Impact of too many decimal places on query performance

[mce]

Hi,

Does anyone know if having too many decimal points will reduce the performance of calculations in TM1?

I have 2 measures in a cube that goes through the same rules and aggregations, and they have records exactly in the same coordinates, that means they have exactly the same number of records in the cube. However one of these is generated through a division operation of the other in a TI process. Therefore it has lots of decimals in most of the records.

I observed that the queries against the TI calculated measure are returning much slower than the other. My hypothesis is that having too many decimals in one of the measures causes it to take more time in calculations, and therefore reducing the query performance. I know I can round the numbers in the TI process after the division operation, but before doing this, I wanted to check comments from the experts.

Any comments will be appreciated.

Regards,

[tomok]

I don't believe your theory holds water and here's why. You have no control over the data format in TM1 except to either say a number or string. For numbers, TM1 uses the double-precision, floating point decimal number format. You can't change that and you also can't control how many decimal places TM1 is going to use. We see that all the time. What you think is 4 is often times 3.9999999 or 4.00000001 in TM1. Issuing a ROUND command in a rule or rounding before loading in a TI will not change that. If you think that it will be faster to do math on rounded numbers then by all means give it a test. I doubt seriously you will find any difference in performance.