[Offload] Introduce dataFence plugin interface. #153793
Merged
+41
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The purpose of this fence is to ensure that any `dataSubmit`s inserted into a queue before a `dataFence` finish before finish before any `dataSubmit`s inserted after it begin. This is a no-op for most queues, since they are in-order, and by design any operations inserted into them occur in order. But the interface is supposed to be functional for out-of-order queues. The addition of the interface means that any operations that rely on such ordering (like ATTACH map-type support in llvm#149036) can invoke it, without worrying about whether the underlying queue is in-order or out-of-order. Once a plugin supports out-of-order queues, the plugins can implement this function, without requiring any change at the libomptarget level.
|
@llvm/pr-subscribers-offload @llvm/pr-subscribers-backend-amdgpu Author: Abhinav Gaba (abhinavgaba) ChangesThe purpose of this fence is to ensure that any This is a no-op for most queues, since they are in-order, and by design But the interface is supposed to be functional for out-of-order queues. The addition of the interface means that any operations that rely on Once a plugin supports out-of-order queues, the plugins can implement Full diff: https://github.com/llvm/llvm-project/pull/153793.diff 5 Files Affected:
diff --git a/offload/plugins-nextgen/amdgpu/src/rtl.cpp b/offload/plugins-nextgen/amdgpu/src/rtl.cpp
index 796182075ff3d..eab0fda9ff0d0 100644
--- a/offload/plugins-nextgen/amdgpu/src/rtl.cpp
+++ b/offload/plugins-nextgen/amdgpu/src/rtl.cpp
@@ -2537,6 +2537,13 @@ struct AMDGPUDeviceTy : public GenericDeviceTy, AMDGenericDeviceTy {
getAgent(), (uint64_t)Size);
}
+ /// Insert a data fence between previous data operations and the following
+ /// operations. This is a no-op for AMDGPU devices as operations inserted into
+ /// a queue are in-order.
+ Error dataFence(__tgt_async_info *Async) override {
+ return Plugin::success();
+ }
+
/// Initialize the async info for interoperability purposes.
Error initAsyncInfoImpl(AsyncInfoWrapperTy &AsyncInfoWrapper) override {
// TODO: Implement this function.
diff --git a/offload/plugins-nextgen/common/include/PluginInterface.h b/offload/plugins-nextgen/common/include/PluginInterface.h
index c9ab34b024b77..1b110f6b74553 100644
--- a/offload/plugins-nextgen/common/include/PluginInterface.h
+++ b/offload/plugins-nextgen/common/include/PluginInterface.h
@@ -944,6 +944,10 @@ struct GenericDeviceTy : public DeviceAllocatorTy {
virtual Error dataRetrieveImpl(void *HstPtr, const void *TgtPtr, int64_t Size,
AsyncInfoWrapperTy &AsyncInfoWrapper) = 0;
+ /// Instert a data fence between previous data operations and the following
+ /// operations if necessary for the device
+ virtual Error dataFence(__tgt_async_info *AsyncInfo) = 0;
+
/// Exchange data between devices (device to device transfer). Calling this
/// function is only valid if GenericPlugin::isDataExchangable() passing the
/// two devices returns true.
@@ -1448,6 +1452,10 @@ struct GenericPluginTy {
int DstDeviceId, void *DstPtr, int64_t Size,
__tgt_async_info *AsyncInfo);
+ /// Places a fence between previous data movements and following data
+ /// movements if necessary on the device
+ int32_t data_fence(int32_t DeviceId, __tgt_async_info *AsyncInfo);
+
/// Begin executing a kernel on the given device.
int32_t launch_kernel(int32_t DeviceId, void *TgtEntryPtr, void **TgtArgs,
ptrdiff_t *TgtOffsets, KernelArgsTy *KernelArgs,
diff --git a/offload/plugins-nextgen/common/src/PluginInterface.cpp b/offload/plugins-nextgen/common/src/PluginInterface.cpp
index 083d41659a469..5a7d8e936313f 100644
--- a/offload/plugins-nextgen/common/src/PluginInterface.cpp
+++ b/offload/plugins-nextgen/common/src/PluginInterface.cpp
@@ -2324,3 +2324,15 @@ int32_t GenericPluginTy::async_barrier(omp_interop_val_t *Interop) {
}
return OFFLOAD_SUCCESS;
}
+
+int32_t GenericPluginTy::data_fence(int32_t DeviceId,
+ __tgt_async_info *AsyncInfo) {
+ auto Err = getDevice(DeviceId).dataFence(AsyncInfo);
+ if (Err) {
+ REPORT("failure to place data fence on device %d: %s\n", DeviceId,
+ toString(std::move(Err)).data());
+ return OFFLOAD_FAIL;
+ }
+
+ return OFFLOAD_SUCCESS;
+}
diff --git a/offload/plugins-nextgen/cuda/src/rtl.cpp b/offload/plugins-nextgen/cuda/src/rtl.cpp
index 82c9f9b706cbd..f3b2229daf372 100644
--- a/offload/plugins-nextgen/cuda/src/rtl.cpp
+++ b/offload/plugins-nextgen/cuda/src/rtl.cpp
@@ -856,6 +856,13 @@ struct CUDADeviceTy : public GenericDeviceTy {
return Plugin::success();
}
+ /// Insert a data fence between previous data operations and the following
+ /// operations. This is a no-op for CUDA devices as operations inserted into
+ /// a queue are in-order.
+ Error dataFence(__tgt_async_info *Async) override {
+ return Plugin::success();
+ }
+
/// Initialize the device info for interoperability purposes.
Error initDeviceInfoImpl(__tgt_device_info *DeviceInfo) override {
assert(Context && "Context is null");
diff --git a/offload/plugins-nextgen/host/src/rtl.cpp b/offload/plugins-nextgen/host/src/rtl.cpp
index ed5213531999d..b88435aa9e02b 100644
--- a/offload/plugins-nextgen/host/src/rtl.cpp
+++ b/offload/plugins-nextgen/host/src/rtl.cpp
@@ -295,6 +295,13 @@ struct GenELF64DeviceTy : public GenericDeviceTy {
"dataExchangeImpl not supported");
}
+ /// Insert a data fence between previous data operations and the following
+ /// operations. This is a no-op for Host devices as operations inserted into
+ /// a queue are in-order.
+ Error dataFence(__tgt_async_info *Async) override {
+ return Plugin::success();
+ }
+
/// All functions are already synchronous. No need to do anything on this
/// synchronization function.
Error synchronizeImpl(__tgt_async_info &AsyncInfo,
|
jhuber6
approved these changes
Aug 15, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The purpose of this fence is to ensure that any
dataSubmits insertedinto a queue before a
dataFencefinish before finish before anydataSubmitsinserted after it begin.
This is a no-op for most queues, since they are in-order, and by design
any operations inserted into them occur in order.
But the interface is supposed to be functional for out-of-order queues.
The addition of the interface means that any operations that rely on
such ordering (like ATTACH map-type support in #149036) can invoke it,
without worrying about whether the underlying queue is in-order or
out-of-order.
Once a plugin supports out-of-order queues, the plugin can implement
this function, without requiring any change at the libomptarget level.