Fix alignment assumptions in DDS loading code. Fixes issue 132.

Merge misc changes from the-witness.

src/nvimage/DirectDrawSurface.cpp

@@ -1404,11 +1404,7 @@ uint DirectDrawSurface::mipmapSize(uint mipmap) const
     {
         nvDebugCheck((header.pf.flags & DDPF_RGB) || (header.pf.flags & DDPF_LUMINANCE));
 
-        // Align pixels to bytes.
-        uint byteCount = (header.pf.bitcount + 7) / 8;
-
-        // Align pitch to 4 bytes.
-        uint pitch = 4 * ((w * byteCount + 3) / 4);
+        uint pitch = computePitch(w, header.pf.bitcount, 8); // Asuming 8 bit alignment, which is the same D3DX expects.
 
         return pitch * h * d;
     }

src/nvimage/FloatImage.cpp

@@ -258,7 +258,7 @@ void FloatImage::exponentiate(uint base_component, uint num, float power)
         float * ptr = this->channel(base_component + c);
 
         for(uint i = 0; i < size; i++) {
-            ptr[i] = pow(ptr[i], power);
+            ptr[i] = powf(max(0.0f, ptr[i]), power);
         }
     }
 }

src/nvimage/nvimage.h

@@ -20,4 +20,14 @@
 #define NVIMAGE_CLASS
 #endif
 
+
+// Some utility functions:
+
+inline uint computePitch(uint w, uint bitsize, uint alignment)
+{
+	return ((w * bitsize +  8 * alignment - 1) / (8 * alignment)) * alignment;
+}
+
+
+
 #endif // NV_IMAGE_H