Is your feature request related to a problem? Please describe.
The HDF5 library is close to releasing (in HDF5-1.13.3) "multi-dataset i/o", a feature allowing users to describe or schedule accesses to several datasets, then fire them all off at once.
The interface is described in the "RFC: New HDF5 API Routines for HPC Applications
Read/Write Multiple Datasets in an HDF5 file" document, and has changed quite a bit over the years (it's been in "proposed" state for close to a decade...)
The 1.13 series is classified as "experimental" and the interface might change (again) before stabilizing in the 1.14 release (currently jan 2023 but I am guessing that will slip)
Is your feature request related to a problem? Please describe. The HDF5 library is close to releasing (in HDF5-1.13.3) "multi-dataset i/o", a feature allowing users to describe or schedule accesses to several datasets, then fire them all off at once.
The interface is described in the "RFC: New HDF5 API Routines for HPC Applications Read/Write Multiple Datasets in an HDF5 file" document, and has changed quite a bit over the years (it's been in "proposed" state for close to a decade...)
(write looks similar, except 'buf' is const and not an output parameter )
Describe the solution you'd like
A way to collect multiple slices and send them to a "write_multi" routine
Describe alternatives you've considered
Is it possible to extract the necessary bits of information from the C api to do this alongside highfive?
Additional context
The HDF group has an RFC describing the benefits of this approach https://www.hdfgroup.org/wp-content/uploads/2022/08/H5HPC_MultiDset_RW_IO_RFC_v7_20220523.pdf
The 1.13 series is classified as "experimental" and the interface might change (again) before stabilizing in the 1.14 release (currently jan 2023 but I am guessing that will slip)