On 1/14/24 16:15, Tobias Burnus wrote:
+@node omp_target_memcpy
+@subsection @code{omp_target_memcpy} -- Copy data between devices
+@table @asis
+@item @emph{Description}:
+This routine tests copies @var{length} of bytes of data from the device
+identified by device number @var{src_device_num} to device
@var{dst_device_num}.
Hmmm, I'm sure it's the train's fault :-) but "tests copies" makes no sense,
and that's cut-and-pasted multiple times. I think you just mean "copies" in
all cases.
+@node omp_target_memcpy_rect
+@subsection @code{omp_target_memcpy_rect} -- Copy a subvolume of data between
devices
+@table @asis
+@item @emph{Description}:
+This routine tests copies a subvolume of data from the device identified by
+device number @var{src_device_num} to device @var{dst_device_num}. The
+subvolume of a multi-dimensional array of array dimension @var{num_dims} and
+each array element has a size of @var{element_size} bytes. The @var{volume}
This is kind of garbled. How about rephrasing that second sentence as
The array has @var{num_dims} and each array element has a size of
@var{element_size} bytes.
+array specifies how many elements per dimension will be copied. The full
s/will be/are/
+array in number of elements is given by the @var{dst_dimensions} and
+@var{src_dimensions} arguments for the array on the destination and source
+device, respectively. The offset per dimension to the first element to
I think we can simplify that sentence, too, like
The full sizes of the destination and source arrays are given by the
@var{dst_dimensions} and @var{src_dimensions} arguments, respectively.
+be copied is given by the @var{dst_offset} and @var{src_offset} arguments.
+The routine returns zero on success and non-zero otherwise.
+
+The OpenMP only requires that @var{num_dims} up to three is supported. In order
s/OpenMP/OpenMP specification/ ?
+to find implementation-specific maximally supported number of dimensions, the
+routine will return this value when invoked with a NULL pointer to both the
s/will return/returns/
either "null pointer" or "@code{NULL}" is preferable to "NULL pointer".
+@var{dst} and @var{src} arguments. As GCC supports arbitrary dimensions, it
+will return INTMAX.
s/will return INTMAX/returns @code{INT_MAX}/
+
+The device-number arguments must be conforming device number, the @var{src} and
s/number,/numbers,/
+@var{dst} must be either both NULL or any of the following must be fulfilled:
same issue with "NULL" here, either "@code{NULL}" or "null pointers".
"any" seems unlikely to be useful. Do you mean "all" of the following
conditions?
+@var{element_size} and @var{num_dims} must be positive, the @var{volume},
offset
+and dimension arrays must have at least @var{num_dims} dimensions.
+Running this routine in a @code{target} region except on the initial device
+is not supported.
The part of the patch for omp_target_memcpy_rect_async has very similar
problems and needs the same fixes.
-Sandra
