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

@@ -48,41 +48,39 @@ CUDA_ztpmqrt( MORSE_enum side, MORSE_enum trans,
         n1 = N;
         ldwork  = IB;
         ldworkc = M;
-        //ws = K * n1;
+        ws = IB * n1;
     }
     else {
         m1 = M;
         n1 = K;
         ldwork  = m1;
         ldworkc = IB;
-        //ws = m1 * K;
+        ws = IB * m1;
     }
-    ws = IB * (side == MorseLeft ? n1 : m1);
 
-    CUDA_ztsmqr( side, trans, m1, n1, M, N, K, IB,
-                 A, LDA, B, LDB, V, LDV, T, LDT,  
-                 WORK, ldwork, WORK + ws, ldworkc,
-                 CUBLAS_STREAM_VALUE );           
     /* TS case */
-    /*if (L == 0) {*/
-    /*    CUDA_ztsmqr( side, trans, m1, n1, M, N, K, IB,*/
-    /*                 A, LDA, B, LDB, V, LDV, T, LDT,  */
-    /*                 WORK, ldwork, WORK + ws, ldworkc,*/
-    /*                 CUBLAS_STREAM_VALUE );           */
-    /*}                                                 */
+    if (L == 0) {
+        CUDA_ztsmqr( side, trans, m1, n1, M, N, K, IB,
+                     A, LDA, B, LDB, V, LDV, T, LDT,
+                     WORK, ldwork, WORK + ws, ldworkc,
+                     CUBLAS_STREAM_VALUE );
+    }
     /* TT case */
-    /*else  if( L == M ) {                              */
-    /*    cudablas_error(-6, "TTMQRT not available on GPU yet\n" );*/
-    /*    return -6;                                               */ 
-        /* 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;                                                */
-    /*    //LAPACKE_ztpmqrt_work( LAPACK_COL_MAJOR, M, N, K, L, IB, V, LDV, T, LDT, A, LDA, B, LDB, WORK );*/
-    /*} */
+    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,
+                     WORK, ldwork, WORK + ws, ldworkc,
+                     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;
 }