Small bug in gumInternal.c

[ciccan]

Code: Select all

Index: gumInternal.c
===================================================================
--- gumInternal.c       (revision 946)
+++ gumInternal.c       (working copy)
@@ -65,7 +65,7 @@
 
 void gumNormalize(ScePspFVector3* v)
 {
-       float l = (v->x*v->x) + (v->y*v->y) + (v->z*v->z);
+       float l = sqrtf((v->x*v->x) + (v->y*v->y) + (v->z*v->z));
        if (l > GUM_EPSILON)
        {
                float il = 1.0f / l;

[chp]

Oops! That's what I get for writing code on a train without ability to test it. :)

Fixed. Thanks!

[ciccan]

Here's a couple more. I think sceGumLookAt is OK with these changes.

Code: Select all

Index: gumInternal.c
===================================================================
--- gumInternal.c       (revision 963)
+++ gumInternal.c       (working copy)
@@ -58,9 +58,9 @@
 
 void gumCrossProduct(ScePspFVector3*r, const ScePspFVector3* a, const ScePspFVector3* b)
 {
-       r->x = (a->y * b->z) + (a->z * b->y);
-       r->y = (a->z * b->x) + (a->x * b->z);
-       r->z = (a->x * b->y) + (a->y * b->x);
+       r->x = (a->y * b->z) - (a->z * b->y);
+       r->y = (a->z * b->x) - (a->x * b->z);
+       r->z = (a->x * b->y) - (a->y * b->x);
 }
 
 void gumNormalize(ScePspFVector3* v)
Index: sceGumLookAt.c
===================================================================
--- sceGumLookAt.c      (revision 963)
+++ sceGumLookAt.c      (working copy)
@@ -37,8 +37,8 @@
   m.y.z = -forward.y;
   m.z.z = -forward.z;
 
-  sceGumMultMatrix(&m);
-
   ieye.x = -eye->x; ieye.y = -eye->y; ieye.z = -eye->z;
   sceGumTranslate(&ieye);
+
+  sceGumMultMatrix(&m);
 }

[chp]

Thanks! Good thing someone tests this code. I sure didn't before submitting. And see how it turned out. :)

[Jim]

Code: Select all

-       float l = (v->x*v->x) + (v->y*v->y) + (v->z*v->z); 
+       float l = sqrtf((v->x*v->x) + (v->y*v->y) + (v->z*v->z)); 
        if (l > GUM_EPSILON) 
        { 
                float il = 1.0f / l;

I think

Code: Select all

        float l = (v->x*v->x) + (v->y*v->y) + (v->z*v->z); 
-        if (l > GUM_EPSILON) 
+        if (l > GUM_EPSILON*GUM_EPSILON) 
         { 
-                float il = 1.0f / l;
+                float il = 1.0f / sqrtf(l);

would be better.

Don't know how often this fragment is called though.

Great work with psp gl!

Jim

[chp]

Well, the if() is only to avoid creating a NaN from the division, so it should always be true as long as you pass it valid data.

Oh, and this isn't for pspgl... :)

[Jim]

You're right, if GUM_EPSILON is very tiny, then GUM_EPSILON * GUM_EPSILON might underflow anyway.

Jim