mtensor

mtensor is a C++/CUDA template library that supports lazy evaluation of Tensor. Unlike Eigen, mtensor focus the tensor structure and supports cuda lazy evalution. mtensor does not provide linear algebra and numerical features.

How to use

tensor

We use the template class tensor to manage basic data structures, and we use it as the following

pointi<2> shape = {10, 20};
tensor<int, 2> cts(shape, runtime::cuda);
tensor<int, 2> ts(shape, runtime::host);

tensor can change value_type and dimensions through template parameters, and specify the memory type o by runtime host or cuda.

lambda_tensor

lambda_tensor is a functional tensor which has strong representatility. we can define a lambda_tensor by a "index->value" function with shape(domain).

pointi<2> shape = {10, 10};
tensor<int, 2> ts(shape);
auto fun = [=](pointi<2> idx) {
    return ts(idx) + ts(idx);
};
lambda_tensor<decltype(fun)> lts(shape, fun);
// will evaluate fun(pointi<2>{2,2}), and return it;
auto value = lts(2, 2);

for_index

for_index is a functional loop, which supports host or cuda execution mode. note that in the cuda runtime, we can only access the cuda tensor.

pointi<2> shape = {10, 10};
tensor<int, 2> cts(shape, runtime::cuda);
auto lts = make_lambda_tensor(shape, [](pointi<2> idx)->int{
    return idx[0] * 10 + idx[1];
});

for_index(shape, [=](pointi<2> idx){
    cts(idx) = lts(idx);
}, cts.runtime()); // in cuda runtime, we can only access cuda runtime tensor

We can extend for_index that implement different execution policy, for example we can extend an openmp for_index to use the CPU's multi-core.

samples build

git submodule update --init .
cmake -B build

How to integrate

Download mtensor.hpp single header file, include it in your project.

#include "mtensor.hpp"

mtensor project is lite. when you find a bug, please fix it and pull a request.