libpressio 0.93.0
Loading...
Searching...
No Matches
IO Modules

by_extension

stability: stable

thread_safety: multiple

a "by_extension" data loader that chooses an appropriate io plugin based on file extension

Options

by_extension:keys

type: char*[]

default: <char*[]> = {, posix, bp, adios2, csv, csv, h5, hdf5, hdf5, hdf5, nc, netcdf, npy, numpy, }

description: extensions to map

by_extension:plugin

type: char*

default: <char*> = "posix"

options: by_extension, csv, cufile, empty, hdf5, iota, mmap, noop, numpy, petsc, posix, select

description: io plugin to delegate to

by_extension:values

type: char*[]

default: <char*[]> = {}

description: io plugins to map to

io:path

type: char*

unset by default

description:

Configuration

other configuration entries:

  • by_extension:plugin
  • pressio:stability
  • pressio:thread_safe

csv

stability: stable

thread_safety: multiple

read CSV files

Options

csv:field_delim

type: char*

default: <char*> = ","

description: delimiter for columns

csv:headers

type: char*[]

default: <char*[]> = {}

description: headers for the CSV file used for writing

csv:line_delim

type: char*

default: <char*> = " "

description: delimiter for rows

csv:skip_rows

type: uint32

default: <uint32> = 0

description: number of rows to skip while reading

io:path

type: char*

default: <char*> = ""

description: path to the file

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

cufile

stability: experimental

thread_safety: multiple

io plugin to load data using NVIDIA's cufile APIs

Options

cufile:buf_register_flags

type: int32

default: <int32> = 0

description: flags passed to cuFileBufRegister

cufile:dev_offset

type: int64

default: <int64> = 0

description: offset to use when reading/writing to device memory

cufile:file_offset

type: int64

default: <int64> = 0

description: offset to use when reading from the file

cufile:open_close_driver

type: int32

default: <int32> = 1

description: should cuFileDriverOpen and cuFileDriverClose be called?

io:path

type: char*

default: <char*> = ""

description: path to read the file from

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

empty

stability: stable

thread_safety: multiple

read in an array of all zeros

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

hdf5

stability: stable

thread_safety: single

read in HDF5 files

Options

hdf5:dataset

type: char*

default: <char*> = "dataset"

description: the name of the dataset to read or write

hdf5:file_block

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: the size of a block in this read or write

hdf5:file_count

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: the number of blocks in this read or write

hdf5:file_extent

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: the extent for the dataset

hdf5:file_start

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: the start of the the read/write

hdf5:file_stride

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: the stride for the read/write

hdf5:mpi_comm

type: void*

default: <void*> = true

description: the MPI communicator to use for reading and writing

hdf5:use_parallel

type: int32

default: <int32> = 0

description: use parallel IO for reading and writing

io:path

type: char*

default: <char*> = ""

description: the path to the file on the disk

Configuration

hdf5:parallel

type: int32

value: <int32> = 1

description: indicates if HDF was built with parallel support

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

iota

stability: stable

thread_safety: multiple

read in a block that counts from 1 to N

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

mmap

stability: stable

thread_safety: single

uses mmap shared mappings to read files

Options

io:file_descriptor

type: int32

unset by default

description: file descriptor for the file on disk

io:path

type: char*

unset by default

description: path to the file on disk

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

noop

stability: stable

thread_safety: multiple

a writer which preforms a noop

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

numpy

stability: stable

thread_safety: single

read Numpy .npy files

Options

io:path

type: char*

default: <char*> = ""

description: path to the file on disk

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

petsc

stability: stable

thread_safety: single

reads PETSc Matrix files

Options

io:path

type: char*

default: <char*> = ""

description: the path to the file on disk

petsc:matrix_format

type: char*

default: <char*> = "seqdense"

description: the petsc matrix format

petsc:viewer_format

type: char*

default: <char*> = "DEFAULT"

description: the petsc viewer format

petsc:viewer_type

type: char*

default: <char*> = "binary"

description: the petsc viewer type

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

posix

stability: stable

thread_safety: single

POSIX io

Options

io:file_descriptor

type: int32

unset by default

description: posix file descriptor to read/write from

io:file_pointer

type: void*

unset by default

description: FILE* to read/write from

io:path

type: char*

unset by default

description: path on the file system to read/write from

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe

select

stability: unstable

thread_safety: single

selects a subset from a buffer read in or written out

Options

select:block

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: size of each block

select:io

type: char*

default: <char*> = "posix"

options: by_extension, csv, cufile, empty, hdf5, iota, mmap, noop, numpy, petsc, posix, select

description: IO to use before selection

select:size

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: size of selection in blocks

select:start

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: starting block for selection

select:stride

type: data

default: <data> = data{ type=uint64_t dims={0, } has_data=false}

description: stride of selection

Configuration

other configuration entries:

  • pressio:stability
  • pressio:thread_safe
  • select:io