Announcement

Collapse
No announcement yet.

RFC: A Preview Of The Phoronix Graphs With Iveland

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    Originally posted by Michael View Post
    Now there's error bars on the static version of Iveland's pts_LineGraph too:
    This is what I thought. Perfect.

    Comment


    • #17
      Originally posted by Michael View Post
      The standard error is already shown as "SE" under each identifier.
      Just spotted it, sorry my mistake.
      However, I don't believe that they are symmetric. :P
      Well, they can be if you run 2 tests only and use linear interpolation. But as far as I know 3 tests are run as default.

      Should be like this: SE: +3.46 -1.18

      Am I missing something?

      Comment


      • #18
        Originally posted by HokTar View Post
        Just spotted it, sorry my mistake.
        However, I don't believe that they are symmetric. :P
        Well, they can be if you run 2 tests only and use linear interpolation. But as far as I know 3 tests are run as default.

        Should be like this: SE: +3.46 -1.18

        Am I missing something?
        Right now the standard error (formula used is below) is calculated from the total number of runs that was done for each individual test.

        Michael Larabel
        http://www.michaellarabel.com/

        Comment


        • #19
          Originally posted by Michael View Post
          Right now the standard error (formula used is below) is calculated from the total number of runs that was done for each individual test.


          Is there a specific reason to use this for computer tests? I think I would just go with the min/max deviation but I'm a mechanical engineer.
          I'm just curious.

          Comment


          • #20
            Originally posted by HokTar View Post
            Is there a specific reason to use this for computer tests? I think I would just go with the min/max deviation but I'm a mechanical engineer.
            I'm just curious.
            From what I have read, this would be the correct formula to use as far as I know.
            Michael Larabel
            http://www.michaellarabel.com/

            Comment


            • #21
              Originally posted by Michael View Post
              From what I have read, this would be the correct formula to use as far as I know.
              You misunderstood. The formula is most probably correct. My question was that why do you think that standard deviation is more interesting/useful than the extrema of the deviation?

              For example I have a machine which manufactures screws.
              In general I want to know when the tool in this machine gets worn out so I measure the screws at an appropriate point and I calculate the standard deviation from the size given on the technical drawing.
              But if for some unknown reason the tool breaks (the standard deviation is still in acceptable range) I would probably start looking at the min/max deviation which may or may not shed some light one what circumstances caused the problem.

              Comment


              • #22
                I suspect Michael went with standard deviation rather than min/max because so many people have asked for standard deviation

                I agree that min/max might actually make more sense but I don't think anyone has asked for it before.

                Comment


                • #23
                  Yeah I went with standard deviation over having the code already written for the calculations in it since it's used elsewhere in PTS, it's simple / easy to understand, and that's what the people have been asking for.
                  Michael Larabel
                  http://www.michaellarabel.com/

                  Comment


                  • #24
                    Yeah, definitely better than before.


                    The logo looks rather bad though:


                    - inconsistent

                    The last vertical line is blurred, but the rest is aliased and blocky

                    - ugly

                    The aliased and blocky
                    The right-bar blur is a bit too strong as well.

                    Comment


                    • #25
                      By dynamic rendering, do you mean client side using js? I hope there'll also be a static view for when js is disabled.

                      Both types look good in any case - much better than the old ones.

                      Comment


                      • #26
                        Originally posted by Otus View Post
                        By dynamic rendering, do you mean client side using js? I hope there'll also be a static view for when js is disabled.

                        Both types look good in any case - much better than the old ones.
                        I plan to have a variety of paths including JS, pure SVG, and PNG... Though when thinking about it yesterday, I may drop the PNG support altogether as all browsers should be fine besides IE with SVG.... Or may drop PNG after PTS 3.2.
                        Michael Larabel
                        http://www.michaellarabel.com/

                        Comment


                        • #27
                          Originally posted by curaga View Post
                          Yeah, definitely better than before.


                          The logo looks rather bad though:


                          - inconsistent

                          The last vertical line is blurred, but the rest is aliased and blocky

                          - ugly

                          The aliased and blocky
                          The right-bar blur is a bit too strong as well.
                          What right bar blur do you refer to?
                          Michael Larabel
                          http://www.michaellarabel.com/

                          Comment


                          • #28

                            Comment


                            • #29
                              Originally posted by curaga View Post
                              Oh with the logo, yeah. Right now the image is being automatically scaled to size, but will end up making it one set size.
                              Michael Larabel
                              http://www.michaellarabel.com/

                              Comment


                              • #30
                                Here's where we're at as of Sunday morning:



                                (From a graph for the Core i7 970 review later in the week.)
                                Michael Larabel
                                http://www.michaellarabel.com/

                                Comment

                                Working...
                                X