix #2424

With a sun lamp shining exact perpendicular to a face, the result was
undefined (black or full lit). Was caused by using acos() without checking
the input range. Float calculus can result in small under or overflows...
therefore: always use the arithb.c function saacos() !!!

Found more acos() usage in rendercore (in blinn and arealight too), removed
all of them. Might solve the white or black dots people were complaining
about?

1 files changed, 9 insertions, 9 deletions

diff --git a/source/blender/render/intern/source/rendercore.c b/source/blender/render/intern/source/rendercore.c
index e73c1a63413..2c77d29421a 100644
--- a/source/blender/render/intern/source/rendercore.c
+++ b/source/blender/render/intern/source/rendercore.c
@@ -643,10 +643,10 @@ static float area_lamp_energy(float *co, float *vn, LampRen *lar)
 	rad[2]= vec[2][0]*vec[3][0]+ vec[2][1]*vec[3][1]+ vec[2][2]*vec[3][2];
 	rad[3]= vec[3][0]*vec[0][0]+ vec[3][1]*vec[0][1]+ vec[3][2]*vec[0][2];
 
-	rad[0]= acos(rad[0]);
-	rad[1]= acos(rad[1]);
-	rad[2]= acos(rad[2]);
-	rad[3]= acos(rad[3]);
+	rad[0]= saacos(rad[0]);
+	rad[1]= saacos(rad[1]);
+	rad[2]= saacos(rad[2]);
+	rad[3]= saacos(rad[3]);
 
 	/* Stoke formula */
 	VECMUL(cross[0], rad[0]);
@@ -791,7 +791,7 @@ float Blinn_Spec(float *n, float *l, float *v, float refrac, float spec_power )
 
 	p = sqrt( (double)((refrac * refrac)+(vh*vh)-1.0) );
 	f = (((p-vh)*(p-vh))/((p+vh)*(p+vh)))*(1+((((vh*(p+vh))-1.0)*((vh*(p+vh))-1.0))/(((vh*(p-vh))+1.0)*((vh*(p-vh))+1.0))));
-	ang = (float)acos((double)(nh));
+	ang = saacos(nh);
 
 	i= f * g * exp((double)(-(ang*ang) / (2.0*spec_power*spec_power)));
 
@@ -812,7 +812,7 @@ float Toon_Spec( float *n, float *l, float *v, float size, float smooth )
 	
 	rslt = h[0]*n[0] + h[1]*n[1] + h[2]*n[2];
 	
-	ang = acos( rslt ); 
+	ang = saacos( rslt ); 
 	
 	if( ang < size ) rslt = 1.0;
 	else if( ang >= (size + smooth) || smooth == 0.0 ) rslt = 0.0;
@@ -828,7 +828,7 @@ float Toon_Diff( float *n, float *l, float *v, float size, float smooth )
 
 	rslt = n[0]*l[0] + n[1]*l[1] + n[2]*l[2];
 
-	ang = acos( (double)(rslt) );
+	ang = saacos( (double)(rslt) );
 
 	if( ang < size ) rslt = 1.0;
 	else if( ang >= (size + smooth) || smooth == 0.0 ) rslt = 0.0;
@@ -865,8 +865,8 @@ static float OrenNayar_Diff_i(float nl, float *n, float *l, float *v, float roug
 	vh= v[0]*h[0]+v[1]*h[1]+v[2]*h[2]; /* Dot product between view vector and halfway vector */
 	if(vh<=0.0) vh= 0.0;
 	
-	Lit_A = acos(realnl);
-	View_A = acos( nv );
+	Lit_A = saacos(realnl);
+	View_A = saacos( nv );
 	
 	Lit_B[0] = l[0] - (realnl * n[0]);
 	Lit_B[1] = l[1] - (realnl * n[1]);