Riza Suminto has uploaded this change for review. (
http://gerrit.cloudera.org:8080/17980
Change subject: IMPALA-10984: Improve performance of TimestampValue::ToString
......................................................................
IMPALA-10984: Improve performance of TimestampValue::ToString
TimestampValue::ToString was implemented by concatenating
boost::gregorian::to_iso_extended_string and
boost::posix_time::to_simple_string using stringstream. This involves
multiple string allocations, copying, and might hit lock within
tcmalloc::CentralFreeList. FROM_UNIXTIME and CAST expression that
touches this function can be inefficient if the expression is being
evaluated for millions of rows.
This patch reimplement TimestampValue::ToString by supplying default
DateTimeFormatContext if no pattern was specified. "yyyy-MM-dd HH:mm:ss"
will be picked as the default format if the time_ component does not
have fractional seconds. Otherwise, "yyyy-MM-dd HH:mm:ss.SSSSSSSSS" will
be picked as the default format. The chosen DateTimeFormatContext then
passed to TimestampParser::Format along with date_ and time_ to be
formatted into the string representation.
This patch gives > 2X performance improvement for CAST timestamp to
string and FROM_UNIXTIME without a date-time pattern, as shown by the
following benchmark (modified from expr-benchmark), before and after the
patch.
Before the patch:
Machine Info: Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz
FromUnixCodegen: Function iters/ms 10%ile 50%ile 90%ile
10%ile 50%ile 90%ile
(relative) (relative) (relative)
---------------------------------------------------------------------------------------------------------
literal 37.4 38.6 39.1
1X 1X 1X
cast(now() as string) 2.2 2.28 2.31
0.0589X 0.0591X 0.0591X
from_unixtime(0,'yyyy-MM-dd HH:mm:ss') 5.94 6.18 6.23
0.159X 0.16X 0.159X
from_unixtime(0,'yyyy-MM-dd') 11.5 11.8 11.9
0.308X 0.305X 0.304X
from_unixtime(0) 2.26 2.35 2.39
0.0606X 0.061X 0.0612X
After the patch:
Machine Info: Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz
FromUnixCodegen: Function iters/ms 10%ile 50%ile 90%ile
10%ile 50%ile 90%ile
(relative) (relative) (relative)
---------------------------------------------------------------------------------------------------------
literal 37.2 37.5 38.1
1X 1X 1X
cast(now() as string) 5.65 5.65 5.67
0.152X 0.15X 0.149X
from_unixtime(0,'yyyy-MM-dd HH:mm:ss') 6.3 6.3 6.37
0.169X 0.168X 0.167X
from_unixtime(0,'yyyy-MM-dd') 11.7 11.9 12
0.315X 0.318X 0.314X
from_unixtime(0) 6.23 6.23 6.3
0.167X 0.166X 0.165X
The literal expression used as the baseline in this benchmark is
"cast('2012-01-01 09:10:11.123456789' as timestamp)".
This patch also updates numbers in expr-benchmark for
BenchmarkTimestampFunctions and tidy up expr-benchmark a bit to clear
its MemPool in between benchmark iteration so that it does not run out
of memory.
Testing:
- Pass core tests.
Change-Id: I4fcb4545d9c9a3fdb38c4db58bb4b1321a429d61
---
M be/src/benchmarks/expr-benchmark.cc
M be/src/exprs/cast-functions-ir.cc
M be/src/exprs/timestamp-functions-ir.cc
M be/src/runtime/datetime-simple-date-format-parser.cc
M be/src/runtime/datetime-simple-date-format-parser.h
M be/src/runtime/timestamp-value.cc
6 files changed, 119 insertions(+), 86 deletions(-)
git pull ssh://gerrit.cloudera.org:29418/Impala-ASF refs/changes/80/17980/1
--
To view, visit http://gerrit.cloudera.org:8080/17980
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: newchange
Gerrit-Change-Id: I4fcb4545d9c9a3fdb38c4db58bb4b1321a429d61
Gerrit-Change-Number: 17980
Gerrit-PatchSet: 1
Gerrit-Owner: Riza Suminto <riza.sumi...@cloudera.com>
