mercury
mercury copied to clipboard
Mercury is a C library for implementing RPC, optimized for HPC.
Something that could be useful in Mercury for upper libraries like Margo is the possibility to inject user-provided file descriptors in the Mercury polling mechanism. Some use-cases would include injecting...
Examples include hg_test_rpc, hg_test_rpc_lat, and hg_test_write_bw, all of which work correctly until the phase of the test that uses multiple handles or concurrent RPCs. This is with libfabric 1.6 and...
When doing bulk transfers using the shared-memory plugin and process_vm_readv/writev calls, performance diminishes as the size of the data increases.
Comment from Jeff: > One thing that occurred to me that may be more efficient is to do an atomic decrement rather than a CAS when releasing the lock. >...
When transferring large data with the NA BMI plugin, Tang is reporting the following error: ``` # NA -- Error -- /global/homes/w/wzhang5/software/mercury/src/na/na_bmi.c:1773 [E 15:01:01.028255] src/io/bmi/bmi_tcp/bmi-tcp.c line 1313: Error: BMI message...
When SELF_FORWARD is disabled, the address returned by HG_Addr_self cannot be used to forward RPCs. Instead, one has to convert it into a string (HG_Addr_to_string) then do a lookup to...
Consider the use-case of creating, and extending, buffer pools of registered memory (ie hg_bulk_t's). Right now, generating `N` bulk handles of size `S` requires `N` calls to HG_Bulk_create, which, depending...
As the title says, would be good to have a way to tell whether an input bulk buffer (with permission HG_READ_ONLY) contains eagerly transmitted data, and further to get a...
Was thinking about the proc issues talked about in #79 and had the following idea: - have a new type of proc operation - HG_GET_SIZE, which increments two (new) size...
In an earlier meeting, there was some confusion about handle management. I.e. when must HG_Destroy be called? Clearing up this confusion amounts to improving documentation on when handle refcounts are...