Yida Wu created IMPALA-11782:
--------------------------------
Summary: Reduce size for each codegen cache entry in NORMAL mode
Key: IMPALA-11782
URL: https://issues.apache.org/jira/browse/IMPALA-11782
Project: IMPALA
Issue Type: Improvement
Components: Backend
Reporter: Yida Wu
We added support for codegen cache in IMPALA-11470, however in NORMAL mode, the
size of each codegen cache entry is large because the key we use is the bitcode
of the whole llvm module which is very large, one possible solution is to use
the query plan plus other elements which could affect the codegen functions to
be the key, from the prototype, it can reduce the cache size significantly, but
one thing we need to confirm is the safety of using such keys.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
