| <html> |
| <head> |
| <meta |
| http-equiv="Content-Type" |
| content="text/html; charset=iso-8859-1"> |
| <style type="text/css"> |
| p, table, td, th { |
| font-family: arial, helvetica, geneva; |
| font-size: 10pt |
| } |
| |
| pre { |
| font-family: "Courier New", Courier, mono; |
| font-size: 10pt |
| } |
| |
| h2 { |
| font-family: arial, helvetica, geneva; |
| font-size: 18pt; |
| font-weight: bold; |
| line-height: 14px |
| } |
| |
| code { |
| font-family: "Courier New", Courier, mono; |
| font-size: 10pt |
| } |
| |
| sup { |
| font-family: arial, helvetica, geneva; |
| font-size: 10px |
| } |
| |
| h3 { |
| font-family: arial, helvetica, geneva; |
| font-size: 14pt; |
| font-weight: bold |
| } |
| |
| li { |
| font-family: arial, helvetica, geneva; |
| font-size: 10pt |
| } |
| |
| h1 { |
| font-family: arial, helvetica, geneva; |
| font-size: 28px; |
| font-weight: bold |
| } |
| |
| body { |
| font-family: arial, helvetica, geneva; |
| font-size: 10pt; |
| clip: rect(); |
| margin-top: 5mm; |
| margin-left: 3mm |
| } |
| |
| .indextop { |
| font-size: x-large;; |
| font-family: Verdana, Arial, Helvetica, sans-serif; |
| font-weight: bold |
| } |
| |
| .indexsub { |
| font-size: xx-small;; |
| font-family: Arial, Helvetica, sans-serif; |
| color: #8080FF |
| } |
| </style> |
| <body> |
| |
| <h1>Table of contents</h1> |
| <ul> |
| <li><a href="#scales">Scales</a> |
| <ul> |
| <li><a href="#percent">Percentage</a></li> |
| <li><a href="#time">Time linear</a></li> |
| <li><a href="#time">Time logarithmic</a></li> |
| </ul></li> |
| <li><a href="#tips">Tips for time scales</a></li> |
| </ul> |
| <h1 id="scales">Scales></h1> |
| Since 3.5, it's possible to see results in fingerprints with three different kinds of scale. |
| <h3 id="percent">Percentage scale</h3> |
| <p>The X axis represents percentage of the variation vs. the given baseline</p> |
| <p> |
| This is the way fingerprints were displayed since the beginning: <img src="images/percentage.png"> |
| </p> |
| <ul> |
| <li>Red bar means a regression, even if it's less than the 10% threshold.</li> |
| <li>Green bar means an improvement</li> |
| <li>Gray bar means an 'explained' regression.</li> |
| </ul> |
| <h3 id="time">Linear and logarithmic time scales</h3> |
| <p> |
| For these scales, the X axis represents the duration time of the test.<br> The colors meanings are the same than for the percentage scale.<br> These kind of graphs give a better idea of |
| time duration for each test. |
| </p> |
| <p>Typically use linear scale if you want to see the tests relativeness for all the component tests:</p> |
| <p> |
| <img src="images/linear.png"> |
| <p>But the logarithmic scale is more appropriate when there are a strong duration differences between tests, hence makes short duration tests easier to survey:</p> |
| <p> |
| <img src="images/log.png"> |
| <p> |
| Each test have two bars: the former is white and shows the baseline result, the latter is colored (red, green or gray) and shows the current build result.<br> The variation between the |
| baseline and the build is displayed as a percentage on top of both bars. |
| </p> |
| <h1 id="tips">Tips for time scales</h1> |
| <p>Tips are almost the same for linear and logarithmic scales:</p> |
| <table style="border: 0" > |
| <tr> |
| <td valign="top"><img src="../images/light.gif"></td> |
| <td><b>Flying over a bar displays its time value</b>:</td> |
| </tr> |
| <tr> |
| <td></td> |
| <td><img src="images/help_time_baseline.png"></td> |
| </tr> |
| <tr> |
| <td></td> |
| <td><img src="images/help_time_current.png"></td> |
| </tr> |
| <tr> |
| <td><br></td> |
| </tr> |
| <tr> |
| <td valign="top"><img src="../images/light.gif"></td> |
| <td><b>For linear scale only, when the error on the time result is noticeable, then the measurement uncertainty is shown in yellow at the end of the bar</b>:</td> |
| </tr> |
| <tr> |
| <td></td> |
| <td><img src="images/help_time_error.png"></td> |
| </tr> |
| <tr> |
| <td><br></td> |
| </tr> |
| <tr> |
| <td valign="top"><img src="../images/light.gif"></td> |
| <td><b>A performance regression may sometimes have a known good reason</b>.<br> In this case, the current build bar is grayed and flying over it also shows the given explanation:</td> |
| </tr> |
| <tr> |
| <td></td> |
| <td><img src="images/help_explained_regression.png"></td> |
| </tr> |
| <tr> |
| <td><br></td> |
| </tr> |
| <tr> |
| <td valign="top"><img src="../images/light.gif"></td> |
| <td><b>Test result may have big error which can make the test result not fully reliable</b>.<br> In this case, a warning icon is shown after the variation value and flying over it |
| gives the offending error value:</td> |
| </tr> |
| <tr> |
| <td></td> |
| <td><img src="images/help_error_warning.png"></td> |
| </tr> |
| <tr> |
| <td><br></td> |
| </tr> |
| <tr> |
| <td valign="top"><img src="../images/light.gif"></td> |
| <td><b>Test may have no result for the used baseline, hence the first available build is used as a reference</b>.<br> In this case, a warning icon is shown after the scenario title and |
| flying over it gives the build ID used to compute the variation:</td> |
| </tr> |
| <tr> |
| <td></td> |
| <td><img src="images/help_no_baseline.png"></td> |
| </tr> |
| </table> |
| </body> |
| </html> |
