Stop use gemm_mt globally

Author: Critsium-xy

source/module_hsolver/kernels/math_kernel_op.cpp

@@ -275,10 +275,16 @@ struct gemm_op<T, base_device::DEVICE_CPU>
                     const int& ldb,
                     const T* beta,
                     T* c,
-                    const int& ldc)
+                    const int& ldc,
+                    bool use_dsp)
     {
 #ifdef __DSP
-        BlasConnector::gemm(transb, transa, n, m, k, *alpha, b, ldb, a, lda, *beta, c, ldc, base_device::AbacusDevice_t::DspDevice);
+        if (use_dsp){
+            BlasConnector::gemm(transb, transa, n, m, k, *alpha, b, ldb, a, lda, *beta, c, ldc, base_device::AbacusDevice_t::DspDevice);
+        }
+        else{
+            BlasConnector::gemm(transb, transa, n, m, k, *alpha, b, ldb, a, lda, *beta, c, ldc);
+        }     
 #else
         BlasConnector::gemm(transb, transa, n, m, k, *alpha, b, ldb, a, lda, *beta, c, ldc);
 #endif

source/module_hsolver/kernels/math_kernel_op.h

@@ -261,7 +261,7 @@ template <typename T, typename Device> struct gemm_op {
   void operator()(const Device *d, const char &transa, const char &transb,
                   const int &m, const int &n, const int &k, const T *alpha,
                   const T *a, const int &lda, const T *b, const int &ldb,
-                  const T *beta, T *c, const int &ldc);
+                  const T *beta, T *c, const int &ldc, bool usd_dsp = false);
 };
 
 template <typename T, typename Device> struct matrixTranspose_op {