chizeta/cnorxz
chizeta/cnorxz
0
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

add performance testing binary

Browse source
This commit is contained in:
Christian Zimmermann 2024-10-22 17:59:26 -07:00
parent df2b6892d9
commit 7754fd084e
2 changed files with 74 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
src/bin/CMakeLists.txt
View file

@ -4,3 +4,11 @@ add_dependencies(cnorxz-config cnorxz)
target_link_libraries(cnorxz-config cnorxz)
install(TARGETS cnorxz-config RUNTIME DESTINATION ${INSTALL_PATH}/bin)
#set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g")
add_executable(performance performance.cc)
add_dependencies(performance cnorxz)
target_link_libraries(performance cnorxz)
install(TARGETS performance RUNTIME DESTINATION ${INSTALL_PATH}/bin)

66
src/bin/performance.cc Normal file
View file

@ -0,0 +1,66 @@
#include <cstdlib>
#include <iostream>
#include "ranges/ranges.h"
#include "operation/operation.h"
#include "array/array.h"
#include <chrono>
#include <random>
namespace
{
using namespace CNORXZ;
}
int main()
{
static std::random_device rd;
static std::mt19937 gen(rd());
std::uniform_real_distribution<> dist(-9.9, 10.1);
auto r0 = CRangeFactory(4).create();
auto r1 = CRangeFactory(8).create();
auto r2 = CRangeFactory(890).create();
auto r3 = CRangeFactory(970).create();
auto i = std::make_shared<SIndex<SizeT,4>>(r0);
auto vi = eindexPtr<1>( std::make_shared<SIndex<SizeT,4>>(r0) );
//auto vi = lindexPtr<1>( std::make_shared<SIndex<SizeT,4>>(r0) );
auto i1i = lindexPtr<2>( std::make_shared<SIndex<SizeT,8>>(r1) );
auto i1j = lindexPtr<3>( std::make_shared<SIndex<SizeT,8>>(r1) );
auto i1k = lindexPtr<4>( std::make_shared<SIndex<SizeT,8>>(r1) );
auto i2i = std::make_shared<CIndex>(r2);
auto i3i = std::make_shared<CIndex>(r3);
MArray<Double> a(r2*r1*r1*r0);
a(i2i*i1i*i1j*i) = operation( [&](auto x) { return x+dist(gen); }, posop(i) );
MArray<Double> b(r3*r1*r1*r0);
b(i3i*i1j*i1k*i) = operation( [&](auto x) { return x+dist(gen); }, posop(i) );
MArray<Double> c(r1*r1*r0);
//const SizeT osize = mindex(i2i*i3i).lmax().val();
const SizeT lsize = mindex(i1i*i1k*i*i2i*i3i*i1j).lmax().val();
const SizeT nflops = lsize*2;
std::cout << " loop size = " << lsize << " => " << nflops/1e9 << " Gflops" << std::endl;
std::cout << " -> start operation" << std::endl;
//auto thexpr = mindexPtr(i1i*i1k*i2i*i3i*i1j*vi)->ifor( operation( [&](auto& t, const auto s) { t += s; }, c(i1i*i1k*vi), a(i2i*i1i*i1j*vi)*b(i3i*i1j*i1k*vi) ), NoF{} );
auto thexpr = mindexPtr(i2i*i3i*i1i*i1j*i1k*vi)->ifor( operation( [&](auto& t, const auto s) { t += s; }, c(i1i*i1k*vi), a(i2i*i1i*i1j*vi)*b(i3i*i1j*i1k*vi) ), NoF{} );
auto tbeg = std::chrono::system_clock::now();
//c(i1i*i1k*vi) += (a(i2i*i1i*i1j*vi)*b(i3i*i1j*i1k*vi)).c(mindexPtr(i2i*i3i*i1j));
//for(SizeT cnt = 0; cnt != )
thexpr();
auto tend = std::chrono::system_clock::now();
Double dt = static_cast<Double>( std::chrono::duration<Double>(tend-tbeg).count() );
std::cout << " -> done: " << dt << " secs" << std::endl;
std::cout << " performance: " << nflops/1e9/dt << " Gflops/sec " << std::endl;
Double s = 0;
for(const auto& x: c){
s += x;
}
std::cout << " res = " << s << std::endl;
return 0;
}