devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.

Browse Source
This commit is contained in:
Mehdi Goli
2016-11-18 16:20:42 +00:00
parent 5159675c33
commit 622805a0c5
13 changed files with 439 additions and 358 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
unsupported/Eigen/CXX11/src/Tensor/TensorSyclRun.h
View File

@@ -37,11 +37,11 @@ void run(Expr &expr, Dev &dev) {
typedef typename internal::createPlaceHolderExpression<Expr>::Type PlaceHolderExpr;
auto functors = internal::extractFunctors(evaluator);
dev.m_queue.submit([&](cl::sycl::handler &cgh) {
dev.sycl_queue().submit([&](cl::sycl::handler &cgh) {
// create a tuple of accessors from Evaluator
auto tuple_of_accessors = internal::createTupleOfAccessors<decltype(evaluator)>(cgh, evaluator);
size_t range, GRange, tileSize;
dev.parallel_for_setup(utility::tuple::get<0>(tuple_of_accessors).get_range()[0], tileSize, range, GRange);
dev.parallel_for_setup(utility::tuple::get<0>(tuple_of_accessors).get_range()[0]/sizeof(typename Expr::Scalar), tileSize, range, GRange);
// run the kernel
cgh.parallel_for<PlaceHolderExpr>( cl::sycl::nd_range<1>(cl::sycl::range<1>(GRange), cl::sycl::range<1>(tileSize)), [=](cl::sycl::nd_item<1> itemID) {
@@ -49,11 +49,11 @@ void run(Expr &expr, Dev &dev) {
auto device_expr =internal::createDeviceExpression<DevExpr, PlaceHolderExpr>(functors, tuple_of_accessors);
auto device_evaluator = Eigen::TensorEvaluator<decltype(device_expr.expr), Eigen::DefaultDevice>(device_expr.expr, Eigen::DefaultDevice());
if (itemID.get_global_linear_id() < range) {
device_evaluator.evalScalar(static_cast<int>(itemID.get_global_linear_id()));
device_evaluator.evalScalar(static_cast<typename DevExpr::Index>(itemID.get_global_linear_id()));
}
});
});
dev.m_queue.throw_asynchronous();
dev.sycl_queue().throw_asynchronous();
}
evaluator.cleanup();