Automapper running extremely slow on mapping 1400 records

Question

I am using AUtomapper which I am very impressed with however, I have a complex object with many nested collections. I'm using Telerik OpenAccess and it returns the 1400 records fast but when I pass it to Automapper and it slows to a ridiculous crawl. Here is my code for reference:

        List<DAL.Event> query = httpContext.Events.Where(e => e.Inactive != true && e.Event_Locations != null).ToList();
        Mapper.CreateMap<DAL.Event, EventDTO>();
        Mapper.CreateMap<DAL.Event_Association, EventAssociationDTO>();
        Mapper.CreateMap<DAL.Event_ExecutingUnit, EventExecutingUnitDTO>();
        Mapper.CreateMap<DAL.Event_Funding, EventFundingDTO>();
        Mapper.CreateMap<DAL.Event_Location, EventLocationDTO>();
        Mapper.CreateMap<DAL.Event_Objective, EventObjectiveDTO>();
        Mapper.CreateMap<DAL.Event_OSR, EventOSRDTO>();
        Mapper.CreateMap<DAL.Event_PaxBreakDown, EventPAXBreakDownDTO>();
        Mapper.CreateMap<DAL.Event_RegionalConsideration, EventRegionalConsiderationDTO>();
        Mapper.CreateMap<DAL.Event_ReviewStatus, EventReviewStatusDTO>();
        Mapper.CreateMap<DAL.Event_SPCalendarClone, EventSPCalendarClonesDTO>();
        Mapper.CreateMap<DAL.Event_Task, EventTasksDTO>();
        Mapper.CreateMap<DAL.Event_TSO, EventTSOsDTO>();
        Mapper.AssertConfigurationIsValid();
        Mapper.AllowNullDestinationValues = true;

        IList<EventDTO> result = Mapper.Map<List<DAL.Event>, List<EventDTO>>(query);
        return result;

HELP!

Answer 1

I benchmarked automapper. A single core of a 2.0GHz Xeon is able to handle 85,000 maps per second, on a small object (3 properties). It was 60 times slower than manually copying the properties. If you'd like, I can benchmark other values for you.

Answer 2

There was a team at my previous job who were also using Automapper but in the end they removed it because of the performance impact.

I think in this specific scenario it's best to write the mapping code yourself, or start replacing them one by one. Maybe one mapping is causing the bad performance?