lmi
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[lmi] wx_test_benchmark_census.cpp


From: Greg Chicares
Subject: [lmi] wx_test_benchmark_census.cpp
Date: Fri, 12 Dec 2014 12:47:02 +0000
User-agent: Mozilla/5.0 (Windows NT 5.1; rv:31.0) Gecko/20100101 Thunderbird/31.3.0

// ERASE THIS BLOCK COMMENT WHEN IMPLEMENTATION COMPLETE. The block
// comment below changes the original specification, and does not
// yet describe the present code. Desired changes:
//  - use "gui_test_path/MSEC*.cns", not a config file list
//  - write timings to stdout only; perform no relative-error
//    calculation, and do not use times stored in a config file
// With these changes, the test should no longer use the config file
// in any way.

/// Measure the speed of various operations on certain census files.
///
/// Comparing the results of this test to a stored touchstone makes it
/// easy to see speed changes, and hence to guard against performance
/// regressions that might otherwise escape timely notice.
///
/// Write timing data to stdout. We had considered storing touchstone
/// timings in a configuration file and calculating relative error
/// here, but found that it's simpler just to print the timings and
/// compare to the results of prior runs.
///
/// These operations are measured because they are the most important:
///   Census | Run case
///   Census | Print case to PDF
///   Census | Print case to spreadsheet
/// We had considered running at least the "Run case" test several
/// times, discarding the first run and reporting an average (probably
/// the mode) of the others; that can be done at a later date if
/// experience demonstrates that it would be useful.
///
/// This test uses all files matching "gui_test_path/MSEC*.cns", which
/// may include proprietary products and should be designed to cover
/// different paths through the code. We had considered specifying the
/// input files in a configuration file, but the chosen way is simpler
/// and makes it even easier to change to change the input set.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]