[GitHub] incubator-quickstep pull request #28: Add option to enable Google Profiler.

[navsan]

Github user navsan commented on a diff in the pull request:

    https://github.com/apache/incubator-quickstep/pull/28#discussion_r66740805
  
    --- Diff: cli/QuickstepCli.cpp ---
    @@ -430,6 +445,13 @@ int main(int argc, char* argv[]) {
             reset_parser = true;
             break;
           }
    +#ifdef QUICKSTEP_ENABLE_GOOGLE_PROFILER
    +      // Profile only if profile_file_name flag is set
    +      if (!started_profiling && 
!quickstep::FLAGS_profile_file_name.empty()) {
    --- End diff --
    
    Good question. It’s because, unless you’ve preloaded the buffer pool 
(not always a good idea), the first run of the query results in disk I/O and 
other overhead that significantly skews the profiling results. It’s the same 
reason we don’t include the first run time in our benchmarking: when 
profiling query execution, it makes more sense to get numbers using a warm 
buffer pool and warm caches. This is not *always* the right thing to do: it’s 
obviously wrong for profiling the TextScan operator. In those cases, you might 
want to put in your own Profiler probes (just follow the start/stop pattern 
used in this PR) or just run quickstep with the CPUPROFILE environment variable 
set (as per gperftools documentation) to get the full profile for the entire 
execution. 
    
    I’ve provided the explanation in the flag definition, CMake comments as 
well as in the PR header comment. 
    
    
    > On Jun 12, 2016, at 22:49, Zuyu ZHANG <notificati...@github.com> wrote:
    > 
    > In cli/QuickstepCli.cpp 
<https://github.com/apache/incubator-quickstep/pull/28#discussion_r66737198>:
    > 
    > > @@ -430,6 +445,13 @@ int main(int argc, char* argv[]) {
    > >          reset_parser = true;
    > >          break;
    > >        }
    > > +#ifdef QUICKSTEP_ENABLE_GOOGLE_PROFILER
    > > +      // Profile only if profile_file_name flag is set
    > > +      if (!started_profiling && 
!quickstep::FLAGS_profile_file_name.empty()) {
    > Now we are using the singleton design pattern.
    > 
    > I still have a question: why we start profiling after one query finishes 
execution, instead of before any query starts? Thanks!
    > 
    > —
    > You are receiving this because you were mentioned.
    > Reply to this email directly, view it on GitHub 
<https://github.com/apache/incubator-quickstep/pull/28/files/00955f19652fe54468377e932365d77a13cc69c8#r66737198>,
 or mute the thread 
<https://github.com/notifications/unsubscribe/ACZB62yfZt1S-zIV8u5HZq-Dryl6gVHLks5qLNNHgaJpZM4Iz4mC>.
    > 
    



---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---