when i run mapd and after that nvidia-smi i always see max memory usage as 2664MiB but i have 12 GB of RAM so where should i change to increase the RAM usage of MapD
The MapD Core database incrementally allocates more RAM on the GPU as needed. You can get a more granular view of how much memory the database has allocated and used by using the
\memory_summary command in
mapdql, documented here.
I am running a select query on the flight data set which is provided by mapd only which has 7,009,728 rows and 56 columns.And you had said after running the query i fire the command \memory_summary from different console .i have approx 11g of VRAM so my query is like select * from flights_2008_7M limit 4709393, which shows 10.24 gb of VRAM allocation but the inuse column is always zero.
I am starting the map by this command $MAPD_PATH/startmapd --data /var/lib/mapd
why my IN-USE is always Zero.
A projection query with no filter like you have executed there does not require the GPU for anything in its execution.
Run a query that requires filter, aggregation or calculations to get the GPU involved.
thanks for this help.