Device abstraction, common interface for various accelerators. More...

Data Types
interface	device_memcpy
	Copy data between host and device (or device and device) More...

interface	device_map
	Map a Fortran array to a device (allocate and associate) More...

interface	device_associate
	Associate a Fortran array to a (allocated) device pointer. More...

interface	device_associated
	Check if a Fortran array is assoicated with a device pointer. More...

interface	device_deassociate
	Deassociate a Fortran array from a device pointer. More...

interface	device_get_ptr
	Return the device pointer for an associated Fortran array. More...

interface	device_sync
	Synchronize a device or stream. More...

Functions/Subroutines
subroutine, public	device_init

subroutine, public	device_finalize

subroutine, public	device_name (name)

subroutine, public	device_alloc (x_d, s)
	Allocate memory on the device. More...

subroutine, public	device_free (x_d)
	Deallocate memory on the device. More...

subroutine	device_memcpy_r1 (x, x_d, n, dir, sync, strm)
	Copy data between host and device (rank 1 arrays) More...

subroutine	device_memcpy_r2 (x, x_d, n, dir, sync, strm)
	Copy data between host and device (rank 2 arrays) More...

subroutine	device_memcpy_r3 (x, x_d, n, dir, sync, strm)
	Copy data between host and device (rank 3 arrays) More...

subroutine	device_memcpy_r4 (x, x_d, n, dir, sync, strm)
	Copy data between host and device (rank 4 arrays) More...

subroutine	device_memcpy_cptr (dst, src, s, dir, sync, strm)
	Copy data between host and device (or device and device) (c-pointers) More...

subroutine, private	device_memcpy_common (ptr_h, x_d, s, dir, sync_device, stream)
	Copy data between host and device. More...

subroutine	device_associate_r1 (x, x_d)
	Associate a Fortran rank 1 array to a (allocated) device pointer. More...

subroutine	device_associate_r2 (x, x_d)
	Associate a Fortran rank 2 array to a (allocated) device pointer. More...

subroutine	device_associate_r3 (x, x_d)
	Associate a Fortran rank 3 array to a (allocated) device pointer. More...

subroutine	device_associate_r4 (x, x_d)
	Associate a Fortran rank 4 array to a (allocated) device pointer. More...

subroutine	device_deassociate_r1 (x)
	Deassociate a Fortran rank 1 array from a device pointer. More...

subroutine	device_deassociate_r2 (x)
	Deassociate a Fortran rank 2 array from a device pointer. More...

subroutine	device_deassociate_r3 (x)
	Deassociate a Fortran rank 3 array from a device pointer. More...

subroutine	device_deassociate_r4 (x)
	Deassociate a Fortran rank 4 array from a device pointer. More...

subroutine	device_map_r1 (x, x_d, n)
	Map a Fortran rank 1 array to a device (allocate and associate) More...

subroutine	device_map_r2 (x, x_d, n)
	Map a Fortran rank 2 array to a device (allocate and associate) More...

subroutine	device_map_r3 (x, x_d, n)
	Map a Fortran rank 3 array to a device (allocate and associate) More...

subroutine	device_map_r4 (x, x_d, n)
	Map a Fortran rank 4 array to a device (allocate and associate) More...

logical function	device_associated_r1 (x)
	Check if a Fortran rank 1 array is assoicated with a device pointer. More...

logical function	device_associated_r2 (x)
	Check if a Fortran rank 2 array is assoicated with a device pointer. More...

logical function	device_associated_r3 (x)
	Check if a Fortran rank 3 array is assoicated with a device pointer. More...

logical function	device_associated_r4 (x)
	Check if a Fortran rank 4 array is assoicated with a device pointer. More...

type(c_ptr) function	device_get_ptr_r1 (x)
	Return the device pointer for an associated Fortran rank 1 array. More...

type(c_ptr) function	device_get_ptr_r2 (x)
	Return the device pointer for an associated Fortran rank 2 array. More...

type(c_ptr) function	device_get_ptr_r3 (x)
	Return the device pointer for an associated Fortran rank 3 array. More...

type(c_ptr) function	device_get_ptr_r4 (x)
	Return the device pointer for an associated Fortran rank 4 array. More...

subroutine	device_sync_device ()
	Synchronize the device. More...

subroutine, public	device_sync_stream (stream)
	Synchronize a device stream. More...

subroutine, public	device_stream_create (stream, flags)
	Create a device stream/command queue. More...

subroutine	device_stream_create_with_priority (stream, flags, prio)
	Create a device stream/command queue with priority. More...

subroutine, public	device_stream_destroy (stream)
	Destroy a device stream/command queue. More...

subroutine, public	device_stream_wait_event (stream, event, flags)
	Synchronize a device stream with an event. More...

subroutine, public	device_profiler_start ()
	Start device profiling. More...

subroutine, public	device_profiler_stop ()
	Stop device profiling. More...

subroutine, public	device_event_create (event, flags)
	Create a device event queue. More...

subroutine, public	device_event_destroy (event)
	Destroy a device event. More...

subroutine, public	device_event_record (event, stream)
	Record a device event. More...

subroutine, public	device_event_sync (event)
	Synchronize an event. More...

Variables
integer, parameter, public	host_to_device = 1

integer, parameter, public	device_to_host = 2

integer, parameter, public	device_to_device = 3

type(htable_cptr_t), private	device_addrtbl
	Table of host to device address mappings. More...

Detailed Description

Device abstraction, common interface for various accelerators.

Function/Subroutine Documentation

◆ device_alloc()

subroutine, public device::device_alloc	(	type(c_ptr), intent(inout)	x_d,
		integer(c_size_t)	s
	)

Allocate memory on the device.

Definition at line 150 of file device.F90.

Here is the caller graph for this function:

◆ device_associate_r1()

subroutine device::device_associate_r1	(	class(*), dimension(:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d
	)

Associate a Fortran rank 1 array to a (allocated) device pointer.

Definition at line 468 of file device.F90.

◆ device_associate_r2()

subroutine device::device_associate_r2	(	class(*), dimension(:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d
	)

Associate a Fortran rank 2 array to a (allocated) device pointer.

Definition at line 493 of file device.F90.

◆ device_associate_r3()

subroutine device::device_associate_r3	(	class(*), dimension(:,:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d
	)

Associate a Fortran rank 3 array to a (allocated) device pointer.

Definition at line 518 of file device.F90.

◆ device_associate_r4()

subroutine device::device_associate_r4	(	class(*), dimension(:,:,:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d
	)

Associate a Fortran rank 4 array to a (allocated) device pointer.

Definition at line 543 of file device.F90.

◆ device_associated_r1()

logical function device::device_associated_r1 ( class(*), dimension(:), intent(inout), target x )

Check if a Fortran rank 1 array is assoicated with a device pointer.

Definition at line 780 of file device.F90.

◆ device_associated_r2()

logical function device::device_associated_r2 ( class(*), dimension(:,:), intent(inout), target x )

Check if a Fortran rank 2 array is assoicated with a device pointer.

Definition at line 807 of file device.F90.

◆ device_associated_r3()

logical function device::device_associated_r3 ( class(*), dimension(:,:,:), intent(inout), target x )

Check if a Fortran rank 3 array is assoicated with a device pointer.

Definition at line 834 of file device.F90.

◆ device_associated_r4()

logical function device::device_associated_r4 ( class(*), dimension(:,:,:,:), intent(inout), target x )

Check if a Fortran rank 4 array is assoicated with a device pointer.

Definition at line 861 of file device.F90.

◆ device_deassociate_r1()

subroutine device::device_deassociate_r1 ( class(*), dimension(:), intent(inout), target x )

Deassociate a Fortran rank 1 array from a device pointer.

Definition at line 568 of file device.F90.

◆ device_deassociate_r2()

subroutine device::device_deassociate_r2 ( class(*), dimension(:,:), intent(inout), target x )

Deassociate a Fortran rank 2 array from a device pointer.

Definition at line 592 of file device.F90.

◆ device_deassociate_r3()

subroutine device::device_deassociate_r3 ( class(*), dimension(:,:,:), intent(inout), target x )

Deassociate a Fortran rank 3 array from a device pointer.

Definition at line 616 of file device.F90.

◆ device_deassociate_r4()

subroutine device::device_deassociate_r4 ( class(*), dimension(:,:,:,:), intent(inout), target x )

Deassociate a Fortran rank 4 array from a device pointer.

Definition at line 640 of file device.F90.

◆ device_event_create()

subroutine, public device::device_event_create	(	type(c_ptr), intent(inout)	event,
		integer, optional	flags
	)

Create a device event queue.

Definition at line 1141 of file device.F90.

Here is the caller graph for this function:

◆ device_event_destroy()

subroutine, public device::device_event_destroy ( type(c_ptr), intent(inout) event )

Destroy a device event.

Definition at line 1171 of file device.F90.

Here is the caller graph for this function:

◆ device_event_record()

subroutine, public device::device_event_record	(	type(c_ptr), intent(in)	event,
		type(c_ptr), intent(in)	stream
	)

Record a device event.

Definition at line 1189 of file device.F90.

Here is the caller graph for this function:

◆ device_event_sync()

subroutine, public device::device_event_sync ( type(c_ptr), intent(in) event )

Synchronize an event.

Definition at line 1208 of file device.F90.

Here is the caller graph for this function:

◆ device_finalize()

subroutine, public device::device_finalize

Definition at line 121 of file device.F90.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ device_free()

subroutine, public device::device_free ( type(c_ptr), intent(inout) x_d )

Deallocate memory on the device.

Definition at line 171 of file device.F90.

Here is the caller graph for this function:

◆ device_get_ptr_r1()

type(c_ptr) function device::device_get_ptr_r1 ( class(*), dimension(:), intent(in), target x )

Return the device pointer for an associated Fortran rank 1 array.

Definition at line 888 of file device.F90.

Here is the caller graph for this function:

◆ device_get_ptr_r2()

type(c_ptr) function device::device_get_ptr_r2 ( class(*), dimension(:,:), intent(in), target x )

Return the device pointer for an associated Fortran rank 2 array.

Definition at line 916 of file device.F90.

Here is the caller graph for this function:

◆ device_get_ptr_r3()

type(c_ptr) function device::device_get_ptr_r3 ( class(*), dimension(:,:,:), intent(in), target x )

Return the device pointer for an associated Fortran rank 3 array.

Definition at line 944 of file device.F90.

Here is the caller graph for this function:

◆ device_get_ptr_r4()

type(c_ptr) function device::device_get_ptr_r4 ( class(*), dimension(:,:,:,:), intent(in), target x )

Return the device pointer for an associated Fortran rank 4 array.

Definition at line 972 of file device.F90.

Here is the caller graph for this function:

◆ device_init()

subroutine, public device::device_init

Definition at line 106 of file device.F90.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ device_map_r1()

subroutine device::device_map_r1	(	class(*), dimension(:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d,
		integer, intent(in)	n
	)

Map a Fortran rank 1 array to a device (allocate and associate)

Definition at line 664 of file device.F90.

◆ device_map_r2()

subroutine device::device_map_r2	(	class(*), dimension(:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d,
		integer, intent(in)	n
	)

Map a Fortran rank 2 array to a device (allocate and associate)

Definition at line 693 of file device.F90.

◆ device_map_r3()

subroutine device::device_map_r3	(	class(*), dimension(:,:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d,
		integer, intent(in)	n
	)

Map a Fortran rank 3 array to a device (allocate and associate)

Definition at line 722 of file device.F90.

◆ device_map_r4()

subroutine device::device_map_r4	(	class(*), dimension(:,:,:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d,
		integer, intent(in)	n
	)

Map a Fortran rank 4 array to a device (allocate and associate)

Definition at line 751 of file device.F90.

◆ device_memcpy_common()

subroutine, private device::device_memcpy_common	(	type(c_ptr), intent(inout)	ptr_h,
		type(c_ptr), intent(inout)	x_d,
		integer(c_size_t), intent(in)	s,
		integer, intent(in), value	dir,
		logical, intent(in)	sync_device,
		type(c_ptr), intent(inout)	stream
	)

private

Copy data between host and device.

Note: For device to device copies, ptr_h is assumed to be the dst device pointer

Definition at line 373 of file device.F90.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ device_memcpy_cptr()

subroutine device::device_memcpy_cptr	(	type(c_ptr), intent(inout)	dst,
		type(c_ptr), intent(inout)	src,
		integer(c_size_t), intent(in)	s,
		integer, intent(in), value	dir,
		logical, optional	sync,
		type(c_ptr), optional	strm
	)

Copy data between host and device (or device and device) (c-pointers)

Note: For host-device copies dst is the host pointer and src is the device pointer (regardless of dir)

Definition at line 344 of file device.F90.

◆ device_memcpy_r1()

subroutine device::device_memcpy_r1	(	class(*), dimension(:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d,
		integer, intent(in)	n,
		integer, intent(in), value	dir,
		logical	sync,
		type(c_ptr), optional	strm
	)

Copy data between host and device (rank 1 arrays)

Definition at line 190 of file device.F90.

◆ device_memcpy_r2()

subroutine device::device_memcpy_r2	(	class(*), dimension(:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d,
		integer, intent(in)	n,
		integer, intent(in), value	dir,
		logical	sync,
		type(c_ptr), optional	strm
	)

Copy data between host and device (rank 2 arrays)

Definition at line 228 of file device.F90.

◆ device_memcpy_r3()

subroutine device::device_memcpy_r3	(	class(*), dimension(:,:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d,
		integer, intent(in)	n,
		integer, intent(in), value	dir,
		logical	sync,
		type(c_ptr), optional	strm
	)

Copy data between host and device (rank 3 arrays)

Definition at line 266 of file device.F90.

◆ device_memcpy_r4()

subroutine device::device_memcpy_r4	(	class(*), dimension(:,:,:,:), intent(inout), target	x,
		type(c_ptr), intent(inout)	x_d,
		integer, intent(in)	n,
		integer, intent(in), value	dir,
		logical	sync,
		type(c_ptr), optional	strm
	)

Copy data between host and device (rank 4 arrays)

Definition at line 304 of file device.F90.

◆ device_name()

subroutine, public device::device_name ( character(len=*), intent(inout) name )

Definition at line 137 of file device.F90.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ device_profiler_start()

subroutine, public device::device_profiler_start

Start device profiling.

Definition at line 1123 of file device.F90.

Here is the caller graph for this function:

◆ device_profiler_stop()

subroutine, public device::device_profiler_stop

Stop device profiling.

Definition at line 1132 of file device.F90.

Here is the caller graph for this function:

◆ device_stream_create()

subroutine, public device::device_stream_create	(	type(c_ptr), intent(inout)	stream,
		integer, optional	flags
	)

Create a device stream/command queue.

Definition at line 1035 of file device.F90.

◆ device_stream_create_with_priority()

subroutine device::device_stream_create_with_priority	(	type(c_ptr), intent(inout)	stream,
		integer, intent(in)	flags,
		integer, intent(in)	prio
	)

Create a device stream/command queue with priority.

Definition at line 1068 of file device.F90.

Here is the caller graph for this function:

◆ device_stream_destroy()

subroutine, public device::device_stream_destroy ( type(c_ptr), intent(inout) stream )

Destroy a device stream/command queue.

Definition at line 1085 of file device.F90.

Here is the caller graph for this function:

◆ device_stream_wait_event()

subroutine, public device::device_stream_wait_event	(	type(c_ptr), intent(in)	stream,
		type(c_ptr), intent(in)	event,
		integer	flags
	)

Synchronize a device stream with an event.

Definition at line 1103 of file device.F90.

Here is the caller graph for this function:

◆ device_sync_device()

subroutine device::device_sync_device

Synchronize the device.

Definition at line 1000 of file device.F90.

◆ device_sync_stream()

subroutine, public device::device_sync_stream ( type(c_ptr), intent(in) stream )

Synchronize a device stream.

Definition at line 1017 of file device.F90.

Here is the caller graph for this function:

Variable Documentation

◆ device_addrtbl

type(htable_cptr_t), private device::device_addrtbl

private

Table of host to device address mappings.

Definition at line 92 of file device.F90.

◆ device_to_device

integer, parameter, public device::device_to_device = 3

Definition at line 47 of file device.F90.

◆ device_to_host

integer, parameter, public device::device_to_host = 2

Definition at line 47 of file device.F90.

◆ host_to_device

integer, parameter, public device::host_to_device = 1

Definition at line 47 of file device.F90.

Data Types

Functions/Subroutines

Variables

Detailed Description

Function/Subroutine Documentation

◆ device_alloc()

◆ device_associate_r1()

◆ device_associate_r2()

◆ device_associate_r3()

◆ device_associate_r4()

◆ device_associated_r1()

◆ device_associated_r2()

◆ device_associated_r3()

◆ device_associated_r4()

◆ device_deassociate_r1()

◆ device_deassociate_r2()

◆ device_deassociate_r3()

◆ device_deassociate_r4()

◆ device_event_create()

◆ device_event_destroy()

◆ device_event_record()

◆ device_event_sync()

◆ device_finalize()

◆ device_free()

◆ device_get_ptr_r1()

◆ device_get_ptr_r2()

◆ device_get_ptr_r3()

◆ device_get_ptr_r4()

◆ device_init()

◆ device_map_r1()

◆ device_map_r2()

◆ device_map_r3()

◆ device_map_r4()

◆ device_memcpy_common()

◆ device_memcpy_cptr()

◆ device_memcpy_r1()

◆ device_memcpy_r2()

◆ device_memcpy_r3()

◆ device_memcpy_r4()

◆ device_name()

◆ device_profiler_start()

◆ device_profiler_stop()

◆ device_stream_create()

◆ device_stream_create_with_priority()

◆ device_stream_destroy()

◆ device_stream_wait_event()

◆ device_sync_device()

◆ device_sync_stream()

Variable Documentation

◆ device_addrtbl

◆ device_to_device

◆ device_to_host

◆ host_to_device