On Mar 1, 2014, at 6:54 PM, Ankesh Anand wrote:
> Although I gave my introduction on IRC before, but here it goes again. I am
> Ankesh Anand, an undergrad from Indian Institute of Technology, Kharagpur and
> love hacking on Python and Javascript.
Hi Ankesh! Refresher introductions are appreciated when there are so many new
people joining in the discussions, especially when it's your first post to the
mailing list. Not all our devs listen in on the IRC channel.
> So the benchmark performance database seemed particularly interesting to me,
> moreover the BRL-CAD benchmark suite has been there for quite some time now,
> and a database and visualization framework on top of it would be really
> useful and exciting to work on.
>
> There was a GSoc project earlier on this as well, I plan to extend that
> project and take the Performance Database and Visualization Framework to
> deployment.
This sounds fantastic. I like to see projects taken to completion (which we
rarely seem to do sometimes, so leveraging the work or extending a prior GSoC
project is a great to see.
> - I ran the benchmark on a couple of systems, and went through the
> documentation to make sense of the results, how the benchmark suite works,
> and how a comparison between different machines is done.
I suggest keeping notes on what specifically you found clear or confusing, what
questions you had, what information was or was not useful, etc. You only a
newbie once.
> - I studied the earlier GSoc project and read the dev-logs and reports. The
> parser and methods to send the log files to server could be re-used.
What cannot be reused?
> - Went through a couple of visualization frameworks to decide what could be
> best for the project. I think d3.js is probably the best library out there
> for interactive data-visualization in the browser. I have personally worked
> with it as well, although I am open to other suggestions as well.
There is no "how" requirement defined for your project idea. That is, we don't
have any specific infrastructure/frameworks that you'd have to use beyond what
helps you leverage progress made in the prior gsoc development project.
> - What is the vision behind the website? It's important for me to understand
> what the community wants so that I can have a clear perspective to move
> forward.
Not enough time to craft a proper "vision statement", but the goal is an
interactive data visualization site where people can view historical
performance metrics in a variety of ways, submit/upload new performance metrics
into the database, and perform queries on the data.
> - What are the different form of analytics that we want? A couple of obvious
> ones that come to mind are comparing the relative performance of processors
> by their VGR ratings, and their performance corresponding to each test image.
I don't think we know this in advance adequately, so the emphasis should be on
infrastructure that is completely flexible and extensible for coming up with
new comparisons easily. You can certainly propose a few with your project to
begin with and/or leverage integrating the ones from the earlier effort.
> - Should we categorize processors? A paper[1] I read measured performances of
> processors across the categories of Desktop, Workstation and Server.
We should categorize absolutely everything we have data for. Looking at a few
sample benchmark log files will show you this variety. Suryajith's work here
was pretty solid, so start there.
Have you gotten his code [1] up and running?
[1] https://bitbucket.org/suryajith/benchmark/src
> - Are there any other deliverables expected from this project other than
> those mentioned on the project page[2]? I remember Morrison mentioning the
> project description needs an update.
The update would be to refer to Suryajith's progress, so any continuation picks
up where he left off and leverages as much as possible.
Cheers!
Sean
------------------------------------------------------------------------------
Subversion Kills Productivity. Get off Subversion & Make the Move to Perforce.
With Perforce, you get hassle-free workflows. Merge that actually works.
Faster operations. Version large binaries. Built-in WAN optimization and the
freedom to use Git, Perforce or both. Make the move to Perforce.
http://pubads.g.doubleclick.net/gampad/clk?id=122218951&iu=/4140/ostg.clktrk
_______________________________________________
BRL-CAD Developer mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/brlcad-devel
