Re:Automatic Generation of a Hybrid Query Execution Engine
by obl on Friday August 17, 2018 @09:43AM
(#17782761)
It's a cute idea to cobble together a meta-tracing JIT for sqlite's query interpreter using source extraction and gcc.
I think it's basically a local minimum for implementation complexity vs performance though. To push that further you probably need to modify the original interpreter or encode more (any) knowledge about the semantics in the JIT, at which point you lose the main advantage of maintaining only one codebase.
Re:Automatic Generation of a Hybrid Query Execution Engine
by hinkley on Friday August 17, 2018 @12:46PM
(#17784112)
It gives you an idea of the sort of gains you might get with a proper JIT. if it had turned out that tracing only gave you a 10% boost then maybe a JIT isn’t the best place to put your time.
Re:Automatic Generation of a Hybrid Query Execution Engine
by ngaut on Friday August 17, 2018 @08:59AM
(#17782447)
"We create a set of automated tools to create a runtime code generation engine and integrate such an engine into SQLite which is one of the most popular relational databases in the world and is used in a large variety of contexts. Speedups of up to 1.7x were observed in microbenchmarks with queries involving a large number of operations"
I think it's basically a local minimum for implementation complexity vs performance though. To push that further you probably need to modify the original interpreter or encode more (any) knowledge about the semantics in the JIT, at which point you lose the main advantage of maintaining only one codebase.