Data Set Options for Relational Databases

BL_CPU_PARALLELISM= Data Set Option

Specifies the number of processes or threads to use when building table objects.

Default value:	none
Valid in:	DATA and PROC steps (when accessing DBMS data using SAS/ACCESS software)
DBMS support:	DB2 under UNIX and PC Hosts

Syntax
Syntax Description
Details
See Also

Syntax

BL_CPU_PARALLELISM=number of processes or threads

Syntax Description

number of processes or threads: specifies the number of processes or threads that the load utility uses to parse, convert, and format data records when building table objects.

Details

To specify this option, you must first set BULKLOAD=YES.

This option exploits intrapartition parallelism and significantly improves load performance. It is particularly useful when loading presorted data, because record order in the source data is preserved.

The maximum number that is allowed is 30. If the value of this parameter is 0 or has not been specified, the load utility selects an intelligent default that is based on the number of available CPUs on the system at run time. If there is insufficient memory to support the specified value, the utility adjusts the value.

When BL_CPU_PARALLELISM is greater than 1, the flushing operations are asynchronous, permitting the loader to exploit the CPU. If tables include either LOB or LONG VARCHAR data, parallelism is not supported and this option is set to 1, regardless of the number of system CPUs or the value that the user specified.

Although use of this parameter is not restricted to symmetric multiprocessor (SMP) hardware, you might not obtain any discernible performance benefit from using it in non-SMP environments.