Loading ...
Sorry, an error occurred while loading the content.

Performance comparioson to linux version

Expand Messages
  • Martin Klingler
    I have tried to compare the Desktop Radiance performance (mainly of rpict) to the linux version. I ended up with a factor of about 1.6 for the Desktop Radiance
    Message 1 of 3 , Jul 22, 2000
    View Source
    • 0 Attachment
      I have tried to compare the Desktop Radiance performance (mainly of
      rpict) to the linux version. I ended up with a factor of about 1.6
      for the Desktop Radiance rpict.
      Does anybody has an explanation why Desktop Radiance rpict is that
      much faster?
      How can I tune the Linux rpict?

      Thank you
      Martin
    • Georg Mischler
      ... Did you make your tests on the same machine? The Microsoft compilers are heavily optimized for the Pentium architecture, especially the newest versions
      Message 2 of 3 , Jul 23, 2000
      View Source
      • 0 Attachment
        Martin Klingler wrote:

        > I have tried to compare the Desktop Radiance performance (mainly of
        > rpict) to the linux version. I ended up with a factor of about 1.6
        > for the Desktop Radiance rpict.
        > Does anybody has an explanation why Desktop Radiance rpict is that
        > much faster?
        > How can I tune the Linux rpict?

        Did you make your tests on the same machine?
        The Microsoft compilers are heavily optimized for the
        Pentium architecture, especially the newest versions
        that LBNL is probably using for Desktop Radiance.
        Gcc, while being a high quality compiler in most
        areas, is designed to give decent results on a
        variety of systems, and may not offer the same
        possibilities on all of them.

        It would be interesting to check the settings in the
        makeall script, and to experiment with some of the CPU
        specific optimization parameters that are probably
        ignored in the defaults from the distribution. If you
        find the right settings, then you should be able to
        improve the linux performance quite a lot. I have never
        tried that myself yet, but would be highly interested
        in your findings!


        Have fun!

        -schorsch

        --
        Georg Mischler -- simulations developer -- schorsch at schorsch.com
        +schorsch.com+ -- lighting design tools -- http://www.schorsch.com/
      • Charles Ehrlich
        Hi Martin, Thank you for independently confirming our findings of 60% speed improvements in Radiance for Windows versus standard Radiance 3.1.8. There was
        Message 3 of 3 , Aug 1, 2000
        View Source
        • 0 Attachment
           
          Hi Martin,

          Thank you for independently confirming our findings of 60% speed improvements in Radiance for Windows versus standard Radiance 3.1.8.  There was significant effort spent on a few optimizations of the code.  These optimizations are not Windows-specific, and in fact were implemented in the Unix environment for the SGI and Cray T3E.  The efficiencies of the Visual C++ compiler can also be credited for some of the speed improvement.

          One optimization that also provides considerable opportunities for feature enhancements involved providing a "hook" into the code for linking pre-compiled C-Code texture and pattern-mapping routines to supplement the interpreted Radiance .CAL routines that have been used exclusively in the past.  The sky brightness mapping routines, the noise function, and a few others have been converted to C.

          We plan to make the source code for this new version available for download under terms similar to the existing Radiance versions.  There are licensing issues to resolve, and there is a problem of multiple source code versions.   If you have a specific application in mind and/or you would like to help us merge these source code versions, then I could make arrangements to provide access to the new source code.  A non-disclosure agreement and a document assigning rights to your work back to LBNL might be necessary.  However, I can make no commitments until I hear what you would like to propose and I get the appropriate approvals.

          -Chas
           

          ===============

          Martin Klingler wrote:

          I have tried to compare the Desktop Radiance performance (mainly of
          rpict) to the linux version. I ended up with a factor of about 1.6
          for the Desktop Radiance rpict.
          Does anybody has an explanation why Desktop Radiance rpict is that
          much faster?
          How can I tune the Linux rpict?

          Thank you
          Martin
           

          To Post a message, send it to:   desktopradiance@...

          To Unsubscribe, send a blank message to: desktopradiance-unsubscribe@...

          To View Recent Messages, connect to:  http://www.eGroups.com/group/desktopradiance

        Your message has been successfully submitted and would be delivered to recipients shortly.