gs_comm.f90 File Reference

Go to the source code of this file.

Data Types
type	gs_comm::gs_comm_t
	Gather-scatter communication method. More...
interface	gs_comm::gs_comm_init
	Abstract interface for initializing a Gather-scatter communication method. More...
interface	gs_comm::gs_comm_free
	Abstract interface for deallocating a Gather-scatter communication method. More...
interface	gs_comm::gs_nbsend
	Abstract interface for initiating non-blocking send operations Sends the values in u(send_dof(send_pe(i))) to each rank send_pe(i) for all ranks in send_pe. More...
interface	gs_comm::gs_nbrecv
	Abstract interface for initiating non-blocking recieve operations Posts non-blocking recieve of values and puts the values into buffers. More...
interface	gs_comm::gs_nbwait
	Abstract interface for waiting on non-blocking operations Waits and checks that data is in buffers and unpacks buffers into correct location in u u(recv_dof(recv_pe(i))) = gs_op(recieve_buffers(recv_pe) for this dof) More...

Modules
module	gs_comm
	Defines a gather-scatter communication method.

Functions/Subroutines
subroutine	gs_comm::init_dofs (this)
subroutine	gs_comm::free_dofs (this)
subroutine	gs_comm::init_order (this, send_pe, recv_pe)
	Obtains which ranks to send and receive data from.
subroutine	gs_comm::free_order (this)
subroutine	gs_comm::take_schedule (this, src)
	Take over the gather-scatter schedule (dof lists and peer order) of src, avoiding a second (expensive) pass over the connectivity. The data is moved rather than copied, so src is left without a schedule and must not be used for communication afterwards (it can still be freed). No communication resources are set up here; complete the handover with init_schedule once src has been freed, so that the two backends never hold their resources at the same time.
subroutine	gs_comm::init_schedule (this)
	Set up this communication method for the schedule taken over by take_schedule. Collective, as init is.
subroutine	gs_comm::gs_nbsend_vec (this, u, n, nc, tag, deps, strm)
	Default fused vector send. Abort unless a backend overrides it.
subroutine	gs_comm::gs_nbrecv_vec (this, tag, nc)
	Default fused vector receive. Abort unless a backend overrides it.
subroutine	gs_comm::gs_nbwait_vec (this, u, n, nc, op, strm)
	Default fused vector wait/reduce. Abort unless a backend overrides it.

Variables
integer, parameter, public	gs_comm::gs_comm_mpi = 1
integer, parameter, public	gs_comm::gs_comm_mpigpu = 2
integer, parameter, public	gs_comm::gs_comm_nccl = 3
integer, parameter, public	gs_comm::gs_comm_nvshmem = 4
integer, parameter, public	gs_comm::gs_comm_openshmem = 5
integer, parameter, public	gs_comm::gs_comm_caf = 6
integer, parameter, public	gs_comm::gs_comm_neighbour = 7
integer, parameter, public	gs_comm::gs_comm_utofu = 8
integer, parameter, public	gs_comm::gs_comm_mpirma = 9
integer, parameter, public	gs_comm::gs_comm_crystal = 10
integer, parameter, public	gs_comm::gs_comm_crystalgpu = 11
integer, parameter, public	gs_comm::gs_vec_nc = 3
	Maximum number of components handled by the fused vector (multi-component) halo exchange used by gs_op_r3. Sizes the backend vector buffers.