Badrish Chandramouli
Badrish Chandramouli
What perf do you see with the fixed code?
Closing as the linked PR has merged.
Better approach is being worked on. You really should help us with a stand alone repro if possible, not just code snippets. Program.cs that we can just dotnet run. Otherwise...
This may still be needed, as the MetadataSize check is different from Length check. A SpanByte may have an 8 byte optional header used for storing logical record expiration time...
Ted's large PR at https://github.com/microsoft/garnet/pull/1186 might clash with this one. We were hoping to get that in without too many more clashes. Any chance this could be revised to work...
Thanks @PaulusParssinen , would also appreciate if you can review that PR https://github.com/microsoft/garnet/pull/1186 (fine to review the draft, as the pending changes right now are focused on larger-than-memory modifications) --...
We _may_ (dont recall 100% at the moment) be simply letting the GC handle the clean up of pages, for the reason that it was overhead to track and ensure...
> the code currently is seemingly relying on the pool to allocate much more behind during alignment than what it expects per the "contract" @PaulusParssinen - can you show an...
Rough prototype of a fix is here - https://github.com/microsoft/garnet/compare/obj-store-wrongtype?expand=1 but this needs to be refined and expanded to cover the surface.
which command did you run? see the list of commands currently supported here: https://microsoft.github.io/garnet/docs/commands/api-compatibility