Merge with HU Chi branch (Don't use this! It's an incomplete version)

Minor error fixed.

source/tensor/core/reduce/ReduceMean.cpp

@@ -42,7 +42,7 @@ void _ReduceMean(const XTensor * input, XTensor * output, int dim)
     int num = input->dimSize[dim];
     int num = input->dimSize[dim];
 
 
     _ReduceSum(input, output, dim);
     _ReduceSum(input, output, dim);
-    _ScaleAndShiftMe(output, (DTYPE)1/num, 0);
+    _ScaleAndShiftMe(output, 1.0F/(DTYPE)(num), 0);
 }
 }
 
 
 /* 
 /* 

source/tensor/core/reduce/ReduceSum.cpp

@@ -134,7 +134,6 @@ void _ReduceSum(const XTensor * input, XTensor * output, int dim, const XTensor
                         VectorBuffer vecBuf[4];
                         VectorBuffer vecBuf[4];
                         for(int k = 0; k < 4; k++){
                         for(int k = 0; k < 4; k++){
                             vecBuf[k] = VectorBuffer::loadu((DTYPE*)(ip) + (j * 4 + k) * 32 / sizeof(DTYPE), isExp, power, bias + k * 32 / sizeof(DTYPE));
                             vecBuf[k] = VectorBuffer::loadu((DTYPE*)(ip) + (j * 4 + k) * 32 / sizeof(DTYPE), isExp, power, bias + k * 32 / sizeof(DTYPE));
-
                         }
                         }
                         for(int k = 1; k < strideNum; k++){
                         for(int k = 1; k < strideNum; k++){
                             DTYPE * ptr = ip + k * stride + (j * 4) * vecBufLength;
                             DTYPE * ptr = ip + k * stride + (j * 4) * vecBufLength;