Parallel I/O workflow

The general workflow for the scda format provided by sc_scda.h

All workflows start with sc_scda_fopen_write or sc_scda_fopen_read that creates a file context sc_scda_fcontext_t that never moves backwards.

Then the user may call any combination of functions out of:

• sc_scda_fwrite_inline,
• sc_scda_fwrite_block,
• sc_scda_fwrite_array and
• sc_scda_fwrite_varray,

for a context obtained from sc_scda_fopen_write.

Alternatively, after using sc_scda_fopen_read the user must call sc_scda_fread_section_header that examines the current file section and then call accordingly to the retrieved file section type one of

• sc_scda_fread_inline_data,
• sc_scda_fread_block_data,
• sc_scda_fread_array_data,
• sc_scda_fread_varray_sizes and
• sc_scda_fread_varray_data.

Finally, the file context is collectively closed and deallocated by sc_scda_fclose.

All above mentioned sc_scda functions are collective and output an error code (cf. sc_scda_ferror_t) that can be translated by sc_scda_ferror_string.

For more details and the option for encoding metadata and data see the functions in sc_scda.h.

Warning

The scda API will change in the next libsc version!