[llvm] [Offload] Implement the remaining initial Offload API (PR #122106)

[Callum Fare via llvm-commits]

================
@@ -245,3 +308,298 @@ ol_impl_result_t olGetDeviceInfoSize_impl(ol_device_handle_t Device,
                                           size_t *PropSizeRet) {
   return olGetDeviceInfoImplDetail(Device, PropName, 0, nullptr, PropSizeRet);
 }
+
+TargetAllocTy convertOlToPluginAllocTy(ol_alloc_type_t Type) {
+  switch (Type) {
+  case OL_ALLOC_TYPE_DEVICE:
+    return TARGET_ALLOC_DEVICE;
+  case OL_ALLOC_TYPE_HOST:
+    return TARGET_ALLOC_HOST;
+  case OL_ALLOC_TYPE_SHARED:
+  default:
+    return TARGET_ALLOC_SHARED;
+  }
+}
+
+ol_impl_result_t olMemAlloc_impl(ol_device_handle_t Device,
+                                 ol_alloc_type_t Type, size_t Size,
+                                 void **AllocationOut) {
+  auto Alloc =
+      Device->Device.dataAlloc(Size, nullptr, convertOlToPluginAllocTy(Type));
+  if (!Alloc) {
+    return {OL_ERRC_OUT_OF_RESOURCES,
+            formatv("Could not create allocation on device {0}", Device).str()};
+  }
+
+  *AllocationOut = *Alloc;
+  return OL_SUCCESS;
+}
+
+ol_impl_result_t olMemFree_impl(ol_device_handle_t Device, ol_alloc_type_t Type,
+                                void *Address) {
+  auto Res = Device->Device.dataDelete(Address, convertOlToPluginAllocTy(Type));
+  if (Res) {
+    return {OL_ERRC_OUT_OF_RESOURCES, "Could not free allocation"};
+  }
+  return OL_SUCCESS;
+}
+
+ol_impl_result_t olCreateQueue_impl(ol_device_handle_t Device,
+                                    ol_queue_handle_t *Queue) {
+  auto CreatedQueue = std::make_unique<ol_queue_handle_t_>();
+  auto Err = Device->Device.initAsyncInfo(&(CreatedQueue->AsyncInfo));
+  if (Err) {
+    return OL_ERRC_OUT_OF_RESOURCES;
+  }
+  // TODO: Check error
+  CreatedQueue->Device = Device;
+  CreatedQueue->RefCount = 1;
+  *Queue = CreatedQueue.release();
+  return OL_SUCCESS;
+}
+
+ol_impl_result_t olRetainQueue_impl(ol_queue_handle_t Queue) {
+  Queue->RefCount++;
+  return OL_SUCCESS;
+}
+
+ol_impl_result_t olReleaseQueue_impl(ol_queue_handle_t Queue) {
+  Queue->RefCount--;
+  return OL_SUCCESS;
+}
+
+ol_impl_result_t olFinishQueue_impl(ol_queue_handle_t Queue) {
+  // Host plugin doesn't have a queue set so it's not safe to call synchronize
+  // on it, but we have nothing to synchronize in that situation anyway.
+  if (Queue->AsyncInfo->Queue) {
+    auto Err = Queue->Device->Device.synchronize(Queue->AsyncInfo);
+    if (Err) {
+      return OL_ERRC_OUT_OF_RESOURCES;
----------------
callumfare wrote:

I've added `OL_ERRC_UNKNOWN` to cover the cases when the underlying cause isn't actually known. I've added info strings to those errors to give some context though. But yes generally the plugins should be returning more meaningful errors which could be converted directly to Offload errors.

https://github.com/llvm/llvm-project/pull/122106