Merge pull request #1542 from martin-frbg/quickdiv64

Avoid out-of-bounds accesses in blas_quickdivide on big X86 systems

Author: martin-frbg

common_x86.h

@@ -178,7 +178,13 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
   result = x/y;
   return result;
 #else
-
+#if (MAX_CPU_NUMBER > 64)
+  if ( y > 64) {
+	  result = x/y;
+	  return result;
+  }
+#endif
+	
   y = blas_quick_divide_table[y];
 
   __asm__ __volatile__  ("mull %0" :"=d" (result) :"a"(x), "0" (y));

common_x86_64.h

@@ -196,6 +196,13 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 
   if (y <= 1) return x;
 
+#if (MAX_CPU_NUMBER > 64)  
+  if (y > 64) { 
+	  result = x / y;
+	  return result;
+  }
+#endif
+	
   y = blas_quick_divide_table[y];
 
   __asm__ __volatile__  ("mull %0" :"=d" (result) :"a"(x), "0" (y));