Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/adaptivecpp/featuresupport


https://github.com/adaptivecpp/featuresupport

Last synced: 5 months ago
JSON representation

Awesome Lists containing this project

README 

          
# AdaptiveCpp feature support



This page summarizes the state of SYCL feature support in the current `develop` branch of AdaptiveCpp. Features that are supported are listed with a link to the pull request where they have been merged.



## SYCL 1.2.1 features

(This list is incomplete and only contains features that are known to be problematic)



| Feature | Supported (PR link) | Caveats | Comments |

| --- | --- | --- | --- |

| Images | :x: | --- | --- |

| OpenCL interop | :x: | --- | --- |

| Hierarchical parallelism | :heavy_check_mark: | HIP/CUDA: Does not limit execution in work group scope to one thread for performance reasons | |



## SYCL 2020 features



| Feature | Supported (PR link) | Caveats | Comments |

| --- | --- | --- | --- |

| Accessor simplifications | :heavy_check_mark: (partial) ([PR](https://github.com/illuhad/hipSYCL/pull/490)) | [6] |  |

| USM: Memory management functions | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/308))| [1] | |

| USM: Queue shortcuts | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/323)) | | |

| USM: Prefetch | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/323)) | [2] | |

| USM: mem_advise | :x: |  | Implementation requires host tasks since backends do not provide async mem advise |

| USM: memcpy | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/323)) | | |

| USM: memset/fill | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/323)) | | |

| host tasks | :x: |  |  |

| Optional lambda naming | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/281)) | | |

| Subgroups | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/282)) | | On CPU, subgroup size is always 1 |

| In-order queues | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/320)) | | |

| Explicit dependencies (`depends_on()`) | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/320)) | | |

| Backend interop API | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/327)) | [3] | |

| Reductions | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/374)) | [4] | |

| Group algorithms | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/391)) | [5] | |

| New device selector API | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/531)) | | |

| Aspect API | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/530)) | | |

| Deduction guides | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/388)) | | |

| `atomic_ref` | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/584)) | | |

| `marray` | :x: | | |

| New `SYCL/sycl.hpp` header | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/216)) | | |

| C++17 by default | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/206)) | | |

| Builtin changes: `ctz()`, `clz()` | :x: | | |

| Remove `*_class` types | :x: | | |

| `const` return type for read accessor `operator[]` | :x: | | |

| Remove buffer API for `unique_ptr` | :x: | | |

| Replace `program` class with `module` | :x: | | |

| Add `kernel_handler` | :x: | | |

| explicit `queue`, `context` constructors | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/328)) | | |

| Only require C++ trivially copyable for shared data | :heavy_check_mark: | | Has always worked thanks to CUDA/HIP toolchain |

| Update group class with new types/member functions | :x: | | |

| Remove `nd_item::barrier()` | :x: | | |

| Replace `mem_fence` with `atomic_fence` | :x: | | |

| Add `vec::operator[]`,unary `+,-`, `static constexpr get_size()/get_count()` | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/415)) | | |

| buffer, local accessor are C++ `ContiguousContainer` | :x: | | |

| Replace `image` with `sampled_image`, `unsampled_image` | :x: | | |

| All accessors are placeholders | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/490)) | | |

| Use single exception type derived from `std::exception` | :x: | | |

| Default asynchronous handler should terminate program  | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/289)) | | |

| Kernel invocation APIs take const reference to kernels, kernels must be immutable | :x: | | |

| Queue constructor accepting both `device` and `context` | :heavy_check_mark: ([PR](https://github.com/illuhad/hipSYCL/pull/447)) | | |

| Simplified `parallel_for` API | :x: | | |

| Clarified names for device specific info queries | :x: | | |

| Address space changes, generic address spaces | :x: | | Partially, we have always had generic address spaces because of CUDA/HIP |

| Updated `multi_ptr` interface | :x: | | |

| Remove OpenCL types, `cl_int` etc | :heavy_check_mark: | | hipSYCL has stopped supporting them a long time ago |



* [1] HIP/ROCm implements unified memory using slow device accessible host memory. This means that hipSYCL's call to `hipMallocManaged` cannot produce efficient shared allocations.

* [2] HIP/ROCm does not provide the required functionality, so hipSYCL cannot expose it. Prefetch calls are ignored at the moment.

* [3] The interop types that backends expose is limited. Native queues can only be obtained using an `interop_handle` because a queue in hipSYCL does not relate to any specific backend object.

* [4] Only scalar reductions are supported. Note that the reduction interface is expected to change slightly with the release of SYCL 2020 final.

* [5] Note that the interface of group algorithms is expected to change slightly with the release of SYCL 2020 final.

* [6] Constructing read-only accessor using `accessor` from non-const `buffer` is not yet supported.