SPD Engine System Options |
Valid in: | configuration file, SAS invocation, OPTIONS statement, System Options window |
Default: | COST |
Syntax | |
Details |
Syntax |
SPDEWHEVAL=COST | EVAL1 | EVAL3EVAL4 |
specifies that the SPD Engine decides which evaluation strategy to use to optimize the WHERE expression. This process calculates the number of threads to be used, which minimizes the overhead of spawning underutilized threads. This is the default.
is a multi-threaded index evaluation strategy that can quickly determine the rows that satisfy the WHERE expression, using multiple threads. The number of threads that are spawned to retrieve the observations is equal to the THREADNUM= value.
is a single-threaded index evaluation strategy that is used for a simple or compound WHERE expression in which all of the key variables have a simple index and no condition tests for non-equality. Multi-threading might be used to retrieve the observations.
Details |
COST, the default setting for SPDEWHEVAL=, analyzes the WHERE expression and any available indexes. Based on the analysis, the SPD Engine chooses an evaluation strategy to optimize the WHERE expression. The evaluation strategy can be EVAL1, EVAL3, EVAL4, or a strategy that sequentially reads the data if no indexes are available, or if the analysis shows that using the index or indexes cannot improve processing time.
COST optimizes the number of threads to use for processing the WHERE expression. COST determines and spawns the number of threads that can be efficiently used. Based on the value of THREADNUM=, COST can save significant processing time by not spawning threads that are underutilized.
COST is the recommended value for SPDEWHEVAL=, unless the WHERE expression exactly meets one of the other evaluation strategy criterion. It is strongly recommended that benchmark tests be used to determine whether a value other than COST is more efficient.
EVAL1 might be more efficient if the WHERE expression is complex and there are multiple indexes for the variables. EVAL1 spawns multiple threads to determine which segments meet the conditions of the WHERE expression. Multiple threads can also be used to retrieve the observations.
Note: In a few situations, COST might not perform the best. To determine whether changing the value to EVAL1 or EVAL3EVAL4 can produce better performance, run a performance benchmark, change the value, and re-run the performance benchmark. Comparing results shows you how the specific data population you are querying responds to rules-based WHERE planning.
Copyright © 2010 by SAS Institute Inc., Cary, NC, USA. All rights reserved.