improve the code of _CudaBLASMatrixMULList

source/tensor/core/arithmetic/XTensorBLAS.cu

@@ -253,14 +253,14 @@ void _CudaBLASMatrixMULList(cublasHandle_t * handle,
 
 
     if (isUniform) {
     if (isUniform) {
         XMem * mem = a0->mem;
         XMem * mem = a0->mem;
-        if (isStrided && a->count > 1) {
+        if (isStrided) {
             _CudaBLASMatrixMULBatchedStrided(handle,
             _CudaBLASMatrixMULBatchedStrided(handle,
-                                            a0->data, transposedA, a0->dataType, strideA / a0->unitSize,
-                                            b0->data, transposedB, b0->dataType, strideB / b0->unitSize,
-                                            c0->data, c0->dataType, strideC / c0->unitSize, a->count,
-                                            a0->dimSize[0], a0->dimSize[1],
-                                            b0->dimSize[0], b0->dimSize[1],
-                                            c0->dimSize[0], c0->dimSize[1], alpha, beta);
+                                             a0->data, transposedA, a0->dataType, strideA / a0->unitSize,
+                                             b0->data, transposedB, b0->dataType, strideB / b0->unitSize,
+                                             c0->data, c0->dataType, strideC / c0->unitSize, a->count,
+                                             a0->dimSize[0], a0->dimSize[1],
+                                             b0->dimSize[0], b0->dimSize[1],
+                                             c0->dimSize[0], c0->dimSize[1], alpha, beta);
         }
         }
         else {
         else {
             DTYPE ** ap = new DTYPE*[a->count];
             DTYPE ** ap = new DTYPE*[a->count];
@@ -324,12 +324,12 @@ void _CudaBLASMatrixMULList(cublasHandle_t * handle,
             XTensor * ci = (XTensor*)c->GetItem(i);
             XTensor * ci = (XTensor*)c->GetItem(i);
 
 
             _CudaBLASMatrixMUL(handle,
             _CudaBLASMatrixMUL(handle,
-                              ai->data, transposedA, ai->dataType,
-                              bi->data, transposedB, bi->dataType,
-                              ci->data, ci->dataType,
-                              ai->dimSize[0], ai->dimSize[1],
-                              bi->dimSize[0], bi->dimSize[1],
-                              ci->dimSize[0], ci->dimSize[1], alpha, beta);
+                               ai->data, transposedA, ai->dataType,
+                               bi->data, transposedB, bi->dataType,
+                               ci->data, ci->dataType,
+                               ai->dimSize[0], ai->dimSize[1],
+                               bi->dimSize[0], bi->dimSize[1],
+                               ci->dimSize[0], ci->dimSize[1], alpha, beta);
         }
         }
     }
     }
 }
 }