[
https://issues.apache.org/jira/browse/IMPALA-11110?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17509985#comment-17509985
]
ASF subversion and git services commented on IMPALA-11110:
----------------------------------------------------------
Commit a5a9b1e3f9310f42785fb2d0459b80571e9fbb80 in impala's branch
refs/heads/master from Joe McDonnell
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=a5a9b1e ]
IMPALA-11110: Switch debug builds to use -Og
GCC's -Og applies optimizations that are compatible with
being debuggable. It is similar to -O1 and results
in a modest speed-up. This modifies the default debug
build to use -Og, so it is now more akin to a fastdebug
mode.
Even though -Og is intended to preserve debuggability,
optimization always impacts debuggability and -Og is
no exception. To enable the old behavior, this adds
a DEBUG_NOOPT build mode that retains the old
non-optimized behavior. Using the -debug_noopt flag
with buildall.sh enables this behavior.
Change-Id: Ie06c149c8181c90572b8668bd01dfd26c0a5971e
Reviewed-on: http://gerrit.cloudera.org:8080/18200
Tested-by: Impala Public Jenkins <[email protected]>
Reviewed-by: Laszlo Gaal (Cloudera) <[email protected]>
Reviewed-by: Csaba Ringhofer <[email protected]>
> Enable basic optimizations for debug builds
> -------------------------------------------
>
> Key: IMPALA-11110
> URL: https://issues.apache.org/jira/browse/IMPALA-11110
> Project: IMPALA
> Issue Type: Improvement
> Components: Infrastructure
> Affects Versions: Impala 4.1.0
> Reporter: Joe McDonnell
> Assignee: Joe McDonnell
> Priority: Major
>
> Currently, Impala debug builds use no optimizations. Some optimizations can
> interfere with debuggability, and the debug builds prioritize debuggability.
> However, running without optimizations means the binary is much slower.
> GCC has a debug-friendly optimization level -Og, which enables optimizations
> that do not interfere with debugging. This may provide a way to get the
> debuggability needed while having faster runtimes. See
> [https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html]
> In testing this, the binaries are definitely faster. Core test runtimes are
> reduced by about 25 minutes. Exhaustive test runtimes are reduced by over an
> hour.
> The extra optimization enables compiler passes that produce more debugging
> information. This means that the binaries have more debug info, which make
> them larger. This extra information is likely related to variable tracking
> (i.e. fvar-tracking-assignments), which is recommended for maximum
> debuggability (but may not be strictly necessary for how we use it). See
> [https://gcc.gnu.org/onlinedocs/gcc/Debugging-Options.html]
> {noformat}
> Binary size for impalad (unstripped)
> Without -Og: 519345880
> With -Og: 709527368
> Binary size for impalad (stripped)
> Without -Og: 164500024
> With -Og: 105873184{noformat}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
