DPC++ Runtime
Runtime libraries for oneAPI DPC++
pi.h
Go to the documentation of this file.
1 //==---------- pi.h - Plugin Interface -------------------------------------==//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
10 // TODO: link to sphinx page
11 
21 
22 #ifndef _PI_H_
23 #define _PI_H_
24 
25 // Every single change in PI API should be accompanied with the minor
26 // version increase (+1). In the cases where backward compatibility is not
27 // maintained there should be a (+1) change to the major version in
28 // addition to the increase of the minor.
29 //
30 // PI version changes log:
31 // -- Version 1.2:
32 // 1. (Binary backward compatibility breaks) Two fields added to the
33 // pi_device_binary_struct structure:
34 // pi_device_binary_property_set PropertySetsBegin;
35 // pi_device_binary_property_set PropertySetsEnd;
36 // 2. A number of types needed to define pi_device_binary_property_set added.
37 // 3. Added new ownership argument to piextContextCreateWithNativeHandle.
38 // 4. Add interoperability interfaces for kernel.
39 // 4.6 Added new ownership argument to piextQueueCreateWithNativeHandle which
40 // changes the API version from 3.5 to 4.6.
41 // 5.7 Added new context and ownership arguments to
42 // piextEventCreateWithNativeHandle
43 // 6.8 Added new ownership argument to piextProgramCreateWithNativeHandle. Added
44 // piQueueFlush function.
45 // 7.9 Added new context and ownership arguments to
46 // piextMemCreateWithNativeHandle.
47 // 8.10 Added new optional device argument to piextQueueCreateWithNativeHandle
48 // 9.11 Use values of OpenCL enums directly, rather than including `<CL/cl.h>`;
49 // NOTE that this results in a changed API for `piProgramGetBuildInfo`.
50 // 10.12 Change enum value PI_MEM_ADVICE_UNKNOWN from 0 to 999, and set enum
51 // PI_MEM_ADVISE_RESET to 0.
52 // 10.13 Added new PI_EXT_ONEAPI_QUEUE_DISCARD_EVENTS queue property.
53 // 10.14 Add PI_EXT_INTEL_DEVICE_INFO_FREE_MEMORY as an extension for
54 // piDeviceGetInfo.
55 // 11.15 piEventCreate creates even in the signalled state now.
56 // 11.16 Add PI_EXT_INTEL_DEVICE_INFO_MEMORY_CLOCK_RATE and
57 // PI_EXT_INTEL_DEVICE_INFO_MEMORY_BUS_WIDTH as an extension for
58 // piDeviceGetInfo.
59 // 11.17 Added new PI_EXT_ONEAPI_QUEUE_PRIORITY_LOW and
60 // PI_EXT_ONEAPI_QUEUE_PRIORITY_HIGH queue properties.
61 // 11.18 Add new parameter name PI_EXT_ONEAPI_QUEUE_INFO_EMPTY to
62 // _pi_queue_info.
63 // 12.19 Add new PI_EXT_INTEL_DEVICE_PARTITION_BY_CSLICE piDevicePartition
64 // scheme. Sub-sub-devices (representing compute slice) creation via
65 // partitioning by affinity domain is disabled by default and can be temporarily
66 // restored via SYCL_PI_LEVEL_ZERO_EXPOSE_CSLICE_IN_AFFINITY_PARTITIONING
67 // environment variable.
68 // 12.20 Added piextQueueCreate API to be used instead of piQueueCreate, also
69 // added PI_EXT_INTEL_DEVICE_INFO_MAX_COMPUTE_QUEUE_INDICES for piDeviceGetInfo.
70 // Both are needed to support sycl_ext_intel_queue_index extension.
71 // 12.21 Added new piextUSMEnqueueFill2D, piextUSMEnqueueMemset2D, and
72 // piextUSMEnqueueMemcpy2D functions. Added new
73 // PI_EXT_ONEAPI_CONTEXT_INFO_USM_FILL2D_SUPPORT,
74 // PI_EXT_ONEAPI_CONTEXT_INFO_USM_MEMSET2D_SUPPORT, and
75 // PI_EXT_ONEAPI_CONTEXT_INFO_USM_MEMCPY2D_SUPPORT context info query
76 // descriptors.
77 // 12.22 Add piGetDeviceAndHostTimer to query device wall-clock timestamp
78 // 12.23 Added new piextEnqueueDeviceGlobalVariableWrite and
79 // piextEnqueueDeviceGlobalVariableRead functions.
80 // 12.24 Added new PI_EXT_KERNEL_EXEC_INFO_CACHE_CONFIG property to the
81 // _pi_kernel_exec_info. Defined _pi_kernel_cache_config enum with values of
82 // the new PI_EXT_KERNEL_EXEC_INFO_CACHE_CONFIG property.
83 // 12.25 Added PI_EXT_DEVICE_INFO_ATOMIC_FENCE_ORDER_CAPABILITIES and
84 // PI_EXT_DEVICE_INFO_ATOMIC_FENCE_SCOPE_CAPABILITIES for piDeviceGetInfo.
85 // 12.26 Added piextEnqueueReadHostPipe and piextEnqueueWriteHostPipe functions.
86 // 12.27 Added properties parameter to piextQueueCreateWithNativeHandle and
87 // changed native handle type of piextQueueCreateWithNativeHandle and
88 // piextQueueGetNativeHandle
89 // 12.28 Added piextMemImageCreateWithNativeHandle for creating images from
90 // native handles.
91 // 12.29 Support PI_EXT_PLATFORM_INFO_BACKEND query in piPlatformGetInfo
92 // 12.30 Added PI_EXT_INTEL_DEVICE_INFO_MEM_CHANNEL_SUPPORT device info query.
93 // 12.31 Added PI_EXT_CODEPLAY_DEVICE_INFO_MAX_REGISTERS_PER_WORK_GROUP device
94 // info query.
95 // 13.32 Removed backwards compatibility of piextQueueCreateWithNativeHandle and
96 // piextQueueGetNativeHandle
97 // 14.33 Added new parameter (memory object properties) to
98 // piextKernelSetArgMemObj
99 // 14.34 Added command-buffer extension methods
100 // 14.35 Added piextEnablePeerAccess, piextDisablePeerAccess,
101 // piextPeerAccessGetInfo, and pi_peer_attr enum.
102 // 14.36 Adding support for experimental bindless images. This includes:
103 // - Added device info queries
104 // - Device queries for bindless image support
105 // - PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_SUPPORT
106 // - PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_SHARED_USM_SUPPORT
107 // - PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_1D_USM_SUPPORT
108 // - PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_2D_USM_SUPPORT
109 // - Device queries for pitched USM allocations
110 // - PI_EXT_ONEAPI_DEVICE_INFO_IMAGE_PITCH_ALIGN
111 // - PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_WIDTH
112 // - PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_HEIGHT
113 // - PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_PITCH
114 // - Device queries for mipmap image support
115 // - PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_SUPPORT
116 // - PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_ANISOTROPY_SUPPORT
117 // - PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_MAX_ANISOTROPY
118 // - PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_LEVEL_REFERENCE_SUPPORT
119 // - Device queries for interop memory support
120 // - PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_MEMORY_IMPORT_SUPPORT
121 // - PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_MEMORY_EXPORT_SUPPORT
122 // - PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_SEMAPHORE_IMPORT_SUPPORT
123 // - PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_SEMAPHORE_EXPORT_SUPPORT
124 // - Added PI_IMAGE_INFO_DEPTH to _pi_image_info
125 // - Added _pi_image_copy_flags enum to determine direction of copy
126 // - Added new extension functions
127 // - piextBindlessImageSamplerCreate
128 // - piextUSMPitchedAlloc
129 // - piextMemUnsampledImageHandleDestroy
130 // - piextMemSampledImageHandleDestroy
131 // - piextMemImageAllocate
132 // - piextMemImageFree
133 // - piextMemUnsampledImageCreate
134 // - piextMemSampledImageCreate
135 // - piextMemImageCopy
136 // - piextMemImageGetInfo
137 // - piextMemMipmapGetLevel
138 // - piextMemMipmapFree
139 // - piextMemImportOpaqueFD
140 // - piextMemMapExternalArray
141 // - piextMemReleaseInterop
142 // - piextImportExternalSemaphoreOpaqueFD
143 // - piextDestroyExternalSemaphore
144 // - piextWaitExternalSemaphore
145 // - piextSignalExternalSemaphore
146 // 14.37 Added piextUSMImportExternalPointer and piextUSMReleaseImportedPointer.
147 // 14.38 Change PI_MEM_ADVICE_* values to flags for use in bitwise operations.
148 // 14.39 Added PI_EXT_INTEL_DEVICE_INFO_ESIMD_SUPPORT device info query.
149 // 14.40 Add HIP _pi_mem_advice alises to match the PI_MEM_ADVICE_CUDA* ones.
150 // 14.41 Added piextCommandBufferMemBufferFill & piextCommandBufferFillUSM
151 // 14.42 Added piextCommandBufferPrefetchUSM and piextCommandBufferAdviseUSM
152 // 15.43 Changed the signature of piextMemGetNativeHandle to also take a
153 // pi_device
154 // 15.44 Add coarse-grain memory advice flag for HIP.
155 // 15.45 Added piextKernelSuggestMaxCooperativeGroupCount and
156 // piextEnqueueCooperativeKernelLaunch.
157 // 15.46 Add piextGetGlobalVariablePointer
158 // 15.47 Added PI_ERROR_FEATURE_UNSUPPORTED.
159 // 15.48 Add CommandBuffer update definitions
160 // 15.49 Added cubemap support:
161 // - Added cubemap image type, PI_MEM_TYPE_IMAGE_CUBEMAP, to _pi_mem_type
162 // - Added cubemap sampling capabilities
163 // - _pi_sampler_cubemap_filter_mode
164 // - PI_SAMPLER_PROPERTIES_CUBEMAP_FILTER_MODE
165 // - Added device queries for cubemap support
166 // - PI_EXT_ONEAPI_DEVICE_INFO_CUBEMAP_SUPPORT
167 // - PI_EXT_ONEAPI_DEVICE_INFO_CUBEMAP_SEAMLESS_FILTERING_SUPPORT
168 
169 #define _PI_H_VERSION_MAJOR 15
170 #define _PI_H_VERSION_MINOR 48
171 
172 #define _PI_STRING_HELPER(a) #a
173 #define _PI_CONCAT(a, b) _PI_STRING_HELPER(a.b)
174 #define _PI_TRIPLE_CONCAT(a, b, c) _PI_STRING_HELPER(a.b.c)
175 
176 // This is the macro that plugins should all use to define their version.
177 // _PI_PLUGIN_VERSION_STRING will be printed when environment variable
178 // SYCL_PI_TRACE is set to 1. PluginVersion should be defined for each plugin
179 // in plugins/*/pi_*.hpp. PluginVersion should be incremented with each change
180 // to the plugin.
181 #define _PI_PLUGIN_VERSION_STRING(PluginVersion) \
182  _PI_TRIPLE_CONCAT(_PI_H_VERSION_MAJOR, _PI_H_VERSION_MINOR, PluginVersion)
183 
184 #define _PI_H_VERSION_STRING \
185  _PI_CONCAT(_PI_H_VERSION_MAJOR, _PI_H_VERSION_MINOR)
186 
187 // This will be used to check the major versions of plugins versus the major
188 // versions of PI.
189 #define _PI_STRING_SUBSTITUTE(X) _PI_STRING_HELPER(X)
190 #define _PI_PLUGIN_VERSION_CHECK(PI_API_VERSION, PI_PLUGIN_VERSION) \
191  if (strncmp(PI_API_VERSION, PI_PLUGIN_VERSION, \
192  sizeof(_PI_STRING_SUBSTITUTE(_PI_H_VERSION_MAJOR))) < 0) { \
193  return PI_ERROR_INVALID_OPERATION; \
194  }
195 
196 // NOTE: This file presents a maping of OpenCL to PI enums, constants and
197 // typedefs. The general approach taken was to replace `CL_` prefix with `PI_`.
198 // Please consider this when adding or modifying values, as the strict value
199 // match is required.
200 // TODO: We should consider re-implementing PI enums and constants and only
201 // perform a mapping of PI to OpenCL in the pi_opencl backend.
202 #include <sycl/detail/export.hpp>
203 
204 #include <cstddef>
205 #include <cstdint>
206 #include <variant>
207 
208 #ifdef __cplusplus
209 extern "C" {
210 #endif // __cplusplus
211 
212 using pi_int32 = int32_t;
213 using pi_uint32 = uint32_t;
214 using pi_uint64 = uint64_t;
217 using pi_native_handle = uintptr_t;
218 
219 //
220 // NOTE: prefer to map 1:1 to OpenCL so that no translation is needed
221 // for PI <-> OpenCL ways. The PI <-> to other BE translation is almost
222 // always needed anyway.
223 //
224 typedef enum {
225 #define _PI_ERRC(NAME, VAL) NAME = VAL,
226 #define _PI_ERRC_WITH_MSG(NAME, VAL, MSG) NAME = VAL,
227 #include <sycl/detail/pi_error.def>
228 #undef _PI_ERRC
229 #undef _PI_ERRC_WITH_MSG
230 } _pi_result;
231 
232 typedef enum {
236  PI_EVENT_QUEUED = 0x3
238 
239 typedef enum {
245  PI_EXT_PLATFORM_INFO_BACKEND = 0x21000 // returns pi_platform_backend
247 
248 typedef enum {
254 
255 typedef enum {
261 
262 typedef enum {
268 
269 // NOTE: this is made 64-bit to match the size of cl_device_type to
270 // make the translation to OpenCL transparent.
271 //
272 typedef enum : pi_uint64 {
274  (1 << 0),
275  PI_DEVICE_TYPE_ALL = 0xFFFFFFFF,
276  PI_DEVICE_TYPE_CPU = (1 << 1),
277  PI_DEVICE_TYPE_GPU = (1 << 2),
278  PI_DEVICE_TYPE_ACC = (1 << 3),
280  PI_DEVICE_TYPE_CUSTOM = (1 << 4)
282 
283 typedef enum {
292 
293 typedef enum {
298 
299 typedef enum {
303 
304 typedef enum {
389  // Intel UUID extension.
391  // These are Intel-specific extensions.
402  // Return true if sub-device should do its own program build
405  // Return 0 if device doesn't have any memory modules. Return the minimum of
406  // the clock rate values if there are several memory modules on the device.
408  // Return 0 if device doesn't have any memory modules. Return the minimum of
409  // the bus width values if there are several memory modules on the device.
411  // Return 1 if the device doesn't have a notion of a "queue index". Otherwise,
412  // return the number of queue indices that are available for this device.
419  // Return whether bfloat16 math functions are supported by device
430  // The number of max registers per block (device specific)
433 
434  // Bindless images, mipmaps, interop
451 
453 
454  // Composite device
457 
458  // Command Buffers
461 
462  // Bindless images cubemaps
466 
467 typedef enum {
478 
479 typedef enum {
485  // Atomics capabilities extensions
490  // Native 2D USM memory operation support
495 
496 typedef enum {
503  // Return 'true' if all commands previously submitted to the queue have
504  // completed, otherwise return 'false'.
507 
508 typedef enum {
516 
517 typedef enum {
524  // The number of registers used by the compiled kernel (device specific)
527 
528 typedef enum {
535  PI_IMAGE_INFO_DEPTH = 0x1116
537 
538 typedef enum {
544 
545 typedef enum {
552 
553 typedef enum {
587 
588 typedef enum {
597 } _pi_mem_type;
598 
599 typedef enum {
600  // Device-specific value opaque in PI API.
614  PI_MEM_ADVICE_UNKNOWN = 0x7FFFFFFF,
616 
617 // HIP _pi_mem_advice aliases
632 static constexpr _pi_mem_advice
639 
640 typedef enum {
657 
658 typedef enum {
675 
676 typedef enum {
681 
683 
684 const pi_bool PI_TRUE = 1;
685 const pi_bool PI_FALSE = 0;
686 
687 typedef enum {
695  PI_SAMPLER_INFO_LOD_MAX = 0x1157
697 
698 typedef enum {
705 
706 typedef enum {
710 
711 typedef enum {
715 
716 using pi_context_properties = intptr_t;
717 
720  (1 << 0);
723 
726  0x1152;
731  0x1156;
732 
739 
746 
747 typedef enum {
753 
754 // NOTE: this is made 64-bit to match the size of cl_mem_flags to
755 // make the translation to OpenCL transparent.
756 // TODO: populate
757 //
759 // Access
762 // Host pointer
766 
767 // flags passed to Map operations
769 constexpr pi_map_flags PI_MAP_READ = (1 << 0);
770 constexpr pi_map_flags PI_MAP_WRITE = (1 << 1);
772 // NOTE: this is made 64-bit to match the size of cl_mem_properties_intel to
773 // make the translation to OpenCL transparent.
777 
778 // NOTE: this is made 64-bit to match the size of cl_mem_properties_intel to
779 // make the translation to OpenCL transparent.
784  (1 << 1);
786 // Hints that the device/shared allocation will not be written on device.
788 
790 
791 // NOTE: queue properties are implemented this way to better support bit
792 // manipulations
796 // clang-format off
806 // clang-format on
807 
808 typedef enum {
809  // No preference for SLM or data cache.
811  // Large SLM size.
813  // Large General Data size.
816 
849 
851 
852 // For compatibility with OpenCL define this not as enum.
855  0x1086;
857  0x1087;
858 static constexpr pi_device_partition_property
860 static constexpr pi_device_partition_property
862 static constexpr pi_device_partition_property
864 
865 // For compatibility with OpenCL define this not as enum.
868  (1 << 0);
870  (1 << 1);
872  (1 << 2);
874  (1 << 3);
876  (1 << 4);
877 static constexpr pi_device_affinity_domain
879 
880 // For compatibility with OpenCL define this not as enum.
882 static constexpr pi_device_fp_config PI_FP_DENORM = (1 << 0);
883 static constexpr pi_device_fp_config PI_FP_INF_NAN = (1 << 1);
884 static constexpr pi_device_fp_config PI_FP_ROUND_TO_NEAREST = (1 << 2);
885 static constexpr pi_device_fp_config PI_FP_ROUND_TO_ZERO = (1 << 3);
886 static constexpr pi_device_fp_config PI_FP_ROUND_TO_INF = (1 << 4);
887 static constexpr pi_device_fp_config PI_FP_FMA = (1 << 5);
888 static constexpr pi_device_fp_config PI_FP_SOFT_FLOAT = (1 << 6);
890  (1 << 7);
891 
892 // For compatibility with OpenCL define this not as enum.
894 static constexpr pi_device_exec_capabilities PI_EXEC_KERNEL = (1 << 0);
896 
897 // Entry type, matches OpenMP for compatibility
899  void *addr;
900  char *name;
901  size_t size;
902  int32_t flags;
903  int32_t reserved;
904 };
905 
907 
908 // A type of a binary image property.
909 typedef enum {
911  PI_PROPERTY_TYPE_UINT32, // 32-bit integer
913  PI_PROPERTY_TYPE_STRING // null-terminated string
915 
916 // Device binary image property.
917 // If the type size of the property value is fixed and is no greater than
918 // 64 bits, then ValAddr is 0 and the value is stored in the ValSize field.
919 // Example - PI_PROPERTY_TYPE_UINT32, which is 32-bit
921  char *Name; // null-terminated property name
922  void *ValAddr; // address of property value
923  uint32_t Type; // _pi_property_type
924  uint64_t ValSize; // size of property value in bytes
925 };
926 
928 
929 // Named array of properties.
931  char *Name; // the name
934 };
935 
937 
939 using pi_device_binary_type = uint8_t;
940 // format is not determined
942 // specific to a device
944 // portable binary types go next
945 // SPIR-V
947 // LLVM bitcode
949 
950 // Device binary descriptor version supported by this library.
951 static const uint16_t PI_DEVICE_BINARY_VERSION = 1;
952 
953 // The kind of offload model the binary employs; must be 4 for SYCL
954 static const uint8_t PI_DEVICE_BINARY_OFFLOAD_KIND_SYCL = 4;
955 
963 #define __SYCL_PI_DEVICE_BINARY_TARGET_UNKNOWN "<unknown>"
965 #define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV32 "spir"
967 #define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64 "spir64"
970 #define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_X86_64 "spir64_x86_64"
971 #define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_GEN "spir64_gen"
972 #define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_FPGA "spir64_fpga"
974 #define __SYCL_PI_DEVICE_BINARY_TARGET_NVPTX64 "nvptx64"
975 #define __SYCL_PI_DEVICE_BINARY_TARGET_AMDGCN "amdgcn"
976 #define __SYCL_PI_DEVICE_BINARY_TARGET_NATIVE_CPU "native_cpu"
977 
981 #define PI_DEVICE_INFO_EXTENSION_DEVICELIB_ASSERT "cl_intel_devicelib_assert"
982 
987 #define __SYCL_PI_PROPERTY_SET_SPEC_CONST_MAP "SYCL/specialization constants"
990 #define __SYCL_PI_PROPERTY_SET_SPEC_CONST_DEFAULT_VALUES_MAP \
991  "SYCL/specialization constants default values"
993 #define __SYCL_PI_PROPERTY_SET_DEVICELIB_REQ_MASK "SYCL/devicelib req mask"
995 #define __SYCL_PI_PROPERTY_SET_KERNEL_PARAM_OPT_INFO "SYCL/kernel param opt"
997 #define __SYCL_PI_PROPERTY_SET_PROGRAM_METADATA "SYCL/program metadata"
999 #define __SYCL_PI_PROPERTY_SET_SYCL_MISC_PROP "SYCL/misc properties"
1001 #define __SYCL_PI_PROPERTY_SET_SYCL_ASSERT_USED "SYCL/assert used"
1003 #define __SYCL_PI_PROPERTY_SET_SYCL_EXPORTED_SYMBOLS "SYCL/exported symbols"
1005 #define __SYCL_PI_PROPERTY_SET_SYCL_DEVICE_GLOBALS "SYCL/device globals"
1007 #define __SYCL_PI_PROPERTY_SET_SYCL_DEVICE_REQUIREMENTS \
1008  "SYCL/device requirements"
1010 #define __SYCL_PI_PROPERTY_SET_SYCL_HOST_PIPES "SYCL/host pipes"
1011 
1014 #define __SYCL_PI_PROGRAM_METADATA_TAG_REQD_WORK_GROUP_SIZE \
1015  "@reqd_work_group_size"
1016 #define __SYCL_PI_PROGRAM_METADATA_GLOBAL_ID_MAPPING "@global_id_mapping"
1017 
1018 #define __SYCL_PI_PROGRAM_METADATA_TAG_NEED_FINALIZATION "Requires finalization"
1019 
1029  uint16_t Version;
1031  uint8_t Kind;
1033  uint8_t Format;
1045  const char *DeviceTargetSpec;
1048  const char *CompileOptions;
1051  const char *LinkOptions;
1053  const char *ManifestStart;
1055  const char *ManifestEnd;
1057  const unsigned char *BinaryStart;
1059  const unsigned char *BinaryEnd;
1063  // Array of preperty sets; e.g. specialization constants symbol-int ID map is
1064  // propagated to runtime with this mechanism.
1067  // TODO Other fields like entries, link options can be propagated using
1068  // the property set infrastructure. This will improve binary compatibility and
1069  // add flexibility.
1070 };
1072 
1073 // pi_buffer_region structure repeats cl_buffer_region, used for sub buffers.
1075  size_t origin;
1076  size_t size;
1077 };
1079 
1080 // pi_buff_rect_offset structure is 3D offset argument passed to buffer rect
1081 // operations (piEnqueueMemBufferCopyRect, etc).
1083  size_t x_bytes;
1084  size_t y_scalar;
1085  size_t z_scalar;
1086 };
1088 
1089 // pi_buff_rect_region structure represents size of 3D region passed to buffer
1090 // rect operations (piEnqueueMemBufferCopyRect, etc).
1092  size_t width_bytes;
1095 };
1097 
1098 // pi_image_offset structure is 3D offset argument passed to image operations
1099 // (piEnqueueMemImageRead, etc).
1101  size_t x;
1102  size_t y;
1103  size_t z;
1104 };
1106 
1107 // pi_image_region structure represents size of 3D region passed to image
1108 // operations (piEnqueueMemImageRead, etc).
1110  size_t width;
1111  size_t height;
1112  size_t depth;
1113 };
1115 
1116 // Offload binaries descriptor version supported by this library.
1117 static const uint16_t PI_DEVICE_BINARIES_VERSION = 1;
1118 
1126  uint16_t Version;
1134 };
1136 
1137 // Opaque types that make reading build log errors easier.
1138 struct _pi_platform;
1139 struct _pi_device;
1140 struct _pi_context;
1141 struct _pi_queue;
1142 struct _pi_mem;
1143 struct _pi_program;
1144 struct _pi_kernel;
1145 struct _pi_event;
1146 struct _pi_sampler;
1147 
1152 using pi_mem = _pi_mem *;
1158 using pi_image_mem_handle = void *;
1161 
1162 typedef struct {
1166 
1167 typedef struct {
1169  size_t image_width;
1171  size_t image_depth;
1178 } _pi_image_desc;
1179 
1182 
1183 typedef enum { PI_MEM_CONTEXT = 0x1106, PI_MEM_SIZE = 0x1102 } _pi_mem_info;
1184 
1185 typedef enum {
1187  0x0,
1190  0x1
1193 
1196 
1197 //
1198 // Following section contains SYCL RT Plugin Interface (PI) functions.
1199 // They are 3 distinct categories:
1200 //
1201 // 1) Ones having direct analogy in OpenCL and needed for the core SYCL
1202 // functionality are started with just "pi" prefix in their names.
1203 // 2) Those having direct analogy in OpenCL but only needed for SYCL
1204 // interoperability with OpenCL are started with "picl" prefix.
1205 // 3) Functions having no direct analogy in OpenCL, started with "piext".
1206 //
1207 // TODO: describe interfaces in Doxygen format
1208 //
1209 
1210 struct _pi_plugin;
1212 
1213 // PI Plugin Initialise.
1214 // Plugin will check the PI version of Plugin Interface,
1215 // populate the PI Version it supports, update targets field and populate
1216 // PiFunctionTable with Supported APIs. The pointers are in a predetermined
1217 // order in pi.def file.
1218 __SYCL_EXPORT pi_result piPluginInit(pi_plugin *plugin_info);
1219 
1220 //
1221 // Platform
1222 //
1223 __SYCL_EXPORT pi_result piPlatformsGet(pi_uint32 num_entries,
1224  pi_platform *platforms,
1225  pi_uint32 *num_platforms);
1226 
1227 __SYCL_EXPORT pi_result piPlatformGetInfo(pi_platform platform,
1228  pi_platform_info param_name,
1229  size_t param_value_size,
1230  void *param_value,
1231  size_t *param_value_size_ret);
1232 
1238  pi_platform platform, pi_native_handle *nativeHandle);
1239 
1246  pi_native_handle nativeHandle, pi_platform *platform);
1247 
1248 __SYCL_EXPORT pi_result piDevicesGet(pi_platform platform,
1250  pi_uint32 num_entries, pi_device *devices,
1251  pi_uint32 *num_devices);
1252 
1253 __SYCL_EXPORT pi_result piextEnablePeerAccess(pi_device command_device,
1254  pi_device peer_device);
1255 __SYCL_EXPORT pi_result piextDisablePeerAccess(pi_device command_device,
1256  pi_device peer_device);
1257 __SYCL_EXPORT pi_result piextPeerAccessGetInfo(
1258  pi_device command_device, pi_device peer_device, pi_peer_attr attr,
1259  size_t param_value_size, void *param_value, size_t *param_value_size_ret);
1260 
1264 __SYCL_EXPORT pi_result piDeviceGetInfo(pi_device device,
1265  pi_device_info param_name,
1266  size_t param_value_size,
1267  void *param_value,
1268  size_t *param_value_size_ret);
1269 
1270 __SYCL_EXPORT pi_result piDeviceRetain(pi_device device);
1271 
1272 __SYCL_EXPORT pi_result piDeviceRelease(pi_device device);
1273 
1274 __SYCL_EXPORT pi_result piDevicePartition(
1275  pi_device device, const pi_device_partition_property *properties,
1276  pi_uint32 num_devices, pi_device *out_devices, pi_uint32 *out_num_devices);
1277 
1282 __SYCL_EXPORT pi_result
1284 
1292  pi_native_handle nativeHandle, pi_platform platform, pi_device *device);
1293 
1297 __SYCL_EXPORT pi_result piextDeviceSelectBinary(pi_device device,
1298  pi_device_binary *binaries,
1299  pi_uint32 num_binaries,
1300  pi_uint32 *selected_binary_ind);
1301 
1314  pi_device device, pi_program program, const char *function_name,
1315  pi_uint64 *function_pointer_ret);
1316 
1318  pi_device Device, pi_program Program, const char *GlobalVariableName,
1319  size_t *GlobalVariableSize, void **GlobalVariablePointerRet);
1320 
1321 //
1322 // Context
1323 //
1325  const pi_context_properties *properties, pi_uint32 num_devices,
1326  const pi_device *devices,
1327  void (*pfn_notify)(const char *errinfo, const void *private_info, size_t cb,
1328  void *user_data),
1329  void *user_data, pi_context *ret_context);
1330 
1331 __SYCL_EXPORT pi_result piContextGetInfo(pi_context context,
1332  pi_context_info param_name,
1333  size_t param_value_size,
1334  void *param_value,
1335  size_t *param_value_size_ret);
1336 
1337 __SYCL_EXPORT pi_result piContextRetain(pi_context context);
1338 
1339 __SYCL_EXPORT pi_result piContextRelease(pi_context context);
1340 
1341 typedef void (*pi_context_extended_deleter)(void *user_data);
1342 
1344  pi_context context, pi_context_extended_deleter func, void *user_data);
1345 
1350 __SYCL_EXPORT pi_result
1352 
1377  pi_native_handle nativeHandle, pi_uint32 numDevices,
1378  const pi_device *devices, bool pluginOwnsNativeHandle, pi_context *context);
1379 
1380 //
1381 // Queue
1382 //
1383 
1384 // TODO: Remove during next ABI break and rename piextQueueCreate to
1385 // piQueueCreate.
1386 __SYCL_EXPORT pi_result piQueueCreate(pi_context context, pi_device device,
1387  pi_queue_properties properties,
1388  pi_queue *queue);
1392 __SYCL_EXPORT pi_result piextQueueCreate(pi_context context, pi_device device,
1393  pi_queue_properties *properties,
1394  pi_queue *queue);
1395 
1396 __SYCL_EXPORT pi_result piQueueGetInfo(pi_queue command_queue,
1397  pi_queue_info param_name,
1398  size_t param_value_size,
1399  void *param_value,
1400  size_t *param_value_size_ret);
1401 
1402 __SYCL_EXPORT pi_result piQueueRetain(pi_queue command_queue);
1403 
1404 __SYCL_EXPORT pi_result piQueueRelease(pi_queue command_queue);
1405 
1406 __SYCL_EXPORT pi_result piQueueFinish(pi_queue command_queue);
1407 
1408 __SYCL_EXPORT pi_result piQueueFlush(pi_queue command_queue);
1409 
1415 __SYCL_EXPORT pi_result piextQueueGetNativeHandle(
1416  pi_queue queue, pi_native_handle *nativeHandle, int32_t *nativeHandleDesc);
1417 
1432  pi_native_handle nativeHandle, int32_t nativeHandleDesc, pi_context context,
1433  pi_device device, bool pluginOwnsNativeHandle,
1434  pi_queue_properties *Properties, pi_queue *queue);
1435 
1436 //
1437 // Memory
1438 //
1439 __SYCL_EXPORT pi_result piMemBufferCreate(
1440  pi_context context, pi_mem_flags flags, size_t size, void *host_ptr,
1441  pi_mem *ret_mem, const pi_mem_properties *properties = nullptr);
1442 
1443 __SYCL_EXPORT pi_result piMemImageCreate(pi_context context, pi_mem_flags flags,
1445  const pi_image_desc *image_desc,
1446  void *host_ptr, pi_mem *ret_mem);
1447 
1448 __SYCL_EXPORT pi_result piMemGetInfo(pi_mem mem, pi_mem_info param_name,
1449  size_t param_value_size, void *param_value,
1450  size_t *param_value_size_ret);
1451 
1452 __SYCL_EXPORT pi_result piMemImageGetInfo(pi_mem image,
1453  pi_image_info param_name,
1454  size_t param_value_size,
1455  void *param_value,
1456  size_t *param_value_size_ret);
1457 
1458 __SYCL_EXPORT pi_result piMemRetain(pi_mem mem);
1459 
1460 __SYCL_EXPORT pi_result piMemRelease(pi_mem mem);
1461 
1462 __SYCL_EXPORT pi_result piMemBufferPartition(
1463  pi_mem buffer, pi_mem_flags flags, pi_buffer_create_type buffer_create_type,
1464  void *buffer_create_info, pi_mem *ret_mem);
1465 
1471 __SYCL_EXPORT pi_result piextMemGetNativeHandle(pi_mem mem, pi_device dev,
1472  pi_native_handle *nativeHandle);
1473 
1483  pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle,
1484  pi_mem *mem);
1485 
1500  pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle,
1501  const pi_image_format *ImageFormat, const pi_image_desc *ImageDesc,
1502  pi_mem *img);
1503 
1504 //
1505 // Program
1506 //
1507 
1508 __SYCL_EXPORT pi_result piProgramCreate(pi_context context, const void *il,
1509  size_t length, pi_program *res_program);
1510 
1527 __SYCL_EXPORT pi_result piProgramCreateWithBinary(
1528  pi_context context, pi_uint32 num_devices, const pi_device *device_list,
1529  const size_t *lengths, const unsigned char **binaries,
1530  size_t num_metadata_entries, const pi_device_binary_property *metadata,
1531  pi_int32 *binary_status, pi_program *ret_program);
1532 
1533 __SYCL_EXPORT pi_result piProgramGetInfo(pi_program program,
1534  pi_program_info param_name,
1535  size_t param_value_size,
1536  void *param_value,
1537  size_t *param_value_size_ret);
1538 
1539 __SYCL_EXPORT pi_result
1540 piProgramLink(pi_context context, pi_uint32 num_devices,
1541  const pi_device *device_list, const char *options,
1542  pi_uint32 num_input_programs, const pi_program *input_programs,
1543  void (*pfn_notify)(pi_program program, void *user_data),
1544  void *user_data, pi_program *ret_program);
1545 
1547  pi_program program, pi_uint32 num_devices, const pi_device *device_list,
1548  const char *options, pi_uint32 num_input_headers,
1549  const pi_program *input_headers, const char **header_include_names,
1550  void (*pfn_notify)(pi_program program, void *user_data), void *user_data);
1551 
1553  pi_program program, pi_uint32 num_devices, const pi_device *device_list,
1554  const char *options,
1555  void (*pfn_notify)(pi_program program, void *user_data), void *user_data);
1556 
1557 __SYCL_EXPORT pi_result piProgramGetBuildInfo(
1558  pi_program program, pi_device device, _pi_program_build_info param_name,
1559  size_t param_value_size, void *param_value, size_t *param_value_size_ret);
1560 
1561 __SYCL_EXPORT pi_result piProgramRetain(pi_program program);
1562 
1563 __SYCL_EXPORT pi_result piProgramRelease(pi_program program);
1564 
1574 __SYCL_EXPORT pi_result
1576  size_t spec_size, const void *spec_value);
1577 
1582 __SYCL_EXPORT pi_result
1584 
1594  pi_native_handle nativeHandle, pi_context context,
1595  bool pluginOwnsNativeHandle, pi_program *program);
1596 
1597 //
1598 // Kernel
1599 //
1600 
1601 typedef enum {
1605  PI_USM_PTRS = 0x4203,
1609 
1611 
1612 __SYCL_EXPORT pi_result piKernelCreate(pi_program program,
1613  const char *kernel_name,
1614  pi_kernel *ret_kernel);
1615 
1616 __SYCL_EXPORT pi_result piKernelSetArg(pi_kernel kernel, pi_uint32 arg_index,
1617  size_t arg_size, const void *arg_value);
1618 
1619 __SYCL_EXPORT pi_result piKernelGetInfo(pi_kernel kernel,
1620  pi_kernel_info param_name,
1621  size_t param_value_size,
1622  void *param_value,
1623  size_t *param_value_size_ret);
1624 
1625 __SYCL_EXPORT pi_result piKernelGetGroupInfo(pi_kernel kernel, pi_device device,
1626  pi_kernel_group_info param_name,
1627  size_t param_value_size,
1628  void *param_value,
1629  size_t *param_value_size_ret);
1630 
1649 __SYCL_EXPORT pi_result piKernelGetSubGroupInfo(
1650  pi_kernel kernel, pi_device device, pi_kernel_sub_group_info param_name,
1651  size_t input_value_size, const void *input_value, size_t param_value_size,
1652  void *param_value, size_t *param_value_size_ret);
1653 
1654 __SYCL_EXPORT pi_result piKernelRetain(pi_kernel kernel);
1655 
1656 __SYCL_EXPORT pi_result piKernelRelease(pi_kernel kernel);
1657 
1665 __SYCL_EXPORT pi_result piextKernelSetArgPointer(pi_kernel kernel,
1666  pi_uint32 arg_index,
1667  size_t arg_size,
1668  const void *arg_value);
1669 
1681 __SYCL_EXPORT pi_result piKernelSetExecInfo(pi_kernel kernel,
1682  pi_kernel_exec_info value_name,
1683  size_t param_value_size,
1684  const void *param_value);
1685 
1696  pi_native_handle nativeHandle, pi_context context, pi_program program,
1697  bool pluginOwnsNativeHandle, pi_kernel *kernel);
1698 
1703 __SYCL_EXPORT pi_result
1705 
1715  pi_kernel kernel, size_t local_work_size, size_t dynamic_shared_memory_size,
1716  pi_uint32 *group_count_ret);
1717 
1718 //
1719 // Events
1720 //
1721 
1726 __SYCL_EXPORT pi_result piEventCreate(pi_context context, pi_event *ret_event);
1727 
1728 __SYCL_EXPORT pi_result piEventGetInfo(pi_event event, pi_event_info param_name,
1729  size_t param_value_size,
1730  void *param_value,
1731  size_t *param_value_size_ret);
1732 
1733 __SYCL_EXPORT pi_result piEventGetProfilingInfo(pi_event event,
1734  pi_profiling_info param_name,
1735  size_t param_value_size,
1736  void *param_value,
1737  size_t *param_value_size_ret);
1738 
1739 __SYCL_EXPORT pi_result piEventsWait(pi_uint32 num_events,
1740  const pi_event *event_list);
1741 
1743  pi_event event, pi_int32 command_exec_callback_type,
1744  void (*pfn_notify)(pi_event event, pi_int32 event_command_status,
1745  void *user_data),
1746  void *user_data);
1747 
1748 __SYCL_EXPORT pi_result piEventSetStatus(pi_event event,
1749  pi_int32 execution_status);
1750 
1751 __SYCL_EXPORT pi_result piEventRetain(pi_event event);
1752 
1753 __SYCL_EXPORT pi_result piEventRelease(pi_event event);
1754 
1759 __SYCL_EXPORT pi_result
1761 
1771  pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle,
1772  pi_event *event);
1773 
1774 //
1775 // Sampler
1776 //
1777 __SYCL_EXPORT pi_result piSamplerCreate(
1778  pi_context context, const pi_sampler_properties *sampler_properties,
1779  pi_sampler *result_sampler);
1780 
1781 __SYCL_EXPORT pi_result piSamplerGetInfo(pi_sampler sampler,
1782  pi_sampler_info param_name,
1783  size_t param_value_size,
1784  void *param_value,
1785  size_t *param_value_size_ret);
1786 
1787 __SYCL_EXPORT pi_result piSamplerRetain(pi_sampler sampler);
1788 
1789 __SYCL_EXPORT pi_result piSamplerRelease(pi_sampler sampler);
1790 
1791 //
1792 // Queue Commands
1793 //
1794 __SYCL_EXPORT pi_result piEnqueueKernelLaunch(
1795  pi_queue queue, pi_kernel kernel, pi_uint32 work_dim,
1796  const size_t *global_work_offset, const size_t *global_work_size,
1797  const size_t *local_work_size, pi_uint32 num_events_in_wait_list,
1798  const pi_event *event_wait_list, pi_event *event);
1799 
1801  pi_queue queue, pi_kernel kernel, pi_uint32 work_dim,
1802  const size_t *global_work_offset, const size_t *global_work_size,
1803  const size_t *local_work_size, pi_uint32 num_events_in_wait_list,
1804  const pi_event *event_wait_list, pi_event *event);
1805 
1806 __SYCL_EXPORT pi_result piEnqueueEventsWait(pi_queue command_queue,
1807  pi_uint32 num_events_in_wait_list,
1808  const pi_event *event_wait_list,
1809  pi_event *event);
1810 
1812  pi_queue command_queue, pi_uint32 num_events_in_wait_list,
1813  const pi_event *event_wait_list, pi_event *event);
1814 
1815 __SYCL_EXPORT pi_result piEnqueueMemBufferRead(
1816  pi_queue queue, pi_mem buffer, pi_bool blocking_read, size_t offset,
1817  size_t size, void *ptr, pi_uint32 num_events_in_wait_list,
1818  const pi_event *event_wait_list, pi_event *event);
1819 
1821  pi_queue command_queue, pi_mem buffer, pi_bool blocking_read,
1822  pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset,
1823  pi_buff_rect_region region, size_t buffer_row_pitch,
1824  size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch,
1825  void *ptr, pi_uint32 num_events_in_wait_list,
1826  const pi_event *event_wait_list, pi_event *event);
1827 
1828 __SYCL_EXPORT pi_result
1829 piEnqueueMemBufferWrite(pi_queue command_queue, pi_mem buffer,
1830  pi_bool blocking_write, size_t offset, size_t size,
1831  const void *ptr, pi_uint32 num_events_in_wait_list,
1832  const pi_event *event_wait_list, pi_event *event);
1833 
1835  pi_queue command_queue, pi_mem buffer, pi_bool blocking_write,
1836  pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset,
1837  pi_buff_rect_region region, size_t buffer_row_pitch,
1838  size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch,
1839  const void *ptr, pi_uint32 num_events_in_wait_list,
1840  const pi_event *event_wait_list, pi_event *event);
1841 
1842 __SYCL_EXPORT pi_result
1843 piEnqueueMemBufferCopy(pi_queue command_queue, pi_mem src_buffer,
1844  pi_mem dst_buffer, size_t src_offset, size_t dst_offset,
1845  size_t size, pi_uint32 num_events_in_wait_list,
1846  const pi_event *event_wait_list, pi_event *event);
1847 
1849  pi_queue command_queue, pi_mem src_buffer, pi_mem dst_buffer,
1850  pi_buff_rect_offset src_origin, pi_buff_rect_offset dst_origin,
1851  pi_buff_rect_region region, size_t src_row_pitch, size_t src_slice_pitch,
1852  size_t dst_row_pitch, size_t dst_slice_pitch,
1853  pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list,
1854  pi_event *event);
1855 
1856 __SYCL_EXPORT pi_result
1857 piEnqueueMemBufferFill(pi_queue command_queue, pi_mem buffer,
1858  const void *pattern, size_t pattern_size, size_t offset,
1859  size_t size, pi_uint32 num_events_in_wait_list,
1860  const pi_event *event_wait_list, pi_event *event);
1861 
1862 __SYCL_EXPORT pi_result piEnqueueMemImageRead(
1863  pi_queue command_queue, pi_mem image, pi_bool blocking_read,
1864  pi_image_offset origin, pi_image_region region, size_t row_pitch,
1865  size_t slice_pitch, void *ptr, pi_uint32 num_events_in_wait_list,
1866  const pi_event *event_wait_list, pi_event *event);
1867 
1868 __SYCL_EXPORT pi_result piEnqueueMemImageWrite(
1869  pi_queue command_queue, pi_mem image, pi_bool blocking_write,
1870  pi_image_offset origin, pi_image_region region, size_t input_row_pitch,
1871  size_t input_slice_pitch, const void *ptr,
1872  pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list,
1873  pi_event *event);
1874 
1875 __SYCL_EXPORT pi_result piEnqueueMemImageCopy(
1876  pi_queue command_queue, pi_mem src_image, pi_mem dst_image,
1877  pi_image_offset src_origin, pi_image_offset dst_origin,
1878  pi_image_region region, pi_uint32 num_events_in_wait_list,
1879  const pi_event *event_wait_list, pi_event *event);
1880 
1881 __SYCL_EXPORT pi_result
1882 piEnqueueMemImageFill(pi_queue command_queue, pi_mem image,
1883  const void *fill_color, const size_t *origin,
1884  const size_t *region, pi_uint32 num_events_in_wait_list,
1885  const pi_event *event_wait_list, pi_event *event);
1886 
1887 __SYCL_EXPORT pi_result piEnqueueMemBufferMap(
1888  pi_queue command_queue, pi_mem buffer, pi_bool blocking_map,
1889  pi_map_flags map_flags, size_t offset, size_t size,
1890  pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list,
1891  pi_event *event, void **ret_map);
1892 
1893 __SYCL_EXPORT pi_result piEnqueueMemUnmap(pi_queue command_queue, pi_mem memobj,
1894  void *mapped_ptr,
1895  pi_uint32 num_events_in_wait_list,
1896  const pi_event *event_wait_list,
1897  pi_event *event);
1898 
1899 #ifndef PI_BIT
1900 #define PI_BIT(_i) (1 << _i)
1901 #endif // PI_BIT
1902 
1903 typedef enum {
1909 typedef uint32_t pi_mem_access_flag;
1910 
1911 typedef enum {
1913  PI_ENUM_FORCE_UINT32 = 0x7fffffff
1916 
1917 typedef struct {
1919  void *pNext;
1923 
1924 // Extension to allow backends to process a PI memory object before adding it
1925 // as an argument for a kernel.
1926 // Note: This is needed by the CUDA backend to extract the device pointer to
1927 // the memory as the kernels uses it rather than the PI object itself.
1928 __SYCL_EXPORT pi_result piextKernelSetArgMemObj(
1929  pi_kernel kernel, pi_uint32 arg_index,
1930  const pi_mem_obj_property *arg_properties, const pi_mem *arg_value);
1931 
1932 // Extension to allow backends to process a PI sampler object before adding it
1933 // as an argument for a kernel.
1934 // Note: This is needed by the CUDA backend to extract the properties of the
1935 // sampler as the kernels uses it rather than the PI object itself.
1936 __SYCL_EXPORT pi_result piextKernelSetArgSampler(pi_kernel kernel,
1937  pi_uint32 arg_index,
1938  const pi_sampler *arg_value);
1939 
1941 // USM
1943 typedef enum {
1950 
1951 typedef enum : pi_bitfield {
1952  PI_USM_ACCESS = (1 << 0),
1957 
1958 typedef enum {
1964 
1965 typedef enum {
1969  PI_MEM_TYPE_SHARED = 0x4199
1971 
1972 // Flag is used for piProgramUSMEnqueuePrefetch. PI_USM_MIGRATION_TBD0 is a
1973 // placeholder for future developments and should not change the behaviour of
1974 // piProgramUSMEnqueuePrefetch
1975 typedef enum : pi_bitfield {
1976  PI_USM_MIGRATION_TBD0 = (1 << 0)
1978 
1984 
1992 __SYCL_EXPORT pi_result piextUSMHostAlloc(void **result_ptr, pi_context context,
1993  pi_usm_mem_properties *properties,
1994  size_t size, pi_uint32 alignment);
1995 
2004 __SYCL_EXPORT pi_result piextUSMDeviceAlloc(void **result_ptr,
2005  pi_context context,
2006  pi_device device,
2007  pi_usm_mem_properties *properties,
2008  size_t size, pi_uint32 alignment);
2009 
2018 __SYCL_EXPORT pi_result piextUSMSharedAlloc(void **result_ptr,
2019  pi_context context,
2020  pi_device device,
2021  pi_usm_mem_properties *properties,
2022  size_t size, pi_uint32 alignment);
2023 
2035 __SYCL_EXPORT pi_result piextUSMPitchedAlloc(
2036  void **result_ptr, size_t *result_pitch, pi_context context,
2037  pi_device device, pi_usm_mem_properties *properties, size_t width_in_bytes,
2038  size_t height, unsigned int element_size_bytes);
2039 
2046 __SYCL_EXPORT pi_result piextUSMFree(pi_context context, void *ptr);
2047 
2059 __SYCL_EXPORT pi_result piextUSMEnqueueMemset(pi_queue queue, void *ptr,
2060  pi_int32 value, size_t count,
2061  pi_uint32 num_events_in_waitlist,
2062  const pi_event *events_waitlist,
2063  pi_event *event);
2064 
2075 __SYCL_EXPORT pi_result piextUSMEnqueueMemcpy(pi_queue queue, pi_bool blocking,
2076  void *dst_ptr,
2077  const void *src_ptr, size_t size,
2078  pi_uint32 num_events_in_waitlist,
2079  const pi_event *events_waitlist,
2080  pi_event *event);
2081 
2091 __SYCL_EXPORT pi_result piextUSMEnqueuePrefetch(
2092  pi_queue queue, const void *ptr, size_t size, pi_usm_migration_flags flags,
2093  pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist,
2094  pi_event *event);
2095 
2103 // USM memadvise API to govern behavior of automatic migration mechanisms
2104 __SYCL_EXPORT pi_result piextUSMEnqueueMemAdvise(pi_queue queue,
2105  const void *ptr, size_t length,
2106  pi_mem_advice advice,
2107  pi_event *event);
2108 
2125 __SYCL_EXPORT pi_result piextUSMGetMemAllocInfo(
2126  pi_context context, const void *ptr, pi_mem_alloc_info param_name,
2127  size_t param_value_size, void *param_value, size_t *param_value_size_ret);
2128 
2141 __SYCL_EXPORT pi_result piextUSMEnqueueFill2D(pi_queue queue, void *ptr,
2142  size_t pitch, size_t pattern_size,
2143  const void *pattern, size_t width,
2144  size_t height,
2145  pi_uint32 num_events_in_waitlist,
2146  const pi_event *events_waitlist,
2147  pi_event *event);
2148 
2160 __SYCL_EXPORT pi_result piextUSMEnqueueMemset2D(
2161  pi_queue queue, void *ptr, size_t pitch, int value, size_t width,
2162  size_t height, pi_uint32 num_events_in_waitlist,
2163  const pi_event *events_waitlist, pi_event *event);
2164 
2179 __SYCL_EXPORT pi_result piextUSMEnqueueMemcpy2D(
2180  pi_queue queue, pi_bool blocking, void *dst_ptr, size_t dst_pitch,
2181  const void *src_ptr, size_t src_pitch, size_t width, size_t height,
2182  pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist,
2183  pi_event *event);
2184 
2190 __SYCL_EXPORT pi_result piextUSMImport(const void *ptr, size_t size,
2191  pi_context context);
2192 
2197 __SYCL_EXPORT pi_result piextUSMRelease(const void *ptr, pi_context context);
2198 
2202 
2217  pi_queue queue, pi_program program, const char *name,
2218  pi_bool blocking_write, size_t count, size_t offset, const void *src,
2219  pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list,
2220  pi_event *event);
2221 
2236  pi_queue queue, pi_program program, const char *name, pi_bool blocking_read,
2237  size_t count, size_t offset, void *dst, pi_uint32 num_events_in_wait_list,
2238  const pi_event *event_wait_list, pi_event *event);
2239 
2244 // Host Pipes
2246 
2265 __SYCL_EXPORT pi_result piextEnqueueReadHostPipe(
2266  pi_queue queue, pi_program program, const char *pipe_symbol,
2267  pi_bool blocking, void *ptr, size_t size, pi_uint32 num_events_in_waitlist,
2268  const pi_event *events_waitlist, pi_event *event);
2269 
2288 __SYCL_EXPORT pi_result piextEnqueueWriteHostPipe(
2289  pi_queue queue, pi_program program, const char *pipe_symbol,
2290  pi_bool blocking, void *ptr, size_t size, pi_uint32 num_events_in_waitlist,
2291  const pi_event *events_waitlist, pi_event *event);
2292 
2299 __SYCL_EXPORT pi_result piextPluginGetOpaqueData(void *opaque_data_param,
2300  void **opaque_data_return);
2301 
2305 __SYCL_EXPORT pi_result piTearDown(void *PluginParameter);
2306 
2318 __SYCL_EXPORT pi_result piPluginGetLastError(char **message);
2319 
2327 __SYCL_EXPORT pi_result piPluginGetBackendOption(pi_platform platform,
2328  const char *frontend_option,
2329  const char **backend_option);
2330 
2341 __SYCL_EXPORT pi_result piGetDeviceAndHostTimer(pi_device Device,
2342  uint64_t *DeviceTime,
2343  uint64_t *HostTime);
2344 
2346 struct _pi_ext_command_buffer;
2347 struct _pi_ext_sync_point;
2348 struct _pi_ext_command_buffer_command;
2349 
2351 using pi_ext_command_buffer_command = _pi_ext_command_buffer_command *;
2353 
2354 typedef enum {
2357 
2360  const void *pNext;
2362 };
2363 
2364 // Command Buffer Update types
2366  uint32_t arg_index;
2369 };
2370 
2372  uint32_t arg_index;
2373  void *new_ptr;
2374 };
2375 
2377  uint32_t arg_index;
2378  uint32_t arg_size;
2379  void *new_value;
2380 };
2381 
2384  uint32_t num_ptr_args;
2385  uint32_t num_value_args;
2386  uint32_t num_work_dim;
2387 
2391 
2395 };
2396 
2403 __SYCL_EXPORT pi_result
2405  const pi_ext_command_buffer_desc *desc,
2406  pi_ext_command_buffer *ret_command_buffer);
2407 
2410 __SYCL_EXPORT pi_result
2412 
2417 __SYCL_EXPORT pi_result
2419 
2423 __SYCL_EXPORT pi_result
2425 
2441  pi_ext_command_buffer command_buffer, pi_kernel kernel, pi_uint32 work_dim,
2442  const size_t *global_work_offset, const size_t *global_work_size,
2443  const size_t *local_work_size, pi_uint32 num_sync_points_in_wait_list,
2444  const pi_ext_sync_point *sync_point_wait_list,
2445  pi_ext_sync_point *sync_point, pi_ext_command_buffer_command *command);
2446 
2458  pi_ext_command_buffer command_buffer, void *dst_ptr, const void *src_ptr,
2459  size_t size, pi_uint32 num_sync_points_in_wait_list,
2460  const pi_ext_sync_point *sync_point_wait_list,
2461  pi_ext_sync_point *sync_point);
2462 
2476  pi_ext_command_buffer command_buffer, pi_mem src_buffer, pi_mem dst_buffer,
2477  size_t src_offset, size_t dst_offset, size_t size,
2478  pi_uint32 num_sync_points_in_wait_list,
2479  const pi_ext_sync_point *sync_point_wait_list,
2480  pi_ext_sync_point *sync_point);
2481 
2499  pi_ext_command_buffer command_buffer, pi_mem src_buffer, pi_mem dst_buffer,
2500  pi_buff_rect_offset src_origin, pi_buff_rect_offset dst_origin,
2501  pi_buff_rect_region region, size_t src_row_pitch, size_t src_slice_pitch,
2502  size_t dst_row_pitch, size_t dst_slice_pitch,
2503  pi_uint32 num_sync_points_in_wait_list,
2504  const pi_ext_sync_point *sync_point_wait_list,
2505  pi_ext_sync_point *sync_point);
2506 
2519  pi_ext_command_buffer command_buffer, pi_mem buffer, size_t offset,
2520  size_t size, void *dst, pi_uint32 num_sync_points_in_wait_list,
2521  const pi_ext_sync_point *sync_point_wait_list,
2522  pi_ext_sync_point *sync_point);
2523 
2541  pi_ext_command_buffer command_buffer, pi_mem buffer,
2542  pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset,
2543  pi_buff_rect_region region, size_t buffer_row_pitch,
2544  size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch,
2545  void *ptr, pi_uint32 num_sync_points_in_wait_list,
2546  const pi_ext_sync_point *sync_point_wait_list,
2547  pi_ext_sync_point *sync_point);
2548 
2561  pi_ext_command_buffer command_buffer, pi_mem buffer, size_t offset,
2562  size_t size, const void *ptr, pi_uint32 num_sync_points_in_wait_list,
2563  const pi_ext_sync_point *sync_point_wait_list,
2564  pi_ext_sync_point *sync_point);
2565 
2583  pi_ext_command_buffer command_buffer, pi_mem buffer,
2584  pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset,
2585  pi_buff_rect_region region, size_t buffer_row_pitch,
2586  size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch,
2587  const void *ptr, pi_uint32 num_sync_points_in_wait_list,
2588  const pi_ext_sync_point *sync_point_wait_list,
2589  pi_ext_sync_point *sync_point);
2590 
2604  pi_ext_command_buffer command_buffer, pi_mem buffer, const void *pattern,
2605  size_t pattern_size, size_t offset, size_t size,
2606  pi_uint32 num_sync_points_in_wait_list,
2607  const pi_ext_sync_point *sync_point_wait_list,
2608  pi_ext_sync_point *sync_point);
2609 
2621 __SYCL_EXPORT pi_result piextCommandBufferFillUSM(
2622  pi_ext_command_buffer command_buffer, void *ptr, const void *pattern,
2623  size_t pattern_size, size_t size, pi_uint32 num_sync_points_in_wait_list,
2624  const pi_ext_sync_point *sync_point_wait_list,
2625  pi_ext_sync_point *sync_point);
2626 
2638  pi_ext_command_buffer command_buffer, const void *ptr, size_t size,
2639  pi_usm_migration_flags flags, pi_uint32 num_sync_points_in_wait_list,
2640  const pi_ext_sync_point *sync_point_wait_list,
2641  pi_ext_sync_point *sync_point);
2642 
2654  pi_ext_command_buffer command_buffer, const void *ptr, size_t length,
2655  pi_mem_advice advice, pi_uint32 num_sync_points_in_wait_list,
2656  const pi_ext_sync_point *sync_point_wait_list,
2657  pi_ext_sync_point *sync_point);
2658 
2668 __SYCL_EXPORT pi_result
2670  pi_uint32 num_events_in_wait_list,
2671  const pi_event *event_wait_list, pi_event *event);
2672 
2680 
2683 __SYCL_EXPORT pi_result
2685 
2689 __SYCL_EXPORT pi_result
2691 
2698  pi_context context, pi_device device, pi_image_handle handle);
2699 
2705  pi_context context, pi_device device, pi_image_handle handle);
2706 
2715 __SYCL_EXPORT pi_result piextMemImageAllocate(pi_context context,
2716  pi_device device,
2718  pi_image_desc *image_desc,
2719  pi_image_mem_handle *ret_mem);
2720 
2728 __SYCL_EXPORT pi_result piextMemMipmapGetLevel(pi_context context,
2729  pi_device device,
2730  pi_image_mem_handle mip_mem,
2731  unsigned int level,
2732  pi_image_mem_handle *ret_mem);
2733 
2739 __SYCL_EXPORT pi_result piextMemImageFree(pi_context context, pi_device device,
2740  pi_image_mem_handle memory_handle);
2741 
2747 __SYCL_EXPORT pi_result piextMemMipmapFree(pi_context context, pi_device device,
2748  pi_image_mem_handle memory_handle);
2749 
2760  pi_context context, pi_device device, pi_image_mem_handle img_mem,
2761  pi_image_format *image_format, pi_image_desc *image_desc, pi_mem *ret_mem,
2762  pi_image_handle *ret_handle);
2763 
2775  pi_context context, pi_device device, pi_image_mem_handle img_mem,
2777  pi_sampler sampler, pi_mem *ret_mem, pi_image_handle *ret_handle);
2778 
2789  pi_context context, const pi_sampler_properties *sampler_properties,
2790  float min_mipmap_level_clamp, float max_mipmap_level_clamp,
2791  float max_anisotropy, pi_sampler *result_sampler);
2792 
2808 __SYCL_EXPORT pi_result piextMemImageCopy(
2809  pi_queue command_queue, void *dst_ptr, void *src_ptr,
2810  const pi_image_format *image_format, const pi_image_desc *image_desc,
2811  const pi_image_copy_flags flags, pi_image_offset src_offset,
2812  pi_image_offset dst_offset, pi_image_region copy_extent,
2813  pi_image_region host_extent, pi_uint32 num_events_in_wait_list,
2814  const pi_event *event_wait_list, pi_event *event);
2815 
2822 __SYCL_EXPORT pi_result piextMemImageGetInfo(
2823  const pi_image_mem_handle mem_handle, pi_image_info param_name,
2824  void *param_value, size_t *param_value_size_ret);
2825 
2834 __SYCL_EXPORT pi_result
2835 piextMemImportOpaqueFD(pi_context context, pi_device device, size_t size,
2836  int file_descriptor, pi_interop_mem_handle *ret_handle);
2837 
2847 __SYCL_EXPORT pi_result piextMemMapExternalArray(
2849  pi_image_desc *image_desc, pi_interop_mem_handle mem_handle,
2850  pi_image_mem_handle *ret_mem);
2851 
2857 __SYCL_EXPORT pi_result piextMemReleaseInterop(
2858  pi_context context, pi_device device, pi_interop_mem_handle memory_handle);
2859 
2868  pi_context context, pi_device device, int file_descriptor,
2869  pi_interop_semaphore_handle *ret_handle);
2870 
2877 __SYCL_EXPORT pi_result
2879  pi_interop_semaphore_handle sem_handle);
2880 
2890  pi_queue command_queue, pi_interop_semaphore_handle sem_handle,
2891  pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list,
2892  pi_event *event);
2893 
2904  pi_queue command_queue, pi_interop_semaphore_handle sem_handle,
2905  pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list,
2906  pi_event *event);
2907 
2908 typedef enum {
2914 
2915 struct _pi_plugin {
2916  // PI version supported by host passed to the plugin. The Plugin
2917  // checks and writes the appropriate Function Pointers in
2918  // PiFunctionTable.
2919  // TODO: Work on version fields and their handshaking mechanism.
2920  // Some choices are:
2921  // - Use of integers to keep major and minor version.
2922  // - Keeping char* Versions.
2923  char PiVersion[20];
2924  // Plugin edits this.
2925  char PluginVersion[20];
2926  char *Targets;
2928 #define _PI_API(api) decltype(::api) *api;
2929 #include <sycl/detail/pi.def>
2931 
2933 };
2934 
2935 #ifdef __cplusplus
2936 } // extern "C"
2937 #endif // __cplusplus
2938 
2939 #endif // _PI_H_
constexpr alignment_key::value_t< K > alignment
multi_ptr< ElementType, access::address_space::ext_intel_global_host_space, IsDecorated > host_ptr
image_format
Definition: image.hpp:93
ValueT length(const ValueT *a, const int len)
Calculate the square root of the input array.
Definition: math.hpp:161
constexpr pi_queue_properties PI_QUEUE_COMPUTE_INDEX
Definition: pi.h:795
pi_result piextUSMEnqueueMemset(pi_queue queue, void *ptr, pi_int32 value, size_t count, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event)
USM Memset API.
Definition: pi_cuda.cpp:905
pi_result piextDisablePeerAccess(pi_device command_device, pi_device peer_device)
Definition: pi_cuda.cpp:1258
_pi_kernel_cache_config
Definition: pi.h:808
@ PI_EXT_KERNEL_EXEC_INFO_CACHE_LARGE_DATA
Definition: pi.h:814
@ PI_EXT_KERNEL_EXEC_INFO_CACHE_DEFAULT
Definition: pi.h:810
@ PI_EXT_KERNEL_EXEC_INFO_CACHE_LARGE_SLM
Definition: pi.h:812
uint32_t pi_mem_access_flag
Definition: pi.h:1909
pi_result piextMemGetNativeHandle(pi_mem mem, pi_device dev, pi_native_handle *nativeHandle)
Gets the native handle of a PI mem object.
Definition: pi_cuda.cpp:236
static constexpr pi_device_fp_config PI_FP_SOFT_FLOAT
Definition: pi.h:888
constexpr pi_mem_flags PI_MEM_ACCESS_READ_ONLY
Definition: pi.h:761
pi_result piKernelCreate(pi_program program, const char *kernel_name, pi_kernel *ret_kernel)
Definition: pi_cuda.cpp:341
pi_result piextEnablePeerAccess(pi_device command_device, pi_device peer_device)
Definition: pi_cuda.cpp:1252
constexpr pi_memory_scope_capabilities PI_MEMORY_SCOPE_WORK_GROUP
Definition: pi.h:743
pi_bitfield pi_device_exec_capabilities
Definition: pi.h:718
int32_t pi_int32
Definition: pi.h:212
constexpr pi_sampler_properties PI_SAMPLER_PROPERTIES_NORMALIZED_COORDS
Definition: pi.h:725
constexpr pi_memory_scope_capabilities PI_MEMORY_SCOPE_WORK_ITEM
Definition: pi.h:741
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_SET_READ_MOSTLY
Definition: pi.h:618
pi_result piQueueFinish(pi_queue command_queue)
Definition: pi_cuda.cpp:186
constexpr pi_usm_mem_properties PI_MEM_ALLOC_INITIAL_PLACEMENT_DEVICE
Definition: pi.h:783
uintptr_t pi_native_handle
Definition: pi.h:217
pi_result piProgramGetBuildInfo(pi_program program, pi_device device, _pi_program_build_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:310
static constexpr pi_device_binary_type PI_DEVICE_BINARY_TYPE_LLVMIR_BITCODE
Definition: pi.h:948
pi_bitfield pi_map_flags
Definition: pi.h:768
pi_result piextDeviceGetNativeHandle(pi_device device, pi_native_handle *nativeHandle)
Gets the native handle of a PI device object.
Definition: pi_cuda.cpp:100
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_UNSET_PREFERRED_LOCATION_HOST
Definition: pi.h:633
pi_result piEnqueueMemBufferCopyRect(pi_queue command_queue, pi_mem src_buffer, pi_mem dst_buffer, pi_buff_rect_offset src_origin, pi_buff_rect_offset dst_origin, pi_buff_rect_region region, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:737
static constexpr pi_device_affinity_domain PI_DEVICE_AFFINITY_DOMAIN_NEXT_PARTITIONABLE
Definition: pi.h:878
_pi_result
Definition: pi.h:224
pi_result piMemImageGetInfo(pi_mem image, pi_image_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:781
pi_result piextEnqueueReadHostPipe(pi_queue queue, pi_program program, const char *pipe_symbol, pi_bool blocking, void *ptr, size_t size, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event)
Plugin.
Definition: pi_cuda.cpp:1013
pi_result piKernelRelease(pi_kernel kernel)
Definition: pi_cuda.cpp:531
constexpr pi_sampler_properties PI_SAMPLER_PROPERTIES_ADDRESSING_MODE
Definition: pi.h:727
pi_uint32 pi_bool
Definition: pi.h:215
static constexpr pi_device_partition_property PI_EXT_INTEL_DEVICE_PARTITION_BY_CSLICE
Definition: pi.h:863
constexpr pi_usm_mem_properties PI_MEM_USM_ALLOC_BUFFER_LOCATION
Definition: pi.h:789
pi_result piextUSMFree(pi_context context, void *ptr)
Indicates that the allocated USM memory is no longer needed on the runtime side.
Definition: pi_cuda.cpp:895
pi_result piextMemImageAllocate(pi_context context, pi_device device, pi_image_format *image_format, pi_image_desc *image_desc, pi_image_mem_handle *ret_mem)
API to allocate memory for bindless images.
Definition: pi_cuda.cpp:374
pi_result piextMemImageCopy(pi_queue command_queue, void *dst_ptr, void *src_ptr, const pi_image_format *image_format, const pi_image_desc *image_desc, const pi_image_copy_flags flags, pi_image_offset src_offset, pi_image_offset dst_offset, pi_image_region copy_extent, pi_image_region host_extent, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
API to copy image data Host to Device or Device to Host.
Definition: pi_cuda.cpp:427
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_UNSET_ACCESSED_BY
Definition: pi.h:628
constexpr pi_queue_properties PI_EXT_QUEUE_FLAG_SUBMISSION_IMMEDIATE
Definition: pi.h:805
pi_bitfield pi_usm_mem_properties
Definition: pi.h:780
_pi_device_info
Definition: pi.h:304
@ PI_DEVICE_INFO_PRINTF_BUFFER_SIZE
Definition: pi.h:374
@ PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_CHAR
Definition: pi.h:315
@ PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_WIDTH
Definition: pi.h:440
@ PI_EXT_ONEAPI_DEVICE_INFO_MAX_WORK_GROUPS_1D
Definition: pi.h:422
@ PI_DEVICE_INFO_MAX_MEM_BANDWIDTH
Definition: pi.h:400
@ PI_DEVICE_INFO_GPU_SLICES
Definition: pi.h:397
@ PI_DEVICE_INFO_PREFERRED_INTEROP_USER_SYNC
Definition: pi.h:375
@ PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_FLOAT
Definition: pi.h:326
@ PI_DEVICE_INFO_BUILD_ON_SUBDEVICE
Definition: pi.h:403
@ PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_MEMORY_IMPORT_SUPPORT
Definition: pi.h:447
@ PI_DEVICE_INFO_MAX_WORK_GROUP_SIZE
Definition: pi.h:310
@ PI_DEVICE_INFO_IMAGE_SRGB
Definition: pi.h:401
@ PI_EXT_ONEAPI_DEVICE_INFO_CUBEMAP_SUPPORT
Definition: pi.h:463
@ PI_DEVICE_INFO_IMAGE3D_MAX_WIDTH
Definition: pi.h:337
@ PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_SHORT
Definition: pi.h:316
@ PI_DEVICE_INFO_IMAGE3D_MAX_HEIGHT
Definition: pi.h:338
@ PI_DEVICE_INFO_MAX_CLOCK_FREQUENCY
Definition: pi.h:329
@ PI_DEVICE_INFO_USM_SINGLE_SHARED_SUPPORT
Definition: pi.h:386
@ PI_DEVICE_INFO_MAX_WORK_ITEM_DIMENSIONS
Definition: pi.h:308
@ PI_DEVICE_INFO_EXECUTION_CAPABILITIES
Definition: pi.h:360
@ PI_EXT_ONEAPI_DEVICE_INFO_COMMAND_BUFFER_UPDATE_SUPPORT
Definition: pi.h:460
@ PI_EXT_INTEL_DEVICE_INFO_ESIMD_SUPPORT
Definition: pi.h:432
@ PI_DEVICE_INFO_QUEUE_PROPERTIES
Definition: pi.h:314
@ PI_DEVICE_INFO_GPU_EU_COUNT
Definition: pi.h:395
@ PI_EXT_ONEAPI_DEVICE_INFO_MAX_WORK_GROUPS_2D
Definition: pi.h:423
@ PI_DEVICE_INFO_COMPILER_AVAILABLE
Definition: pi.h:358
@ PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_PITCH
Definition: pi.h:442
@ PI_DEVICE_INFO_IMAGE_SUPPORT
Definition: pi.h:332
@ PI_DEVICE_INFO_ERROR_CORRECTION_SUPPORT
Definition: pi.h:353
@ PI_EXT_INTEL_DEVICE_INFO_FREE_MEMORY
Definition: pi.h:404
@ PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_MEMORY_EXPORT_SUPPORT
Definition: pi.h:448
@ PI_DEVICE_INFO_QUEUE_ON_DEVICE_PROPERTIES
Definition: pi.h:361
@ PI_DEVICE_INFO_PARTITION_MAX_SUB_DEVICES
Definition: pi.h:378
@ PI_DEVICE_INFO_BUILT_IN_KERNELS
Definition: pi.h:363
@ PI_DEVICE_INFO_UUID
Definition: pi.h:390
@ PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_FLOAT
Definition: pi.h:319
@ PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_HEIGHT
Definition: pi.h:441
@ PI_DEVICE_INFO_MAX_NUM_SUB_GROUPS
Definition: pi.h:381
@ PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_SUPPORT
Definition: pi.h:443
@ PI_DEVICE_INFO_AVAILABLE
Definition: pi.h:357
@ PI_DEVICE_INFO_HALF_FP_CONFIG
Definition: pi.h:312
@ PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_ANISOTROPY_SUPPORT
Definition: pi.h:444
@ PI_DEVICE_INFO_PARENT_DEVICE
Definition: pi.h:376
@ PI_DEVICE_INFO_SUB_GROUP_SIZES_INTEL
Definition: pi.h:383
@ PI_DEVICE_INFO_GPU_EU_SIMD_WIDTH
Definition: pi.h:396
@ PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_SHORT
Definition: pi.h:323
@ PI_DEVICE_INFO_QUEUE_ON_HOST_PROPERTIES
Definition: pi.h:362
@ PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_DOUBLE
Definition: pi.h:327
@ PI_DEVICE_INFO_GLOBAL_MEM_CACHE_TYPE
Definition: pi.h:345
@ PI_DEVICE_INFO_DRIVER_VERSION
Definition: pi.h:369
@ PI_DEVICE_INFO_MAX_PARAMETER_SIZE
Definition: pi.h:343
@ PI_DEVICE_INFO_PARTITION_PROPERTIES
Definition: pi.h:377
@ PI_EXT_DEVICE_INFO_ATOMIC_MEMORY_ORDER_CAPABILITIES
Definition: pi.h:415
@ PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_INT
Definition: pi.h:317
@ PI_DEVICE_INFO_USM_CROSS_SHARED_SUPPORT
Definition: pi.h:387
@ PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_HALF
Definition: pi.h:321
@ PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_SUPPORT
Definition: pi.h:435
@ PI_DEVICE_INFO_VERSION
Definition: pi.h:371
@ PI_EXT_INTEL_DEVICE_INFO_MEM_CHANNEL_SUPPORT
Definition: pi.h:429
@ PI_DEVICE_INFO_IMAGE3D_MAX_DEPTH
Definition: pi.h:339
@ PI_DEVICE_INFO_MAX_COMPUTE_UNITS
Definition: pi.h:307
@ PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_CHAR
Definition: pi.h:322
@ PI_DEVICE_INFO_IMAGE2D_MAX_HEIGHT
Definition: pi.h:336
@ PI_DEVICE_INFO_VENDOR_ID
Definition: pi.h:306
@ PI_DEVICE_INFO_GLOBAL_MEM_CACHELINE_SIZE
Definition: pi.h:346
@ PI_EXT_DEVICE_INFO_ATOMIC_FENCE_SCOPE_CAPABILITIES
Definition: pi.h:428
@ PI_DEVICE_INFO_GPU_HW_THREADS_PER_EU
Definition: pi.h:417
@ PI_DEVICE_INFO_BACKEND_VERSION
Definition: pi.h:418
@ PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_LONG
Definition: pi.h:318
@ PI_DEVICE_INFO_PLATFORM
Definition: pi.h:364
@ PI_EXT_ONEAPI_DEVICE_INFO_CUBEMAP_SEAMLESS_FILTERING_SUPPORT
Definition: pi.h:464
@ PI_DEVICE_INFO_LOCAL_MEM_SIZE
Definition: pi.h:352
@ PI_DEVICE_INFO_REFERENCE_COUNT
Definition: pi.h:365
@ PI_DEVICE_INFO_MAX_SAMPLERS
Definition: pi.h:342
@ PI_DEVICE_INFO_DEVICE_ID
Definition: pi.h:393
@ PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_SEMAPHORE_EXPORT_SUPPORT
Definition: pi.h:450
@ PI_DEVICE_INFO_PROFILE
Definition: pi.h:370
@ PI_DEVICE_INFO_MAX_WORK_ITEM_SIZES
Definition: pi.h:309
@ PI_DEVICE_INFO_EXTENSIONS
Definition: pi.h:373
@ PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_INT
Definition: pi.h:324
@ PI_DEVICE_INFO_HOST_UNIFIED_MEMORY
Definition: pi.h:354
@ PI_DEVICE_INFO_PCI_ADDRESS
Definition: pi.h:394
@ PI_DEVICE_INFO_MAX_MEM_ALLOC_SIZE
Definition: pi.h:331
@ PI_DEVICE_INFO_GLOBAL_MEM_CACHE_SIZE
Definition: pi.h:347
@ PI_DEVICE_INFO_IMAGE_MAX_ARRAY_SIZE
Definition: pi.h:341
@ PI_DEVICE_INFO_LINKER_AVAILABLE
Definition: pi.h:359
@ PI_EXT_ONEAPI_DEVICE_INFO_MAX_WORK_GROUPS_3D
Definition: pi.h:424
@ PI_EXT_ONEAPI_DEVICE_INFO_CUDA_ASYNC_BARRIER
Definition: pi.h:425
@ PI_DEVICE_INFO_ADDRESS_BITS
Definition: pi.h:330
@ PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_DOUBLE
Definition: pi.h:320
@ PI_DEVICE_INFO_ATOMIC_64
Definition: pi.h:414
@ PI_DEVICE_INFO_USM_HOST_SUPPORT
Definition: pi.h:384
@ PI_DEVICE_INFO_USM_DEVICE_SUPPORT
Definition: pi.h:385
@ PI_EXT_INTEL_DEVICE_INFO_MAX_COMPUTE_QUEUE_INDICES
Definition: pi.h:413
@ PI_DEVICE_INFO_SUB_GROUP_INDEPENDENT_FORWARD_PROGRESS
Definition: pi.h:382
@ PI_DEVICE_INFO_GPU_EU_COUNT_PER_SUBSLICE
Definition: pi.h:399
@ PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_LEVEL_REFERENCE_SUPPORT
Definition: pi.h:446
@ PI_DEVICE_INFO_MAX_WRITE_IMAGE_ARGS
Definition: pi.h:334
@ PI_EXT_DEVICE_INFO_ATOMIC_MEMORY_SCOPE_CAPABILITIES
Definition: pi.h:416
@ PI_EXT_CODEPLAY_DEVICE_INFO_MAX_REGISTERS_PER_WORK_GROUP
Definition: pi.h:431
@ PI_DEVICE_INFO_IL_VERSION
Definition: pi.h:366
@ PI_EXT_CODEPLAY_DEVICE_INFO_SUPPORTS_FUSION
Definition: pi.h:426
@ PI_DEVICE_INFO_PROFILING_TIMER_RESOLUTION
Definition: pi.h:355
@ PI_DEVICE_INFO_VENDOR
Definition: pi.h:368
@ PI_DEVICE_INFO_MEM_BASE_ADDR_ALIGN
Definition: pi.h:344
@ PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_LONG
Definition: pi.h:325
@ PI_DEVICE_INFO_MAX_READ_IMAGE_ARGS
Definition: pi.h:333
@ PI_DEVICE_INFO_PARTITION_AFFINITY_DOMAIN
Definition: pi.h:379
@ PI_DEVICE_INFO_USM_SYSTEM_SHARED_SUPPORT
Definition: pi.h:388
@ PI_EXT_ONEAPI_DEVICE_INFO_MAX_GLOBAL_WORK_GROUPS
Definition: pi.h:421
@ PI_DEVICE_INFO_IMAGE_MAX_BUFFER_SIZE
Definition: pi.h:340
@ PI_EXT_INTEL_DEVICE_INFO_MEMORY_CLOCK_RATE
Definition: pi.h:407
@ PI_DEVICE_INFO_DOUBLE_FP_CONFIG
Definition: pi.h:313
@ PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_SEMAPHORE_IMPORT_SUPPORT
Definition: pi.h:449
@ PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_SHARED_USM_SUPPORT
Definition: pi.h:436
@ PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_1D_USM_SUPPORT
Definition: pi.h:437
@ PI_EXT_ONEAPI_DEVICE_INFO_MATRIX_COMBINATIONS
Definition: pi.h:452
@ PI_EXT_INTEL_DEVICE_INFO_MEMORY_BUS_WIDTH
Definition: pi.h:410
@ PI_DEVICE_INFO_MAX_CONSTANT_BUFFER_SIZE
Definition: pi.h:349
@ PI_DEVICE_INFO_PARTITION_TYPE
Definition: pi.h:380
@ PI_DEVICE_INFO_TYPE
Definition: pi.h:305
@ PI_EXT_ONEAPI_DEVICE_INFO_COMPONENT_DEVICES
Definition: pi.h:455
@ PI_DEVICE_INFO_SINGLE_FP_CONFIG
Definition: pi.h:311
@ PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_HALF
Definition: pi.h:328
@ PI_DEVICE_INFO_GLOBAL_MEM_SIZE
Definition: pi.h:348
@ PI_DEVICE_INFO_NAME
Definition: pi.h:367
@ PI_EXT_ONEAPI_DEVICE_INFO_COMPOSITE_DEVICE
Definition: pi.h:456
@ PI_DEVICE_INFO_LOCAL_MEM_TYPE
Definition: pi.h:351
@ PI_DEVICE_INFO_MAX_CONSTANT_ARGS
Definition: pi.h:350
@ PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_MAX_ANISOTROPY
Definition: pi.h:445
@ PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_2D_USM_SUPPORT
Definition: pi.h:438
@ PI_EXT_ONEAPI_DEVICE_INFO_COMMAND_BUFFER_SUPPORT
Definition: pi.h:459
@ PI_EXT_DEVICE_INFO_ATOMIC_FENCE_ORDER_CAPABILITIES
Definition: pi.h:427
@ PI_EXT_ONEAPI_DEVICE_INFO_IMAGE_PITCH_ALIGN
Definition: pi.h:439
@ PI_DEVICE_INFO_OPENCL_C_VERSION
Definition: pi.h:372
@ PI_DEVICE_INFO_ENDIAN_LITTLE
Definition: pi.h:356
@ PI_DEVICE_INFO_IMAGE2D_MAX_WIDTH
Definition: pi.h:335
@ PI_DEVICE_INFO_GPU_SUBSLICES_PER_SLICE
Definition: pi.h:398
@ PI_EXT_ONEAPI_DEVICE_INFO_IP_VERSION
Definition: pi.h:392
@ PI_EXT_ONEAPI_DEVICE_INFO_BFLOAT16_MATH_FUNCTIONS
Definition: pi.h:420
pi_uint64 pi_bitfield
Definition: pi.h:216
static constexpr pi_device_binary_type PI_DEVICE_BINARY_TYPE_NATIVE
Definition: pi.h:943
static constexpr pi_device_fp_config PI_FP_DENORM
Definition: pi.h:882
pi_result piextUSMEnqueueMemAdvise(pi_queue queue, const void *ptr, size_t length, pi_mem_advice advice, pi_event *event)
USM Memadvise API.
Definition: pi_cuda.cpp:934
constexpr pi_mem_flags PI_MEM_FLAGS_HOST_PTR_ALLOC
Definition: pi.h:765
pi_result piextMemImportOpaqueFD(pi_context context, pi_device device, size_t size, int file_descriptor, pi_interop_mem_handle *ret_handle)
API to import external memory in the form of a file descriptor.
Definition: pi_cuda.cpp:459
_pi_program_build_status
Definition: pi.h:255
@ PI_PROGRAM_BUILD_STATUS_SUCCESS
Definition: pi.h:258
@ PI_PROGRAM_BUILD_STATUS_ERROR
Definition: pi.h:257
@ PI_PROGRAM_BUILD_STATUS_IN_PROGRESS
Definition: pi.h:259
@ PI_PROGRAM_BUILD_STATUS_NONE
Definition: pi.h:256
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_UNSET_ACCESSED_BY_HOST
Definition: pi.h:637
static constexpr pi_device_affinity_domain PI_DEVICE_AFFINITY_DOMAIN_L2_CACHE
Definition: pi.h:873
constexpr pi_usm_mem_properties PI_MEM_ALLOC_FLAGS
Definition: pi.h:781
_pi_queue_info
Definition: pi.h:496
@ PI_QUEUE_INFO_DEVICE_DEFAULT
Definition: pi.h:499
@ PI_QUEUE_INFO_PROPERTIES
Definition: pi.h:500
@ PI_QUEUE_INFO_DEVICE
Definition: pi.h:498
@ PI_QUEUE_INFO_CONTEXT
Definition: pi.h:497
@ PI_QUEUE_INFO_REFERENCE_COUNT
Definition: pi.h:501
@ PI_QUEUE_INFO_SIZE
Definition: pi.h:502
@ PI_EXT_ONEAPI_QUEUE_INFO_EMPTY
Definition: pi.h:505
_pi_mem_advice
Definition: pi.h:599
@ PI_MEM_ADVICE_CUDA_UNSET_ACCESSED_BY_HOST
Definition: pi.h:611
@ PI_MEM_ADVICE_CUDA_UNSET_READ_MOSTLY
Definition: pi.h:603
@ PI_MEM_ADVICE_CUDA_UNSET_PREFERRED_LOCATION
Definition: pi.h:605
@ PI_MEM_ADVICE_CUDA_SET_READ_MOSTLY
Definition: pi.h:602
@ PI_MEM_ADVICE_RESET
Definition: pi.h:601
@ PI_MEM_ADVICE_CUDA_SET_PREFERRED_LOCATION_HOST
Definition: pi.h:608
@ PI_MEM_ADVICE_HIP_SET_COARSE_GRAINED
Definition: pi.h:612
@ PI_MEM_ADVICE_UNKNOWN
Definition: pi.h:614
@ PI_MEM_ADVICE_HIP_UNSET_COARSE_GRAINED
Definition: pi.h:613
@ PI_MEM_ADVICE_CUDA_UNSET_ACCESSED_BY
Definition: pi.h:607
@ PI_MEM_ADVICE_CUDA_SET_ACCESSED_BY_HOST
Definition: pi.h:610
@ PI_MEM_ADVICE_CUDA_UNSET_PREFERRED_LOCATION_HOST
Definition: pi.h:609
@ PI_MEM_ADVICE_CUDA_SET_PREFERRED_LOCATION
Definition: pi.h:604
@ PI_MEM_ADVICE_CUDA_SET_ACCESSED_BY
Definition: pi.h:606
pi_result piEnqueueMemImageRead(pi_queue command_queue, pi_mem image, pi_bool blocking_read, pi_image_offset origin, pi_image_region region, size_t row_pitch, size_t slice_pitch, void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:789
static constexpr pi_device_affinity_domain PI_DEVICE_AFFINITY_DOMAIN_L3_CACHE
Definition: pi.h:871
pi_result piextPluginGetOpaqueData(void *opaque_data_param, void **opaque_data_return)
API to get Plugin internal data, opaque to SYCL RT.
Definition: pi_cuda.cpp:1238
_pi_image_channel_type
Definition: pi.h:658
@ PI_IMAGE_CHANNEL_TYPE_FLOAT
Definition: pi.h:673
@ PI_IMAGE_CHANNEL_TYPE_UNORM_SHORT_565
Definition: pi.h:663
@ PI_IMAGE_CHANNEL_TYPE_UNORM_INT8
Definition: pi.h:661
@ PI_IMAGE_CHANNEL_TYPE_SIGNED_INT16
Definition: pi.h:667
@ PI_IMAGE_CHANNEL_TYPE_SNORM_INT16
Definition: pi.h:660
@ PI_IMAGE_CHANNEL_TYPE_UNSIGNED_INT8
Definition: pi.h:669
@ PI_IMAGE_CHANNEL_TYPE_SIGNED_INT8
Definition: pi.h:666
@ PI_IMAGE_CHANNEL_TYPE_UNORM_INT_101010
Definition: pi.h:665
@ PI_IMAGE_CHANNEL_TYPE_UNSIGNED_INT32
Definition: pi.h:671
@ PI_IMAGE_CHANNEL_TYPE_UNSIGNED_INT16
Definition: pi.h:670
@ PI_IMAGE_CHANNEL_TYPE_SNORM_INT8
Definition: pi.h:659
@ PI_IMAGE_CHANNEL_TYPE_UNORM_SHORT_555
Definition: pi.h:664
@ PI_IMAGE_CHANNEL_TYPE_SIGNED_INT32
Definition: pi.h:668
@ PI_IMAGE_CHANNEL_TYPE_HALF_FLOAT
Definition: pi.h:672
@ PI_IMAGE_CHANNEL_TYPE_UNORM_INT16
Definition: pi.h:662
_pi_kernel_sub_group_info
Definition: pi.h:538
@ PI_KERNEL_COMPILE_NUM_SUB_GROUPS
Definition: pi.h:541
@ PI_KERNEL_MAX_NUM_SUB_GROUPS
Definition: pi.h:540
@ PI_KERNEL_COMPILE_SUB_GROUP_SIZE_INTEL
Definition: pi.h:542
@ PI_KERNEL_MAX_SUB_GROUP_SIZE
Definition: pi.h:539
pi_result piextProgramCreateWithNativeHandle(pi_native_handle nativeHandle, pi_context context, bool pluginOwnsNativeHandle, pi_program *program)
Creates PI program object from a native handle.
Definition: pi_cuda.cpp:333
pi_bitfield pi_mem_properties
Definition: pi.h:774
pi_result piextEnqueueWriteHostPipe(pi_queue queue, pi_program program, const char *pipe_symbol, pi_bool blocking, void *ptr, size_t size, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event)
Write to pipe of a given name.
Definition: pi_cuda.cpp:1033
pi_result piKernelSetArg(pi_kernel kernel, pi_uint32 arg_index, size_t arg_size, const void *arg_value)
Definition: pi_cuda.cpp:347
pi_result piEnqueueEventsWaitWithBarrier(pi_queue command_queue, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:665
pi_result piEnqueueKernelLaunch(pi_queue queue, pi_kernel kernel, pi_uint32 work_dim, const size_t *global_work_offset, const size_t *global_work_size, const size_t *local_work_size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:537
_pi_kernel_group_info
Definition: pi.h:517
@ PI_KERNEL_GROUP_INFO_WORK_GROUP_SIZE
Definition: pi.h:519
@ PI_KERNEL_GROUP_INFO_COMPILE_WORK_GROUP_SIZE
Definition: pi.h:520
@ PI_KERNEL_GROUP_INFO_GLOBAL_WORK_SIZE
Definition: pi.h:518
@ PI_KERNEL_GROUP_INFO_PRIVATE_MEM_SIZE
Definition: pi.h:523
@ PI_KERNEL_GROUP_INFO_LOCAL_MEM_SIZE
Definition: pi.h:521
@ PI_KERNEL_GROUP_INFO_PREFERRED_WORK_GROUP_SIZE_MULTIPLE
Definition: pi.h:522
@ PI_KERNEL_GROUP_INFO_NUM_REGS
Definition: pi.h:525
pi_result piEnqueueMemBufferRead(pi_queue queue, pi_mem buffer, pi_bool blocking_read, size_t offset, size_t size, void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:674
constexpr pi_queue_properties PI_EXT_QUEUE_FLAG_SUBMISSION_NO_IMMEDIATE
Definition: pi.h:804
constexpr pi_memory_order_capabilities PI_MEMORY_ORDER_ACQUIRE
Definition: pi.h:735
pi_result piProgramRetain(pi_program program)
Definition: pi_cuda.cpp:320
pi_uint64 pi_image_handle
Definition: pi.h:1157
pi_result piextWaitExternalSemaphore(pi_queue command_queue, pi_interop_semaphore_handle sem_handle, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
API to instruct the queue with a non-blocking wait on an external semaphore.
Definition: pi_cuda.cpp:492
const pi_bool PI_TRUE
Definition: pi.h:684
pi_result piextCommandBufferMemBufferCopy(pi_ext_command_buffer command_buffer, pi_mem src_buffer, pi_mem dst_buffer, size_t src_offset, size_t dst_offset, size_t size, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a mem buffer copy command to the command-buffer.
Definition: pi_cuda.cpp:1108
_pi_ext_command_buffer_command * pi_ext_command_buffer_command
Definition: pi.h:2351
pi_result piextUSMImport(const void *ptr, size_t size, pi_context context)
Import host system memory into USM.
Definition: pi_cuda.cpp:985
pi_result piMemImageCreate(pi_context context, pi_mem_flags flags, const pi_image_format *image_format, const pi_image_desc *image_desc, void *host_ptr, pi_mem *ret_mem)
Definition: pi_cuda.cpp:227
pi_property_type
Definition: pi.h:909
@ PI_PROPERTY_TYPE_UINT32
Definition: pi.h:911
@ PI_PROPERTY_TYPE_UNKNOWN
Definition: pi.h:910
@ PI_PROPERTY_TYPE_BYTE_ARRAY
Definition: pi.h:912
@ PI_PROPERTY_TYPE_STRING
Definition: pi.h:913
pi_result piextCommandBufferRetainCommand(pi_ext_command_buffer_command command)
API to increment the reference count of a command-buffer command.
Definition: pi_cuda.cpp:1229
pi_result piextUSMEnqueuePrefetch(pi_queue queue, const void *ptr, size_t size, pi_usm_migration_flags flags, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event)
Hint to migrate memory to the device.
Definition: pi_cuda.cpp:924
_pi_kernel_exec_info
Definition: pi.h:1601
@ PI_USM_INDIRECT_ACCESS
indicates that the kernel might access data through USM ptrs
Definition: pi.h:1603
@ PI_EXT_KERNEL_EXEC_INFO_CACHE_CONFIG
provides the preferred cache configuration (large slm or large data)
Definition: pi.h:1607
@ PI_USM_PTRS
provides an explicit list of pointers that the kernel will access
Definition: pi.h:1605
pi_result piextPlatformGetNativeHandle(pi_platform platform, pi_native_handle *nativeHandle)
Gets the native handle of a PI platform object.
Definition: pi_cuda.cpp:42
pi_result piProgramRelease(pi_program program)
Definition: pi_cuda.cpp:324
pi_result piextEnqueueCommandBuffer(pi_ext_command_buffer command_buffer, pi_queue queue, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
API to submit the command-buffer to queue for execution, returns an error if the command-buffer is no...
Definition: pi_cuda.cpp:1213
_pi_platform_info
Definition: pi.h:239
@ PI_PLATFORM_INFO_VENDOR
Definition: pi.h:243
@ PI_PLATFORM_INFO_EXTENSIONS
Definition: pi.h:240
@ PI_PLATFORM_INFO_PROFILE
Definition: pi.h:242
@ PI_EXT_PLATFORM_INFO_BACKEND
Definition: pi.h:245
@ PI_PLATFORM_INFO_NAME
Definition: pi.h:241
@ PI_PLATFORM_INFO_VERSION
Definition: pi.h:244
pi_result piextKernelSetArgSampler(pi_kernel kernel, pi_uint32 arg_index, const pi_sampler *arg_value)
Definition: pi_cuda.cpp:360
pi_result piextMemMapExternalArray(pi_context context, pi_device device, pi_image_format *image_format, pi_image_desc *image_desc, pi_interop_mem_handle mem_handle, pi_image_mem_handle *ret_mem)
API to map an interop memory handle to an image memory handle.
Definition: pi_cuda.cpp:465
static constexpr pi_device_partition_property PI_DEVICE_PARTITION_BY_COUNTS
Definition: pi.h:856
pi_result piSamplerRetain(pi_sampler sampler)
Definition: pi_cuda.cpp:649
pi_result piextMemSampledImageCreate(pi_context context, pi_device device, pi_image_mem_handle img_mem, pi_image_format *image_format, pi_image_desc *image_desc, pi_sampler sampler, pi_mem *ret_mem, pi_image_handle *ret_handle)
API to create sampled bindless image handles.
Definition: pi_cuda.cpp:391
constexpr pi_queue_properties PI_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE
Definition: pi.h:797
pi_result piMemGetInfo(pi_mem mem, pi_mem_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:217
pi_result piextUSMHostAlloc(void **result_ptr, pi_context context, pi_usm_mem_properties *properties, size_t size, pi_uint32 alignment)
Allocates host memory accessible by the device.
Definition: pi_cuda.cpp:888
pi_result piDeviceGetInfo(pi_device device, pi_device_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Returns requested info for provided native device Return PI_DEVICE_INFO_EXTENSION_DEVICELIB_ASSERT fo...
Definition: pi_cuda.cpp:78
pi_result piKernelRetain(pi_kernel kernel)
Definition: pi_cuda.cpp:526
pi_result piPlatformsGet(pi_uint32 num_entries, pi_platform *platforms, pi_uint32 *num_platforms)
Definition: pi_cuda.cpp:30
pi_result piextEnqueueDeviceGlobalVariableWrite(pi_queue queue, pi_program program, const char *name, pi_bool blocking_write, size_t count, size_t offset, const void *src, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Device global variable.
Definition: pi_cuda.cpp:993
pi_result piextKernelSuggestMaxCooperativeGroupCount(pi_kernel kernel, size_t local_work_size, size_t dynamic_shared_memory_size, pi_uint32 *group_count_ret)
Gets the max work group count for a cooperative kernel.
Definition: pi_cuda.cpp:572
pi_result piextQueueCreate(pi_context context, pi_device device, pi_queue_properties *properties, pi_queue *queue)
Definition: pi_cuda.cpp:167
pi_result piextDestroyExternalSemaphore(pi_context context, pi_device device, pi_interop_semaphore_handle sem_handle)
API to destroy the external semaphore handle.
Definition: pi_cuda.cpp:487
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_UNSET_READ_MOSTLY
Definition: pi.h:620
_pi_image_info
Definition: pi.h:528
@ PI_IMAGE_INFO_HEIGHT
Definition: pi.h:534
@ PI_IMAGE_INFO_SLICE_PITCH
Definition: pi.h:532
@ PI_IMAGE_INFO_ROW_PITCH
Definition: pi.h:531
@ PI_IMAGE_INFO_WIDTH
Definition: pi.h:533
@ PI_IMAGE_INFO_FORMAT
Definition: pi.h:529
@ PI_IMAGE_INFO_ELEMENT_SIZE
Definition: pi.h:530
@ PI_IMAGE_INFO_DEPTH
Definition: pi.h:535
pi_result piextEventGetNativeHandle(pi_event event, pi_native_handle *nativeHandle)
Gets the native handle of a PI event object.
Definition: pi_cuda.cpp:621
pi_result piEnqueueEventsWait(pi_queue command_queue, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:657
_pi_usm_capabilities
Definition: pi.h:1951
@ PI_USM_ATOMIC_ACCESS
Definition: pi.h:1953
@ PI_USM_ACCESS
Definition: pi.h:1952
@ PI_USM_CONCURRENT_ACCESS
Definition: pi.h:1954
@ PI_USM_CONCURRENT_ATOMIC_ACCESS
Definition: pi.h:1955
pi_uint32 pi_ext_sync_point
Definition: pi.h:2352
_pi_image_channel_order
Definition: pi.h:640
@ PI_IMAGE_CHANNEL_ORDER_RGB
Definition: pi.h:645
@ PI_IMAGE_CHANNEL_ORDER_RA
Definition: pi.h:644
@ PI_IMAGE_CHANNEL_ORDER_RGBA
Definition: pi.h:646
@ PI_IMAGE_CHANNEL_ORDER_INTENSITY
Definition: pi.h:650
@ PI_IMAGE_CHANNEL_ORDER_R
Definition: pi.h:642
@ PI_IMAGE_CHANNEL_ORDER_ABGR
Definition: pi.h:649
@ PI_IMAGE_CHANNEL_ORDER_RGBx
Definition: pi.h:654
@ PI_IMAGE_CHANNEL_ORDER_RG
Definition: pi.h:643
@ PI_IMAGE_CHANNEL_ORDER_BGRA
Definition: pi.h:647
@ PI_IMAGE_CHANNEL_ORDER_Rx
Definition: pi.h:652
@ PI_IMAGE_CHANNEL_ORDER_A
Definition: pi.h:641
@ PI_IMAGE_CHANNEL_ORDER_sRGBA
Definition: pi.h:655
@ PI_IMAGE_CHANNEL_ORDER_LUMINANCE
Definition: pi.h:651
@ PI_IMAGE_CHANNEL_ORDER_RGx
Definition: pi.h:653
@ PI_IMAGE_CHANNEL_ORDER_ARGB
Definition: pi.h:648
constexpr pi_memory_order_capabilities PI_MEMORY_ORDER_ACQ_REL
Definition: pi.h:737
_pi_profiling_info
Definition: pi.h:747
@ PI_PROFILING_INFO_COMMAND_END
Definition: pi.h:751
@ PI_PROFILING_INFO_COMMAND_START
Definition: pi.h:750
@ PI_PROFILING_INFO_COMMAND_SUBMIT
Definition: pi.h:749
@ PI_PROFILING_INFO_COMMAND_QUEUED
Definition: pi.h:748
static constexpr pi_device_partition_property PI_DEVICE_PARTITION_BY_COUNTS_LIST_END
Definition: pi.h:859
_pi_usm_type
Definition: pi.h:1965
@ PI_MEM_TYPE_SHARED
Definition: pi.h:1969
@ PI_MEM_TYPE_UNKNOWN
Definition: pi.h:1966
@ PI_MEM_TYPE_DEVICE
Definition: pi.h:1968
@ PI_MEM_TYPE_HOST
Definition: pi.h:1967
_pi_sampler_cubemap_filter_mode
Definition: pi.h:711
@ PI_SAMPLER_CUBEMAP_FILTER_MODE_SEAMLESS
Definition: pi.h:713
@ PI_SAMPLER_CUBEMAP_FILTER_MODE_DISJOINTED
Definition: pi.h:712
_pi_usm_migration_flags
Definition: pi.h:1975
@ PI_USM_MIGRATION_TBD0
Definition: pi.h:1976
pi_result piTearDown(void *PluginParameter)
API to notify that the plugin should clean up its resources.
Definition: pi_cuda.cpp:1243
_pi_device_type
Definition: pi.h:272
@ PI_DEVICE_TYPE_ACC
A PI device that is a dedicated accelerator.
Definition: pi.h:278
@ PI_DEVICE_TYPE_ALL
All devices available in the PI plugin.
Definition: pi.h:275
@ PI_DEVICE_TYPE_DEFAULT
The default device available in the PI plugin.
Definition: pi.h:273
@ PI_DEVICE_TYPE_GPU
A PI device that is a GPU.
Definition: pi.h:277
@ PI_DEVICE_TYPE_CUSTOM
A PI device that is a custom device.
Definition: pi.h:280
@ PI_DEVICE_TYPE_CPU
A PI device that is the host processor.
Definition: pi.h:276
static constexpr pi_device_partition_property PI_DEVICE_PARTITION_EQUALLY
Definition: pi.h:854
pi_result piextUSMRelease(const void *ptr, pi_context context)
Release host system memory from USM.
Definition: pi_cuda.cpp:989
pi_result piContextRetain(pi_context context)
Definition: pi_cuda.cpp:152
#define PI_BIT(_i)
Definition: pi.h:1900
_pi_device_mem_cache_type
Definition: pi.h:293
@ PI_DEVICE_MEM_CACHE_TYPE_READ_WRITE_CACHE
Definition: pi.h:296
@ PI_DEVICE_MEM_CACHE_TYPE_READ_ONLY_CACHE
Definition: pi.h:295
@ PI_DEVICE_MEM_CACHE_TYPE_NONE
Definition: pi.h:294
uint64_t pi_uint64
Definition: pi.h:214
pi_bitfield pi_queue_properties
Definition: pi.h:793
pi_result piEventCreate(pi_context context, pi_event *ret_event)
Create PI event object in a signalled/completed state.
Definition: pi_cuda.cpp:579
pi_result piextCommandBufferNDRangeKernel(pi_ext_command_buffer command_buffer, pi_kernel kernel, pi_uint32 work_dim, const size_t *global_work_offset, const size_t *global_work_size, const size_t *local_work_size, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point, pi_ext_command_buffer_command *command)
API to append a kernel execution command to the command-buffer.
Definition: pi_cuda.cpp:1087
pi_result piextPlatformCreateWithNativeHandle(pi_native_handle nativeHandle, pi_platform *platform)
Creates PI platform object from a native handle.
Definition: pi_cuda.cpp:47
uint8_t pi_device_binary_type
Types of device binary.
Definition: pi.h:939
pi_result piMemRetain(pi_mem mem)
Definition: pi_cuda.cpp:223
_pi_image_copy_flags
Definition: pi.h:676
@ PI_IMAGE_COPY_DEVICE_TO_DEVICE
Definition: pi.h:679
@ PI_IMAGE_COPY_DEVICE_TO_HOST
Definition: pi.h:678
@ PI_IMAGE_COPY_HOST_TO_DEVICE
Definition: pi.h:677
pi_result piextCommandBufferFinalize(pi_ext_command_buffer command_buffer)
API to stop command-buffer recording such that no more commands can be appended, and makes the comman...
Definition: pi_cuda.cpp:1083
pi_result piKernelGetSubGroupInfo(pi_kernel kernel, pi_device device, pi_kernel_sub_group_info param_name, size_t input_value_size, const void *input_value, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
API to query information from the sub-group from a kernel.
Definition: pi_cuda.cpp:516
pi_result piextProgramSetSpecializationConstant(pi_program prog, pi_uint32 spec_id, size_t spec_size, const void *spec_value)
Sets a specialization constant to a specific value.
Definition: pi_cuda.cpp:1060
const pi_bool PI_FALSE
Definition: pi.h:685
pi_result piEventGetInfo(pi_event event, pi_event_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:583
pi_result piextCommandBufferMemBufferReadRect(pi_ext_command_buffer command_buffer, pi_mem buffer, pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset, pi_buff_rect_region region, size_t buffer_row_pitch, size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, void *ptr, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a rectangular mem buffer read command to the command-buffer.
Definition: pi_cuda.cpp:1139
pi_result piDevicesGet(pi_platform platform, pi_device_type device_type, pi_uint32 num_entries, pi_device *devices, pi_uint32 *num_devices)
Definition: pi_cuda.cpp:63
pi_result piQueueRelease(pi_queue command_queue)
Definition: pi_cuda.cpp:182
constexpr pi_queue_properties PI_EXT_ONEAPI_QUEUE_FLAG_PRIORITY_LOW
Definition: pi.h:802
pi_result piextCommandBufferMemBufferRead(pi_ext_command_buffer command_buffer, pi_mem buffer, size_t offset, size_t size, void *dst, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a mem buffer read command to the command-buffer.
Definition: pi_cuda.cpp:1130
pi_result piextUSMEnqueueMemset2D(pi_queue queue, void *ptr, size_t pitch, int value, size_t width, size_t height, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event)
USM 2D Memset API.
Definition: pi_cuda.cpp:954
constexpr pi_memory_order_capabilities PI_MEMORY_ORDER_RELEASE
Definition: pi.h:736
pi_bitfield pi_sampler_properties
Definition: pi.h:724
_pi_usm_capability_query
Definition: pi.h:1943
@ PI_USM_SYSTEM_SHARED_SUPPORT
Definition: pi.h:1948
@ PI_USM_SINGLE_SHARED_SUPPORT
Definition: pi.h:1946
@ PI_USM_DEVICE_SUPPORT
Definition: pi.h:1945
@ PI_USM_HOST_SUPPORT
Definition: pi.h:1944
@ PI_USM_CROSS_SHARED_SUPPORT
Definition: pi.h:1947
constexpr pi_queue_properties PI_QUEUE_FLAG_ON_DEVICE_DEFAULT
Definition: pi.h:800
pi_result piextCommandBufferMemBufferFill(pi_ext_command_buffer command_buffer, pi_mem buffer, const void *pattern, size_t pattern_size, size_t offset, size_t size, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a mem buffer fill command to the command-buffer.
Definition: pi_cuda.cpp:1174
constexpr pi_queue_properties PI_EXT_ONEAPI_QUEUE_FLAG_DISCARD_EVENTS
Definition: pi.h:801
pi_result piextContextCreateWithNativeHandle(pi_native_handle nativeHandle, pi_uint32 numDevices, const pi_device *devices, bool pluginOwnsNativeHandle, pi_context *context)
Creates PI context object from a native handle.
Definition: pi_cuda.cpp:143
pi_result piextMemImageFree(pi_context context, pi_device device, pi_image_mem_handle memory_handle)
API to free memory for bindless images.
Definition: pi_cuda.cpp:417
constexpr pi_queue_properties PI_QUEUE_FLAG_ON_DEVICE
Definition: pi.h:799
pi_result piEventGetProfilingInfo(pi_event event, pi_profiling_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:590
pi_result piEnqueueMemBufferWrite(pi_queue command_queue, pi_mem buffer, pi_bool blocking_write, size_t offset, size_t size, const void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:700
pi_result piEnqueueMemBufferCopy(pi_queue command_queue, pi_mem src_buffer, pi_mem dst_buffer, size_t src_offset, size_t dst_offset, size_t size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:726
constexpr pi_mem_flags PI_MEM_FLAGS_HOST_PTR_COPY
Definition: pi.h:764
pi_result piEnqueueMemBufferWriteRect(pi_queue command_queue, pi_mem buffer, pi_bool blocking_write, pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset, pi_buff_rect_region region, size_t buffer_row_pitch, size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, const void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:712
_pi_buffer_create_type
Definition: pi.h:682
@ PI_BUFFER_CREATE_TYPE_REGION
Definition: pi.h:682
static constexpr pi_device_affinity_domain PI_DEVICE_AFFINITY_DOMAIN_L4_CACHE
Definition: pi.h:869
pi_result piEnqueueMemUnmap(pi_queue command_queue, pi_mem memobj, void *mapped_ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:773
pi_result piKernelSetExecInfo(pi_kernel kernel, pi_kernel_exec_info value_name, size_t param_value_size, const void *param_value)
API to set attributes controlling kernel execution.
Definition: pi_cuda.cpp:1053
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_SET_ACCESSED_BY
Definition: pi.h:626
static constexpr pi_device_fp_config PI_FP_ROUND_TO_NEAREST
Definition: pi.h:884
pi_result piProgramGetInfo(pi_program program, pi_program_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:272
pi_result piProgramBuild(pi_program program, pi_uint32 num_devices, const pi_device *device_list, const char *options, void(*pfn_notify)(pi_program program, void *user_data), void *user_data)
uint32_t pi_uint32
Definition: pi.h:213
_pi_device_binary_property_set_struct * pi_device_binary_property_set
Definition: pi.h:936
static const uint16_t PI_DEVICE_BINARY_VERSION
Definition: pi.h:951
static constexpr pi_device_partition_property PI_DEVICE_PARTITION_BY_AFFINITY_DOMAIN
Definition: pi.h:861
constexpr pi_memory_order_capabilities PI_MEMORY_ORDER_RELAXED
Definition: pi.h:734
_pi_event_info
Definition: pi.h:545
@ PI_EVENT_INFO_COMMAND_EXECUTION_STATUS
Definition: pi.h:549
@ PI_EVENT_INFO_REFERENCE_COUNT
Definition: pi.h:550
@ PI_EVENT_INFO_COMMAND_TYPE
Definition: pi.h:548
@ PI_EVENT_INFO_CONTEXT
Definition: pi.h:547
@ PI_EVENT_INFO_COMMAND_QUEUE
Definition: pi.h:546
constexpr pi_queue_properties PI_EXT_ONEAPI_QUEUE_FLAG_PRIORITY_HIGH
Definition: pi.h:803
pi_result piextEventCreateWithNativeHandle(pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle, pi_event *event)
Creates PI event object from a native handle.
Definition: pi_cuda.cpp:627
constexpr pi_usm_mem_properties PI_MEM_ALLOC_INITIAL_PLACEMENT_HOST
Definition: pi.h:785
static constexpr pi_device_fp_config PI_FP_ROUND_TO_INF
Definition: pi.h:886
pi_result piextMemUnsampledImageCreate(pi_context context, pi_device device, pi_image_mem_handle img_mem, pi_image_format *image_format, pi_image_desc *image_desc, pi_mem *ret_mem, pi_image_handle *ret_handle)
API to create bindless image handles.
Definition: pi_cuda.cpp:383
pi_result piextGetGlobalVariablePointer(pi_device Device, pi_program Program, const char *GlobalVariableName, size_t *GlobalVariableSize, void **GlobalVariablePointerRet)
Definition: pi_cuda.cpp:851
pi_result piextMemSampledImageHandleDestroy(pi_context context, pi_device device, pi_image_handle handle)
API to destroy bindless sampled image handles.
Definition: pi_cuda.cpp:445
intptr_t pi_context_properties
Definition: pi.h:716
pi_result piEventsWait(pi_uint32 num_events, const pi_event *event_list)
Definition: pi_cuda.cpp:598
pi_bitfield pi_mem_flags
Definition: pi.h:758
constexpr pi_usm_mem_properties PI_MEM_ALLOC_WRTITE_COMBINED
Definition: pi.h:782
constexpr pi_queue_properties PI_QUEUE_FLAGS
Definition: pi.h:794
pi_result piContextCreate(const pi_context_properties *properties, pi_uint32 num_devices, const pi_device *devices, void(*pfn_notify)(const char *errinfo, const void *private_info, size_t cb, void *user_data), void *user_data, pi_context *ret_context)
pi_result piEnqueueMemBufferMap(pi_queue command_queue, pi_mem buffer, pi_bool blocking_map, pi_map_flags map_flags, size_t offset, size_t size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event, void **ret_map)
Definition: pi_cuda.cpp:762
pi_result piextCommandBufferRelease(pi_ext_command_buffer command_buffer)
API to decrement the reference count of the command-buffer.
Definition: pi_cuda.cpp:1079
pi_result piEventSetCallback(pi_event event, pi_int32 command_exec_callback_type, void(*pfn_notify)(pi_event event, pi_int32 event_command_status, void *user_data), void *user_data)
static constexpr pi_device_binary_type PI_DEVICE_BINARY_TYPE_SPIRV
Definition: pi.h:946
pi_result piextUSMEnqueueFill2D(pi_queue queue, void *ptr, size_t pitch, size_t pattern_size, const void *pattern, size_t width, size_t height, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event)
USM 2D fill API.
Definition: pi_cuda.cpp:941
constexpr pi_queue_properties PI_QUEUE_FLAG_PROFILING_ENABLE
Definition: pi.h:798
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_SET_PREFERRED_LOCATION_HOST
Definition: pi.h:630
pi_result piextCommandBufferMemBufferCopyRect(pi_ext_command_buffer command_buffer, pi_mem src_buffer, pi_mem dst_buffer, pi_buff_rect_offset src_origin, pi_buff_rect_offset dst_origin, pi_buff_rect_region region, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a rectangular mem buffer copy command to the command-buffer.
Definition: pi_cuda.cpp:1118
pi_result piQueueFlush(pi_queue command_queue)
Definition: pi_cuda.cpp:188
pi_result piMemRelease(pi_mem mem)
Definition: pi_cuda.cpp:225
constexpr pi_mem_flags PI_MEM_FLAGS_HOST_PTR_USE
Definition: pi.h:763
void(* pi_context_extended_deleter)(void *user_data)
Definition: pi.h:1341
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_SET_ACCESSED_BY_HOST
Definition: pi.h:635
constexpr pi_map_flags PI_MAP_WRITE
Definition: pi.h:770
constexpr pi_map_flags PI_MAP_WRITE_INVALIDATE_REGION
Definition: pi.h:771
_pi_mem_type
Definition: pi.h:588
@ PI_MEM_TYPE_BUFFER
Definition: pi.h:589
@ PI_MEM_TYPE_IMAGE_CUBEMAP
Definition: pi.h:596
@ PI_MEM_TYPE_IMAGE1D_BUFFER
Definition: pi.h:595
@ PI_MEM_TYPE_IMAGE1D
Definition: pi.h:593
@ PI_MEM_TYPE_IMAGE1D_ARRAY
Definition: pi.h:594
@ PI_MEM_TYPE_IMAGE2D
Definition: pi.h:590
@ PI_MEM_TYPE_IMAGE2D_ARRAY
Definition: pi.h:592
@ PI_MEM_TYPE_IMAGE3D
Definition: pi.h:591
pi_result piSamplerCreate(pi_context context, const pi_sampler_properties *sampler_properties, pi_sampler *result_sampler)
Definition: pi_cuda.cpp:635
constexpr pi_usm_mem_properties PI_MEM_ALLOC_DEVICE_READ_ONLY
Definition: pi.h:787
pi_result piEnqueueMemImageCopy(pi_queue command_queue, pi_mem src_image, pi_mem dst_image, pi_image_offset src_origin, pi_image_offset dst_origin, pi_image_region region, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:815
_pi_device_binary_property_struct * pi_device_binary_property
Definition: pi.h:927
pi_result piextCommandBufferMemcpyUSM(pi_ext_command_buffer command_buffer, void *dst_ptr, const void *src_ptr, size_t size, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a USM memcpy command to the command-buffer.
Definition: pi_cuda.cpp:1099
_pi_sampler_info
Definition: pi.h:687
@ PI_SAMPLER_INFO_CONTEXT
Definition: pi.h:689
@ PI_SAMPLER_INFO_MIP_FILTER_MODE
Definition: pi.h:693
@ PI_SAMPLER_INFO_NORMALIZED_COORDS
Definition: pi.h:690
@ PI_SAMPLER_INFO_FILTER_MODE
Definition: pi.h:692
@ PI_SAMPLER_INFO_LOD_MIN
Definition: pi.h:694
@ PI_SAMPLER_INFO_LOD_MAX
Definition: pi.h:695
@ PI_SAMPLER_INFO_REFERENCE_COUNT
Definition: pi.h:688
@ PI_SAMPLER_INFO_ADDRESSING_MODE
Definition: pi.h:691
pi_result piProgramLink(pi_context context, pi_uint32 num_devices, const pi_device *device_list, const char *options, pi_uint32 num_input_programs, const pi_program *input_programs, void(*pfn_notify)(pi_program program, void *user_data), void *user_data, pi_program *ret_program)
pi_result piQueueCreate(pi_context context, pi_device device, pi_queue_properties properties, pi_queue *queue)
Definition: pi_cuda.cpp:161
pi_result piextMemUnsampledImageHandleDestroy(pi_context context, pi_device device, pi_image_handle handle)
API to destroy bindless unsampled image handles.
Definition: pi_cuda.cpp:440
pi_result piEventSetStatus(pi_event event, pi_int32 execution_status)
Definition: pi_cuda.cpp:611
pi_result piextQueueGetNativeHandle(pi_queue queue, pi_native_handle *nativeHandle, int32_t *nativeHandleDesc)
Gets the native handle of a PI queue object.
Definition: pi_cuda.cpp:190
static constexpr pi_device_fp_config PI_FP_CORRECTLY_ROUNDED_DIVIDE_SQRT
Definition: pi.h:889
pi_result piextEnqueueCooperativeKernelLaunch(pi_queue queue, pi_kernel kernel, pi_uint32 work_dim, const size_t *global_work_offset, const size_t *global_work_size, const size_t *local_work_size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:547
constexpr pi_map_flags PI_MAP_READ
Definition: pi.h:769
pi_result piextContextGetNativeHandle(pi_context context, pi_native_handle *nativeHandle)
Gets the native handle of a PI context object.
Definition: pi_cuda.cpp:138
pi_result piextMemMipmapFree(pi_context context, pi_device device, pi_image_mem_handle memory_handle)
API to free mipmap memory for bindless images.
Definition: pi_cuda.cpp:422
static constexpr pi_device_affinity_domain PI_DEVICE_AFFINITY_DOMAIN_L1_CACHE
Definition: pi.h:875
static constexpr pi_device_affinity_domain PI_DEVICE_AFFINITY_DOMAIN_NUMA
Definition: pi.h:867
pi_bitfield pi_memory_scope_capabilities
Definition: pi.h:740
pi_result piPluginGetLastError(char **message)
API to get Plugin specific warning and error messages.
Definition: pi_cuda.cpp:52
pi_result piProgramCreate(pi_context context, const void *il, size_t length, pi_program *res_program)
Definition: pi_cuda.cpp:248
pi_result piEnqueueMemBufferFill(pi_queue command_queue, pi_mem buffer, const void *pattern, size_t pattern_size, size_t offset, size_t size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:750
pi_result piPluginInit(pi_plugin *plugin_info)
Definition: pi_cuda.cpp:1275
pi_result piextMemMipmapGetLevel(pi_context context, pi_device device, pi_image_mem_handle mip_mem, unsigned int level, pi_image_mem_handle *ret_mem)
API to retrieve individual image from mipmap.
Definition: pi_cuda.cpp:409
pi_result piextDeviceCreateWithNativeHandle(pi_native_handle nativeHandle, pi_platform platform, pi_device *device)
Creates PI device object from a native handle.
Definition: pi_cuda.cpp:106
pi_result piDeviceRetain(pi_device device)
Definition: pi_cuda.cpp:70
pi_result piDeviceRelease(pi_device device)
Definition: pi_cuda.cpp:74
_pi_sanitize_type
Definition: pi.h:2908
@ _PI_SANITIZE_TYPE_ADDRESS
Definition: pi.h:2910
@ _PI_SANITIZE_TYPE_THREAD
Definition: pi.h:2912
@ _PI_SANITIZE_TYPE_MEMORY
Definition: pi.h:2911
@ _PI_SANITIZE_TYPE_NONE
Definition: pi.h:2909
pi_result piProgramCompile(pi_program program, pi_uint32 num_devices, const pi_device *device_list, const char *options, pi_uint32 num_input_headers, const pi_program *input_headers, const char **header_include_names, void(*pfn_notify)(pi_program program, void *user_data), void *user_data)
pi_result piMemBufferPartition(pi_mem buffer, pi_mem_flags flags, pi_buffer_create_type buffer_create_type, void *buffer_create_info, pi_mem *ret_mem)
Definition: pi_cuda.cpp:836
pi_bitfield pi_device_fp_config
Definition: pi.h:881
pi_result piextUSMDeviceAlloc(void **result_ptr, pi_context context, pi_device device, pi_usm_mem_properties *properties, size_t size, pi_uint32 alignment)
Allocates device memory.
Definition: pi_cuda.cpp:860
constexpr pi_sampler_properties PI_SAMPLER_PROPERTIES_FILTER_MODE
Definition: pi.h:728
pi_result piextCommandBufferMemBufferWrite(pi_ext_command_buffer command_buffer, pi_mem buffer, size_t offset, size_t size, const void *ptr, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a mem buffer write command to the command-buffer.
Definition: pi_cuda.cpp:1152
constexpr pi_mem_flags PI_MEM_FLAGS_ACCESS_RW
Definition: pi.h:760
_pi_mem_obj_access
Definition: pi.h:1903
@ PI_ACCESS_READ_WRITE
Definition: pi.h:1904
@ PI_ACCESS_READ_ONLY
Definition: pi.h:1906
@ PI_ACCESS_WRITE_ONLY
Definition: pi.h:1905
_pi_kernel_info
Definition: pi.h:508
@ PI_KERNEL_INFO_PROGRAM
Definition: pi.h:513
@ PI_KERNEL_INFO_FUNCTION_NAME
Definition: pi.h:509
@ PI_KERNEL_INFO_REFERENCE_COUNT
Definition: pi.h:511
@ PI_KERNEL_INFO_CONTEXT
Definition: pi.h:512
@ PI_KERNEL_INFO_ATTRIBUTES
Definition: pi.h:514
@ PI_KERNEL_INFO_NUM_ARGS
Definition: pi.h:510
constexpr pi_device_exec_capabilities PI_DEVICE_EXEC_CAPABILITIES_NATIVE_KERNEL
Definition: pi.h:722
constexpr pi_memory_scope_capabilities PI_MEMORY_SCOPE_DEVICE
Definition: pi.h:744
constexpr pi_sampler_properties PI_SAMPLER_PROPERTIES_CUBEMAP_FILTER_MODE
Definition: pi.h:730
_pi_command_type
Definition: pi.h:553
@ PI_COMMAND_TYPE_SVM_UNMAP
Definition: pi.h:582
@ PI_COMMAND_TYPE_MARKER
Definition: pi.h:572
@ PI_COMMAND_TYPE_COPY_IMAGE_TO_BUFFER
Definition: pi.h:570
@ PI_COMMAND_TYPE_COPY_BUFFER_TO_IMAGE
Definition: pi.h:569
@ PI_COMMAND_TYPE_RELEASE_GL_OBJECTS
Definition: pi.h:574
@ PI_COMMAND_TYPE_MIGRATE_MEM_OBJECTS
Definition: pi.h:576
@ PI_COMMAND_TYPE_IMAGE_WRITE
Definition: pi.h:566
@ PI_COMMAND_TYPE_SVM_FREE
Definition: pi.h:578
@ PI_COMMAND_TYPE_MEM_BUFFER_COPY_RECT
Definition: pi.h:562
@ PI_COMMAND_TYPE_SVM_MAP
Definition: pi.h:581
@ PI_COMMAND_TYPE_DEVICE_GLOBAL_VARIABLE_WRITE
Definition: pi.h:585
@ PI_COMMAND_TYPE_MEM_BUFFER_READ
Definition: pi.h:555
@ PI_COMMAND_TYPE_DEVICE_GLOBAL_VARIABLE_READ
Definition: pi.h:584
@ PI_COMMAND_TYPE_MEM_BUFFER_UNMAP
Definition: pi.h:559
@ PI_COMMAND_TYPE_IMAGE_READ
Definition: pi.h:565
@ PI_COMMAND_TYPE_IMAGE_COPY
Definition: pi.h:567
@ PI_COMMAND_TYPE_MEM_BUFFER_READ_RECT
Definition: pi.h:560
@ PI_COMMAND_TYPE_MEM_BUFFER_COPY
Definition: pi.h:557
@ PI_COMMAND_TYPE_MEM_BUFFER_MAP
Definition: pi.h:558
@ PI_COMMAND_TYPE_MEM_BUFFER_WRITE
Definition: pi.h:556
@ PI_COMMAND_TYPE_ACQUIRE_GL_OBJECTS
Definition: pi.h:573
@ PI_COMMAND_TYPE_BARRIER
Definition: pi.h:575
@ PI_COMMAND_TYPE_NDRANGE_KERNEL
Definition: pi.h:554
@ PI_COMMAND_TYPE_NATIVE_KERNEL
Definition: pi.h:568
@ PI_COMMAND_TYPE_MEM_BUFFER_WRITE_RECT
Definition: pi.h:561
@ PI_COMMAND_TYPE_EXT_COMMAND_BUFFER
Definition: pi.h:583
@ PI_COMMAND_TYPE_SVM_MEMFILL
Definition: pi.h:580
@ PI_COMMAND_TYPE_MEM_BUFFER_FILL
Definition: pi.h:564
@ PI_COMMAND_TYPE_USER
Definition: pi.h:563
@ PI_COMMAND_TYPE_FILL_IMAGE
Definition: pi.h:577
@ PI_COMMAND_TYPE_MAP_IMAGE
Definition: pi.h:571
@ PI_COMMAND_TYPE_SVM_MEMCPY
Definition: pi.h:579
pi_result piGetDeviceAndHostTimer(pi_device Device, uint64_t *DeviceTime, uint64_t *HostTime)
Queries device for it's global timestamp in nanoseconds, and updates HostTime with the value of the h...
Definition: pi_cuda.cpp:1247
pi_result piextSignalExternalSemaphore(pi_queue command_queue, pi_interop_semaphore_handle sem_handle, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
API to instruct the queue to signal the external semaphore handle once all previous commands have com...
Definition: pi_cuda.cpp:500
pi_result piextMemReleaseInterop(pi_context context, pi_device device, pi_interop_mem_handle memory_handle)
API to destroy interop memory.
Definition: pi_cuda.cpp:473
pi_result piextPeerAccessGetInfo(pi_device command_device, pi_device peer_device, pi_peer_attr attr, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:1264
pi_result piextBindlessImageSamplerCreate(pi_context context, const pi_sampler_properties *sampler_properties, float min_mipmap_level_clamp, float max_mipmap_level_clamp, float max_anisotropy, pi_sampler *result_sampler)
API to create samplers for bindless images.
Definition: pi_cuda.cpp:400
pi_result piSamplerGetInfo(pi_sampler sampler, pi_sampler_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:641
static const uint8_t PI_DEVICE_BINARY_OFFLOAD_KIND_SYCL
Definition: pi.h:954
_pi_mem_alloc_info
Definition: pi.h:1958
@ PI_MEM_ALLOC_BASE_PTR
Definition: pi.h:1960
@ PI_MEM_ALLOC_TYPE
Definition: pi.h:1959
@ PI_MEM_ALLOC_DEVICE
Definition: pi.h:1962
@ PI_MEM_ALLOC_SIZE
Definition: pi.h:1961
pi_result piextCommandBufferFillUSM(pi_ext_command_buffer command_buffer, void *ptr, const void *pattern, size_t pattern_size, size_t size, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a USM fill command to the command-buffer.
Definition: pi_cuda.cpp:1184
pi_result piextKernelGetNativeHandle(pi_kernel kernel, pi_native_handle *nativeHandle)
Gets the native handle of a PI kernel object.
Definition: pi_cuda.cpp:567
_pi_device_local_mem_type
Definition: pi.h:299
@ PI_DEVICE_LOCAL_MEM_TYPE_GLOBAL
Definition: pi.h:301
@ PI_DEVICE_LOCAL_MEM_TYPE_LOCAL
Definition: pi.h:300
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_UNSET_PREFERRED_LOCATION
Definition: pi.h:624
constexpr pi_memory_scope_capabilities PI_MEMORY_SCOPE_SYSTEM
Definition: pi.h:745
constexpr pi_memory_scope_capabilities PI_MEMORY_SCOPE_SUB_GROUP
Definition: pi.h:742
pi_result piEnqueueMemImageWrite(pi_queue command_queue, pi_mem image, pi_bool blocking_write, pi_image_offset origin, pi_image_region region, size_t input_row_pitch, size_t input_slice_pitch, const void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:801
_pi_context_info
Definition: pi.h:479
@ PI_EXT_CONTEXT_INFO_ATOMIC_MEMORY_ORDER_CAPABILITIES
Definition: pi.h:486
@ PI_CONTEXT_INFO_PROPERTIES
Definition: pi.h:483
@ PI_CONTEXT_INFO_PLATFORM
Definition: pi.h:481
@ PI_CONTEXT_INFO_NUM_DEVICES
Definition: pi.h:482
@ PI_EXT_ONEAPI_CONTEXT_INFO_USM_MEMSET2D_SUPPORT
Definition: pi.h:492
@ PI_EXT_ONEAPI_CONTEXT_INFO_USM_MEMCPY2D_SUPPORT
Definition: pi.h:493
@ PI_EXT_CONTEXT_INFO_ATOMIC_FENCE_ORDER_CAPABILITIES
Definition: pi.h:488
@ PI_CONTEXT_INFO_DEVICES
Definition: pi.h:480
@ PI_CONTEXT_INFO_REFERENCE_COUNT
Definition: pi.h:484
@ PI_EXT_CONTEXT_INFO_ATOMIC_FENCE_SCOPE_CAPABILITIES
Definition: pi.h:489
@ PI_EXT_ONEAPI_CONTEXT_INFO_USM_FILL2D_SUPPORT
Definition: pi.h:491
@ PI_EXT_CONTEXT_INFO_ATOMIC_MEMORY_SCOPE_CAPABILITIES
Definition: pi.h:487
pi_result piextKernelCreateWithNativeHandle(pi_native_handle nativeHandle, pi_context context, pi_program program, bool pluginOwnsNativeHandle, pi_kernel *kernel)
Creates PI kernel object from a native handle.
Definition: pi_cuda.cpp:557
pi_result piMemBufferCreate(pi_context context, pi_mem_flags flags, size_t size, void *host_ptr, pi_mem *ret_mem, const pi_mem_properties *properties=nullptr)
Definition: pi_cuda.cpp:210
pi_result piextImportExternalSemaphoreOpaqueFD(pi_context context, pi_device device, int file_descriptor, pi_interop_semaphore_handle *ret_handle)
API to import an external semaphore in the form of a file descriptor.
Definition: pi_cuda.cpp:479
_pi_sampler_filter_mode
Definition: pi.h:706
@ PI_SAMPLER_FILTER_MODE_NEAREST
Definition: pi.h:707
@ PI_SAMPLER_FILTER_MODE_LINEAR
Definition: pi.h:708
static const uint16_t PI_DEVICE_BINARIES_VERSION
Definition: pi.h:1117
pi_result piEventRelease(pi_event event)
Definition: pi_cuda.cpp:617
pi_result piPluginGetBackendOption(pi_platform platform, const char *frontend_option, const char **backend_option)
API to get backend specific option.
Definition: pi_cuda.cpp:56
static constexpr pi_device_exec_capabilities PI_EXEC_NATIVE_KERNEL
Definition: pi.h:895
pi_result piQueueGetInfo(pi_queue command_queue, pi_queue_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:172
pi_result piextCommandBufferAdviseUSM(pi_ext_command_buffer command_buffer, const void *ptr, size_t length, pi_mem_advice advice, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a USM Advise command to the command-buffer.
Definition: pi_cuda.cpp:1204
pi_result piextUSMEnqueueMemcpy2D(pi_queue queue, pi_bool blocking, void *dst_ptr, size_t dst_pitch, const void *src_ptr, size_t src_pitch, size_t width, size_t height, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event)
USM 2D Memcpy API.
Definition: pi_cuda.cpp:966
pi_result piContextRelease(pi_context context)
Definition: pi_cuda.cpp:157
_pi_event_status
Definition: pi.h:232
@ PI_EVENT_SUBMITTED
Definition: pi.h:235
@ PI_EVENT_QUEUED
Definition: pi.h:236
@ PI_EVENT_COMPLETE
Definition: pi.h:233
@ PI_EVENT_RUNNING
Definition: pi.h:234
pi_result piextCommandBufferRetain(pi_ext_command_buffer command_buffer)
API to increment the reference count of the command-buffer.
Definition: pi_cuda.cpp:1075
constexpr pi_mem_properties PI_MEM_PROPERTIES_ALLOC_BUFFER_LOCATION
Definition: pi.h:776
pi_result piEnqueueMemImageFill(pi_queue command_queue, pi_mem image, const void *fill_color, const size_t *origin, const size_t *region, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:824
pi_result piProgramCreateWithBinary(pi_context context, pi_uint32 num_devices, const pi_device *device_list, const size_t *lengths, const unsigned char **binaries, size_t num_metadata_entries, const pi_device_binary_property *metadata, pi_int32 *binary_status, pi_program *ret_program)
Creates a PI program for a context and loads the given binary into it.
Definition: pi_cuda.cpp:253
pi_result piextProgramGetNativeHandle(pi_program program, pi_native_handle *nativeHandle)
Gets the native handle of a PI program object.
Definition: pi_cuda.cpp:328
pi_result piextUSMSharedAlloc(void **result_ptr, pi_context context, pi_device device, pi_usm_mem_properties *properties, size_t size, pi_uint32 alignment)
Allocates memory accessible on both host and device.
Definition: pi_cuda.cpp:869
pi_result piextUSMEnqueueMemcpy(pi_queue queue, pi_bool blocking, void *dst_ptr, const void *src_ptr, size_t size, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event)
USM Memcpy API.
Definition: pi_cuda.cpp:913
pi_result piextMemImageGetInfo(const pi_image_mem_handle mem_handle, pi_image_info param_name, void *param_value, size_t *param_value_size_ret)
API to query an image memory handle for specific properties.
Definition: pi_cuda.cpp:450
static constexpr pi_device_fp_config PI_FP_ROUND_TO_ZERO
Definition: pi.h:885
void * pi_image_mem_handle
Definition: pi.h:1158
static constexpr pi_device_fp_config PI_FP_FMA
Definition: pi.h:887
pi_result piEnqueueMemBufferReadRect(pi_queue command_queue, pi_mem buffer, pi_bool blocking_read, pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset, pi_buff_rect_region region, size_t buffer_row_pitch, size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
Definition: pi_cuda.cpp:686
_pi_sampler_addressing_mode
Definition: pi.h:698
@ PI_SAMPLER_ADDRESSING_MODE_MIRRORED_REPEAT
Definition: pi.h:699
@ PI_SAMPLER_ADDRESSING_MODE_CLAMP
Definition: pi.h:702
@ PI_SAMPLER_ADDRESSING_MODE_NONE
Definition: pi.h:703
@ PI_SAMPLER_ADDRESSING_MODE_REPEAT
Definition: pi.h:700
@ PI_SAMPLER_ADDRESSING_MODE_CLAMP_TO_EDGE
Definition: pi.h:701
_pi_platform_backend
Definition: pi.h:283
@ PI_EXT_PLATFORM_BACKEND_OPENCL
The backend is OpenCL.
Definition: pi.h:286
@ PI_EXT_PLATFORM_BACKEND_NATIVE_CPU
The backend is NATIVE_CPU.
Definition: pi.h:290
@ PI_EXT_PLATFORM_BACKEND_UNKNOWN
The backend is not a recognized one.
Definition: pi.h:284
@ PI_EXT_PLATFORM_BACKEND_LEVEL_ZERO
The backend is Level Zero.
Definition: pi.h:285
@ PI_EXT_PLATFORM_BACKEND_ESIMD
The backend is ESIMD.
Definition: pi.h:289
@ PI_EXT_PLATFORM_BACKEND_HIP
The backend is HIP.
Definition: pi.h:288
@ PI_EXT_PLATFORM_BACKEND_CUDA
The backend is CUDA.
Definition: pi.h:287
pi_result piextKernelSetArgPointer(pi_kernel kernel, pi_uint32 arg_index, size_t arg_size, const void *arg_value)
Sets up pointer arguments for CL kernels.
Definition: pi_cuda.cpp:900
_pi_program_info
Definition: pi.h:467
@ PI_PROGRAM_INFO_NUM_DEVICES
Definition: pi.h:470
@ PI_PROGRAM_INFO_SOURCE
Definition: pi.h:472
@ PI_PROGRAM_INFO_BINARY_SIZES
Definition: pi.h:473
@ PI_PROGRAM_INFO_KERNEL_NAMES
Definition: pi.h:476
@ PI_PROGRAM_INFO_CONTEXT
Definition: pi.h:469
@ PI_PROGRAM_INFO_REFERENCE_COUNT
Definition: pi.h:468
@ PI_PROGRAM_INFO_BINARIES
Definition: pi.h:474
@ PI_PROGRAM_INFO_NUM_KERNELS
Definition: pi.h:475
@ PI_PROGRAM_INFO_DEVICES
Definition: pi.h:471
pi_result piKernelGetInfo(pi_kernel kernel, pi_kernel_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:366
static constexpr _pi_mem_advice PI_MEM_ADVICE_HIP_SET_PREFERRED_LOCATION
Definition: pi.h:622
intptr_t pi_device_partition_property
Definition: pi.h:853
static constexpr pi_device_exec_capabilities PI_EXEC_KERNEL
Definition: pi.h:894
pi_result piKernelGetGroupInfo(pi_kernel kernel, pi_device device, pi_kernel_group_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:508
pi_result piextUSMGetMemAllocInfo(pi_context context, const void *ptr, pi_mem_alloc_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
API to query information about USM allocated pointers Valid Queries: PI_MEM_ALLOC_TYPE returns host/d...
Definition: pi_cuda.cpp:977
pi_result piextCommandBufferUpdateKernelLaunch(pi_ext_command_buffer_command command, pi_ext_command_buffer_update_kernel_launch_desc *desc)
API to update a kernel launch command inside of a command-buffer.
Definition: pi_cuda.cpp:1222
pi_result piPlatformGetInfo(pi_platform platform, pi_platform_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:35
constexpr pi_sampler_properties PI_SAMPLER_PROPERTIES_MIP_FILTER_MODE
Definition: pi.h:729
pi_result piextMemCreateWithNativeHandle(pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle, pi_mem *mem)
Creates PI mem object from a native handle.
Definition: pi_cuda.cpp:241
pi_result piextGetDeviceFunctionPointer(pi_device device, pi_program program, const char *function_name, pi_uint64 *function_pointer_ret)
Retrieves a device function pointer to a user-defined function.
Definition: pi_cuda.cpp:844
pi_result piextQueueCreateWithNativeHandle(pi_native_handle nativeHandle, int32_t nativeHandleDesc, pi_context context, pi_device device, bool pluginOwnsNativeHandle, pi_queue_properties *Properties, pi_queue *queue)
Creates PI queue object from a native handle.
Definition: pi_cuda.cpp:198
pi_result piContextGetInfo(pi_context context, pi_context_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret)
Definition: pi_cuda.cpp:124
pi_result piextDeviceSelectBinary(pi_device device, pi_device_binary *binaries, pi_uint32 num_binaries, pi_uint32 *selected_binary_ind)
Selects the most appropriate device binary based on runtime information and the IR characteristics.
Definition: pi_cuda.cpp:93
constexpr pi_device_exec_capabilities PI_DEVICE_EXEC_CAPABILITIES_KERNEL
Definition: pi.h:719
constexpr pi_memory_order_capabilities PI_MEMORY_ORDER_SEQ_CST
Definition: pi.h:738
pi_ext_structure_type
Definition: pi.h:2354
@ PI_EXT_STRUCTURE_TYPE_COMMAND_BUFFER_DESC
Definition: pi.h:2355
static constexpr pi_device_fp_config PI_FP_INF_NAN
Definition: pi.h:883
_pi_mem_info
Definition: pi.h:1183
@ PI_MEM_SIZE
Definition: pi.h:1183
@ PI_MEM_CONTEXT
Definition: pi.h:1183
pi_result piQueueRetain(pi_queue command_queue)
Definition: pi_cuda.cpp:180
pi_result piextMemImageCreateWithNativeHandle(pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle, const pi_image_format *ImageFormat, const pi_image_desc *ImageDesc, pi_mem *img)
Creates PI image object from a native handle.
Definition: pi_cuda.cpp:264
pi_result piextUSMPitchedAlloc(void **result_ptr, size_t *result_pitch, pi_context context, pi_device device, pi_usm_mem_properties *properties, size_t width_in_bytes, size_t height, unsigned int element_size_bytes)
Allocates memory accessible on device.
Definition: pi_cuda.cpp:878
pi_result piextEnqueueDeviceGlobalVariableRead(pi_queue queue, pi_program program, const char *name, pi_bool blocking_read, size_t count, size_t offset, void *dst, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event)
API reading data from a device global variable to host.
Definition: pi_cuda.cpp:1002
_pi_program_binary_type
Definition: pi.h:262
@ PI_PROGRAM_BINARY_TYPE_LIBRARY
Definition: pi.h:265
@ PI_PROGRAM_BINARY_TYPE_EXECUTABLE
Definition: pi.h:266
@ PI_PROGRAM_BINARY_TYPE_COMPILED_OBJECT
Definition: pi.h:264
@ PI_PROGRAM_BINARY_TYPE_NONE
Definition: pi.h:263
pi_uint64 pi_interop_semaphore_handle
Definition: pi.h:1160
_pi_peer_attr
Definition: pi.h:1185
@ PI_PEER_ACCESS_SUPPORTED
returns a uint32_t: 1 if P2P Access is supported otherwise P2P Access is not supported.
Definition: pi.h:1186
@ PI_PEER_ATOMICS_SUPPORTED
returns a uint32_t: 1 if Atomic operations are supported over the P2P link, otherwise such operations...
Definition: pi.h:1189
pi_result piextContextSetExtendedDeleter(pi_context context, pi_context_extended_deleter func, void *user_data)
Definition: pi_cuda.cpp:132
pi_result piextCommandBufferMemBufferWriteRect(pi_ext_command_buffer command_buffer, pi_mem buffer, pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset, pi_buff_rect_region region, size_t buffer_row_pitch, size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, const void *ptr, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a rectangular mem buffer write command to the command-buffer.
Definition: pi_cuda.cpp:1161
static constexpr pi_device_binary_type PI_DEVICE_BINARY_TYPE_NONE
Definition: pi.h:941
pi_result piextCommandBufferPrefetchUSM(pi_ext_command_buffer command_buffer, const void *ptr, size_t size, pi_usm_migration_flags flags, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point)
API to append a USM Prefetch command to the command-buffer.
Definition: pi_cuda.cpp:1195
pi_bitfield pi_device_affinity_domain
Definition: pi.h:866
pi_result piSamplerRelease(pi_sampler sampler)
Definition: pi_cuda.cpp:653
pi_result piextCommandBufferReleaseCommand(pi_ext_command_buffer_command command)
API to decrement the reference count of a command-buffer command.
Definition: pi_cuda.cpp:1234
pi_uint64 pi_interop_mem_handle
Definition: pi.h:1159
pi_result piDevicePartition(pi_device device, const pi_device_partition_property *properties, pi_uint32 num_devices, pi_device *out_devices, pi_uint32 *out_num_devices)
Definition: pi_cuda.cpp:85
pi_bitfield pi_memory_order_capabilities
Definition: pi.h:733
_pi_mem_obj_property_type
Definition: pi.h:1911
@ PI_ENUM_FORCE_UINT32
Definition: pi.h:1913
@ PI_KERNEL_ARG_MEM_OBJ_ACCESS
Definition: pi.h:1912
pi_result piextKernelSetArgMemObj(pi_kernel kernel, pi_uint32 arg_index, const pi_mem_obj_property *arg_properties, const pi_mem *arg_value)
Definition: pi_cuda.cpp:353
pi_result piEventRetain(pi_event event)
Definition: pi_cuda.cpp:615
constexpr pi_mem_properties PI_MEM_PROPERTIES_CHANNEL
Definition: pi.h:775
_pi_program_build_info
Definition: pi.h:248
@ PI_PROGRAM_BUILD_INFO_LOG
Definition: pi.h:251
@ PI_PROGRAM_BUILD_INFO_OPTIONS
Definition: pi.h:250
@ PI_PROGRAM_BUILD_INFO_BINARY_TYPE
Definition: pi.h:252
@ PI_PROGRAM_BUILD_INFO_STATUS
Definition: pi.h:249
pi_result piextCommandBufferCreate(pi_context context, pi_device device, const pi_ext_command_buffer_desc *desc, pi_ext_command_buffer *ret_command_buffer)
API to create a command-buffer.
Definition: pi_cuda.cpp:1068
pi_device_binary_property PropertiesBegin
Definition: pi.h:932
pi_device_binary_property PropertiesEnd
Definition: pi.h:933
size_t image_slice_pitch
Definition: pi.h:1174
pi_uint32 num_mip_levels
Definition: pi.h:1175
size_t image_height
Definition: pi.h:1170
size_t image_row_pitch
Definition: pi.h:1173
pi_uint32 num_samples
Definition: pi.h:1176
size_t image_depth
Definition: pi.h:1171
pi_mem buffer
Definition: pi.h:1177
size_t image_width
Definition: pi.h:1169
pi_mem_type image_type
Definition: pi.h:1168
size_t image_array_size
Definition: pi.h:1172
pi_image_channel_type image_channel_data_type
Definition: pi.h:1164
pi_image_channel_order image_channel_order
Definition: pi.h:1163
pi_mem_obj_property_type type
Definition: pi.h:1918
pi_mem_access_flag mem_access
Definition: pi.h:1920
void * pNext
Definition: pi.h:1919
Definition: pi.h:898
char * name
Definition: pi.h:900
size_t size
Definition: pi.h:901
int32_t reserved
Definition: pi.h:903
int32_t flags
Definition: pi.h:902
void * addr
Definition: pi.h:899
char PluginVersion[20]
Definition: pi.h:2925
_pi_sanitize_type SanitizeType
Definition: pi.h:2932
struct _pi_plugin::FunctionPointers PiFunctionTable
char PiVersion[20]
Definition: pi.h:2923
char * Targets
Definition: pi.h:2926
This struct is a record of all the device code that may be offloaded.
Definition: pi.h:1122
uint16_t Version
version of this structure - for backward compatibility; all modifications which change order/type/off...
Definition: pi.h:1126
uint16_t NumDeviceBinaries
Number of device binaries in this descriptor.
Definition: pi.h:1128
pi_device_binary DeviceBinaries
Device binaries data.
Definition: pi.h:1130
_pi_offload_entry * HostEntriesEnd
Definition: pi.h:1133
_pi_offload_entry * HostEntriesBegin
the offload entry table (not used, for compatibility with OpenMP)
Definition: pi.h:1132
This struct is a record of the device binary information.
Definition: pi.h:1025
_pi_offload_entry EntriesEnd
Definition: pi.h:1062
const char * LinkOptions
a null-terminated string; target- and compiler-specific options which are suggested to use to "link" ...
Definition: pi.h:1051
_pi_offload_entry EntriesBegin
the offload entry table
Definition: pi.h:1061
const char * CompileOptions
a null-terminated string; target- and compiler-specific options which are suggested to use to "compil...
Definition: pi.h:1048
const unsigned char * BinaryEnd
Pointer to the target code end.
Definition: pi.h:1059
const char * ManifestStart
Pointer to the manifest data start.
Definition: pi.h:1053
uint16_t Version
version of this structure - for backward compatibility; all modifications which change order/type/off...
Definition: pi.h:1029
const char * DeviceTargetSpec
null-terminated string representation of the device's target architecture which holds one of: __SYCL_...
Definition: pi.h:1045
pi_device_binary_property_set PropertySetsEnd
Definition: pi.h:1066
pi_device_binary_property_set PropertySetsBegin
Definition: pi.h:1065
uint8_t Format
format of the binary data - SPIR-V, LLVM IR bitcode,...
Definition: pi.h:1033
const char * ManifestEnd
Pointer to the manifest data end.
Definition: pi.h:1055
uint8_t Kind
the type of offload model the binary employs; must be 4 for SYCL
Definition: pi.h:1031
const unsigned char * BinaryStart
Pointer to the target code start.
Definition: pi.h:1057
pi_ext_structure_type stype
Definition: pi.h:2359
const void * pNext
Definition: pi.h:2360
pi_ext_command_buffer_update_pointer_arg_desc_t * ptr_arg_list
Definition: pi.h:2389
pi_ext_command_buffer_update_memobj_arg_desc_t * mem_obj_arg_list
Definition: pi.h:2388
pi_ext_command_buffer_update_value_arg_desc_t * value_arg_list
Definition: pi.h:2390
const pi_mem_obj_property * properties
Definition: pi.h:2367