wgtmac commented on code in PR #39818:
URL: https://github.com/apache/arrow/pull/39818#discussion_r1468728673
##########
cpp/src/parquet/column_reader_benchmark.cc:
##########
@@ -219,5 +251,19 @@ BENCHMARK(RecordReaderReadRecords)
->Args({2, 1000, true})
->Args({2, 1000, false});
+BENCHMARK(RecordReaderReadAndSkipRecords)
+ ->ArgNames({"Repetition", "BatchSize"})
Review Comment:
BTW, should the batch size be power of two which is a common case?
##########
cpp/src/parquet/column_reader.cc:
##########
@@ -70,6 +70,8 @@ namespace {
// The minimum number of repetition/definition levels to decode at a time, for
// better vectorized performance when doing many smaller record reads
constexpr int64_t kMinLevelBatchSize = 1024;
+// The max buffer size of validility bitmap for skipping buffered levels.
+constexpr int64_t kMaxSkipLevelBufferSize = 128;
Review Comment:
Did you get a chance to benchmark different value for
kMaxSkipLevelBufferSize?
##########
cpp/src/parquet/column_reader_benchmark.cc:
##########
@@ -219,5 +251,19 @@ BENCHMARK(RecordReaderReadRecords)
->Args({2, 1000, true})
->Args({2, 1000, false});
+BENCHMARK(RecordReaderReadAndSkipRecords)
+ ->ArgNames({"Repetition", "BatchSize"})
Review Comment:
nit: add name for levels_per_page?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
