Unexpected performance finding with MOVE-CORRESPONDING between table and structure variant

[florian_martin4]

During some tests with a huge amount of data a a colleague of me Dirk Freundel gets unexpected performance finding between:

1. MOVE-CORRESPONDING struc1 TO struc2


2. MOVE-CORRESPONDING itab1 TO itab2

He put 1.000.000 lines of a table with 27 columns into a internal table with 16 columns. Variant 1 needed about 179 ms less time than variant 2.



Test code:

DO 10 TIMES.



*   Init

    FREE lt_middle.



*   Runtime

    GET RUN TIME FIELD lv_runtime_a.



*   LOOP

    LOOP AT lt_database ASSIGNING <fs_database>.   



        MOVE-CORRESPONDING <fs_database> TO ls_middle.     

        INSERT ls_middle INTO TABLE lt_middle.



    ENDLOOP.



*   OR TABLE

    "MOVE-CORRESPONDING lt_database TO lt_middle.



*   Runtime

    GET RUN TIME FIELD lv_runtime_b.

    lv_runtime_b = ( lv_runtime_b - lv_runtime_a ) / 1000.

    WRITE: / SY-INDEX, ' Runtime [msec]: ', lv_runtime_b.



ENDDO.

Has anybody else observed the same performance finding, Please add a comment.