Michael Ho has posted comments on this change. Change subject: IMPALA-3066: Lazy materialization of LLVM module bitcode. ......................................................................
Patch Set 1: (2 comments) http://gerrit.cloudera.org:8080/#/c/3220/1/be/src/codegen/llvm-codegen.cc File be/src/codegen/llvm-codegen.cc: Line 76: DEFINE_bool(disable_lazy_materialization, false, > I spent a little time trying to think of corner cases that we might be miss Thanks. I just feel that we don't have very strict control of the IR generated from our cross-compiled code. For instance, LikePredicate() can generate alias (http://llvm.org/docs/LangRef.html#aliases) in the old way it was written. The alias destructor generated in the IR doesn't exist in the impalad binary, only in the IR. Not sure if there are flags which we can pass to clang to prevent surprise like this ? Line 565: if (called_fn != NULL) MaterializeFunctionHelper(called_fn); > Oh, so isMaterializable() is false if the function is already materialized? They should rename the function isMaterialized(). -- To view, visit http://gerrit.cloudera.org:8080/3220 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I6ed7862fc5e86005ecea83fa2ceb489e737d66b2 Gerrit-PatchSet: 1 Gerrit-Project: Impala Gerrit-Branch: cdh5-trunk Gerrit-Owner: Michael Ho <[email protected]> Gerrit-Reviewer: Michael Ho <[email protected]> Gerrit-Reviewer: Mostafa Mokhtar <[email protected]> Gerrit-Reviewer: Tim Armstrong <[email protected]> Gerrit-HasComments: Yes
