Making parcel coalescing functional #1942
Conversation
- add parcelhandler::put_parcels - fixed parcelport_impl::put_parcels not to send parcels separately - add test for put_parcels - flyby change: improved argument handling in coalescing_message_handler
- fixing parcelport::put_parcels API - adding test
|
@justwagle, @quantomb: please try this branch in order to check whether parcel coalescing has any effect on your application |
- adding a new API for discovering performance counters used in the tests
| { | ||
| HPX_THROW_EXCEPTION(bad_parameter, | ||
| "parcelhandler::put_parcels", | ||
| "mismatched number of parcels and handlers"); |
There was a problem hiding this comment.
There's a typo here. It should probably read: "mismatched destinations" or something along the lines.
There was a problem hiding this comment.
Right, thanks for spotting.
|
@sithhell the review comments have been addressed |
- clean up API - adding additional error checks - avoid calling into parcel-send operation if not necessary - fixing error message
| @@ -184,6 +191,49 @@ namespace hpx { namespace parcelset | |||
|
|
|||
| } | |||
|
|
|||
| protected: | |||
| void put_parcel_await( | |||
| locality const & dest, parcel p, write_handler_type f, bool trigger | |||
There was a problem hiding this comment.
The boolean trigger argument seems to be redundant now. it is always true.
There was a problem hiding this comment.
I noticed that, in hindsight it was always true only. I left it as I thought you were planning for it to be available for any of the upcoming parcel port implementations. But I'm fine to remove it.
There was a problem hiding this comment.
At one point, the put_parcels set it to false initially. but that's not needed anymore. It can certainly be removed.
|
@justwagle, @quantomb: have you tried this? does it work for you? |
Making parcel coalescing functional
This fixes #1941