Step 1: check the datasheet for your chosen fpga to see what the allowed input frequencies are for the DCM.
Step 2: cascade 2 DCM's to obtain larger division ratios. First divide by 5 (50 MHz in ==> 10 MHz out), then divide by 10 (10 MHz in==> 0.5 MHz out)
I'm assuming here that the fpga you did not specify will be able to handle 10 MHz input on the DCM. And you may want to check that, because some Xilinx fpga can take 10 MHz input on the DCMN, and some don't. Older spartan-3's only take 18 MHz or higher. So if you have one of those you're boned. 18 MHz / 16 ... higher than 0.5 MHz.
As an alternative to using DCM for clock division you can use clock enable signals. If you don't know clock enables, google for "xilinx clock enable CE" for some ideas.
- - - Updated - - -
Hey wait, I just noted the logical progression of your posts. Well, alogical progression really, but who cares.
First you used some gated clock as divider, then K-J suggested to use clock enables, then <something>, and then you want to use DCM's for clock division. I interpret that as
"I tried clock enables but failed, so now I will try DCMs". If that is the case, I suggest you try harder to get clock enables to work. If only because for this particular problem (*) it is the right solution.
*) assuming this thread is about the exact same thing as your other thread "VHDL clock divider warning". But of course it will be the exact same thing because humans are predictable that way.