Fix for the TPMQRT routine

Fixes on zgeqrf_qdwh: fix maxmt computing in pztpqrt.c and pztpgqrt. Fixes on cuda_ztpmqrt.c: call CUDA_ztsmqr for TS and TT case for now (TODO: once we have CUDA_zttmqr, we will use it for TT case) : change workspace shift ws

cudablas/compute/cuda_ztpmqrt.c

@@ -47,7+47,7 @@
         m1 = K;
         n1 = N;
         ldwork  = IB;
         ldworkc = M;
         ws = IB * n1;
     }
     else {
@@ -61,7+61,7 @@
     /* TS case */
     if (L == 0) {
         CUDA_ztsmqr( side, trans, m1, n1, M, N, K, IB,
-                     A, LDA, B, LDB, V, LDV, T, LDT,  
+                     A, LDA, B, LDB, V, LDV, T, LDT,
                      WORK, ldwork, WORK + ws, ldworkc,
-                     CUBLAS_STREAM_VALUE );           
-    }                                                 
+                     CUBLAS_STREAM_VALUE );      
+    }
     /* TT case */
-    else  if( L == M ) {                              
+    else  if( L == M ) {
+        // CUDA_zttmqr( side, trans, m1, n1, M, N, K, IB,
+        //              A, LDA, B, LDB, V, LDV, T, LDT,
+        //              WORK, ldwork );
         CUDA_ztsmqr( side, trans, m1, n1, M, N, K, IB,
-                     A, LDA, B, LDB, V, LDV, T, LDT,  
+                     A, LDA, B, LDB, V, LDV, T, LDT,
                      WORK, ldwork, WORK + ws, ldworkc,
-                     CUBLAS_STREAM_VALUE );           
-        morse_warning("The lower triangular part must be set to 0 for to be correct");
-        // CUDA_zttmqr( side, trans, m1, n1, M, N, K, IB, 
-        //              A, LDA, B, LDB, V, LDV, T, LDT, 
-        //              WORK, ldwork ); 
-    }                                 
-    else {                            
-        cudablas_error(-6, "TPMQRT not available on GPU yet\n" );
-        return -6;                                                
+                     CUBLAS_STREAM_VALUE );
+        morse_warning("TPMQRT is not available on GPU yet: The lower triangular part must be set to 0 for to be correct");
+    }
+    else {
         //LAPACKE_ztpmqrt_work( LAPACK_COL_MAJOR, M, N, K, L, IB, V, LDV, T, LDT, A, LDA, B, LDB, WORK );
-    } 
+        cudablas_error(-6, "TPMQRT not available on GPU yet\n" );
+        return -6;
+    }
 
     return MORSE_SUCCESS;
 }