[Openmp-commits] [openmp] [Libomptarget] Rework Record & Replay to be a plugin member (#88928) (PR #89097)
Joseph Huber via Openmp-commits
openmp-commits at lists.llvm.org
Wed Apr 17 15:29:13 PDT 2024
https://github.com/jhuber6 updated https://github.com/llvm/llvm-project/pull/89097
>From c0b49d88244c15868c1c92004a123614fd73ae50 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn at outlook.com>
Date: Tue, 16 Apr 2024 14:19:12 -0500
Subject: [PATCH] [Libomptarget] Rework Record & Replay to be a plugin member
(#88928)
Summary:
Previously, the R&R support was global state initialized by a global
constructor. This is bad because it prevents us from adequately
constraining the lifetime of the library. Additionally, we want to
minimize the amount of global state floating around.
This patch moves the R&R support into a plugin member like everything
else. This means there will be multiple copies of the R&R implementation
floating around, but this was already the case given the fact that we
currently handle everything with dynamic libraries.
---
.../common/include/PluginInterface.h | 16 +++++---
.../common/src/PluginInterface.cpp | 37 ++++++++++++-------
2 files changed, 34 insertions(+), 19 deletions(-)
diff --git a/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h b/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
index 79e8464bfda5c1..703a457fb0f4ee 100644
--- a/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
+++ b/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
@@ -45,6 +45,8 @@
#include "llvm/Support/raw_ostream.h"
#include "llvm/TargetParser/Triple.h"
+struct RecordReplayTy;
+
namespace llvm {
namespace omp {
namespace target {
@@ -1031,6 +1033,12 @@ struct GenericPluginTy {
return *RPCServer;
}
+ /// Get a reference to the R&R interface for this plugin.
+ RecordReplayTy &getRecordAndReplay() const {
+ assert(RecordReplay && "R&R not initialized");
+ return *RecordReplay;
+ }
+
/// Get the OpenMP requires flags set for this plugin.
int64_t getRequiresFlags() const { return RequiresFlags; }
@@ -1220,6 +1228,9 @@ struct GenericPluginTy {
/// The interface between the plugin and the GPU for host services.
RPCServerTy *RPCServer;
+
+ /// The interface into the record-and-replay functionality.
+ RecordReplayTy *RecordReplay;
};
namespace Plugin {
@@ -1286,11 +1297,6 @@ class PluginTy {
static Error deinit() {
assert(SpecificPlugin && "Plugin no longer valid");
- for (int32_t DevNo = 0, NumDev = SpecificPlugin->getNumDevices();
- DevNo < NumDev; ++DevNo)
- if (auto Err = SpecificPlugin->deinitDevice(DevNo))
- return Err;
-
// Deinitialize the plugin.
if (auto Err = SpecificPlugin->deinit())
return Err;
diff --git a/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp b/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
index b5f3c45c835fdb..1d6fe247fabb5e 100644
--- a/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
+++ b/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
@@ -362,8 +362,6 @@ struct RecordReplayTy {
}
};
-static RecordReplayTy RecordReplay;
-
// Extract the mapping of host function pointers to device function pointers
// from the entry table. Functions marked as 'indirect' in OpenMP will have
// offloading entries generated for them which map the host's function pointer
@@ -473,7 +471,8 @@ GenericKernelTy::getKernelLaunchEnvironment(
// Ctor/Dtor have no arguments, replaying uses the original kernel launch
// environment. Older versions of the compiler do not generate a kernel
// launch environment.
- if (isCtorOrDtor() || RecordReplay.isReplaying() ||
+ if (isCtorOrDtor() ||
+ GenericDevice.Plugin.getRecordAndReplay().isReplaying() ||
Version < OMP_KERNEL_ARG_MIN_VERSION_WITH_DYN_PTR)
return nullptr;
@@ -562,6 +561,7 @@ Error GenericKernelTy::launch(GenericDeviceTy &GenericDevice, void **ArgPtrs,
// Record the kernel description after we modified the argument count and num
// blocks/threads.
+ RecordReplayTy &RecordReplay = GenericDevice.Plugin.getRecordAndReplay();
if (RecordReplay.isRecording()) {
RecordReplay.saveImage(getName(), getImage());
RecordReplay.saveKernelInput(getName(), getImage());
@@ -839,8 +839,8 @@ Error GenericDeviceTy::deinit(GenericPluginTy &Plugin) {
delete MemoryManager;
MemoryManager = nullptr;
- if (RecordReplay.isRecordingOrReplaying())
- RecordReplay.deinit();
+ if (Plugin.getRecordAndReplay().isRecordingOrReplaying())
+ Plugin.getRecordAndReplay().deinit();
if (RPCServer)
if (auto Err = RPCServer->deinitDevice(*this))
@@ -858,6 +858,7 @@ Error GenericDeviceTy::deinit(GenericPluginTy &Plugin) {
return deinitImpl();
}
+
Expected<DeviceImageTy *>
GenericDeviceTy::loadBinary(GenericPluginTy &Plugin,
const __tgt_device_image *InputTgtImage) {
@@ -892,7 +893,8 @@ GenericDeviceTy::loadBinary(GenericPluginTy &Plugin,
return std::move(Err);
// Setup the global device memory pool if needed.
- if (!RecordReplay.isReplaying() && shouldSetupDeviceMemoryPool()) {
+ if (!Plugin.getRecordAndReplay().isReplaying() &&
+ shouldSetupDeviceMemoryPool()) {
uint64_t HeapSize;
auto SizeOrErr = getDeviceHeapSize(HeapSize);
if (SizeOrErr) {
@@ -1307,8 +1309,8 @@ Expected<void *> GenericDeviceTy::dataAlloc(int64_t Size, void *HostPtr,
TargetAllocTy Kind) {
void *Alloc = nullptr;
- if (RecordReplay.isRecordingOrReplaying())
- return RecordReplay.alloc(Size);
+ if (Plugin.getRecordAndReplay().isRecordingOrReplaying())
+ return Plugin.getRecordAndReplay().alloc(Size);
switch (Kind) {
case TARGET_ALLOC_DEFAULT:
@@ -1344,7 +1346,7 @@ Expected<void *> GenericDeviceTy::dataAlloc(int64_t Size, void *HostPtr,
Error GenericDeviceTy::dataDelete(void *TgtPtr, TargetAllocTy Kind) {
// Free is a noop when recording or replaying.
- if (RecordReplay.isRecordingOrReplaying())
+ if (Plugin.getRecordAndReplay().isRecordingOrReplaying())
return Plugin::success();
int Res;
@@ -1396,6 +1398,7 @@ Error GenericDeviceTy::launchKernel(void *EntryPtr, void **ArgPtrs,
ptrdiff_t *ArgOffsets,
KernelArgsTy &KernelArgs,
__tgt_async_info *AsyncInfo) {
+ RecordReplayTy &RecordReplay = Plugin.getRecordAndReplay();
AsyncInfoWrapperTy AsyncInfoWrapper(
*this, RecordReplay.isRecordingOrReplaying() ? nullptr : AsyncInfo);
@@ -1495,6 +1498,9 @@ Error GenericPluginTy::init() {
RPCServer = new RPCServerTy(*this);
assert(RPCServer && "Invalid RPC server");
+ RecordReplay = new RecordReplayTy();
+ assert(RecordReplay && "Invalid Record and Replay handler");
+
return Plugin::success();
}
@@ -1515,6 +1521,9 @@ Error GenericPluginTy::deinit() {
if (RPCServer)
delete RPCServer;
+ if (RecordReplay)
+ delete RecordReplay;
+
// Perform last deinitializations on the plugin.
return deinitImpl();
}
@@ -1630,12 +1639,12 @@ int32_t GenericPluginTy::initialize_record_replay(int32_t DeviceId,
isRecord ? RecordReplayTy::RRStatusTy::RRRecording
: RecordReplayTy::RRStatusTy::RRReplaying;
- if (auto Err = RecordReplay.init(&Device, MemorySize, VAddr, Status,
- SaveOutput, ReqPtrArgOffset)) {
+ if (auto Err = getRecordAndReplay().init(&Device, MemorySize, VAddr, Status,
+ SaveOutput, ReqPtrArgOffset)) {
REPORT("WARNING RR did not intialize RR-properly with %lu bytes"
"(Error: %s)\n",
MemorySize, toString(std::move(Err)).data());
- RecordReplay.setStatus(RecordReplayTy::RRStatusTy::RRDeactivated);
+ RecordReplay->setStatus(RecordReplayTy::RRStatusTy::RRDeactivated);
if (!isRecord) {
return OFFLOAD_FAIL;
@@ -1984,8 +1993,8 @@ int32_t GenericPluginTy::get_global(__tgt_device_binary Binary, uint64_t Size,
assert(DevicePtr && "Invalid device global's address");
// Save the loaded globals if we are recording.
- if (RecordReplay.isRecording())
- RecordReplay.addEntry(Name, Size, *DevicePtr);
+ if (getRecordAndReplay().isRecording())
+ getRecordAndReplay().addEntry(Name, Size, *DevicePtr);
return OFFLOAD_SUCCESS;
}
More information about the Openmp-commits
mailing list