* Auto-detect whether to produce colorized output
Rename --color_print to --benchmark_color for consistency with the other
flags (and Google Test). Old flag name is kept around for compatibility.
The --benchmark_color/--color_print flag takes a third option, "auto",
which is the new default. In this mode, we attempt to auto-detect
whether to produce colorized output. (The logic for deciding whether to
use colorized output was lifted from GTest:
<https://github.com/google/googletest/blob/master/googletest/src/gtest.cc#L2925>.)
* Update CONTRIBUTORS, AUTHORS
If a reporter's output stream isn't line-buffered (e.g. it's not writing
to a terminal) then it can be some time before a write to it becomes
visible.
This is problematic if, say, you're wanting to use tail -f to view the
file written to via --benchmark_out. Or if the application crashes,
leaving you with no results.
Addressed by flushing the reporters' output streams whenever we invoke
methods that may write to them.
* Refactor benchmark.cc into benchmark_register.cc and benchmark_run.cc
The benchmark.cc file is getting really big and it contains a bunch of
unrelated components. This patch separates the files into two separate
parts. The "runtime" parts and the "registration" parts.
This patch also removes the PIMPL used by Benchmark. Previously we couldn't
have STL types in the interface but now we can. Therefore there is no reason
to keep BenchmarkImp.
* add missing include
* rework windows timers again
* Guard timespec on older Windows versions
* Remove old thread safety annotation workarounds
* Change to using per-thread timers
* fix bad assertions
* fix copy paste error on windows
* Fix thread safety annotations
* Make null-log thread safe
* remove remaining globals
* use chrono for walltime since it is thread safe
* consolidate timer functions
* Add missing ctime include
* Rename to be consistent with Google style
* Format patch using clang-format
* cleanup -Wthread-safety configuration
* Don't trust _POSIX_FEATURE macros because OS X lies.
* Fix OS X thread timings
* attempt to fix mingw build
* Attempt to make mingw work again
* Revert old mingw workaround
* improve diagnostics
* Drastically improve OS X measurements
* Use average real time instead of max
In the `Ranges(...)` generation code a "control" vector which stores
the current index for each range passed to `Ranges`. Previously this vector
was incorrectly initialized to the size of the subranges not the number
of subranges.
Additionally this patch suppresses unused warnings generated by
`stream_init_anchor`.
The benchmark library internals write to std::cout/std::cerr during program
startup. This can cause segfaults when the user doesn't include <iostream> in
the benchmark (which init's the streams). This patch fixes this by emitting
a dynamic initializer in every TU which initializes the streams.
Previously the FittingCurve functions for n^2 and n^3 did the calculation
using int types. This can overflow and cause UB. This patch changes the
calculations to use std::pow to prevent this.
Also re-enable VC 2013 appveyor bot since I *hope* this is what was causing
the failures.
* Support multiple ranges in the benchmark
google-benchmark library allows to provide up to two ranges to the
benchmark method (range_x and range_y). However, in many cases it's not
sufficient. The patch introduces multi-range features, so user can easily
define multiple ranges by passing a vector of integers, and access values
through the method range(i).
* Remove redundant API
Functions State::range_x() and State::range_y() have been removed. They should
be replaced by State::range(0) and State::range(1).
Functions Benchmark::ArgPair() and Benchmark::RangePair() have been removed.
They should be replaced by Benchmark::Args() and Benchmark::Ranges().
These options allow you to write the output of a benchmark to the specified
file and with the specified format. The goal of this change is to help support
tooling.
Without these, clang reorders these instructions as if they were
regular loads/stores which causes SIGILL from the kernel because
it performs all the loads before it starts testing the values.
* Move ComputeStats call out of the reporters
* Cleanup adjusted time calculations in reporters
* Move ComputeBigO call out of reporters
* Remove ReportComplexity interface using ReportRuns instead
* Factor out reporting of basic context information
* Attempt to fix GCC 4.6 build errors
* Move ComputeStats to complexity.cc
* Add test for reporter output.
* setup err_stream tests
* Fix warnings in tests
* whitespace
* Fix build errors caused by super pedantic compilers
* Pass streams by pointer not non-const reference
Some benchmarks may run a few milliseconds which makes it kind of hard to visually compare, since the currently only available nanoseconds numbers can get very large in this case. Therefore this commit adds an optional command line flag --benchmark_time_unit which lets the user choose between ns and ms time units for displaying the mean execution time.
Having access to the thread count from within a benchmark is useful,
for when one wants to distribute a workload dynamically among the
benchmarks running in parallel e.g when using ThreadRange() or
ThreadPerCpu().
Using `0` as a null pointer is illegal when `-Wzero-as-null-pointer-constant`
is given to G++. To avoid the warning `zero-as-null-pointer-constant`,
`nullptr` (C++11 keyword) instead of `0` is used in the `sysctl` invocation.
When the library is created as a *.dll on Windows it is treated like a
runtime object so we must proivde the destination for the runtime
objects in our install command
For cross platform and cross compiler portability we use the
standard integer type for a 64-bit integer. MinGW on Windows doesn't
have the definition for `int64`.
We use the SHGetValueA on Windows to retrieve the MHz of the processor
but this requires the shlwapi library. Previous to this patch the
library was linked with a MSVC specific pragma but there is no
guarantee that on Windows we will be using MSVC. Therefore, it is much
compile agnostic to use the standard CMAKE library linking mechanism
to provide the definition of SHGetValueA
This patch adopts a new internal structure for how timings are performed.
Currently every iteration of a benchmark checks to see if it has been running
for an appropriate amount of time. Checking the clock introduces noise into
the timings and this can cause inconsistent output from each benchmark.
Now every iteration of a benchmark only checks an iteration count to see if it
should stop running. The iteration count is determined before hand by testing
the benchmark on a series of increasing iteration counts until a suitable count
is found. This increases the amount of time it takes to run the actual benchmarks
but it also greatly increases the accuracy of the results.
This patch introduces some breaking changes. The notable breaking changes are:
1. Benchmarks run on multiple threads no generate a report per thread. Instead
only a single report is generated.
2. ::benchmark::UseRealTime() was removed and replaced with State::UseRealTime().
