Hello, Although we originally provided hooks deep in the kernel to enrich it with accelerators, experienced has shown over the years that 1) it wasn't used often, 2) it creates a low-level maintenance headache, and 3) complex datastructures and accelerators can often be modelled using ordinary BAT as representations and judicious use of the BAT operators. For volatile structures, it sometimes pays to create an in-memory structure yourself, enriched with operators of interest, and a scheme to map them to BATs for persistency. This approach is a hybrid approach. The new crackers and database partitioning modules follow this approach. The choice for a C-data structure is often chosen when the programmer is not trained to properly use the GDK (BAT) data structures directly. The MonetDB 5 approach also brings program transformations into view as a scheme to exploit. If you have runtime or application knowledge, you may e.g. replace sequences produced by the SQL compiler with cheaper ones. Or you might even produce plans that embody multiple alternatives, which are chosen at runtime by an application specific scheduler. The src/scheduler/mal_memo.mx illustrates how this might work. (.../Tests/memo*.mal) regards, Martin Agustin Schapira wrote:
Hi all,
We have created a new Monet data type for one of our projects, and now want to write a set of accelerators for it. Is there any good documentation/examples on how to do that (for 4.16)?
Thanks a lot,
-- Agustin
------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Monetdb-developers mailing list Monetdb-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-developers