[SYCL][NFC] Refactor #includes (#16030)

This patch is a collection of various cleanups made in public headers:
- Cleaned up many unnecessary includes. It doesn't change total amount
of header files we use in total by `sycl.hpp`, but makes our code
cleaner
- Made it so there are no headers (except for
`backend/%backend_name%.hpp`) depend on `backend.hpp` and it is (almost)
only included by `sycl.hpp`, so that we can make it an opt-in header
- Removed `types.hpp` in favor of direct use of `vector.hpp`
- Added missing includes and forward-declarations to places where we
relied on implicit includes
- Moved certain helper function declarations/definitions to better
places (common utils to utils headers, library-only declarations to
library headers, etc.)

Author: AlexeySachkov

libdevice/nativecpu_utils.cpp

@@ -17,7 +17,7 @@
 #include "device.h"
 #include <cstdint>
 #include <sycl/__spirv/spirv_ops.hpp>
-#include <sycl/types.hpp>
+#include <sycl/vector.hpp>
 
 // including state definition from Native CPU UR adapter
 #include "nativecpu_state.hpp"

sycl/include/sycl/aliases.hpp

@@ -9,7 +9,6 @@
 #pragma once
 
 #include <sycl/detail/defines_elementary.hpp> // for __SYCL2020_DEPRECATED
-#include <sycl/half_type.hpp>                 // for half
 
 #include <cstdint> // for uint8_t, int16_t, int32_t
 

sycl/include/sycl/backend.hpp

@@ -15,21 +15,17 @@
 #include <sycl/buffer.hpp>                    // for buffer_allocator
 #include <sycl/context.hpp>                   // for context, get_na...
 #include <sycl/detail/backend_traits.hpp>     // for InteropFeatureS...
-#include <sycl/detail/cl.h>                   // for _cl_event
 #include <sycl/detail/defines_elementary.hpp> // for __SYCL_DEPRECATED
 #include <sycl/detail/export.hpp>             // for __SYCL_EXPORT
 #include <sycl/detail/impl_utils.hpp>         // for createSyclObjFr...
 #include <sycl/device.hpp>                    // for device, get_native
 #include <sycl/event.hpp>                     // for event, get_native
 #include <sycl/exception.hpp>                 // for make_error_code
 #include <sycl/feature_test.hpp>              // for SYCL_BACKEND_OP...
-#include <sycl/handler.hpp>                   // for buffer
 #include <sycl/image.hpp>                     // for image, image_al...
-#include <sycl/kernel.hpp>                    // for kernel, get_native
 #include <sycl/kernel_bundle.hpp>             // for kernel_bundle
 #include <sycl/kernel_bundle_enums.hpp>       // for bundle_state
 #include <sycl/platform.hpp>                  // for platform, get_n...
-#include <sycl/property_list.hpp>             // for property_list
 #include <sycl/queue.hpp>                     // for queue, get_native
 #include <ur_api.h>                           // for ur_native_handle_t
 
@@ -60,13 +56,12 @@
 namespace sycl {
 inline namespace _V1 {
 
+class property_list;
+
 namespace detail {
 // TODO each backend can have its own custom errc enumeration
 // but the details for this are not fully specified yet
 enum class backend_errc : unsigned int {};
-
-// Convert from UR backend to SYCL backend enum
-backend convertUrBackend(ur_platform_backend_t UrBackend);
 } // namespace detail
 
 template <backend Backend> class backend_traits {
@@ -78,14 +73,6 @@ template <backend Backend> class backend_traits {
   using return_type = typename detail::BackendReturn<Backend, T>::type;
 };
 
-template <backend Backend, typename SyclType>
-using backend_input_t =
-    typename backend_traits<Backend>::template input_type<SyclType>;
-
-template <backend Backend, typename SyclType>
-using backend_return_t =
-    typename backend_traits<Backend>::template return_type<SyclType>;
-
 namespace detail {
 template <backend Backend, typename DataT, int Dimensions, typename AllocatorT>
 struct BufferInterop {

sycl/include/sycl/builtins_esimd.hpp

@@ -11,7 +11,7 @@
 #include <sycl/detail/common.hpp>
 #include <sycl/detail/generic_type_traits.hpp>
 #include <sycl/ext/intel/esimd/detail/math_intrin.hpp>
-#include <sycl/types.hpp>
+#include <sycl/vector.hpp>
 
 // TODO Decide whether to mark functions with this attribute.
 #define __NOEXC /*noexcept*/

sycl/include/sycl/builtins_utils_vec.hpp

@@ -13,8 +13,9 @@
 #include <sycl/builtins_utils_scalar.hpp>
 
 #include <sycl/detail/type_traits.hpp>
+#include <sycl/detail/vector_convert.hpp>
 #include <sycl/marray.hpp> // for marray
-#include <sycl/types.hpp>  // for vec
+#include <sycl/vector.hpp> // for vec
 
 namespace sycl {
 inline namespace _V1 {

sycl/include/sycl/detail/backend_traits.hpp

@@ -8,10 +8,9 @@
 
 #pragma once
 
-#include <sycl/backend_types.hpp>
-
 namespace sycl {
 inline namespace _V1 {
+enum class backend : char;
 namespace detail {
 
 template <backend Backend, typename SYCLObjectT> struct interop;

sycl/include/sycl/detail/device_filter.hpp

@@ -8,7 +8,6 @@
 
 #pragma once
 
-#include <sycl/backend_types.hpp>
 #include <sycl/detail/defines.hpp>
 #include <sycl/info/info_desc.hpp>
 
@@ -18,6 +17,7 @@
 
 namespace sycl {
 inline namespace _V1 {
+enum class backend : char;
 namespace detail {
 
 // ---------------------------------------

sycl/include/sycl/detail/image_accessor_util.hpp

@@ -23,7 +23,7 @@
 #include <sycl/image.hpp>                      // for image_channel_type
 #include <sycl/range.hpp>                      // for range
 #include <sycl/sampler.hpp>                    // for addressing_mode, coor...
-#include <sycl/types.hpp>                      // for vec, operator*, round...
+#include <sycl/vector.hpp>                     // for vec, operator*, round...
 
 #include <cstdint>     // for int32_t, uint16_t
 #include <stddef.h>    // for size_t

sycl/include/sycl/detail/ur.hpp

@@ -14,7 +14,6 @@
 
 #pragma once
 
-#include <sycl/backend_types.hpp>
 #include <sycl/detail/export.hpp>
 #include <sycl/detail/os_util.hpp>
 #include <ur_api.h>
@@ -42,6 +41,7 @@ struct trace_event_data_t;
 namespace sycl {
 inline namespace _V1 {
 
+enum class backend : char;
 class context;
 
 namespace detail {

sycl/include/sycl/detail/util.hpp

@@ -15,6 +15,9 @@
 
 #include <cstring>
 #include <mutex>
+#include <string>
+#include <type_traits>
+#include <utility>
 #include <vector>
 
 namespace sycl {
@@ -83,6 +86,40 @@ template <> struct ABINeutralT<std::vector<std::string>> {
 };
 
 template <typename T> using ABINeutralT_t = typename ABINeutralT<T>::type;
+
+template <typename ParamT> auto convert_to_abi_neutral(ParamT &&Info) {
+  using ParamNoRef = std::remove_reference_t<ParamT>;
+  if constexpr (std::is_same_v<ParamNoRef, std::string>) {
+    return detail::string{Info};
+  } else if constexpr (std::is_same_v<ParamNoRef, std::vector<std::string>>) {
+    std::vector<detail::string> Res;
+    Res.reserve(Info.size());
+    for (std::string &Str : Info) {
+      Res.push_back(detail::string{Str});
+    }
+    return Res;
+  } else {
+    return std::forward<ParamT>(Info);
+  }
+}
+
+template <typename ParamT> auto convert_from_abi_neutral(ParamT &&Info) {
+  using ParamNoRef = std::remove_reference_t<ParamT>;
+  if constexpr (std::is_same_v<ParamNoRef, detail::string>) {
+    return Info.c_str();
+  } else if constexpr (std::is_same_v<ParamNoRef,
+                                      std::vector<detail::string>>) {
+    std::vector<std::string> Res;
+    Res.reserve(Info.size());
+    for (detail::string &Str : Info) {
+      Res.push_back(Str.c_str());
+    }
+    return Res;
+  } else {
+    return std::forward<ParamT>(Info);
+  }
+}
+
 } // namespace detail
 } // namespace _V1
 } // namespace sycl

sycl/include/sycl/device.hpp

@@ -8,7 +8,6 @@
 
 #pragma once
 
-#include <sycl/aspects.hpp>
 #include <sycl/backend_types.hpp>
 #include <sycl/detail/defines_elementary.hpp>
 #include <sycl/detail/export.hpp>
@@ -21,9 +20,12 @@
 #include <sycl/ext/oneapi/experimental/device_architecture.hpp>
 #include <sycl/info/info_desc.hpp>
 #include <sycl/kernel_bundle_enums.hpp>
-#include <sycl/platform.hpp>
 #include <ur_api.h>
 
+#ifdef __SYCL_INTERNAL_API
+#include <sycl/detail/cl.h>
+#endif
+
 #include <cstddef>
 #include <memory>
 #include <string>
@@ -35,7 +37,7 @@
 namespace sycl {
 inline namespace _V1 {
 // Forward declarations
-class device_selector;
+class platform;
 template <backend BackendName, class SyclObjectT>
 auto get_native(const SyclObjectT &Obj)
     -> backend_return_t<BackendName, SyclObjectT>;

sycl/include/sycl/ext/intel/esimd/detail/memory_intrin.hpp

@@ -24,7 +24,7 @@
 #include <sycl/ext/intel/esimd/common.hpp>
 #include <sycl/ext/intel/esimd/detail/types.hpp>
 #include <sycl/ext/intel/esimd/detail/util.hpp>
-#include <sycl/types.hpp>
+#include <sycl/vector.hpp>
 
 #include <cstdint>
 

sycl/include/sycl/ext/intel/fpga_device_selector.hpp

@@ -10,6 +10,7 @@
 
 #include <sycl/device.hpp>
 #include <sycl/device_selector.hpp>
+#include <sycl/platform.hpp>
 
 #include <string>
 #include <string_view>

sycl/include/sycl/ext/oneapi/backend/hip.hpp

@@ -8,7 +8,7 @@
 
 #pragma once
 
-#include <sycl/backend.hpp>
+#include <sycl/backend_types.hpp>
 #include <sycl/detail/backend_traits_hip.hpp>
 
 namespace sycl {

sycl/include/sycl/ext/oneapi/backend/level_zero.hpp

@@ -9,8 +9,7 @@
 #pragma once
 
 #include <sycl/async_handler.hpp>                           // for async_han...
-#include <sycl/backend.hpp>                                 // for backend_i...
-#include <sycl/backend_types.hpp>                           // for backend
+#include <sycl/backend.hpp>                                 // for backend
 #include <sycl/buffer.hpp>                                  // for buffer_al...
 #include <sycl/buffer.hpp>                                  // for buffer
 #include <sycl/context.hpp>                                 // for context

sycl/include/sycl/ext/oneapi/dot_product.hpp

@@ -11,7 +11,7 @@
 #pragma once
 
 #include <sycl/detail/defines_elementary.hpp>
-#include <sycl/types.hpp>
+#include <sycl/vector.hpp>
 
 namespace sycl {
 inline namespace _V1 {

sycl/include/sycl/ext/oneapi/experimental/bfloat16_math.hpp

@@ -8,9 +8,10 @@
 
 #pragma once
 
-#include <sycl/builtins.hpp>            // for ceil, cos, exp, exp10, exp2
-#include <sycl/builtins_utils_vec.hpp>  // For simplify_if_swizzle, is_swizzle
-#include <sycl/detail/memcpy.hpp>       // sycl::detail::memcpy
+#include <sycl/builtins.hpp>           // for ceil, cos, exp, exp10, exp2
+#include <sycl/builtins_utils_vec.hpp> // For simplify_if_swizzle, is_swizzle
+#include <sycl/detail/memcpy.hpp>      // sycl::detail::memcpy
+#include <sycl/detail/vector_convert.hpp>
 #include <sycl/ext/oneapi/bfloat16.hpp> // for bfloat16, bfloat16ToBits
 #include <sycl/marray.hpp>              // for marray
 

sycl/include/sycl/ext/oneapi/experimental/builtins.hpp

@@ -15,7 +15,7 @@
 #include <sycl/detail/generic_type_traits.hpp> // for is_svgenfloath, is_sv...
 #include <sycl/detail/memcpy.hpp>              // detail::memcpy
 #include <sycl/marray.hpp>                     // for marray
-#include <sycl/types.hpp>                      // for vec
+#include <sycl/vector.hpp>                     // for vec
 
 #include <cstring>     // for size_t
 #include <stdio.h>     // for printf

sycl/include/sycl/ext/oneapi/experimental/cuda/builtins.hpp

@@ -10,7 +10,7 @@
 
 #define SYCL_EXT_ONEAPI_CUDA_TEX_CACHE_READ 1
 
-#include <sycl/types.hpp>
+#include <sycl/vector.hpp>
 
 #if defined(_WIN32) || defined(_WIN64)
 #define ATTRIBUTE_EXT_VEC_TYPE(N) __declspec(ext_vector_type(N))

sycl/include/sycl/ext/oneapi/experimental/non_uniform_groups.hpp

@@ -10,7 +10,7 @@
 
 #include <sycl/ext/oneapi/sub_group_mask.hpp> // for sub_group_mask
 #include <sycl/marray.hpp>                    // for marray
-#include <sycl/types.hpp>                     // for vec
+#include <sycl/vector.hpp>                    // for vec
 
 #include <stddef.h> // for size_t
 #include <stdint.h> // for uint32_t

sycl/include/sycl/ext/oneapi/experimental/prefetch.hpp

@@ -10,7 +10,7 @@
 
 #include <sycl/__spirv/spirv_ops.hpp>
 #include <sycl/ext/oneapi/properties/properties.hpp>
-#include <sycl/types.hpp>
+#include <sycl/vector.hpp>
 
 namespace sycl {
 inline namespace _V1 {

sycl/include/sycl/ext/oneapi/sub_group_mask.hpp

@@ -14,7 +14,7 @@
 #include <sycl/feature_test.hpp>   // for SYCL_EXT_ONEAPI_SUB_GROUP_MASK
 #include <sycl/id.hpp>             // for id
 #include <sycl/marray.hpp>         // for marray
-#include <sycl/types.hpp>          // for vec
+#include <sycl/vector.hpp>         // for vec
 
 #include <assert.h>     // for assert
 #include <climits>      // for CHAR_BIT

sycl/include/sycl/group_algorithm.hpp

@@ -20,7 +20,7 @@
 #include <sycl/known_identity.hpp>     // for known_identity_v
 #include <sycl/nd_item.hpp>            // for nd_item
 #include <sycl/range.hpp>              // for range
-#include <sycl/types.hpp>              // for vec
+#include <sycl/vector.hpp>             // for vec
 
 #ifdef __SYCL_DEVICE_ONLY__
 #include <sycl/ext/oneapi/functional.hpp>

sycl/include/sycl/image.hpp

@@ -28,7 +28,7 @@
 #include <sycl/property_list.hpp>                     // for property_list
 #include <sycl/range.hpp>                             // for range, rangeTo...
 #include <sycl/sampler.hpp>                           // for image_sampler
-#include <sycl/types.hpp>                             // for vec
+#include <sycl/vector.hpp>                            // for vec
 #include <ur_api.h>                                   // for ur_native_hand...
 
 #include <cstddef>     // for size_t, nullptr_t

sycl/include/sycl/interop_handle.hpp

@@ -11,16 +11,12 @@
 #include <sycl/access/access.hpp>     // for target, mode, place...
 #include <sycl/accessor.hpp>          // for AccessorBaseHost
 #include <sycl/backend_types.hpp>     // for backend, backend_re...
-#include <sycl/context.hpp>           // for context
+#include <sycl/buffer.hpp>            // for buffer
 #include <sycl/detail/export.hpp>     // for __SYCL_EXPORT
-#include <sycl/detail/helpers.hpp>    // for context_impl
 #include <sycl/detail/impl_utils.hpp> // for getSyclObjImpl
-#include <sycl/device.hpp>            // for device, device_impl
 #include <sycl/exception.hpp>
-#include <sycl/exception_list.hpp>                    // for queue_impl
 #include <sycl/ext/oneapi/accessor_property_list.hpp> // for accessor_property_list
 #include <sycl/image.hpp>                             // for image
-#include <sycl/properties/buffer_properties.hpp>      // for buffer
 #include <ur_api.h> // for ur_mem_handle_t, ur...
 
 #include <memory>      // for shared_ptr

sycl/include/sycl/kernel.hpp

@@ -8,28 +8,26 @@
 
 #pragma once
 
-#include <cstddef>                            // for size_t
-#include <memory>                             // for shared_ptr, hash, opera...
 #include <sycl/backend_types.hpp>             // for backend, backend_return_t
-#include <sycl/context.hpp>                   // for context
 #include <sycl/detail/defines_elementary.hpp> // for __SYCL2020_DEPRECATED
 #include <sycl/detail/export.hpp>             // for __SYCL_EXPORT
 #include <sycl/detail/info_desc_helpers.hpp>  // for is_kernel_device_specif...
 #include <sycl/detail/owner_less_base.hpp>    // for OwnerLessBase
-#include <sycl/detail/string.hpp>
-#include <sycl/detail/string_view.hpp>
 #include <sycl/detail/util.hpp>
-#include <sycl/device.hpp>              // for device
 #include <sycl/kernel_bundle_enums.hpp> // for bundle_state
-#include <sycl/range.hpp>               // for range
 #include <ur_api.h>                     // for ur_native_handle_t
-#include <variant>                      // for hash
+
+#include <cstddef> // for size_t
+#include <memory>  // for shared_ptr, hash, opera...
+#include <variant> // for hash
 
 namespace sycl {
 inline namespace _V1 {
 // Forward declaration
 class context;
 class queue;
+class device;
+template <int Dimensions> class range;
 template <backend Backend> class backend_traits;
 template <bundle_state State> class kernel_bundle;
 template <backend BackendName, class SyclObjectT>