commit d50bda0ed5ef2620d88cdd2b54d76e7d3ca72be7 Author: Olaf Mersmann Date: 2013-02-19 13:47:14 +0100 Remove superfluous LICENSE file. commit b9c2925114da4b56784b8d24e02846d292cda8c7 Author: Olaf Mersmann Date: 2013-02-19 13:45:27 +0100 Update DESCRIPTION. Updated the email address of the primary author and maintainer. Also changed from the FreeBSD to the equivalent BSD license. commit fe8c553307e9e19aa766c0bc08ce2e716c928246 Author: Olaf Mersmann Date: 2013-02-19 13:39:45 +0100 Add OpenBSD support. Patch-By: Zé Loff commit 4a5ffc9d3626a96fe74eaa6d2f8415516df51bf2 Author: Olaf Mersmann Date: 2012-11-16 21:14:08 +0100 Add a new 'unit' option to microbenchmark. There is now a way to set the unit used to present the results when calling microbenchmark(). You can also set the unit used globally by setting the option 'microbenchmark.unit'. This commit also adds FreeBSD support thanks to a patch from Rainer Hurling. Patch-by: Rainer Hurling (FreeBSD support) commit c8e546a572452a85ac9eacbc7a14ba8a19cbf189 Author: Olaf Mersmann Date: 2012-11-02 12:29:00 +0100 Multiple enhancements suggested by Hadley Wickham microbenchmark() now retains the order of the expressions as they are passed in. print.microbenchmark() has a new unit 'relative' and a new argument 'order'. The 'relative' unit rescales all other times relative to the time of the quickest w.r.t the median time. Using the 'order' argument, you can reorder the results according to any of the columns displayed. Apart from these major changes there are also several smaller bugfixes in the handling of expressions and their representation as returned by deparse(). Suggested-By: Hadley Wickham commit e44fe8c7f39a57e1d137dc3cfa87903474d479c0 Author: Olaf Mersmann Date: 2012-08-23 10:19:13 +0200 Upgrade to roxygen2. commit f1c0358b0a593c0c3ca8fc6a0c80dab08d637ccd Author: Olaf Mersmann Date: 2012-08-23 10:15:22 +0200 Add autplot method for microbenchmark objects The ggplot2 package as a new(ish) method named autoplot which is used "to draw a particular plot for an object of a particular class in a single command." This commit adds an implementation for microbenchmark objects that draws a violin plot of the observed runtimes on log scale. commit 8928057053ad43e1d72fdb1ca443a7db8fc83a84 Author: Olaf Mersmann Date: 2012-08-22 20:33:55 +0200 Cleanup documentation. commit 79d4a23db9fd79ccb8c9d60b33eb045fac11c4a7 Author: Olaf Mersmann Date: 2012-02-28 00:25:06 +0100 Fix typos in documentation. Fix several typos and spelling errors in the documentation. Thanks go to Petr Savicky for reporting these. Reported-By: Petr Savicky commit e879c03a4d2da7ad053441da99bb696c892a6a9c Author: Olaf Mersmann Date: 2011-10-18 13:46:40 +0200 Fix signed/unsigned and C99 issues of nanotime_t. commit 7e6881256dee11b19a38fb3f46cebb989a29ec39 Author: Olaf Mersmann Date: 2011-10-17 15:53:30 +0200 Language / spelling fixes. commit 4b79bd205f1f791e85f06c61c63b37af527afdfe Author: Olaf Mersmann Date: 2011-10-17 15:12:14 +0200 Set required _POSIX_C_SOURCE to 200112L. Brian Ripley commented that it is invalid to compile a c99 source file using versions prior to POSIX.1-2001. This should fix that. commit e09b9c74c7b1cabcad52bc02c766ca4971c517f6 Author: Olaf Mersmann Date: 2011-10-10 10:23:41 +0200 Fix use of uninitialized variable. commit ff091e8eac5ee5bb9eeb255a3a9901d7863e58ce Author: Olaf Mersmann Date: 2011-10-07 09:07:31 +0200 Remove unused variables in C code. commit fba26ef775317fe44ef64956e40cbfcef6983b1b Author: Olaf Mersmann Date: 2011-10-07 08:58:47 +0200 Add code 't' and 'f' units which choose an SI prefix automatically. The new units 't' and 'f' (time and frequency) can be used instead of explicitly stating the units used to display the runtimes. The code guessed the correct SI prefix based on the magnitute of the minium observerd runtime. Patch by Claudia Beleites. commit 00d1fd850277ef16c671ac6a23f2478f96d40aa6 Author: Olaf Mersmann Date: 2011-08-16 21:04:10 +0200 Explicitly set _POSIX_C_SOURCE for better compatibility with c99 compilers. commit 1b1ac21ef9562db092466806ffb26b9357be2ace Author: Olaf Mersmann Date: 2011-04-07 16:25:23 +0200 Many small improvements. * microbenchmark() now returns a data.frame instead of a matrix. * Ordering of the expression evaluation is user controllable. * A new (experimental) function is added to estimate the timing overhead and the granularity of the underlying timer. * The documentation has improved. commit 0d968a6545bddfc526ddbecab15472cf3fd3ef14 Author: Olaf Mersmann Date: 2011-01-25 19:06:47 +0100 General package improvements * The documentation has been extended and clarified in many places. * The length of the warmup phase is now user controllable using a 'control' argument to microbenchmark(). * Milliseconds are now actual milliseconds and not microseconds. commit d4ab38ebc911e1da2bd9d9ce08b1070c6cb0ad51 Author: Olaf Mersmann Date: 2011-01-24 19:24:29 +0100 Change wording of warning on underflow. commit 5591ab3da8faf42766ee86894877ef8228dbad3e Author: Olaf Mersmann Date: 2011-01-17 09:51:12 +0100 Clarify license. commit 1a6214f8d75c9e5f940f0e6ef3630110a9db6e80 Author: Olaf Mersmann Date: 2011-01-16 23:21:48 +0100 Change Title and Description in DESCRIPTION. commit adf929d49ca6ae6ad2d7809b1ae8ec3d8a1fb186 Author: Olaf Mersmann Date: 2011-01-16 22:49:20 +0100 Deal with (too) fast evaluation timings. If the evaluation of an expression is sufficiently fast and the overhead is overestimated sufficiently, the estimate of the true evaluation time could be grossly exagerated. USER VISIBLE CHANGE: The microbenchmark object (a matrix) may now contain 'NA' values if the measured time of evaluation is smaller than the estimated evaluation overhead. RATIONAL: This is mainly a bugfix. In the past, this could lead to very large values in the result object. WHAT TO CHANGE: You may need to adjust your analysis to deal with the 'NA' values. commit aa5b4374bdbb13c28d949baf3cce36d4aa9ee6e9 Author: Olaf Mersmann Date: 2011-01-16 22:43:57 +0100 Cleanup Linux source code. commit 302fe1035f76a5ad35c750010d31fa2e93a49f97 Author: Olaf Mersmann Date: 2011-01-16 02:40:49 +0100 Inital commit.