ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Update Kokkos library in LAMMPS to v3.6.0

Browse Source
This commit is contained in:
Stan Gerald Moore
2022-05-05 11:44:47 -06:00
parent bd4bbbddbe
commit b79c0bc7b4
380 changed files with 41928 additions and 8786 deletions
Download Patch File Download Diff File Expand all files Collapse all files

100
lib/kokkos/core/unit_test/TestMathematicalFunctions.hpp
View File

@ -59,6 +59,12 @@
#define MATHEMATICAL_FUNCTIONS_HAVE_LONG_DOUBLE_OVERLOADS
#endif
// WORKAROUND icpx changing default FP model when optimization level is >= 1
// using -fp-model=precise works too
#if defined(__INTEL_LLVM_COMPILER)
#define KOKKOS_IMPL_WORKAROUND_INTEL_LLVM_DEFAULT_FLOATING_POINT_MODEL
#endif
// clang-format off
template <class>
struct math_unary_function_return_type;
@ -342,7 +348,13 @@ DEFINE_UNARY_FUNCTION_EVAL(asinh, 4);
DEFINE_UNARY_FUNCTION_EVAL(acosh, 2);
DEFINE_UNARY_FUNCTION_EVAL(atanh, 2);
#if defined(__APPLE__)
// Apple's standard library implementation seems to have a poor implementation
DEFINE_UNARY_FUNCTION_EVAL(erf, 5);
#else
DEFINE_UNARY_FUNCTION_EVAL(erf, 2);
#endif
DEFINE_UNARY_FUNCTION_EVAL(erfc, 5);
// has a larger error due to some impls doing integer exact.
// We cast always to double leading to larger difference when comparing our
@ -415,8 +427,6 @@ struct TestMathUnaryFunction : FloatingPointComparison {
Arg val_[N];
Ret res_[N];
TestMathUnaryFunction(const Arg (&val)[N]) {
std::cout << math_function_name<Func>::name << "("
<< type_helper<Arg>::name() << ")\n";
std::copy(val, val + N, val_);
std::transform(val, val + N, res_,
[](auto x) { return Func::eval_std(x); });
@ -425,7 +435,9 @@ struct TestMathUnaryFunction : FloatingPointComparison {
void run() {
int errors = 0;
Kokkos::parallel_reduce(Kokkos::RangePolicy<Space>(0, N), *this, errors);
ASSERT_EQ(errors, 0);
ASSERT_EQ(errors, 0) << "Failed check no error for "
<< math_function_name<Func>::name << "("
<< type_helper<Arg>::name() << ")";
}
KOKKOS_FUNCTION void operator()(int i, int& e) const {
bool ar = compare(Func::eval(val_[i]), res_[i], Func::ulp_factor());
@ -456,15 +468,15 @@ struct TestMathBinaryFunction : FloatingPointComparison {
Ret res_;
TestMathBinaryFunction(Arg1 val1, Arg2 val2)
: val1_(val1), val2_(val2), res_(Func::eval_std(val1, val2)) {
std::cout << math_function_name<Func>::name << "("
<< type_helper<Arg1>::name() << ", " << type_helper<Arg2>::name()
<< ")\n";
run();
}
void run() {
int errors = 0;
Kokkos::parallel_reduce(Kokkos::RangePolicy<Space>(0, 1), *this, errors);
ASSERT_EQ(errors, 0);
ASSERT_EQ(errors, 0) << "Failed check no error for "
<< math_function_name<Func>::name << "("
<< type_helper<Arg1>::name() << ", "
<< type_helper<Arg2>::name() << ")";
}
KOKKOS_FUNCTION void operator()(int, int& e) const {
bool ar = compare(Func::eval(val1_, val2_), res_, Func::ulp_factor());
@ -917,9 +929,7 @@ struct TestAbsoluteValueFunction {
using Kokkos::Experimental::isinf;
using Kokkos::Experimental::isnan;
if (abs(-0.) != 0.
// WORKAROUND icpx changing default FP model when optimization level is >= 1
// using -fp-model=precise works too
#ifndef __INTEL_LLVM_COMPILER
#ifndef KOKKOS_IMPL_WORKAROUND_INTEL_LLVM_DEFAULT_FLOATING_POINT_MODEL
|| !isinf(abs(-INFINITY)) || !isnan(abs(-NAN))
#endif
) {
@ -942,3 +952,73 @@ struct TestAbsoluteValueFunction {
TEST(TEST_CATEGORY, mathematical_functions_absolute_value) {
TestAbsoluteValueFunction<TEST_EXECSPACE>();
}
template <class Space>
struct TestIsNaN {
TestIsNaN() { run(); }
void run() const {
int errors = 0;
Kokkos::parallel_reduce(Kokkos::RangePolicy<Space>(0, 1), *this, errors);
ASSERT_EQ(errors, 0);
}
KOKKOS_FUNCTION void operator()(int, int& e) const {
using Kokkos::Experimental::isnan;
using Kokkos::Experimental::quiet_NaN;
using Kokkos::Experimental::signaling_NaN;
if (isnan(1) || isnan(INT_MAX)) {
++e;
KOKKOS_IMPL_DO_NOT_USE_PRINTF("failed isnan(integral)\n");
}
if (isnan(2.f)
#ifndef KOKKOS_IMPL_WORKAROUND_INTEL_LLVM_DEFAULT_FLOATING_POINT_MODEL
|| !isnan(quiet_NaN<float>::value) ||
!isnan(signaling_NaN<float>::value)
#endif
) {
++e;
KOKKOS_IMPL_DO_NOT_USE_PRINTF("failed isnan(float)\n");
}
if (isnan(3.)
#ifndef KOKKOS_IMPL_WORKAROUND_INTEL_LLVM_DEFAULT_FLOATING_POINT_MODEL
|| !isnan(quiet_NaN<double>::value) ||
!isnan(signaling_NaN<double>::value)
#endif
) {
++e;
KOKKOS_IMPL_DO_NOT_USE_PRINTF("failed isnan(double)\n");
}
#ifdef MATHEMATICAL_FUNCTIONS_HAVE_LONG_DOUBLE_OVERLOADS
if (isnan(4.l)
#ifndef KOKKOS_IMPL_WORKAROUND_INTEL_LLVM_DEFAULT_FLOATING_POINT_MODEL
|| !isnan(quiet_NaN<long double>::value) ||
!isnan(signaling_NaN<long double>::value)
#endif
) {
++e;
KOKKOS_IMPL_DO_NOT_USE_PRINTF("failed isnan(long double)\n");
}
#endif
// special values
if (isnan(INFINITY)
#ifndef KOKKOS_IMPL_WORKAROUND_INTEL_LLVM_DEFAULT_FLOATING_POINT_MODEL
|| !isnan(NAN)
#endif
) {
++e;
KOKKOS_IMPL_DO_NOT_USE_PRINTF(
"failed isnan(floating_point) special values\n");
}
static_assert(std::is_same<decltype(isnan(1)), bool>::value, "");
static_assert(std::is_same<decltype(isnan(2.f)), bool>::value, "");
static_assert(std::is_same<decltype(isnan(3.)), bool>::value, "");
#ifdef MATHEMATICAL_FUNCTIONS_HAVE_LONG_DOUBLE_OVERLOADS
static_assert(std::is_same<decltype(isnan(4.l)), bool>::value, "");
#endif
}
};
TEST(TEST_CATEGORY, mathematical_functions_isnan) {
TestIsNaN<TEST_EXECSPACE>();
}