[SYCL] Implement sycl_ext_oneapi_address_cast #12382
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,46 @@ | ||
| //==----------- address_cast.hpp - sycl_ext_oneapi_address_cast ------------==// | ||
| // | ||
| // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. | ||
| // See https://llvm.org/LICENSE.txt for license information. | ||
| // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception | ||
| // | ||
| //===----------------------------------------------------------------------===// | ||
|
|
||
| #pragma once | ||
| #include <sycl/multi_ptr.hpp> | ||
|
|
||
| namespace sycl { | ||
| inline namespace _V1 { | ||
| namespace ext { | ||
| namespace oneapi { | ||
| namespace experimental { | ||
|
|
||
| template <access::address_space Space, access::decorated DecorateAddress, | ||
| typename ElementType> | ||
| multi_ptr<ElementType, Space, DecorateAddress> | ||
| static_address_cast(ElementType *Ptr) { | ||
| #ifdef __SYCL_DEVICE_ONLY__ | ||
| auto CastPtr = sycl::detail::spirv::GenericCastToPtr<Space>(Ptr); | ||
| return multi_ptr<ElementType, Space, DecorateAddress>(CastPtr); | ||
| #else | ||
| return multi_ptr<ElementType, Space, DecorateAddress>(Ptr); | ||
| #endif | ||
| } | ||
|
|
||
| template <access::address_space Space, access::decorated DecorateAddress, | ||
| typename ElementType> | ||
| multi_ptr<ElementType, Space, DecorateAddress> | ||
| dynamic_address_cast(ElementType *Ptr) { | ||
| #ifdef __SYCL_DEVICE_ONLY__ | ||
| auto CastPtr = sycl::detail::spirv::GenericCastToPtrExplicit<Space>(Ptr); | ||
| return multi_ptr<ElementType, Space, DecorateAddress>(CastPtr); | ||
| #else | ||
| return multi_ptr<ElementType, Space, DecorateAddress>(Ptr); | ||
| #endif | ||
| } | ||
|
|
||
| } // namespace experimental | ||
| } // namespace oneapi | ||
| } // namespace ext | ||
| } // namespace _V1 | ||
| } // namespace sycl |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,112 @@ | ||
| //==--------- dynamic_address_cast.cpp - dynamic address_cast test ---------==// | ||
| // | ||
| // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. | ||
| // See https://llvm.org/LICENSE.txt for license information. | ||
| // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception | ||
| // | ||
| //===----------------------------------------------------------------------===// | ||
|
|
||
| // Issue with OpenCL CPU runtime implementation of OpGenericCastToPtrExplicit | ||
| // OpGenericCastToPtr* intrinsics not implemented on AMD or NVIDIA | ||
| // UNSUPPORTED: cpu, hip, cuda | ||
| // RUN: %{build} -o %t.out | ||
| // RUN: %{run} %t.out | ||
| #include <sycl/sycl.hpp> | ||
|
|
||
| int main() { | ||
|
|
||
| sycl::queue Queue; | ||
|
|
||
| sycl::range<1> NItems{1}; | ||
|
|
||
| sycl::buffer<int, 1> GlobalBuffer{NItems}; | ||
| sycl::buffer<bool, 1> ResultBuffer{NItems}; | ||
|
|
||
| Queue | ||
| .submit([&](sycl::handler &cgh) { | ||
| auto GlobalAccessor = | ||
| GlobalBuffer.get_access<sycl::access::mode::read_write>(cgh); | ||
| auto LocalAccessor = sycl::local_accessor<int>(1, cgh); | ||
| auto ResultAccessor = | ||
| ResultBuffer.get_access<sycl::access::mode::write>(cgh); | ||
| cgh.parallel_for<class Kernel>( | ||
| sycl::nd_range<1>(NItems, 1), [=](sycl::nd_item<1> Item) { | ||
| bool Success = true; | ||
| size_t Index = Item.get_global_id(0); | ||
|
|
||
| int *RawGlobalPointer = &GlobalAccessor[Index]; | ||
| { | ||
| auto GlobalPointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::global_space, | ||
| sycl::access::decorated::no>(RawGlobalPointer); | ||
| auto LocalPointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::local_space, | ||
| sycl::access::decorated::no>(RawGlobalPointer); | ||
| auto PrivatePointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::private_space, | ||
| sycl::access::decorated::no>(RawGlobalPointer); | ||
| Success &= reinterpret_cast<size_t>(RawGlobalPointer) == | ||
| reinterpret_cast<size_t>(GlobalPointer.get_raw()); | ||
| Success &= LocalPointer.get_raw() == nullptr; | ||
| Success &= PrivatePointer.get_raw() == nullptr; | ||
|
Comment on lines
+53
to
+54
There was a problem hiding this comment. @mmoadeli for your awareness, your patch is going to be required to enable this test on cuda and hip |
||
| } | ||
|
|
||
| int *RawLocalPointer = &LocalAccessor[0]; | ||
| { | ||
| auto GlobalPointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::global_space, | ||
| sycl::access::decorated::no>(RawLocalPointer); | ||
| auto LocalPointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::local_space, | ||
| sycl::access::decorated::no>(RawLocalPointer); | ||
| auto PrivatePointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::private_space, | ||
| sycl::access::decorated::no>(RawLocalPointer); | ||
| Success &= GlobalPointer.get_raw() == nullptr; | ||
| Success &= reinterpret_cast<size_t>(RawLocalPointer) == | ||
| reinterpret_cast<size_t>(LocalPointer.get_raw()); | ||
| Success &= PrivatePointer.get_raw() == nullptr; | ||
| } | ||
|
|
||
| int PrivateVariable = 0; | ||
| int *RawPrivatePointer = &PrivateVariable; | ||
| { | ||
| auto GlobalPointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::global_space, | ||
| sycl::access::decorated::no>(RawPrivatePointer); | ||
| auto LocalPointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::local_space, | ||
| sycl::access::decorated::no>(RawPrivatePointer); | ||
| auto PrivatePointer = | ||
| sycl::ext::oneapi::experimental::dynamic_address_cast< | ||
| sycl::access::address_space::private_space, | ||
| sycl::access::decorated::no>(RawPrivatePointer); | ||
| Success &= GlobalPointer.get_raw() == nullptr; | ||
| Success &= LocalPointer.get_raw() == nullptr; | ||
| Success &= reinterpret_cast<size_t>(RawPrivatePointer) == | ||
| reinterpret_cast<size_t>(PrivatePointer.get_raw()); | ||
| } | ||
|
|
||
| ResultAccessor[Index] = Success; | ||
| }); | ||
| }) | ||
| .wait(); | ||
|
|
||
| bool Success = true; | ||
| { | ||
| auto ResultAccessor = sycl::host_accessor(ResultBuffer); | ||
| for (int i = 0; i < NItems.size(); ++i) { | ||
| Success &= ResultAccessor[i]; | ||
| }; | ||
| } | ||
|
|
||
| return (Success) ? 0 : -1; | ||
| } | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,75 @@ | ||
| //==---------- static_address_cast.cpp - static address_cast test ----------==// | ||
| // | ||
| // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. | ||
| // See https://llvm.org/LICENSE.txt for license information. | ||
| // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception | ||
| // | ||
| //===----------------------------------------------------------------------===// | ||
|
|
||
| // OpGenericCastToPtr* intrinsics not implemented on AMD or NVIDIA | ||
| // UNSUPPORTED: hip, cuda | ||
| // RUN: %{build} -o %t.out | ||
| // RUN: %{run} %t.out | ||
| #include <sycl/sycl.hpp> | ||
|
|
||
| int main() { | ||
|
|
||
| sycl::queue Queue; | ||
|
|
||
| sycl::range<1> NItems{1}; | ||
|
|
||
| sycl::buffer<int, 1> GlobalBuffer{NItems}; | ||
| sycl::buffer<bool, 1> ResultBuffer{NItems}; | ||
|
|
||
| Queue | ||
| .submit([&](sycl::handler &cgh) { | ||
| auto GlobalAccessor = | ||
| GlobalBuffer.get_access<sycl::access::mode::read_write>(cgh); | ||
| auto LocalAccessor = sycl::local_accessor<int>(1, cgh); | ||
| auto ResultAccessor = | ||
| ResultBuffer.get_access<sycl::access::mode::write>(cgh); | ||
| cgh.parallel_for<class Kernel>( | ||
| sycl::nd_range<1>(NItems, 1), [=](sycl::nd_item<1> Item) { | ||
| bool Success = true; | ||
| size_t Index = Item.get_global_id(0); | ||
|
|
||
| int *RawGlobalPointer = &GlobalAccessor[Index]; | ||
| auto GlobalPointer = | ||
| sycl::ext::oneapi::experimental::static_address_cast< | ||
| sycl::access::address_space::global_space, | ||
| sycl::access::decorated::no>(RawGlobalPointer); | ||
| Success &= reinterpret_cast<size_t>(RawGlobalPointer) == | ||
| reinterpret_cast<size_t>(GlobalPointer.get_raw()); | ||
|
|
||
| int *RawLocalPointer = &LocalAccessor[0]; | ||
| auto LocalPointer = | ||
| sycl::ext::oneapi::experimental::static_address_cast< | ||
| sycl::access::address_space::local_space, | ||
| sycl::access::decorated::no>(RawLocalPointer); | ||
| Success &= reinterpret_cast<size_t>(RawLocalPointer) == | ||
| reinterpret_cast<size_t>(LocalPointer.get_raw()); | ||
|
|
||
| int PrivateVariable = 0; | ||
| int *RawPrivatePointer = &PrivateVariable; | ||
| auto PrivatePointer = | ||
| sycl::ext::oneapi::experimental::static_address_cast< | ||
| sycl::access::address_space::private_space, | ||
| sycl::access::decorated::no>(RawPrivatePointer); | ||
| Success &= reinterpret_cast<size_t>(RawPrivatePointer) == | ||
| reinterpret_cast<size_t>(PrivatePointer.get_raw()); | ||
|
|
||
| ResultAccessor[Index] = Success; | ||
| }); | ||
| }) | ||
| .wait(); | ||
|
|
||
| bool Success = true; | ||
| { | ||
| auto ResultAccessor = sycl::host_accessor(ResultBuffer); | ||
| for (int i = 0; i < NItems.size(); ++i) { | ||
| Success &= ResultAccessor[i]; | ||
| }; | ||
| } | ||
|
|
||
| return (Success) ? 0 : -1; | ||
| } |
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
minor: no need for cast, the returned type should be already correctly typed
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Good point! I just copied the existing definitions of
__SYCL_GenericCastToPtrExplicithere.I agree it makes sense to change this, but I'd prefer for it to be done as part of a follow-up PR. I don't really want to touch the existing definitions, in case something breaks unexpectedly.