[UR][L0] Add support for passing device list to urProgramBuild/Link/Compile

piProgramBuild receives a list of devices, while urProgramBuild does not. This produces a series of issues when a UR program needs to be created for a specific device.

So define a new API, called urProgramBuildExp to pass this list.

Requires related patch in Unified Runtime Adapters here: oneapi-src/unified-runtime#934

Signed-off-by: Spruit, Neil R <[email protected]>

Author: nrspruit

sycl/plugins/unified_runtime/CMakeLists.txt

@@ -55,14 +55,14 @@ endif()
 if(SYCL_PI_UR_USE_FETCH_CONTENT)
   include(FetchContent)
 
-  set(UNIFIED_RUNTIME_REPO "https://github.com/oneapi-src/unified-runtime.git")
+  set(UNIFIED_RUNTIME_REPO "https://github.com/nrspruit/unified-runtime.git")
   # commit 659d3f469faa99a886fa680a3d6d20449b109578
   # Merge: 192e9404 f94550b4
   # Author: Kenneth Benzie (Benie) <[email protected]>
   # Date:   Tue Nov 14 16:45:24 2023 +0000
   #     Merge pull request #1059 from martygrant/martin/moveNativeCPUAdapterToUR
   #     [NATIVECPU] Move Native CPU adapter to UR.
-  set(UNIFIED_RUNTIME_TAG 659d3f469faa99a886fa680a3d6d20449b109578)
+  set(UNIFIED_RUNTIME_TAG 0790bf8fd34f5d41b5ef4ffc120a6a2acc84b454)
 
   if(SYCL_PI_UR_OVERRIDE_FETCH_CONTENT_REPO)
     set(UNIFIED_RUNTIME_REPO "${SYCL_PI_UR_OVERRIDE_FETCH_CONTENT_REPO}")

sycl/plugins/unified_runtime/pi2ur.hpp

@@ -1985,10 +1985,17 @@ piProgramLink(pi_context Context, pi_uint32 NumDevices,
   ur_program_handle_t *UrProgram =
       reinterpret_cast<ur_program_handle_t *>(RetProgram);
 
-  HANDLE_ERRORS(urProgramLink(UrContext, NumInputPrograms, UrInputPrograms,
-                              Options, UrProgram));
+  auto UrDevices = reinterpret_cast<ur_device_handle_t *>(
+      const_cast<pi_device *>(DeviceList));
 
-  return PI_SUCCESS;
+  auto urResult =
+      urProgramLinkExp(UrContext, NumDevices, UrDevices, NumInputPrograms,
+                       UrInputPrograms, Options, UrProgram);
+  if (urResult == UR_RESULT_ERROR_UNSUPPORTED_FEATURE) {
+    HANDLE_ERRORS(urProgramLink(UrContext, NumInputPrograms, UrInputPrograms,
+                                Options, UrProgram));
+  }
+  return ur2piResult(urResult);
 }
 
 inline pi_result piProgramCompile(
@@ -2017,9 +2024,15 @@ inline pi_result piProgramCompile(
   HANDLE_ERRORS(urProgramGetInfo(UrProgram, PropName, sizeof(&UrContext),
                                  &UrContext, nullptr));
 
-  HANDLE_ERRORS(urProgramCompile(UrContext, UrProgram, Options));
+  auto UrDevices = reinterpret_cast<ur_device_handle_t *>(
+      const_cast<pi_device *>(DeviceList));
 
-  return PI_SUCCESS;
+  auto urResult =
+      urProgramCompileExp(UrProgram, NumDevices, UrDevices, Options);
+  if (urResult == UR_RESULT_ERROR_UNSUPPORTED_FEATURE) {
+    HANDLE_ERRORS(urProgramCompile(UrContext, UrProgram, Options));
+  }
+  return ur2piResult(urResult);
 }
 
 inline pi_result
@@ -2050,9 +2063,14 @@ piProgramBuild(pi_program Program, pi_uint32 NumDevices,
   HANDLE_ERRORS(urProgramGetInfo(UrProgram, PropName, sizeof(&UrContext),
                                  &UrContext, nullptr));
 
-  HANDLE_ERRORS(urProgramBuild(UrContext, UrProgram, Options));
+  auto UrDevices = reinterpret_cast<ur_device_handle_t *>(
+      const_cast<pi_device *>(DeviceList));
 
-  return PI_SUCCESS;
+  auto urResult = urProgramBuildExp(UrProgram, NumDevices, UrDevices, Options);
+  if (urResult == UR_RESULT_ERROR_UNSUPPORTED_FEATURE) {
+    HANDLE_ERRORS(urProgramBuild(UrContext, UrProgram, Options));
+  }
+  return ur2piResult(urResult);
 }
 
 inline pi_result piextProgramSetSpecializationConstant(pi_program Program,