Time Intelligence: TOTALMTD vs DATESMTD vs DATEADD

If you want to understand time intelligence better in DAX, read this excellent blog post.

In short here is the behavior of the three functions you mentioned.

TOTALMTD(): This is only syntactic sugar, all it does is give you the following:

CALCULATE(
    [measure]
    ,DATESMTD(DimDate[Date])
)

DATESMTD(): This works in a date dimension, which must have contiguous, nonrepeating dates from January 1 of the first year you have data to December 31 of the last year you have data. The function returns a 1 column table made up of dates between the first of the month of the current date in context and the current date in context.

DATEADD(): This essentially gives you a range of dates (one column table) based on the number of intervals you've requested in either direction. It does not behave intuitively compared to a DATEADD() function in any other language, and I am not aware of any circumstances we (we being a Microsoft BI consultancy with a number of DAX experts) have decided this is the right function to use for any of our work. There is a good description in the linked blog post at the top of my reply.