Merge optimizations from squish.

src/nvtt/squish/weightedclusterfit.cpp

@@ -181,10 +181,10 @@ void WeightedClusterFit::Compress3( void* block )
 			a = Truncate( MultiplyAdd( grid, a, half ) ) * gridrcp;
 			b = Truncate( MultiplyAdd( grid, b, half ) ) * gridrcp;
 			
-			// compute the error (we skip the constant xxsum)
-			Vec4 e1 = MultiplyAdd( a*a, alpha2_sum, b*b*beta2_sum );
-			Vec4 e2 = NegativeMultiplySubtract( a, alphax_sum, a*b*alphabeta_sum );
-			Vec4 e3 = NegativeMultiplySubtract( b, betax_sum, e2 );
+			// compute the error (we skip the constant xxsum)
+			Vec4 e1 = MultiplyAdd( a*a, alpha2_sum, b*b*beta2_sum );
+			Vec4 e2 = NegativeMultiplySubtract( a, alphax_sum, a*b*alphabeta_sum );
+			Vec4 e3 = NegativeMultiplySubtract( b, betax_sum, e2 );
 			Vec4 e4 = MultiplyAdd( two, e3, e1 );
 
 			// apply the metric to the error term
@@ -299,10 +299,10 @@ void WeightedClusterFit::Compress4( void* block )
 				a = Truncate( MultiplyAdd( grid, a, half ) ) * gridrcp;
 				b = Truncate( MultiplyAdd( grid, b, half ) ) * gridrcp;
 				
-				// compute the error (we skip the constant xxsum)
-				Vec4 e1 = MultiplyAdd( a*a, alpha2_sum, b*b*beta2_sum );
-				Vec4 e2 = NegativeMultiplySubtract( a, alphax_sum, a*b*alphabeta_sum );
-				Vec4 e3 = NegativeMultiplySubtract( b, betax_sum, e2 );
+				// compute the error (we skip the constant xxsum)
+				Vec4 e1 = MultiplyAdd( a*a, alpha2_sum, b*b*beta2_sum );
+				Vec4 e2 = NegativeMultiplySubtract( a, alphax_sum, a*b*alphabeta_sum );
+				Vec4 e3 = NegativeMultiplySubtract( b, betax_sum, e2 );
 				Vec4 e4 = MultiplyAdd( two, e3, e1 );
 
 				// apply the metric to the error term