I want to run a google/benchmark on float, double and long double.
Given the BENCHMARK_TEMPLATE example, I attempted the following:
#include <cmath>
#include <ostream>
#include <random>
#include <benchmark/benchmark.h>
template<typename Real>
BM_PowTemplated(benchmark::State& state) {
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_real_distribution<Real> dis(1, 10);
auto s = dis(gen);
auto t = dis(gen);
Real y;
while (state.KeepRunning()) {
benchmark::DoNotOptimize(y = std::pow(s, t));
}
std::ostream cnull(0);
cnull << y;
}
BENCHMARK_TEMPLATE(BM_PowTemplated, float, double, long double);
BENCHMARK_MAIN();
I imagined that this would then create three benchmarks for float, double and long double, but instead it doesn't compile!
What is the proper syntax for creating templated google benchmarks? Is my mental model for how BENCHMARK_TEMPLATE should work correct and if so how can I fix this code?
You use BENCHMARK_TEMPLATE in wrong way with your
BENCHMARK_TEMPLATE(BM_PowTemplated, float, double, long double);
The readme file of https://github.com/google/benchmark/ says
template <class Q> int BM_Sequential(benchmark::State& state) { .. }
BENCHMARK_TEMPLATE(BM_Sequential, WaitQueue<int>)->Range(1<<0, 1<<10);
Three macros are provided for adding benchmark templates.
#define BENCHMARK_TEMPLATE(func, ...) // Takes any number of parameters.
#define BENCHMARK_TEMPLATE1(func, arg1)
#define BENCHMARK_TEMPLATE2(func, arg1, arg2)
So, BENCHMARK_TEMPLATE with arg1 is used for functions with one template parameter, with arg1 and arg2 for functions with two template parameters. Your BM_PowTemplated has only one parameter, so you can't use BENCHMARK_TEMPLATE with 3 args.
Check test/cxx03_test.cc of google/benchmark for example: https://github.com/google/benchmark/blob/b2e734087532897b7bb4c51a6b4f503060c9a20f/test/cxx03_test.cc
template <class T, class U>
void BM_template2(benchmark::State& state) {
BM_empty(state);
}
BENCHMARK_TEMPLATE2(BM_template2, int, long);
template <class T>
void BM_template1(benchmark::State& state) {
BM_empty(state);
}
BENCHMARK_TEMPLATE(BM_template1, long);
BENCHMARK_TEMPLATE1(BM_template1, int);
PS: definition of macro:
https://github.com/google/benchmark/blob/2d088a9f2d41acb77afc99d045f669e1a21b61ef/include/benchmark/benchmark_api.h#L684
We can see that a (a, b) arguments of macro goes inside < and >, and they are used as f< a,b >:
// template<int arg>
// void BM_Foo(int iters);
//
// BENCHMARK_TEMPLATE(BM_Foo, 1);
//
// will register BM_Foo<1> as a benchmark.
#define BENCHMARK_TEMPLATE1(n, a) \
BENCHMARK_PRIVATE_DECLARE(n) = \
(::benchmark::internal::RegisterBenchmarkInternal( \
new ::benchmark::internal::FunctionBenchmark(#n "<" #a ">", n<a>)))
#define BENCHMARK_TEMPLATE2(n, a, b) \
BENCHMARK_PRIVATE_DECLARE(n) = \
(::benchmark::internal::RegisterBenchmarkInternal( \
new ::benchmark::internal::FunctionBenchmark( \
#n "<" #a "," #b ">", n<a, b>)))
#define BENCHMARK_TEMPLATE(n, ...) \
BENCHMARK_PRIVATE_DECLARE(n) = \
(::benchmark::internal::RegisterBenchmarkInternal( \
new ::benchmark::internal::FunctionBenchmark( \
#n "<" #__VA_ARGS__ ">", n<__VA_ARGS__>)))
So, BENCHMARK_TEMPLATE can't iterate over several variants and define several variants of you function from one line.
Related
I'm a bit simple it seems as I'm not quite able to see clearly the cause of this error on the line marked error below.
std::sort and boost::sort picks up the default predicate, but ranges-v3 doesn't for some reason. This is ranges-v3 0.36. Similar error on clang 6/7 & gcc 7/8.
#include <range/v3/all.hpp>
#include <algorithm>
#include <boost/hana.hpp>
#include <utility>
#include <vector>
#include <boost/range/algorithm.hpp>
namespace hana = boost::hana;
template< typename T = int>
struct point_t {
BOOST_HANA_DEFINE_STRUCT(point_t<T>, (T, x), (T, y));
constexpr bool operator<(const point_t<T> &b) const noexcept {
return hana::less(hana::to_tuple(*this), hana::to_tuple(b));
};
};
int main() {
std::vector<point_t<point_t<>>> all;
boost::sort(all); // OK
std::sort(std::begin(all), std::end(all)); //OK
ranges::sort(all, std::less<point_t<point_t<>>>()); // OK
ranges::sort(all, hana::less); // OK
ranges::sort(all); // error no matching function for call to object of type 'const with_braced_init_args<ranges::v3::sort_fn>'
return 0;
}
Casey answered quickly via the range-v3 issue list.
Here is his comment as text as requested instead of the original image I placed here:
ranges::sort without a comparator argument requires the type to be
sorted to model the StrictTotallyOrdered concept. That means
the type must define all of ==, !=, <, >, <=, and >= with
consistent semantcs.
To which I replied there:
Thank you for getting back so quickly. I understand now. I must say
it's a little disappointing it is incompatible with std::sort and
boost::sort requirements. That is the price we pay for range-v3
loveliness I guess.
Thanks again, --Matt.
It is unfortunate the requirements are higher than std::sort and boost::sort so that code will not just work. I understand the motivation.
For the curious, std::rel_ops and boost/operators seemed to interfere with my goal for aggregate initialisation support for introspectable structs, so I ended up resorting to macros (similar to below), sadly.
I will play some more and look for a better static polymorphic solution.
Kind regards,
--Matt.
#define JEST_STRUCT(T) \
constexpr bool operator==(const T &b) const noexcept { \
return hana::equal(hana::to_tuple(*this), hana::to_tuple(b)); \
}; \
\
constexpr bool operator!=(const T &b) const noexcept { \
return hana::not_equal(hana::to_tuple(*this), hana::to_tuple(b)); \
}; \
\
constexpr bool operator<(const T &b) const noexcept { \
return hana::less(hana::to_tuple(*this), hana::to_tuple(b)); \
}; \
\
constexpr bool operator<=(const T &b) const noexcept { \
return hana::less_equal(hana::to_tuple(*this), hana::to_tuple(b)); \
}; \
\
constexpr bool operator>(const T &b) const noexcept { \
return hana::greater(hana::to_tuple(*this), hana::to_tuple(b)); \
}; \
\
constexpr bool operator>=(const T &b) const noexcept { \
return hana::greater_equal(hana::to_tuple(*this), hana::to_tuple(b)); \
}
This question relates to an earlier one I asked regarding implementing something akin to Qt's signal/slots in C++11.
Consider the following (very simplified signal dispatcher, that in this example does nothing of any use, it's just to demonstrate the pattern/problem):
template< typename... TYPES >
class Signal
{
public:
Signal() = default;
~Signal() = default;
template< typename... PARAMETERS >
void broadcast( PARAMETERS &&... p )
{
// static_assert to confirm PARAMETERS can map to TYPES
}
};
This works well enough, but there's some unwanted type conversion going on in practice. e.g.;
// acceptable use.
Signal< int, unsigned, float, char >().broadcast( 1, 2u, 0.f, 'a' );
// should fail compilation, first parameter is a float, 4th is an int.
Signal< int, unsigned, float, char >().broadcast( 0.f, 0, 0.f, 0 );
// acceptable use, first parameter is const, but it's convertible.
const int i = 3;
Signal< int, unsigned, float, char >().broadcast( i, 2u, 0.f, 'a');
// acceptable use, first parameter is const &, but it's convertible.
const int & j = i;
Signal< int, unsigned, float, char >().broadcast( j, 2u, 0.f, 'a');
There should be no silent float to int conversion. Conversion of const/const & in this instance should be possible (the format of TYPES should not have const or & as all data should be passed by value).
I'd like to prevent compilation where such unwanted type conversion happens. I thought to wrap up both TYPES and PARAMETERS in tuples, iterate over the tuple and confirm that each type in a given tuple parameter index matches (including using std::decay), but then I couldn't see a way to do that at compile time so that it could go in a static_assert.
For reference, compilers of choice are clang (latest on OS X 7.3 (clang-703.0.31)) and vc14.
Is what I want to do possible and, if so, can anyone offer any pointers?
Using (once again) the all_true bool pack trick from Columbo:
template <bool...> struct bool_pack;
template <bool... v>
using all_true = std::is_same<bool_pack<true, v...>, bool_pack<v..., true>>;
template <class... Args>
struct Signal {
template <class... Dargs, class = typename std::enable_if<all_true<
std::is_same<Args, typename std::decay<Dargs>::type>{}...
>{}>::type>
void broadcast(Dargs &&...) {}
};
This SFINAE's away the function if the parameters don't match exactly.
Here is a metaprogram I quickly came up with. It is a bit coarse, but can be implemented in a more better way. You should probably use the decayed type (std::decay) in the metaprogram to get correct result.
#include <iostream>
#include <type_traits>
template <typename... T> struct param_pack {};
template <typename, typename> struct is_all_same_impl;
template <>
struct is_all_same_impl<param_pack<>, param_pack<>>
{
static bool const value = true;
};
template <typename T, typename S, typename... Rest, typename... SRest>
struct is_all_same_impl<param_pack<T, Rest...>, param_pack<S, SRest...>>
{
static bool const value = false;
};
template <typename T, typename... Rest, typename... SRest>
struct is_all_same_impl<param_pack<T, Rest...>, param_pack<T, SRest...>>
{
static bool const value = is_all_same_impl<param_pack<Rest...>, param_pack<SRest...>>::value;
};
template <typename, typename>
struct is_all_same;
template <typename... FSet, typename... SSet>
struct is_all_same<param_pack<FSet...>, param_pack<SSet...>>: is_all_same_impl<param_pack<FSet...>, param_pack<SSet...>> {};
int main() {
std::cout << is_all_same<param_pack<int, char, float>, param_pack<int, char, int>>::value << std::endl;
return 0;
}
UPDATE :: More simpler version
template <typename... T> struct param_pack {};
int main() {
std::cout << std::is_same<param_pack<int, float, int>, param_pack<int,float,int>>::value << std::endl;
return 0;
}
So you can do something like:
static_assert( is_same<param_pack<Args...>, param_pack<std::decay_t<Dargs>...>>::value, "Parameters do not sufficiently match." );
I realize this has been asked before more than once on SO but I couldn't find a question explicitly looking for a current solution to this issue with C++11, so here we go again..
Can we conveniently get the string value of an enum with C++11?
I.e. is there (now) any built-in functionality in C++11 that allows us to get a string representation of enum types as in
typedef enum {Linux, Apple, Windows} OS_type;
OS_type myOS = Linux;
cout << myOS
that would print Linux on the console?
The longstanding and unnecessary lack of a generic enum-to-string feature in C++ (and C) is a painful one. C++11 didn't address this, and as far as I know neither will C++14.
Personally I'd solve this problem using code generation. The C preprocessor is one way--you can see some other answers linked in the comments here for that. But really I prefer to just write my own code generation specifically for enums. It can then easily generate to_string (char*), from_string, ostream operator<<, istream operator<<, is_valid, and more methods as needed. This approach can be very flexible and powerful, yet it enforces absolute consistency across many enums in a project, and it incurs no runtime cost.
Do it using Python's excellent "mako" package, or in Lua if you're into lightweight, or the CPP if you're against dependencies, or CMake's own facilities for generating code. Lots of ways, but it all comes down to the same thing: you need to generate the code yourself--C++ won't do this for you (unfortunately).
In my opinion, the most maintainable approach is to write a helper function:
const char* get_name(OS_type os) {
switch (os) {
case Linux: return "Linux";
case Apple: return "Apple";
case Windows: return "Windows";
}
}
It is a good idea not to implement the "default" case, since doing so will ensure that you get a compiler warning if you forget to implement a case (with the right compiler and compiler settings).
I like a hack using the C preprocessor, which I first saw here:
http://blogs.msdn.com/b/vcblog/archive/2008/04/30/enums-macros-unicode-and-token-pasting.aspx .
It uses the token-pasting operator # .
// This code defines the enumerated values:
#define MY_ENUM(x) x,
enum Fruit_Type {
MY_ENUM(Banana)
MY_ENUM(Apple)
MY_ENUM(Orange)
};
#undef MY_ENUM
// and this code defines an array of string literals for them:
#define MY_ENUM(x) #x,
const char* const fruit_name[] = {
MY_ENUM(Banana)
MY_ENUM(Apple)
MY_ENUM(Orange)
};
#undef MY_ENUM
// Finally, here is some client code:
std::cout << fruit_name[Banana] << " is enum #" << Banana << "\n";
// In practice, those three "MY_ENUM" macro calls will be inside an #include file.
Frankly, it's ugly and. but you end up typing your enums exactly ONCE in an include file, which is more maintainable.
BTW, on that MSDN blog link (see above) a user made a comment with a trick that makes the whole thing much prettier, and avoids #includes:
#define Fruits(FOO) \
FOO(Apple) \
FOO(Banana) \
FOO(Orange)
#define DO_DESCRIPTION(e) #e,
#define DO_ENUM(e) e,
char* FruitDescription[] = {
Fruits(DO_DESCRIPTION)
};
enum Fruit_Type {
Fruits(DO_ENUM)
};
// Client code:
std::cout << FruitDescription[Banana] << " is enum #" << Banana << "\n";
(I just noticed that 0x17de's answer also uses the token-pasting operator)
Here is a simple example using namespaces and structs.
A class is created for each enum item. In this example i chose int as the type for the id.
#include <iostream>
using namespace std;
#define ENUMITEM(Id, Name) \
struct Name {\
static constexpr const int id = Id;\
static constexpr const char* name = #Name;\
};
namespace Food {
ENUMITEM(1, Banana)
ENUMITEM(2, Apple)
ENUMITEM(3, Orange)
}
int main() {
cout << Food::Orange::id << ":" << Food::Orange::name << endl;
return 0;
}
Output:
3:Orange
== Update ==
Using:
#define STARTENUM() constexpr const int enumStart = __LINE__;
#define ENUMITEM(Name) \
struct Name {\
static constexpr const int id = __LINE__ - enumStart - 1;\
static constexpr const char* name = #Name;\
};
and using it once before the first usage of ENUMITEM the ids would not be needed anymore.
namespace Food {
STARTENUM()
ENUMITEM(Banana)
ENUMITEM(Apple)
ENUMITEM(Orange)
}
The variable enumStart is only accessible through the namespace - so still multiple enums can be used.
You can use macro to solve this problem:
#define MAKE_ENUM(name, ...) enum class name { __VA_ARGS__}; \
static std::vector<std::string> Enum_##name##_init(){\
const std::string content = #__VA_ARGS__; \
std::vector<std::string> str;\
size_t len = content.length();\
std::ostringstream temp;\
for(size_t i = 0; i < len; i ++) {\
if(isspace(content[i])) continue;\
else if(content[i] == ',') {\
str.push_back(temp.str());\
temp.str(std::string());}\
else temp<< content[i];}\
str.push_back(temp.str());\
return str;}\
static const std::vector<std::string> Enum_##name##_str_vec = Enum_##name##_init();\
static std::string to_string(name val){\
return Enum_##name##_str_vec[static_cast<size_t>(val)];\
}\
static std::string print_all_##name##_enum(){\
int count = 0;\
std::string ans;\
for(auto& item:Enum_##name##_str_vec)\
ans += std::to_string(count++) + ':' + item + '\n';\
return ans;\
}
As the static variable can only be initialized once, so the Enum_##name##_str_vec will use the Enum_##name##_init() function to initialize itself at first.
The sample code is as below:
MAKE_ENUM(Analysis_Time_Type,
UNKNOWN,
REAL_TIME,
CLOSSING_TIME
);
Then you can use below sentence to print an enum value:
to_string(Analysis_Time_Type::UNKNOWN)
And use below sentence to print all enum as string:
print_all_Analysis_Time_Type_enum()
As mentioned, there is no standard way to do this. But with a little preprocessor magic (similar to AlejoHausner's second contribution) and some template magic, it can be fairly elegant.
Include this code once:
#include <string>
#include <algorithm>
#define ENUM_VALS( name ) name,
#define ENUM_STRINGS( name ) # name,
/** Template function to return the enum value for a given string
* Note: assumes enums are all upper or all lowercase,
* that they are contiguous/default-ordered,
* and that the first value is the default
* #tparam ENUM type of the enum to retrieve
* #tparam ENUMSIZE number of elements in the enum (implicit; need not be passed in)
* #param valStr string version of enum value to convert; may be any capitalization (capitalization may be modified)
* #param enumStrs array of strings corresponding to enum values, assumed to all be in lower/upper case depending upon
* enumsUpper
* #param enumsUpper true if the enum values are in all uppercase, false if in all lowercase (mixed case not supported)
* #return enum value corresponding to valStr, or the first enum value if not found
*/
template <typename ENUM, size_t ENUMSIZE>
static inline ENUM fromString(std::string &valStr, const char *(&enumStrs)[ENUMSIZE], bool enumsUpper = true) {
ENUM e = static_cast< ENUM >(0); // by default, first value
// convert valStr to lower/upper-case
std::transform(valStr.begin(), valStr.end(), valStr.begin(), enumsUpper ? ::toupper : ::tolower);
for (size_t i = 0; i< ENUMSIZE; i++) {
if (valStr == std::string(enumStrs[i])) {
e = static_cast< ENUM >(i);
break;
}
}
return e;
}
Then define each enum like so:
//! Define ColorType enum with array for converting to/from strings
#define ColorTypes(ENUM) \
ENUM(BLACK) \
ENUM(RED) \
ENUM(GREEN) \
ENUM(BLUE)
enum ColorType {
ColorTypes(ENUM_VALS)
};
static const char* colorTypeNames[] = {
ColorTypes(ENUM_STRINGS)
};
You only have to enumerate the enum values once and the code to define it is fairly compact and intuitive.
Values will necessarily be numbered in the default way (ie, 0,1,2,...). The code of fromString() assumes that enum values are in either all uppercase or all lowercase (for converting from strings) that the default value is first, but you can of course change how these things are handled.
Here is how you get the string value:
ColorType c = ColorType::BLUE;
std::cout << colorTypeNames[c]; // BLUE
Here is how you set the enum from a string value:
ColorType c2 = fromString<ColorType>("Green", colorTypeNames); // == ColorType::GREEN
I'm having trouble with std::initializer_list. I reduced it down to a simple example:
#include <initializer_list>
#include <cstdio>
class Test {
public:
template <typename type> Test(const std::initializer_list<type>& args) {}
};
int main(int argc, char* argv[]) {
Test({1,2});
getchar();
return 0;
}
When compiled using g++ test_initializer.cpp -std=c++0x, it compiles and runs well. However, if line 11 is changed to Test({1,2.0});, one gets:
ian#<host>:~/Desktop$ g++ test_initializer.cpp -std=c++0x
test_initializer.cpp: In function ‘int main(int, char**)’:
test_initializer.cpp:11:14: error: no matching function for call to ‘Test::Test(<brace-enclosed initializer list>)’
test_initializer.cpp:11:14: note: candidates are:
test_initializer.cpp:7:28: note: template<class type> Test::Test(const std::initializer_list<_Tp>&)
test_initializer.cpp:5:7: note: constexpr Test::Test(const Test&)
test_initializer.cpp:5:7: note: no known conversion for argument 1 from ‘<brace-enclosed initializer list>’ to ‘const Test&’
test_initializer.cpp:5:7: note: constexpr Test::Test(Test&&)
test_initializer.cpp:5:7: note: no known conversion for argument 1 from ‘<brace-enclosed initializer list>’ to ‘Test&&’
I suspect this happens because the compiler can't figure out what type to make the initializer list. Is there a way to fix the example so that it works with different types (and still uses initializer lists)?
An std::initializer_list takes only one type. If you need different types, you can use variadic templates:
template<typename... Args>
Test(Args&&... args);
/* ... */
int main()
{
Test(1, 2.0);
}
Would a std::tuple<int.double> work for the OP? If the code will always have a int followed by a double, then the OP could get strict type-checking for all arguments, which the variable arguments solution does not allow. The std::tuple<>, however, would not work for any number or order of values, so may not be appropriate for all use cases.
Let the initializer_list hold the most arbitrary pointers, void*, and do your own casting from there. Here is an example.
#include <initializer_list>
#include <iostream>
using std::initializer_list;
using std::cout;
using std::endl;
class Person {
private:
string _name;
int _age;
public:
Person(initializer_list<void*> init_list) {
auto it = init_list.begin();
_name = *((string*)(*it));
it++;
_age = *((int*)(*it));
}
void print() {
cout << "name: " << _name << ". age: " << _age << endl;
}
};
int main(void) {
string name{"Vanderbutenburg};
int age{23};
Person p{&name,&age};
p.print(); // "name: Vanderbutenburg. age: 23"
return 0;
}
Is there any way to get behavior like this?
// Some definition(s) of operator "" _my_str
// Some definition of function or macro MY_STR_LEN
using T1 = MY_STR_LEN("ape"_my_str);
// T1 is std::integral_constant<std::size_t, 3U>.
using T2 = MY_STR_LEN("aardvark"_my_str);
// T2 is std::integral_constant<std::size_t, 8U>.
It seems not, since the string literals are passed immediately to some_return_type operator "" _my_str(const char*, std::size_t); and never to a literal operator template (2.14.8/5). That size function parameter can't be used as a template argument, even though it will almost always be a constant expression.
But it seems like there ought to be some way to do this.
Update: The accepted answer, that this is not possible without an extra definition per literal, is accurate for C++11 as asked, and also C++14 and C++17. C++20 allows the exact result asked for:
#include <cstdlib>
#include <type_traits>
#include <string_view>
struct cexpr_str {
const char* ptr;
std::size_t len;
template <std::size_t Len>
constexpr cexpr_str(const char (&str)[Len]) noexcept
: ptr(str), len(Len) {}
};
// Essentially the same as
// std::literals::string_view_literals::operator""sv :
template <cexpr_str Str>
constexpr std::string_view operator "" _my_str () noexcept
{
return std::string_view(Str.ptr, Str.len);
}
#define MY_STR_LEN(sv) \
std::integral_constant<std::size_t, (sv).size()>
Reading C++11 2.14.8 carefully reveals that the "literal operator template" is only considered for numeric literals, but not for string and character literals.
However, the following approach seems to give you constexpr access to the string length (but not the pointer):
struct MyStr
{
char const * str;
unsigned int len;
constexpr MyStr(char const * p, unsigned int n) : str(p), len(n) {}
};
constexpr MyStr operator "" _xyz (char const * s, unsigned int len)
{
return MyStr(s, len);
}
constexpr auto s = "Hello"_xyz;
Test:
#include <array>
using atype = std::array<int, s.len>; // OK