[Offload] Replace "EventOut" parameters with olCreateEvent

offload/liboffload/API/Event.td

@@ -10,6 +10,19 @@
 //
 //===----------------------------------------------------------------------===//
 
+def : Function {
+    let name = "olCreateEvent";
+    let desc = "Enqueue an event to `Queue` and return it.";
+    let details = [
+      "This event can be used with `olSyncEvent` and `olWaitEvents` and will be complete once all enqueued work prior to the `olCreateEvent` call is complete.",
+    ];
+    let params = [
+        Param<"ol_queue_handle_t", "Queue", "queue to create the event for", PARAM_IN>,
+        Param<"ol_event_handle_t*", "Event", "output pointer for the created event", PARAM_OUT>
+    ];
+    let returns = [];
+}
+
 def : Function {
     let name = "olDestroyEvent";
     let desc = "Destroy the event and free all underlying resources.";

offload/liboffload/API/Kernel.td

@@ -35,10 +35,8 @@ def : Function {
         Param<"const void*", "ArgumentsData", "pointer to the kernel argument struct", PARAM_IN_OPTIONAL>,
         Param<"size_t", "ArgumentsSize", "size of the kernel argument struct", PARAM_IN>,
         Param<"const ol_kernel_launch_size_args_t*", "LaunchSizeArgs", "pointer to the struct containing launch size parameters", PARAM_IN>,
-        Param<"ol_event_handle_t*", "EventOut", "optional recorded event for the enqueued operation", PARAM_OUT_OPTIONAL>
     ];
     let returns = [
-        Return<"OL_ERRC_INVALID_ARGUMENT", ["`Queue == NULL && EventOut != NULL`"]>,
         Return<"OL_ERRC_INVALID_ARGUMENT", ["`ArgumentsSize > 0 && ArgumentsData == NULL`"]>,
         Return<"OL_ERRC_INVALID_DEVICE", ["If Queue is non-null but does not belong to Device"]>,
         Return<"OL_ERRC_SYMBOL_KIND", ["The provided symbol is not a kernel"]>,

offload/liboffload/API/Memory.td

@@ -60,9 +60,6 @@ def : Function {
         Param<"const void*", "SrcPtr", "pointer to copy from", PARAM_IN>,
         Param<"ol_device_handle_t", "SrcDevice", "device that SrcPtr belongs to", PARAM_IN>,
         Param<"size_t", "Size", "size in bytes of data to copy", PARAM_IN>,
-        Param<"ol_event_handle_t*", "EventOut", "optional recorded event for the enqueued operation", PARAM_OUT_OPTIONAL>
-    ];
-    let returns = [
-        Return<"OL_ERRC_INVALID_ARGUMENT", ["`Queue == NULL && EventOut != NULL`"]>
     ];
+    let returns = [];
 }

offload/liboffload/src/OffloadImpl.cpp

@@ -569,26 +569,21 @@ Error olGetEventInfoSize_impl(ol_event_handle_t Event, ol_event_info_t PropName,
   return olGetEventInfoImplDetail(Event, PropName, 0, nullptr, PropSizeRet);
 }
 
-ol_event_handle_t makeEvent(ol_queue_handle_t Queue) {
-  auto EventImpl = std::make_unique<ol_event_impl_t>(nullptr, Queue);
-  if (auto Res = Queue->Device->Device->createEvent(&EventImpl->EventInfo)) {
-    llvm::consumeError(std::move(Res));
-    return nullptr;
-  }
+Error olCreateEvent_impl(ol_queue_handle_t Queue, ol_event_handle_t *EventOut) {
+  *EventOut = new ol_event_impl_t(nullptr, Queue);
+  if (auto Res = Queue->Device->Device->createEvent(&(*EventOut)->EventInfo))
+    return Res;
 
-  if (auto Res = Queue->Device->Device->recordEvent(EventImpl->EventInfo,
-                                                    Queue->AsyncInfo)) {
-    llvm::consumeError(std::move(Res));
-    return nullptr;
-  }
+  if (auto Res = Queue->Device->Device->recordEvent((*EventOut)->EventInfo,
+                                                    Queue->AsyncInfo))
+    return Res;
 
-  return EventImpl.release();
+  return Plugin::success();
 }
 
 Error olMemcpy_impl(ol_queue_handle_t Queue, void *DstPtr,
                     ol_device_handle_t DstDevice, const void *SrcPtr,
-                    ol_device_handle_t SrcDevice, size_t Size,
-                    ol_event_handle_t *EventOut) {
+                    ol_device_handle_t SrcDevice, size_t Size) {
   auto Host = OffloadContext::get().HostDevice();
   if (DstDevice == Host && SrcDevice == Host) {
     if (!Queue) {
@@ -619,9 +614,6 @@ Error olMemcpy_impl(ol_queue_handle_t Queue, void *DstPtr,
       return Res;
   }
 
-  if (EventOut)
-    *EventOut = makeEvent(Queue);
-
   return Error::success();
 }
 
@@ -668,8 +660,7 @@ Error olDestroyProgram_impl(ol_program_handle_t Program) {
 Error olLaunchKernel_impl(ol_queue_handle_t Queue, ol_device_handle_t Device,
                           ol_symbol_handle_t Kernel, const void *ArgumentsData,
                           size_t ArgumentsSize,
-                          const ol_kernel_launch_size_args_t *LaunchSizeArgs,
-                          ol_event_handle_t *EventOut) {
+                          const ol_kernel_launch_size_args_t *LaunchSizeArgs) {
   auto *DeviceImpl = Device->Device;
   if (Queue && Device != Queue->Device) {
     return createOffloadError(
@@ -707,9 +698,6 @@ Error olLaunchKernel_impl(ol_queue_handle_t Queue, ol_device_handle_t Device,
   if (Err)
     return Err;
 
-  if (EventOut)
-    *EventOut = makeEvent(Queue);
-
   return Error::success();
 }
 

offload/unittests/OffloadAPI/CMakeLists.txt

@@ -9,6 +9,7 @@ add_offload_unittest("device"
     device/olGetDeviceInfoSize.cpp)
 
 add_offload_unittest("event"
+    event/olCreateEvent.cpp
     event/olDestroyEvent.cpp
     event/olSyncEvent.cpp
     event/olGetEventInfo.cpp

offload/unittests/OffloadAPI/common/Fixtures.hpp

@@ -171,16 +171,8 @@ struct OffloadQueueTest : OffloadDeviceTest {
 struct OffloadEventTest : OffloadQueueTest {
   void SetUp() override {
     RETURN_ON_FATAL_FAILURE(OffloadQueueTest::SetUp());
-    // Get an event from a memcpy. We can still use it in olGetEventInfo etc
-    // after it has been waited on.
-    void *Alloc;
-    uint32_t Value = 42;
-    ASSERT_SUCCESS(
-        olMemAlloc(Device, OL_ALLOC_TYPE_DEVICE, sizeof(Value), &Alloc));
-    ASSERT_SUCCESS(
-        olMemcpy(Queue, Alloc, Device, &Value, Host, sizeof(Value), &Event));
-    ASSERT_SUCCESS(olSyncEvent(Event));
-    ASSERT_SUCCESS(olMemFree(Alloc));
+    ASSERT_SUCCESS(olCreateEvent(Queue, &Event));
+    ASSERT_SUCCESS(olSyncQueue(Queue));
   }
 
   void TearDown() override {

offload/unittests/OffloadAPI/event/olCreateEvent.cpp

@@ -0,0 +1,29 @@
+//===------- Offload API tests - olCreateEvent ----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "../common/Fixtures.hpp"
+#include <OffloadAPI.h>
+#include <gtest/gtest.h>
+
+using olCreateEventTest = OffloadQueueTest;
+OFFLOAD_TESTS_INSTANTIATE_DEVICE_FIXTURE(olCreateEventTest);
+
+TEST_P(olCreateEventTest, Success) {
+  ol_event_handle_t Event = nullptr;
+  ASSERT_SUCCESS(olCreateEvent(Queue, &Event));
+  ASSERT_NE(Event, nullptr);
+}
+
+TEST_P(olCreateEventTest, InvalidNullQueue) {
+  ol_event_handle_t Event;
+  ASSERT_ERROR(OL_ERRC_INVALID_NULL_HANDLE, olCreateEvent(nullptr, &Event));
+}
+
+TEST_P(olCreateEventTest, InvalidNullDest) {
+  ASSERT_ERROR(OL_ERRC_INVALID_NULL_POINTER, olCreateEvent(Queue, nullptr));
+}

offload/unittests/OffloadAPI/event/olDestroyEvent.cpp

@@ -14,14 +14,8 @@ using olDestroyEventTest = OffloadQueueTest;
 OFFLOAD_TESTS_INSTANTIATE_DEVICE_FIXTURE(olDestroyEventTest);
 
 TEST_P(olDestroyEventTest, Success) {
-  uint32_t Src = 42;
-  void *DstPtr;
-
   ol_event_handle_t Event = nullptr;
-  ASSERT_SUCCESS(
-      olMemAlloc(Device, OL_ALLOC_TYPE_DEVICE, sizeof(uint32_t), &DstPtr));
-  ASSERT_SUCCESS(
-      olMemcpy(Queue, DstPtr, Device, &Src, Host, sizeof(Src), &Event));
+  ASSERT_SUCCESS(olCreateEvent(Queue, &Event));
   ASSERT_NE(Event, nullptr);
   ASSERT_SUCCESS(olSyncQueue(Queue));
   ASSERT_SUCCESS(olDestroyEvent(Event));

offload/unittests/OffloadAPI/event/olSyncEvent.cpp

@@ -14,14 +14,8 @@ using olSyncEventTest = OffloadQueueTest;
 OFFLOAD_TESTS_INSTANTIATE_DEVICE_FIXTURE(olSyncEventTest);
 
 TEST_P(olSyncEventTest, Success) {
-  uint32_t Src = 42;
-  void *DstPtr;
-
   ol_event_handle_t Event = nullptr;
-  ASSERT_SUCCESS(
-      olMemAlloc(Device, OL_ALLOC_TYPE_DEVICE, sizeof(uint32_t), &DstPtr));
-  ASSERT_SUCCESS(
-      olMemcpy(Queue, DstPtr, Device, &Src, Host, sizeof(Src), &Event));
+  ASSERT_SUCCESS(olCreateEvent(Queue, &Event));
   ASSERT_NE(Event, nullptr);
   ASSERT_SUCCESS(olSyncEvent(Event));
   ASSERT_SUCCESS(olDestroyEvent(Event));
@@ -31,15 +25,9 @@ TEST_P(olSyncEventTest, InvalidNullEvent) {
   ASSERT_ERROR(OL_ERRC_INVALID_NULL_HANDLE, olSyncEvent(nullptr));
 }
 
-TEST_P(olSyncEventTest, SuccessMultipleWait) {
-  uint32_t Src = 42;
-  void *DstPtr;
-
+TEST_P(olSyncEventTest, SuccessMultipleSync) {
   ol_event_handle_t Event = nullptr;
-  ASSERT_SUCCESS(
-      olMemAlloc(Device, OL_ALLOC_TYPE_DEVICE, sizeof(uint32_t), &DstPtr));
-  ASSERT_SUCCESS(
-      olMemcpy(Queue, DstPtr, Device, &Src, Host, sizeof(Src), &Event));
+  ASSERT_SUCCESS(olCreateEvent(Queue, &Event));
   ASSERT_NE(Event, nullptr);
 
   for (size_t I = 0; I < 10; I++)

offload/unittests/OffloadAPI/kernel/olLaunchKernel.cpp

@@ -91,8 +91,8 @@ TEST_P(olLaunchKernelFooTest, Success) {
     void *Mem;
   } Args{Mem};
 
-  ASSERT_SUCCESS(olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args),
-                                &LaunchArgs, nullptr));
+  ASSERT_SUCCESS(
+      olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args), &LaunchArgs));
 
   ASSERT_SUCCESS(olSyncQueue(Queue));
 
@@ -106,7 +106,7 @@ TEST_P(olLaunchKernelFooTest, Success) {
 
 TEST_P(olLaunchKernelNoArgsTest, Success) {
   ASSERT_SUCCESS(
-      olLaunchKernel(Queue, Device, Kernel, nullptr, 0, &LaunchArgs, nullptr));
+      olLaunchKernel(Queue, Device, Kernel, nullptr, 0, &LaunchArgs));
 
   ASSERT_SUCCESS(olSyncQueue(Queue));
 }
@@ -121,7 +121,7 @@ TEST_P(olLaunchKernelFooTest, SuccessSynchronous) {
   } Args{Mem};
 
   ASSERT_SUCCESS(olLaunchKernel(nullptr, Device, Kernel, &Args, sizeof(Args),
-                                &LaunchArgs, nullptr));
+                                &LaunchArgs));
 
   uint32_t *Data = (uint32_t *)Mem;
   for (uint32_t i = 0; i < 64; i++) {
@@ -144,8 +144,8 @@ TEST_P(olLaunchKernelLocalMemTest, Success) {
     void *Mem;
   } Args{Mem};
 
-  ASSERT_SUCCESS(olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args),
-                                &LaunchArgs, nullptr));
+  ASSERT_SUCCESS(
+      olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args), &LaunchArgs));
 
   ASSERT_SUCCESS(olSyncQueue(Queue));
 
@@ -167,8 +167,8 @@ TEST_P(olLaunchKernelLocalMemReductionTest, Success) {
     void *Mem;
   } Args{Mem};
 
-  ASSERT_SUCCESS(olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args),
-                                &LaunchArgs, nullptr));
+  ASSERT_SUCCESS(
+      olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args), &LaunchArgs));
 
   ASSERT_SUCCESS(olSyncQueue(Queue));
 
@@ -190,8 +190,8 @@ TEST_P(olLaunchKernelLocalMemStaticTest, Success) {
     void *Mem;
   } Args{Mem};
 
-  ASSERT_SUCCESS(olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args),
-                                &LaunchArgs, nullptr));
+  ASSERT_SUCCESS(
+      olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args), &LaunchArgs));
 
   ASSERT_SUCCESS(olSyncQueue(Queue));
 
@@ -210,11 +210,11 @@ TEST_P(olLaunchKernelGlobalTest, Success) {
     void *Mem;
   } Args{Mem};
 
-  ASSERT_SUCCESS(olLaunchKernel(Queue, Device, Kernels[0], nullptr, 0,
-                                &LaunchArgs, nullptr));
+  ASSERT_SUCCESS(
+      olLaunchKernel(Queue, Device, Kernels[0], nullptr, 0, &LaunchArgs));
   ASSERT_SUCCESS(olSyncQueue(Queue));
   ASSERT_SUCCESS(olLaunchKernel(Queue, Device, Kernels[1], &Args, sizeof(Args),
-                                &LaunchArgs, nullptr));
+                                &LaunchArgs));
   ASSERT_SUCCESS(olSyncQueue(Queue));
 
   uint32_t *Data = (uint32_t *)Mem;
@@ -229,9 +229,8 @@ TEST_P(olLaunchKernelGlobalTest, InvalidNotAKernel) {
   ol_symbol_handle_t Global = nullptr;
   ASSERT_SUCCESS(
       olGetSymbol(Program, "global", OL_SYMBOL_KIND_GLOBAL_VARIABLE, &Global));
-  ASSERT_ERROR(
-      OL_ERRC_SYMBOL_KIND,
-      olLaunchKernel(Queue, Device, Global, nullptr, 0, &LaunchArgs, nullptr));
+  ASSERT_ERROR(OL_ERRC_SYMBOL_KIND,
+               olLaunchKernel(Queue, Device, Global, nullptr, 0, &LaunchArgs));
 }
 
 TEST_P(olLaunchKernelGlobalCtorTest, Success) {
@@ -242,8 +241,8 @@ TEST_P(olLaunchKernelGlobalCtorTest, Success) {
     void *Mem;
   } Args{Mem};
 
-  ASSERT_SUCCESS(olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args),
-                                &LaunchArgs, nullptr));
+  ASSERT_SUCCESS(
+      olLaunchKernel(Queue, Device, Kernel, &Args, sizeof(Args), &LaunchArgs));
   ASSERT_SUCCESS(olSyncQueue(Queue));
 
   uint32_t *Data = (uint32_t *)Mem;
@@ -259,6 +258,6 @@ TEST_P(olLaunchKernelGlobalDtorTest, Success) {
   // find/implement a way, update this test. For now we just check that nothing
   // crashes
   ASSERT_SUCCESS(
-      olLaunchKernel(Queue, Device, Kernel, nullptr, 0, &LaunchArgs, nullptr));
+      olLaunchKernel(Queue, Device, Kernel, nullptr, 0, &LaunchArgs));
   ASSERT_SUCCESS(olSyncQueue(Queue));
 }