sysperfana/perfinvestigator/com.nokia.carbide.cpp.pi.doc.user/html/reference/analyzer/view_performance_counters.htm
changeset 5 844b047e260d
child 12 ae255c9aa552
equal deleted inserted replaced
4:615035072f7e 5:844b047e260d
       
     1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
       
     2 "http://www.w3.org/TR/html4/loose.dtd">
       
     3 <html>
       
     4 <head>
       
     5   <meta http-equiv="content-type" content="text/html; charset=UTF-8">
       
     6   <title>Performance Counters View</title>
       
     7   <link href="../../../book.css" rel="stylesheet" type="text/css">
       
     8 </head>
       
     9 
       
    10 <body>
       
    11 <h2>Performance Counters View</h2>
       
    12 
       
    13 <p>Use the <strong>Performance Counters</strong> view to display information on
       
    14 performance counter traces. By default, the Performance Counters view opens up
       
    15 as minimised on the PIAnalyzer main view. To view the graphs, click the restore
       
    16 button <img src="images\minimize_restore_graph.png" width="16" height="16"> or
       
    17 the maximize button <img src="images\maximize_graph.png" width="16"
       
    18 height="16"> on the Title bar.</p>
       
    19 
       
    20 <p>The Performance Counters displayed in the view include two user selected
       
    21 traces (A and B in Figure 1) and CPU cycles (C in Figure 1). A MIPS (Million
       
    22 Instructions Per Second) counter is also available if <b>Instructions
       
    23 executed</b> has been selected as one of the traces.</p>
       
    24 
       
    25 <p>You can select which traces are displayed in the graphs view by clicking the
       
    26 check box next to the trace name in the legend table below the graphs. When
       
    27 several traces are selected, a separate graph line is displayed for each trace,
       
    28 as shown in Figure 1. </p>
       
    29 
       
    30 <p>In the top left-hand corner of each graph line, the following information
       
    31 displayed:</p>
       
    32 <ul>
       
    33   <li>type of the trace,</li>
       
    34   <li>the sampling interval (1 sample per millisecond in Figure 1),</li>
       
    35   <li>the minimum and maximum values over the whole duration of the trace.
       
    36     (Note that the legend table lists the minimum and maximum values only for
       
    37     the selected timeframe.) </li>
       
    38 </ul>
       
    39 
       
    40 <p>Note that the scaling of the graphs changes according to the maximum values
       
    41 for each graph. You can see the vertical axis values for each graph on the left
       
    42 side of the graph.</p>
       
    43 
       
    44 <p>A legend table is displayed below the graphs.</p>
       
    45 
       
    46 <p align="center"><img
       
    47 src="images/performance_counters_view_all_graphs.png"></p>
       
    48 
       
    49 <p class="figure">Figure 1. Performance Counters view with all trace graphs
       
    50 displayed</p>
       
    51 
       
    52 <p></p>
       
    53 
       
    54 <p>Description of the table columns: </p>
       
    55 <ul>
       
    56   <li>The <b>Char</b> column contains an identifier for each trace.</li>
       
    57   <li>The <b>Name</b> column contains the type of performance counter that was
       
    58     traced.</li>
       
    59   <li>The <b>Average (1/ms)</b> column contains the average occurrence of the
       
    60     event that was traced, within the selected timeframe (for example, the
       
    61     average number of cache misses per millisecond, as shown in Figure 1).</li>
       
    62   <li>The <b>Sum</b> column contains the total number of event occurrences
       
    63     within the selected timeframe. (not included for MIPS)</li>
       
    64   <li>The <b>Min (1/ms)</b> column contains the minimum occurrence of the event
       
    65     that was traced, within the selected timeframe (for example, minimum number
       
    66     of data stalls per millisecond, as show in Figure 1).</li>
       
    67   <li>The <b>Max (1/ms)</b> column contains the maximum occurrence of the event
       
    68     that was traced, within the selected timeframe (for example, maximum number
       
    69     of data stalls per millisecond, as show in Figure 1).</li>
       
    70   <li>The <b>Per A</b> / <b>Per B</b> / <b>Per C</b> columns show the ratio of
       
    71     the values of the different traces (For example, Per A shows the number of
       
    72     events in the sum column of each trace divided by the sum of trace A). (not
       
    73     included for MIPS)</li>
       
    74 </ul>
       
    75 
       
    76 <p>Check the boxes at the right side of the table to select the traces for the
       
    77 graphs view.</p>
       
    78 
       
    79 <p>Columns may be reordered, resized, and sorted by clicking or dragging column
       
    80 headers.</p>
       
    81 
       
    82 <p>You can right-click on the table to open a context menu, where you can
       
    83 select to copy either selected lines or the entire table to clipboard; or to
       
    84 save the table to CSV format. </p>
       
    85 <br>
       
    86 
       
    87 
       
    88 <p><strong>MIPS counter and graph</strong></p>
       
    89 
       
    90 <p>The Million Instructions Per Second (MIPS) graph gives you a general idea of
       
    91 CPU speed in a measured use case. The basic use case for the MIPS analysis
       
    92 would be to check any low areas in the graph since they may reveal unoptimized
       
    93 features, for example, in terms of memory access (reads and writes). The higher
       
    94 the MIPS values, the better and the faster the CPU is in executing
       
    95 instructions.</p>
       
    96 
       
    97 <p>A MIPS (Million Instructions Per Millisecond) graph is displayed in the
       
    98 Performance Counters view, when <b>Instructions executed</b> has been selected
       
    99 as one of the counters. If <b>Instructions executed</b> has been selected, the
       
   100 dialog below is displayed during PIAnalyser file import:</p>
       
   101 
       
   102 <p align="center"><img src="images/MIPS_speed_dialog.png"></p>
       
   103 
       
   104 <p class="figure">Figure 1. MIPS Graph Generation</p>
       
   105 
       
   106 <p>The processor speed information is needed for calculating the data for the
       
   107 MIPS graph. The rest of the data for the calculation is included in the
       
   108 <b>Instructions executed</b> and <b>CPU cycles</b> counters.</p>
       
   109 
       
   110 <p>The MIPS counter information is displayed in the graphs view and the legend
       
   111 table together with the other performance counters. Note, however, that for
       
   112 MIPS only <b>Average (1/ms)</b>, <b>Min (1/ms)</b>, and <b>Max (1/ms)</b>
       
   113 information is displayed in the legend table.</p>
       
   114 
       
   115 <p><br>
       
   116 </p>
       
   117 
       
   118 <p><strong>Related references </strong></p>
       
   119 <ul>
       
   120   <li><a href="../../reference/analyzer/traceable_events.htm">Traceable Event
       
   121     Types</a></li>
       
   122   <li><a href="../../reference/profiler/Prof_counter_settings.htm">PIProfiler
       
   123     Performance Counter Settings</a></li>
       
   124   <li><a href="main_view.htm">Analyzer View</a></li>
       
   125 </ul>
       
   126 
       
   127 <div id="footer">
       
   128 Copyright © 2010 Nokia Corporation and/or its subsidiary(-ies). All rights
       
   129 reserved. <br>
       
   130 License: <a
       
   131 href="http://www.eclipse.org/legal/epl-v10.html">http://www.eclipse.org/legal/epl-v10.html</a></div>
       
   132 </body>
       
   133 </html>