- Mar 10, 2022
-
-
PAUL M CARPENTER authored
-
PAUL M CARPENTER authored
There was a node(2) clause, now changed to node(0)
-
PAUL M CARPENTER authored
This test was fixed by d57d49191 Disable disableReadPropagationToNext The bug previously existed since at least f3a0bc0140ae0958ce3769af44a32d1c3aab4780 Fix issues with WriteID
-
PAUL M CARPENTER authored
This tests ea7e7c711 Allow tasks to do lmallocs.
-
- Mar 09, 2022
-
-
PAUL M CARPENTER authored
-
PAUL M CARPENTER authored
Makes the tests run very slightly faster
-
PAUL M CARPENTER authored
-
- Oct 04, 2021
-
-
PAUL M CARPENTER authored
This tests "Fix setting of location and WriteID after taskwait".
-
- Sep 19, 2021
-
-
PAUL M CARPENTER authored
This tests "Accesses cannot be merged if Write IDs are different.
-
- Sep 02, 2021
-
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
-
- Aug 25, 2021
-
-
PAUL M CARPENTER authored
Fix the pragma annotations in manual_simple, which were: out(arr1[0;SIZE]) for initialization of arr1 - CORRECT out(arr1[0;SIZE]) for initialization of arr2 - WRONG out(arr2[0;SIZE]) for initialization of sum - WRONG Naively you might think that even with wrong pragmas it should work, because the wrong pragmas still mean that all arrays must have been initialized before any of the strong tasks can execute. Also (1) the location of arr2 will be correct (node 0, since all initialization tasks execute on node 0) and (2) the location of sum will be wrong, but it doesn't matter because memory starts off as zeros on all the nodes anyway. The problem is caused by eager weak fetch. If satisfiability of arr2 (actually meaning that sum has been initialized) reaches a weak task before that weak task has been scheduled and before arr2 is initialized, then the weak task will fetch zeros for arr2. It will then set the location to the current node and pass that to its descendants and eventually to the strong tasks. This is what was happening, some of the calculated values were 3+0 = 3 rather than 3+4 = 7.
-
- Aug 23, 2021
-
-
PAUL M CARPENTER authored
This test needs allmemory support, so it will fail without the correct Nanos6 configuration variables, which are only set for the allmemory tests. Note: do not squash this commit as the relevant commit in Nanos6 refers to the test by its old name, nasty/test3_fail17.
-
PAUL M CARPENTER authored
This test case is fixed by "Fix occasional hang caused by passing namespace information". A similar hang was happening in nasty/test3_stress but this new test case is a bit simpler (it is still over 3600 lines long). Note: you need to run it several times to exhibit the hang.
-
PAUL M CARPENTER authored
This tests "fixup! Fix unregistration of local regions with all memory". It is the example described in the comment in unregisterLocalAccess before setting the location to the directory. Tasks func18 and func20 correspond to tasks A and B in the comment, respectively.
-
PAUL M CARPENTER authored
This tests "Fix race condition when constructing an offloaded task"
-
PAUL M CARPENTER authored
Several tests checking basic allmemory support, enabled by default, as they should all work. Note: the allmemory tests have special configuration variables, so always run them with ctest not mpirun directly.
-
PAUL M CARPENTER authored
Several basic tests of commutative and concurrent tasks, enabled by default, since they should now work. commutative7 tests "Fix propagation of concurrent and commutative satisfiability".
-
PAUL M CARPENTER authored
Taskfor support in Nanos6 is now quite stable and should always be tested.
-
- Jul 07, 2021
-
-
PAUL M CARPENTER authored
-
- Jun 25, 2021
-
-
PAUL M CARPENTER authored
These tests check "Fix dependencies of variables on the stack".
-
PAUL M CARPENTER authored
This tests "Also call releasePendingAccesses in disposeTask". Without this change, this test hangs roughly one in five times running on two physical MN4 nodes.
-
- Jun 24, 2021
-
-
PAUL M CARPENTER authored
Test for "Do not namespace propagate write to pseudowrite"
-
PAUL M CARPENTER authored
This tests "Fix "Offloaded in accesses use pseudowrite""
-
PAUL M CARPENTER authored
This checks "disableReadPropagationToNext only if not weak and not offloaded"
-
PAUL M CARPENTER authored
-
PAUL M CARPENTER authored
-
- Jun 21, 2021
-
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
Adds the .xml and trace.sh files to get extrae traces from the tests when desired. There is not any command to force cmake to generate extrae traces.
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
Include unistd.
-
Jimmy Aguilar Mena authored
-
Jimmy Aguilar Mena authored
-
- Jun 17, 2021
-
-
PAUL M CARPENTER authored
-