| 1 | |
|---|
| 2 | <br><font size=2 face="sans-serif">Hi guys, just picking off some failures |
|---|
| 3 | in the IntelANL MPI test suite (2006) against MPICH2 1.1. It appears that |
|---|
| 4 | all the failures are MPI 2.1 vs MPI 2.0 changes. </font> |
|---|
| 5 | <br> |
|---|
| 6 | <br><font size=2 face="sans-serif">However (and this might be a forum question, |
|---|
| 7 | but I figured I'd ask here anyway), it appears there might either be an |
|---|
| 8 | ambiguity in MPI_Cart_map and MPI_Graph_create, or the MPICH2 code is "ahead" |
|---|
| 9 | of the standard as I read it.</font> |
|---|
| 10 | <br> |
|---|
| 11 | <br><font size=2 face="sans-serif">Basically, this test:</font> |
|---|
| 12 | <br> |
|---|
| 13 | <br><font size=2 face="sans-serif">num_dims = 0;</font> |
|---|
| 14 | <br><font size=2 face="sans-serif">MPI_Cart_create(MPI_COMM_WORLD, num_dims, |
|---|
| 15 | dim_size, periods, reorder, &comm_cart);</font> |
|---|
| 16 | <br> |
|---|
| 17 | <br><font size=2 face="sans-serif">"fails" since it is expecting |
|---|
| 18 | the MPI 2.0 behavior where num_dims = 0 was un(under?)defined. In MPI 2.1, |
|---|
| 19 | we updated the description with "If ndims is zero, then a zero-dimensional |
|---|
| 20 | Cartesian topology is created", so I'm happy marking this as an invalid |
|---|
| 21 | test.</font> |
|---|
| 22 | <br> |
|---|
| 23 | <br><font size=2 face="sans-serif">However, this test:</font> |
|---|
| 24 | <br><font size=2 face="sans-serif">num_dims = 0;</font> |
|---|
| 25 | <br><font size=2 face="sans-serif">MPI_Cart_map(comm_cart, num_dims, dim_size, |
|---|
| 26 | periods, &newrank);</font> |
|---|
| 27 | <br> |
|---|
| 28 | <br><font size=2 face="sans-serif">also "fails" since it doesn't |
|---|
| 29 | report an error. I don't see any comments in the 2.1 standard or 2.2 issues |
|---|
| 30 | that suggest MPI_Cart_map() changed in the same way that MPI_Cart_create() |
|---|
| 31 | did, but it seems logical. So, my question is, is this correct behavior |
|---|
| 32 | for MPI_Cart_map() and if so, should we add a comment in the 2.2 standard |
|---|
| 33 | (or errata since I believe new tickets are closed?) that says num_dims=0 |
|---|
| 34 | is a valid argument?</font> |
|---|
| 35 | <br> |
|---|
| 36 | <br><font size=2 face="sans-serif">Finally, the third test that "fails" |
|---|
| 37 | is doing this:</font> |
|---|
| 38 | <br> |
|---|
| 39 | <br><font size=2 face="sans-serif">edges[0] = 0; edges[1] = 3; edges[2] |
|---|
| 40 | = 0; edges[3] = 3; edges[4] = 0; edges[5] = 2;</font> |
|---|
| 41 | <br><font size=2 face="sans-serif">MPI_Graph_create(MPI_COMM_WORLD, nnodes, |
|---|
| 42 | index, edges, reorder, &comm_graph)</font> |
|---|
| 43 | <br> |
|---|
| 44 | <br><font size=2 face="sans-serif">Again, I don't see any comments on null |
|---|
| 45 | edges for MPI_Graph_create() in 2.1 or 2.2 tickets. So is this correct |
|---|
| 46 | behavior?</font> |
|---|
| 47 | <br> |
|---|
| 48 | <br><font size=2 face="sans-serif">Thanks.</font> |
|---|
| 49 | <br> |
|---|
| 50 | <br><font size=2 face="sans-serif"><br> |
|---|
| 51 | Brian Smith<br> |
|---|
| 52 | BlueGene MPI Development<br> |
|---|
| 53 | IBM Rochester<br> |
|---|
| 54 | Phone: 507 253 4717<br> |
|---|
| 55 | smithbr@us.ibm.com<br> |
|---|
| 56 | </font> |
|---|
