Michael Ho has posted comments on this change. Change subject: IMPALA-2548: Codegen Tuple::MaterializeExprs() and use in TopN node ......................................................................
Patch Set 8: (6 comments) http://gerrit.cloudera.org:8080/#/c/1901/8/be/src/codegen/codegen-anyval.cc File be/src/codegen/codegen-anyval.cc: Line 585: WriteToSlot Have you considered refactoring the code so there is more sharing with TextConverter::CodegenWriteSlot() ? It seems that both of them are trying to achieve similar things. Line 597: // Create new block after conditional blocks if necessary : if (insert_before == NULL) insert_before = BasicBlock::Create(context, "end_write", fn); Just wondering why this statement isn't before the creation of the two basic blocks "null" and "non_null" which use "insert_before" ? http://gerrit.cloudera.org:8080/#/c/1901/8/be/src/codegen/llvm-codegen.h File be/src/codegen/llvm-codegen.h: Line 428: CodegenMemset Not related to this change but just wondering why we never need one for memcmp ? It appears that @_ZNK10impala_udf9StringValeqERKS0_ in the IR is calling the non-intrinsic memcmp(). http://gerrit.cloudera.org:8080/#/c/1901/8/be/src/runtime/tuple.cc File be/src/runtime/tuple.cc: Line 331: >GetPtrType( Can you please update the patch to converge on using only one of GetPtrType() or getPointerTo() ? http://gerrit.cloudera.org:8080/#/c/1901/8/be/src/runtime/tuple.h File be/src/runtime/tuple.h: Line 148: MemPool* pool, Can we drop this parameter and update CodegenMaterializeExprs() to generate different signature for the two versions of the functions ? In this way, at least llvm will flag an error if the call sites replacement screws up. Line 164: MemPool* pool, Can we drop this parameter ? -- To view, visit http://gerrit.cloudera.org:8080/1901 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: Ib422a8d50303c21c6a228675157bf867e8619444 Gerrit-PatchSet: 8 Gerrit-Project: Impala Gerrit-Branch: cdh5-trunk Gerrit-Owner: Skye Wanderman-Milne <[email protected]> Gerrit-Reviewer: Michael Ho <[email protected]> Gerrit-Reviewer: Skye Wanderman-Milne <[email protected]> Gerrit-HasComments: Yes
