| | If set, do not attempt to open the input GVCF file paths in
Java. This can only be used together with
`--bypass_feature_reader`, and allows operating on file systems
which GenomicsDB understands how to open but GATK does not.
This disables many sanity checks and cannot be used together
with `--variant` (only with `--sample_name_map`). GATK also
requires `--header` to be set whenever `--avoid_nio` is used.
|
| | Batch size controls the number of samples for which readers are
open at once and therefore provides a way to minimize memory
consumption. However, it can take longer to complete. Use
`--consolidate` if more than a hundred batches were used, to
improve feature read time. A value of `0` means no batching
(i.e. readers for all samples will be opened at once).
Default: 0.
|
| | If set, use htslib to read input GVCFs instead of GATK's
FeatureReader. This reduces memory usage and can speed up the
import. Lower memory requirements may also enable parallelism
through `--max_num_intervals_to_import_in_parallel`. To enable
this option, GVCFs must be normalized, block-compressed and
indexed.
|
| | If set, merge all fragments into one. If importing data in
batches, a new fragment is created for each batch. In case
thousands of fragments are created, GenomicsDB feature readers
will try to open ~20x as many files, and internally GenomicsDB
would consume more memory to maintain bookkeeping data from all
fragments. Merging can potentially improve read performance,
however this has no effect if only one batch is used.
|
--genomicsdb_segment_size | | Buffer size in bytes allocated for GenomicsDB attributes during
import. Should be large enough to hold data from one site.
Default: 1048576.
|
--genomicsdb_shared_posixfs_optimizations | | If set, allow for optimizations to improve the usability and
performance for shared POSIX filesystems (e.g. NFS, Lustre).
File level locking is disabled and file system writes are
minimized by keeping a higher number of file descriptors open
for longer periods of time. Use with `--batch_size` if keeping a
large number of file descriptors open is an issue.
|
--genomicsdb_vcf_buffer_size | | Buffer size in bytes to store variant contexts. Larger values
are better, as smaller values cause frequent disk writes.
Default: 16384.
|
| | VCF file to use instead of reading and combining headers from
the input GVCFs.
Cannot be used together with an incremental import into an
existing GenomicsDB workspace.
|
--max_num_intervals_to_import_in_parallel | | Maximum number of intervals to import in parallel. Higher
values may improve performance but require more memory and a
higher number of file descriptors open at the same time.
Default: 1.
|
--merge_contigs_into_num_partitions | | Number of GenomicsDB arrays to merge input intervals into. A
value of `0` disables this merging. This option can only be
used if entire contigs are specified as intervals: the tool
will not split a contig into multiple arrays, so the actual
number of partitions may be less than what is specified here.
This can improve performance when importing a very large number
of contigs (more than 100). Default: 0.
|
| | If set, import all data in between intervals. Improves
performance using large lists of intervals, as in exome
sequencing, especially if GVCF data only exists for specified
intervals.
|
--overwrite_existing_genomicsdb_workspace | | If set, overwrite `--genomicsdb_workspace_path` if it already
exists. Otherwise a new workspace is created.
|
| | How many simultaneous threads to use when opening VCFs in
batches. Higher values may improve performance when network
latency is an issue. Multiple reader threads are not supported
when running with multiple intervals. Default: 1.
|
--validate_sample_name_map | | If set, enable checks on the `--sample_name_map` file. Verifies
that feature readers are valid and shows a warning if sample
names do not match with the headers.
|