Fixing integer divides to make clBLAS work when building with python3

Author: pavanky

src/library/blas/AutoGemm/KernelOpenCL.py

@@ -351,11 +351,11 @@ def makeOpenCLKernelString(kernel):
   kStr += endLine
   kStr += "    /* load global -> local */" + endLine
   numALoads  = (kernel.workGroupNumRows*kernel.microTileNumRows*kernel.unroll) \
-      / (kernel.workGroupNumRows*kernel.workGroupNumCols)
+      // (kernel.workGroupNumRows*kernel.workGroupNumCols) # // -- integer divide
   numALoadsR = (kernel.workGroupNumRows*kernel.microTileNumRows*kernel.unroll) \
       % (kernel.workGroupNumRows*kernel.workGroupNumCols)
   numBLoads  = (kernel.workGroupNumCols*kernel.microTileNumCols*kernel.unroll) \
-      / (kernel.workGroupNumRows*kernel.workGroupNumCols)
+      // (kernel.workGroupNumRows*kernel.workGroupNumCols) # // - integer divide
   numBLoadsR = (kernel.workGroupNumCols*kernel.microTileNumCols*kernel.unroll) \
       % (kernel.workGroupNumRows*kernel.workGroupNumCols)
 

src/library/blas/AutoGemm/KernelParameters.py

@@ -89,11 +89,11 @@ def isValid(self):
     return True
     """
     numALoads = (self.workGroupNumRows*self.microTileNumRows*self.unroll) \
-        / (self.workGroupNumRows*self.workGroupNumCols)
+        // (self.workGroupNumRows*self.workGroupNumCols)
     numALoadsR = (self.workGroupNumRows*self.microTileNumRows*self.unroll) \
         % (self.workGroupNumRows*self.workGroupNumCols)
     numBLoads = (self.workGroupNumCols*self.microTileNumCols*self.unroll) \
-        / (self.workGroupNumRows*self.workGroupNumCols)
+        // (self.workGroupNumRows*self.workGroupNumCols)
     numBLoadsR = (self.workGroupNumCols*self.microTileNumCols*self.unroll) \
         % (self.workGroupNumRows*self.workGroupNumCols)
     if (numALoads>0 and numALoadsR>0):