James Osborn

Feature/sycl

1

not a real PR, just for illustrating changes currently in SYCL port

This is an initial set of changes to start tagging kernel operations that require special resources (e.g. shared memory) or other treatment (e.g. block synchronization). This initial version only enables...

Wrong data with MPI send/recv and pipelining on Intel GPUs

2

We're getting incorrect results in application code when using MPIR_CVAR_CH4_OFI_ENABLE_GPU_PIPELINE=1 if the buffer size MPIR_CVAR_CH4_OFI_GPU_PIPELINE_BUFFER_SZ isn't set large enough. Setting it larger seems to work, but MPI should still give...

Hang with MPI send/recv and pipelining on Intel GPUs

8

An application is hanging on Aurora with pipelining on. The hang is dependent on the relative message sizes being sent between nodes and within a node. This issue was mentioned...

The test suite sets `scale` for the staggered gauge field parameters (https://github.com/lattice/quda/blob/develop/tests/utils/set_params.cpp#L100) which then gets set for both the fat and long link gauge fields. These are initially floating point...

The issue reported in #1535 seems to be back in develop. It appears in the `invert_test_mobius_sym` and `invert_test_mobius_asym` tests. They exit with something like: ``` [ RUN ] SchwarzNormal/InvertTest.verify/double_double_pcg_mat_pc_dag_mat_pc_normop_pc_additive_schwarz_cg_half_l2 Computed...