[PBDS] Rejuvenation

classic Classic list List threaded Threaded
12 messages Options
Reply | Threaded
Open this post in threaded view
|

[PBDS] Rejuvenation

Tadeus Prastowo
Hi Alex,

As Jonathan suggested in
https://gcc.gnu.org/ml/libstdc++/2019-07/msg00066.html, what do you
think if we start with: "regenerate the graphs in the documentation
that show poor binary heap performance, as mentioned in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62045"?

--
Best regards,
Tadeus
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Alexander Kulkov
Hi, Tadeus!

I'll take a look in everything Jonathan mentioned and reply in few days.

Regards,
Alex
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Jonathan Wakely-4
In reply to this post by Tadeus Prastowo
On Wed, 24 Jul 2019 at 12:58, Tadeus Prastowo <[hidden email]> wrote:
>
> Hi Alex,
>
> As Jonathan suggested in
> https://gcc.gnu.org/ml/libstdc++/2019-07/msg00066.html, what do you
> think if we start with: "regenerate the graphs in the documentation
> that show poor binary heap performance, as mentioned in
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62045"?

Sounds good to me. The data in those graphs are produced by the tests
in testsuite/performance/ext/pb_ds/*.cc but I'm not sure how the XML
output of those tests is turned into the PDF and PNG files that are
included in doc/xml/manual/test_policy_data_structures.xml (and it
doesn't help that the filenames listed in the docs are wrong, e.g. the
docs say:

      <para>
        It uses the test file:
        <filename>performance/ext/pb_ds/priority_queue_random_int_push_timing.cc</filename>
      </para>

but actually that should be performance/ext/pb_ds/priority_queue_int_push.cc

You can find the right test file names by grepping in the
testsuite/performance/ext/pb_ds/ directory for the name in the docs,
because the test files contain the same name as used in the docs (it
just doesn't match the actual filename):

master% libstdc++-v3$ git grep priority_queue_random_int_push_timing
testsuite/performance/ext/pb_ds/
testsuite/performance/ext/pb_ds/priority_queue_int_push.cc: * @file
priority_queue_random_int_push_timing_test.cpp
testsuite/performance/ext/pb_ds/priority_queue_int_push.cc:  cerr <<
"usage: priority_queue_random_int_push_timing_test <vn> <vs> <vm>" <<
master% libstdc++-v3$ git grep
priority_queue_random_int_push_pop_timing
testsuite/performance/ext/pb_ds/
testsuite/performance/ext/pb_ds/priority_queue_int_push_pop.cc: *
@file priority_queue_random_int_push_pop_timing_test.cpp
testsuite/performance/ext/pb_ds/priority_queue_int_push_pop.cc:
std::cerr << "usage: priority_queue_random_int_push_pop_timing_test
<vn> <vs> <vm>" <<
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Jonathan Wakely-3
On 24/07/19 13:18 +0100, Jonathan Wakely wrote:

>On Wed, 24 Jul 2019 at 12:58, Tadeus Prastowo <[hidden email]> wrote:
>>
>> Hi Alex,
>>
>> As Jonathan suggested in
>> https://gcc.gnu.org/ml/libstdc++/2019-07/msg00066.html, what do you
>> think if we start with: "regenerate the graphs in the documentation
>> that show poor binary heap performance, as mentioned in
>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62045"?
>
>Sounds good to me. The data in those graphs are produced by the tests
>in testsuite/performance/ext/pb_ds/*.cc but I'm not sure how the XML
>output of those tests is turned into the PDF and PNG files that are
>included in doc/xml/manual/test_policy_data_structures.xml (and it
>doesn't help that the filenames listed in the docs are wrong, e.g. the
>docs say:
>
>      <para>
>        It uses the test file:
>        <filename>performance/ext/pb_ds/priority_queue_random_int_push_timing.cc</filename>
>      </para>
>
>but actually that should be performance/ext/pb_ds/priority_queue_int_push.cc
>
>You can find the right test file names by grepping in the
>testsuite/performance/ext/pb_ds/ directory for the name in the docs,
>because the test files contain the same name as used in the docs (it
>just doesn't match the actual filename):
>
>master% libstdc++-v3$ git grep priority_queue_random_int_push_timing
>testsuite/performance/ext/pb_ds/
>testsuite/performance/ext/pb_ds/priority_queue_int_push.cc: * @file
>priority_queue_random_int_push_timing_test.cpp
>testsuite/performance/ext/pb_ds/priority_queue_int_push.cc:  cerr <<
>"usage: priority_queue_random_int_push_timing_test <vn> <vs> <vm>" <<
>master% libstdc++-v3$ git grep
>priority_queue_random_int_push_pop_timing
>testsuite/performance/ext/pb_ds/
>testsuite/performance/ext/pb_ds/priority_queue_int_push_pop.cc: *
>@file priority_queue_random_int_push_pop_timing_test.cpp
>testsuite/performance/ext/pb_ds/priority_queue_int_push_pop.cc:
>std::cerr << "usage: priority_queue_random_int_push_pop_timing_test
><vn> <vs> <vm>" <<

That would be easier to read without gmail's auto-wrapping:

master% libstdc++-v3$ git grep priority_queue_random_int_push_timing testsuite/performance/ext/pb_ds/
testsuite/performance/ext/pb_ds/priority_queue_int_push.cc: * @file priority_queue_random_int_push_timing_test.cpp
testsuite/performance/ext/pb_ds/priority_queue_int_push.cc:  cerr << "usage: priority_queue_random_int_push_timing_test <vn> <vs> <vm>" <<
master% libstdc++-v3$ git grep priority_queue_random_int_push_pop_timing testsuite/performance/ext/pb_ds/
testsuite/performance/ext/pb_ds/priority_queue_int_push_pop.cc: * @file priority_queue_random_int_push_pop_timing_test.cpp
testsuite/performance/ext/pb_ds/priority_queue_int_push_pop.cc:  std::cerr << "usage: priority_queue_random_int_push_pop_timing_test <vn> <vs> <vm>" <<



Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Tadeus Prastowo
In reply to this post by Jonathan Wakely-4
On Wed, Jul 24, 2019 at 2:18 PM Jonathan Wakely <[hidden email]> wrote:

>
> On Wed, 24 Jul 2019 at 12:58, Tadeus Prastowo <[hidden email]> wrote:
> >
> > Hi Alex,
> >
> > As Jonathan suggested in
> > https://gcc.gnu.org/ml/libstdc++/2019-07/msg00066.html, what do you
> > think if we start with: "regenerate the graphs in the documentation
> > that show poor binary heap performance, as mentioned in
> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62045"?
>
> Sounds good to me. The data in those graphs are produced by the tests
> in testsuite/performance/ext/pb_ds/*.cc but I'm not sure how the XML
> output of those tests is turned into the PDF and PNG files that are
> included in doc/xml/manual/test_policy_data_structures.xml (and it
> doesn't help that the filenames listed in the docs are wrong, e.g. the
> docs say:
>
>       <para>
>         It uses the test file:
>         <filename>performance/ext/pb_ds/priority_queue_random_int_push_timing.cc</filename>
>       </para>
>
> but actually that should be performance/ext/pb_ds/priority_queue_int_push.cc
>
> You can find the right test file names by grepping in the
> testsuite/performance/ext/pb_ds/ directory for the name in the docs,
> because the test files contain the same name as used in the docs (it
> just doesn't match the actual filename):
>
> master% libstdc++-v3$ git grep priority_queue_random_int_push_timing
> testsuite/performance/ext/pb_ds/
> testsuite/performance/ext/pb_ds/priority_queue_int_push.cc: * @file
> priority_queue_random_int_push_timing_test.cpp
> testsuite/performance/ext/pb_ds/priority_queue_int_push.cc:  cerr <<
> "usage: priority_queue_random_int_push_timing_test <vn> <vs> <vm>" <<
> master% libstdc++-v3$ git grep
> priority_queue_random_int_push_pop_timing
> testsuite/performance/ext/pb_ds/
> testsuite/performance/ext/pb_ds/priority_queue_int_push_pop.cc: *
> @file priority_queue_random_int_push_pop_timing_test.cpp
> testsuite/performance/ext/pb_ds/priority_queue_int_push_pop.cc:
> std::cerr << "usage: priority_queue_random_int_push_pop_timing_test
> <vn> <vs> <vm>" <<

Thank you very much for the lead.  I agree that rejuvenating the PBDS
doc is a good first step.  I will wait for Alex, and then we can
propose the patchset.

--
Best regards,
Tadeus
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Tadeus Prastowo
In reply to this post by Alexander Kulkov
Hi Alex,

On Wed, Jul 24, 2019 at 2:17 PM Alexander Kulkov <[hidden email]> wrote:
>
> Hi, Tadeus!
>
> I'll take a look in everything Jonathan mentioned and reply in few days.

Sure, take your time.

> Regards,
> Alex

--
Best regards,
Tadeus
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Alexander Kulkov
Hi! It's just a gentle self-ping and reminder to myself that I'm going to
answer till the end of the week :)

Regards, Alex

ср, 24 июл. 2019 г., 15:25 Tadeus Prastowo <[hidden email]>:

> Hi Alex,
>
> On Wed, Jul 24, 2019 at 2:17 PM Alexander Kulkov <[hidden email]>
> wrote:
> >
> > Hi, Tadeus!
> >
> > I'll take a look in everything Jonathan mentioned and reply in few days.
>
> Sure, take your time.
>
> > Regards,
> > Alex
>
> --
> Best regards,
> Tadeus
>
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Alexander Kulkov
Okay, it may take me forever to properly prepare since I currently have
much of stuff going on in real life.

So, probably the best thing for me to do now is just start doing at least
something and hopefully I will catch things on fly.
That being said, I'd like to start with two questions:
- Is there any way to contact you outside e-mail threads? I mean, like for
instant messaging. I may assume that e-mail is the main means
of communication here, but everything might get very slow with it.
On the other hand I might have *lots* of stupid technical questions which I
may be too embarrassed to raise publicly :-)
So, if that is possible I'd prefer to keep e-mail for some major stuff and
contact in some other way for details.
- Do you have any particular suggestions on what should we do right now? I
still feel a bit lost..

Regards,
Alex
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Tadeus Prastowo
On Sun, Aug 11, 2019 at 11:07 PM Alexander Kulkov <[hidden email]> wrote:
>
> Okay, it may take me forever to properly prepare since I currently have much of stuff going on in real life.

And, I will have a summer break.  Don't worry.

> So, probably the best thing for me to do now is just start doing at least something and hopefully I will catch things on fly.

Okay, then I will start it after summer.

> That being said, I'd like to start with two questions:
> - Is there any way to contact you outside e-mail threads? I mean, like for instant messaging. I may assume that e-mail is the main means of communication here, but everything might get very slow with it.

Unfortunately, I rarely use instant messaging.  Let's start with the
real work here, and then, if things really get slow, we can
temporarily move to, for example, an IRC channel.

> On the other hand I might have lots of stupid technical questions which I may be too embarrassed to raise publicly :-)
> So, if that is possible I'd prefer to keep e-mail for some major stuff and contact in some other way for details.

Just send a private e-mail?

> - Do you have any particular suggestions on what should we do right now? I still feel a bit lost..

So, in September I will propose a patch to fix the documentation
problem that Jonathan suggested last time.  As I told Jonathan,
currently I could only contribute from the technical side.  I need you
to contribute from the user side: performance testing, feature
requests, and the like.

> Regards,
> Alex

--
Best regards,
Tadeus
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Alexander Kulkov
Hi there!

It's late September so I wanted to check up on how are things going here

пн, 12 авг. 2019 г. в 10:06, Tadeus Prastowo <[hidden email]>:

> On Sun, Aug 11, 2019 at 11:07 PM Alexander Kulkov <[hidden email]>
> wrote:
> >
> > Okay, it may take me forever to properly prepare since I currently have
> much of stuff going on in real life.
>
> And, I will have a summer break.  Don't worry.
>
> > So, probably the best thing for me to do now is just start doing at
> least something and hopefully I will catch things on fly.
>
> Okay, then I will start it after summer.
>
> > That being said, I'd like to start with two questions:
> > - Is there any way to contact you outside e-mail threads? I mean, like
> for instant messaging. I may assume that e-mail is the main means of
> communication here, but everything might get very slow with it.
>
> Unfortunately, I rarely use instant messaging.  Let's start with the
> real work here, and then, if things really get slow, we can
> temporarily move to, for example, an IRC channel.
>
> > On the other hand I might have lots of stupid technical questions which
> I may be too embarrassed to raise publicly :-)
> > So, if that is possible I'd prefer to keep e-mail for some major stuff
> and contact in some other way for details.
>
> Just send a private e-mail?
>
> > - Do you have any particular suggestions on what should we do right now?
> I still feel a bit lost..
>
> So, in September I will propose a patch to fix the documentation
> problem that Jonathan suggested last time.  As I told Jonathan,
> currently I could only contribute from the technical side.  I need you
> to contribute from the user side: performance testing, feature
> requests, and the like.
>
> > Regards,
> > Alex
>
> --
> Best regards,
> Tadeus
>
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Tadeus Prastowo
> Hi there!

Hi Alex!

> It's late September so I wanted to check up on how are things going here

Sorry that currently I am tied to completing my dissertation.  So,
please ping me again in early November.  Thanks.

--
Best regards,
Tadeus
Reply | Threaded
Open this post in threaded view
|

Re: [PBDS] Rejuvenation

Alexander Kulkov
Okay, good luck with your dissertation!

пт, 27 сент. 2019 г. в 00:40, Tadeus Prastowo <[hidden email]>:

> > Hi there!
>
> Hi Alex!
>
> > It's late September so I wanted to check up on how are things going here
>
> Sorry that currently I am tied to completing my dissertation.  So,
> please ping me again in early November.  Thanks.
>
> --
> Best regards,
> Tadeus
>