Ken Raffenetti
Ken Raffenetti
Currently, there is no way to exclude cvars from ch3 which building ch4 and vice-versa. A nice features for users would be to only expose the variables which are supported...
## Pull Request Description Use flags to control the completion behavior of send ops. First try immediate completion. If that fails, allocate a request and ensure it is passed to...
A user suggested this for heterogeneous systems with CPU and GPU partitions. Some processes may be restricted from seeing GPU devices, in which case we should gracefully disable the GPU...
https://github.com/pmodels/mpich/blob/9b6343287be3b93a40b68b33e01eccb7a99ed289/src/mpid/ch4/netmod/ofi/ofi_iovec_util.h#L387-L388 `fi_rma_iov` uses `uint64_t` for addr field as opposed to `void *` in `struct iovec`, which causes a mismatch when `sizeof(void *) < 8`. https://github.com/ofiwg/libfabric/blob/master/include/rdma/fi_rma.h#L43-L47
These specific failures were evinced during the addition of GPU support to the testsuite. There is a segfault during buffer initialization inside DTPools. 32-bit CentOS, ch3/tcp configuration. ``` ./pt2pt/sendrecv1 2...
The PMIx put/get usage in MPICH converts values to string representation despite PMIx supporting binary blob data. It should be OK to remove this extra step without any loss of...
## Pull Request Description If there are still messages using a communicator context_id in the unexpected queue at destruction time, they could disturb message matching if the context_id is reused....
## Pull Request Description If a blocking send, ssend, or rsend operation returns a request, it should be visible in the message queues until it completes. Fixes pmodels/mpich#6131 ## Author...
## Pull Request Description C++20 removed support for == and != on complex numbers. Instead, compare the individual parts. Fixes compilation with Intel oneAPI C++ (icpx). ## Author Checklist *...
## Pull Request Description There is a bug in older versions that will cause this path to segfault with GPU buffers. Avoid it unless we are using a known good...