SUMMA GEMM

Add BLAS toutines with SUMMA algorithm to reduce the communications.

timing/time_zhemm_tile.c

@@ -33,6 +33,7 @@ RunTest(int *iparam, double *dparam, chameleon_time_t *t_)
     CHAMELEON_Complex64_t alpha, beta;
     PASTE_CODE_IPARAM_LOCALS( iparam );
 
+    LDA = chameleon_max(M, iparam[IPARAM_LDA]);
     LDB = chameleon_max(M, iparam[IPARAM_LDB]);
     LDC = chameleon_max(M, iparam[IPARAM_LDC]);