search

antonmks / Alenka

GPU database engine
Other
1.17k stars 120 forks source link

zone map question #16

Closed sam1988 closed 11 years ago

sam1988 commented 11 years ago

when doing filter ，zone map is used to filter some block that is not between the number of [min,max] ,it has a great help in a sorted column,but seems has little affect on column that not sorted and the number is random? and also when doing filter, it seems transfer the needed compressed column to device and also to decompress the data to do fiter operation, will it cost some time,i mean whether alenka filter operation has increase speed over same cpu filter operation ?

antonmks commented 11 years ago

Yes, this is correct. Zone maps are a bit like indexes. Alenka checks if a segments mix and max column values fall within a filter expression. This way we can skip entire segments. And yes, often it requires sorting data based on filter columns. About the speed of CPU and GPU operations - I believe that many operations are faster on a GPU even when taking into consideration the cost of data transfers and decompression.