56fd68e7dd
CCCL (CUDA C++ Core Libraries) provides: - CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk) - Thrust: high-level parallel algorithms (transform_reduce, sort, scan) - libcudacxx: CUDA C++ standard library (atomics, barriers, memory) - cudax: experimental features (memory resources, allocators) - Tuning policies: per-SM hardware-specific algorithm parameters Competition optimization vectors mapped to CCCL: - Output TPS (83% weight): warp_reduce, block_reduce, device_topk - Input TPS (14% weight): device_scan, block_load, prefetch - Cache TPS (3% weight): prefix caching strategy patterns - Memory (0.9 util): pooled/cached/buddy allocators Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only) License: Apache-2.0
87 lines
5.2 KiB
Plaintext
87 lines
5.2 KiB
Plaintext
#include <thrust/functional.h>
|
|
#include <thrust/transform.h>
|
|
|
|
#include <algorithm>
|
|
#include <functional>
|
|
|
|
#include <unittest/unittest.h>
|
|
|
|
const size_t NUM_SAMPLES = 10000;
|
|
|
|
template <class InputVector, class OutputVector, class Operator, class ReferenceOperator>
|
|
void TestBinaryFunctional()
|
|
{
|
|
using InputType = typename InputVector::value_type;
|
|
using OutputType = typename OutputVector::value_type;
|
|
|
|
thrust::host_vector<InputType> std_input1 = unittest::random_samples<InputType>(NUM_SAMPLES);
|
|
thrust::host_vector<InputType> std_input2 = unittest::random_samples<InputType>(NUM_SAMPLES);
|
|
thrust::host_vector<OutputType> std_output(NUM_SAMPLES);
|
|
|
|
// Replace zeros to avoid divide by zero exceptions
|
|
std::replace(std_input2.begin(), std_input2.end(), (InputType) 0, (InputType) 1);
|
|
|
|
InputVector input1 = std_input1;
|
|
InputVector input2 = std_input2;
|
|
OutputVector output(NUM_SAMPLES);
|
|
|
|
thrust::transform(input1.begin(), input1.end(), input2.begin(), output.begin(), Operator());
|
|
thrust::transform(std_input1.begin(), std_input1.end(), std_input2.begin(), std_output.begin(), ReferenceOperator());
|
|
|
|
// Note: FP division is not bit-equal, even when nvcc is invoked with --prec-div
|
|
ASSERT_ALMOST_EQUAL(output, std_output);
|
|
}
|
|
|
|
// XXX add bool to list
|
|
// Instantiate a macro for all integer-like data types
|
|
#define INSTANTIATE_INTEGER_TYPES(Macro, vector_type, operator_name) \
|
|
Macro(vector_type, operator_name, unittest::int8_t) Macro(vector_type, operator_name, unittest::uint8_t) \
|
|
Macro(vector_type, operator_name, unittest::int16_t) Macro(vector_type, operator_name, unittest::uint16_t) \
|
|
Macro(vector_type, operator_name, unittest::int32_t) Macro(vector_type, operator_name, unittest::uint32_t) \
|
|
Macro(vector_type, operator_name, unittest::int64_t) Macro(vector_type, operator_name, unittest::uint64_t)
|
|
|
|
// Instantiate a macro for all integer and floating point data types
|
|
#define INSTANTIATE_ALL_TYPES(Macro, vector_type, operator_name) \
|
|
INSTANTIATE_INTEGER_TYPES(Macro, vector_type, operator_name) \
|
|
Macro(vector_type, operator_name, float)
|
|
|
|
// op(T,T) -> T
|
|
#define INSTANTIATE_BINARY_ARITHMETIC_FUNCTIONAL_TEST(vector_type, operator_name, data_type) \
|
|
TestBinaryFunctional<thrust::vector_type<data_type>, \
|
|
thrust::vector_type<data_type>, \
|
|
::cuda::std::operator_name<data_type>, \
|
|
::std::operator_name<data_type>>();
|
|
// op(T,T) -> T
|
|
#define DECLARE_BINARY_ARITHMETIC_FUNCTIONAL_UNITTEST(operator_name, OperatorName) \
|
|
void Test##OperatorName##FunctionalHost() \
|
|
{ \
|
|
INSTANTIATE_ALL_TYPES(INSTANTIATE_BINARY_ARITHMETIC_FUNCTIONAL_TEST, host_vector, operator_name); \
|
|
} \
|
|
DECLARE_UNITTEST(Test##OperatorName##FunctionalHost); \
|
|
void Test##OperatorName##FunctionalDevice() \
|
|
{ \
|
|
INSTANTIATE_ALL_TYPES(INSTANTIATE_BINARY_ARITHMETIC_FUNCTIONAL_TEST, device_vector, operator_name); \
|
|
} \
|
|
DECLARE_UNITTEST(Test##OperatorName##FunctionalDevice);
|
|
|
|
// op(T,T) -> T (for integer T only)
|
|
#define DECLARE_BINARY_INTEGER_ARITHMETIC_FUNCTIONAL_UNITTEST(operator_name, OperatorName) \
|
|
void Test##OperatorName##FunctionalHost() \
|
|
{ \
|
|
INSTANTIATE_INTEGER_TYPES(INSTANTIATE_BINARY_ARITHMETIC_FUNCTIONAL_TEST, host_vector, operator_name); \
|
|
} \
|
|
DECLARE_UNITTEST(Test##OperatorName##FunctionalHost); \
|
|
void Test##OperatorName##FunctionalDevice() \
|
|
{ \
|
|
INSTANTIATE_INTEGER_TYPES(INSTANTIATE_BINARY_ARITHMETIC_FUNCTIONAL_TEST, device_vector, operator_name); \
|
|
} \
|
|
DECLARE_UNITTEST(Test##OperatorName##FunctionalDevice);
|
|
|
|
// Create the unit tests
|
|
DECLARE_BINARY_ARITHMETIC_FUNCTIONAL_UNITTEST(plus, Plus);
|
|
DECLARE_BINARY_ARITHMETIC_FUNCTIONAL_UNITTEST(minus, Minus);
|
|
DECLARE_BINARY_ARITHMETIC_FUNCTIONAL_UNITTEST(multiplies, Multiplies);
|
|
DECLARE_BINARY_ARITHMETIC_FUNCTIONAL_UNITTEST(divides, Divides);
|
|
|
|
DECLARE_BINARY_INTEGER_ARITHMETIC_FUNCTIONAL_UNITTEST(modulus, Modulus);
|