-
When we zero out the shadow for the shmem, we currently do it on each lane,
we should split the work across multiple threads.
-
I see that the `fair::mq::shmem::Manager::SendHeartbeats` method raises a number of exceptions, which are apparently caught and ignored. Is this normal?
```
frame #9: 0x000000010ec68017 libFairM…
-
The shmem notifier code calls shm_open on a predictable filename. On glibc, at least, shm_open has countermeasures to the obvious symlink attacks, but it's still a bit dubious. If nothing else, anot…
-
This is a follow-on work from #4548.
Shared memory usage in array reductions is currently only checked in `Cuda` backend.
Since it has been noted that shmem limit differs per backend according to #4…
ldh4 updated
5 months ago
-
### 🐛 Describe the bug
```RuntimeError: shmem_size
-
While working on #1114:
- https://github.com/DaemonEngine/Daemon/pull/1114
I noticed that message size of synchrnous IPC based trap calls have huge impact on performance.
I wonder if it would…
-
The documentation for the function `shmem_wait` seems inconsistent with the function signatures as it describes parameters `target`, `pe`, and `value` that are actually not part of the interface. The …
-
### Issue
shmem_ptr is currently not supported with OFI: enable supporting PE's that use the same node
(currently supported by Portals4 transport)
### Solution Ideas
1) look into extending m…
-
While doing some profiling I just noticed that the fair::mq::shmem::Message has a fQueued boolean in the middle of it, introducing some unneeded extra padding. If I naively move it to the MetaHeader c…
-
Hi namhyung & honggyu
I'm using for performance optimizing for some app
```
This app start from systemd as below
-/usr/bin/uftrace -d /tmp/uftrace.data --no-libcall record "my app path"
```
When…