[L0] Rebased against phase 2 of cb events

winstonzhang-intel · winstonzhang-intel · commit 6582a8a0ea06 · 2024-11-05T10:21:17.000-08:00
CI in LLVM/SYCL: intel/llvm#14536 Signed-off-by: Winston Zhang <winston.zhang@intel.com>
diff --git a/source/adapters/level_zero/common.hpp b/source/adapters/level_zero/common.hpp
@@ -240,6 +240,19 @@ enum {
       2, // blocking UR calls, where supported (usually in enqueue commands)
 };
 
+typedef uint32_t ur_event_flags_t;
+typedef enum ur_event_flag_t {
+  USING_IMM_CMDLIST = UR_BIT(0),
+  HOST_VISIBLE = UR_BIT(1),
+  COUNTER_BASED = UR_BIT(2),
+  ENABLE_PROFILER = UR_BIT(3),
+  MULTIDEVICE = UR_BIT(4),
+  /// @cond
+  UR_EVENT_FLAG_FORCE_UINT32 = 0x7FFFFFFF
+  /// @endcond
+
+} ur_event_flag_t;
+
 static const uint32_t UrL0Serialize = [] {
   const char *ZeSerializeMode = std::getenv("ZE_SERIALIZE");
   const char *UrL0SerializeMode = std::getenv("UR_L0_SERIALIZE");
diff --git a/source/adapters/level_zero/context.cpp b/source/adapters/level_zero/context.cpp
@@ -473,9 +473,8 @@ static const uint32_t MaxNumEventsPerPool = [] {
 }();
 
 ur_result_t ur_context_handle_t_::getFreeSlotInExistingOrNewPool(
-    ze_event_pool_handle_t &Pool, size_t &Index, bool HostVisible,
-    bool ProfilingEnabled, ur_device_handle_t Device,
-    bool CounterBasedEventEnabled, bool UsingImmCmdList) {
+    ze_event_pool_handle_t &Pool, size_t &Index, ur_event_flags_t Flags,
+    ur_device_handle_t Device) {
   // Lock while updating event pool machinery.
   std::scoped_lock<ur_mutex> Lock(ZeEventPoolCacheMutex);
 
@@ -485,8 +484,7 @@ ur_result_t ur_context_handle_t_::getFreeSlotInExistingOrNewPool(
     ZeDevice = Device->ZeDevice;
   }
   std::list<ze_event_pool_handle_t> *ZePoolCache =
-      getZeEventPoolCache(HostVisible, ProfilingEnabled,
-                          CounterBasedEventEnabled, UsingImmCmdList, ZeDevice);
+      getZeEventPoolCache(Flags, ZeDevice);
 
   if (!ZePoolCache->empty()) {
     if (NumEventsAvailableInEventPool[ZePoolCache->front()] == 0) {
@@ -517,14 +515,14 @@ ur_result_t ur_context_handle_t_::getFreeSlotInExistingOrNewPool(
     ZeEventPoolDesc.count = MaxNumEventsPerPool;
     ZeEventPoolDesc.flags = 0;
     ZeEventPoolDesc.pNext = nullptr;
-    if (HostVisible)
+    if (Flags & HOST_VISIBLE)
       ZeEventPoolDesc.flags |= ZE_EVENT_POOL_FLAG_HOST_VISIBLE;
-    if (ProfilingEnabled)
+    if (Flags & ENABLE_PROFILER)
       ZeEventPoolDesc.flags |= ZE_EVENT_POOL_FLAG_KERNEL_TIMESTAMP;
     logger::debug("ze_event_pool_desc_t flags set to: {}",
                   ZeEventPoolDesc.flags);
-    if (CounterBasedEventEnabled) {
-      if (UsingImmCmdList) {
+    if (Flags & COUNTER_BASED) {
+      if (Flags & USING_IMM_CMDLIST) {
         counterBasedExt.flags = ZE_EVENT_POOL_COUNTER_BASED_EXP_FLAG_IMMEDIATE;
       } else {
         counterBasedExt.flags =
@@ -558,19 +556,24 @@ ur_result_t ur_context_handle_t_::getFreeSlotInExistingOrNewPool(
   return UR_RESULT_SUCCESS;
 }
 
-ur_event_handle_t ur_context_handle_t_::getEventFromContextCache(
-    bool HostVisible, bool WithProfiling, ur_device_handle_t Device,
-    bool CounterBasedEventEnabled) {
+ur_event_handle_t
+ur_context_handle_t_::getEventFromContextCache(ur_event_flags_t Flags,
+                                               ur_device_handle_t Device) {
   std::scoped_lock<ur_mutex> Lock(EventCacheMutex);
-  auto Cache = getEventCache(HostVisible, WithProfiling, Device);
+  auto Cache = getEventCache(Flags, Device);
+  if (Flags & COUNTER_BASED) {
+    Cache = getCounterBasedEventCache(Flags, Device);
+  }
   if (Cache->empty())
     return nullptr;
 
   auto It = Cache->begin();
   ur_event_handle_t Event = *It;
-  if (Event->CounterBasedEventsEnabled != CounterBasedEventEnabled) {
+  if (Event->CounterBasedEventsEnabled !=
+      static_cast<bool>(Flags & COUNTER_BASED)) {
     return nullptr;
   }
+
   Cache->erase(It);
   // We have to reset event before using it.
   Event->reset();
@@ -584,10 +587,19 @@ void ur_context_handle_t_::addEventToContextCache(ur_event_handle_t Event) {
   if (!Event->IsMultiDevice && Event->UrQueue) {
     Device = Event->UrQueue->Device;
   }
-
-  auto Cache = getEventCache(Event->isHostVisible(),
-                             Event->isProfilingEnabled(), Device);
-  Cache->emplace_back(Event);
+  ur_event_flags_t Flags = 0;
+  if (Event->isHostVisible())
+    Flags |= HOST_VISIBLE;
+  if (Event->isProfilingEnabled())
+    Flags |= ENABLE_PROFILER;
+
+  if (Event->CounterBasedEventsEnabled) {
+    auto Cache = getCounterBasedEventCache(Flags, Device);
+    Cache->emplace_back(Event);
+  } else {
+    auto Cache = getEventCache(Flags, Device);
+    Cache->emplace_back(Event);
+  }
 }
 
 ur_result_t
@@ -609,9 +621,17 @@ ur_context_handle_t_::decrementUnreleasedEventsInPool(ur_event_handle_t Event) {
     ZeDevice = Event->UrQueue->Device->ZeDevice;
   }
 
-  std::list<ze_event_pool_handle_t> *ZePoolCache = getZeEventPoolCache(
-      Event->isHostVisible(), Event->isProfilingEnabled(),
-      Event->CounterBasedEventsEnabled, UsingImmediateCommandlists, ZeDevice);
+  ur_event_flags_t Flags = 0;
+  if (Event->isHostVisible())
+    Flags |= HOST_VISIBLE;
+  if (Event->isProfilingEnabled())
+    Flags |= ENABLE_PROFILER;
+  if (Event->CounterBasedEventsEnabled)
+    Flags |= COUNTER_BASED;
+  if (UsingImmediateCommandlists)
+    Flags |= USING_IMM_CMDLIST;
+  std::list<ze_event_pool_handle_t> *ZePoolCache =
+      getZeEventPoolCache(Flags, ZeDevice);
 
   // Put the empty pool to the cache of the pools.
   if (NumEventsUnreleasedInEventPool[Event->ZeEventPool] == 0)
diff --git a/source/adapters/level_zero/context.hpp b/source/adapters/level_zero/context.hpp
@@ -204,75 +204,49 @@ struct ur_context_handle_t_ : _ur_object {
   // slot for a host-visible event. The ProfilingEnabled tells is we need a
   // slot for an event with profiling capabilities.
   ur_result_t getFreeSlotInExistingOrNewPool(ze_event_pool_handle_t &, size_t &,
-                                             bool HostVisible,
-                                             bool ProfilingEnabled,
-                                             ur_device_handle_t Device,
-                                             bool CounterBasedEventEnabled,
-                                             bool UsingImmCmdList);
+                                             ur_event_flags_t Flags,
+                                             ur_device_handle_t Device);
 
   // Get ur_event_handle_t from cache.
-  ur_event_handle_t getEventFromContextCache(bool HostVisible,
-                                             bool WithProfiling,
-                                             ur_device_handle_t Device,
-                                             bool CounterBasedEventEnabled);
+  ur_event_handle_t getEventFromContextCache(ur_event_flags_t Flags,
+                                             ur_device_handle_t Device);
 
   // Add ur_event_handle_t to cache.
   void addEventToContextCache(ur_event_handle_t);
 
-  enum EventPoolCacheType {
-    HostVisibleCacheType,
-    HostInvisibleCacheType,
-    HostVisibleCounterBasedRegularCacheType,
-    HostInvisibleCounterBasedRegularCacheType,
-    HostVisibleCounterBasedImmediateCacheType,
-    HostInvisibleCounterBasedImmediateCacheType
+  enum EventCacheType {
+    HostVisibleProfilingCacheType,
+    HostVisibleRegularCacheType,
+    HostInvisibleProfilingCacheType,
+    HostInvisibleRegularCacheType,
+    CounterBasedImmediateCacheType,
+    CounterBasedRegularCacheType,
+    CounterBasedImmediateProfilingCacheType,
+    CounterBasedRegularProfilingCacheType,
+    EventCacheTypeCount
   };
-
   std::list<ze_event_pool_handle_t> *
-  getZeEventPoolCache(bool HostVisible, bool WithProfiling,
-                      bool CounterBasedEventEnabled, bool UsingImmediateCmdList,
-                      ze_device_handle_t ZeDevice) {
-    EventPoolCacheType CacheType;
+  getZeEventPoolCache(ur_event_flags_t Flags, ze_device_handle_t ZeDevice) {
+    bool WithProfiling = Flags & ENABLE_PROFILER;
+    Flags &= ~ENABLE_PROFILER;
+    Flags &= ~MULTIDEVICE;
+    int index = static_cast<int>(Flags);
+    (Flags & COUNTER_BASED) ? index /= 2 : (index - 2);
 
-    calculateCacheIndex(HostVisible, CounterBasedEventEnabled,
-                        UsingImmediateCmdList, CacheType);
     if (ZeDevice) {
       auto ZeEventPoolCacheMap =
-          WithProfiling ? &ZeEventPoolCacheDeviceMap[CacheType * 2]
-                        : &ZeEventPoolCacheDeviceMap[CacheType * 2 + 1];
+          WithProfiling ? &ZeEventPoolCacheDeviceMap[index * 2]
+                        : &ZeEventPoolCacheDeviceMap[index * 2 + 1];
       if (ZeEventPoolCacheMap->find(ZeDevice) == ZeEventPoolCacheMap->end()) {
         ZeEventPoolCache.emplace_back();
         ZeEventPoolCacheMap->insert(
             std::make_pair(ZeDevice, ZeEventPoolCache.size() - 1));
       }
       return &ZeEventPoolCache[(*ZeEventPoolCacheMap)[ZeDevice]];
     } else {
-      return WithProfiling ? &ZeEventPoolCache[CacheType * 2]
-                           : &ZeEventPoolCache[CacheType * 2 + 1];
-    }
-  }
-
-  ur_result_t calculateCacheIndex(bool HostVisible,
-                                  bool CounterBasedEventEnabled,
-                                  bool UsingImmediateCmdList,
-                                  EventPoolCacheType &CacheType) {
-    if (CounterBasedEventEnabled && HostVisible && !UsingImmediateCmdList) {
-      CacheType = HostVisibleCounterBasedRegularCacheType;
-    } else if (CounterBasedEventEnabled && !HostVisible &&
-               !UsingImmediateCmdList) {
-      CacheType = HostInvisibleCounterBasedRegularCacheType;
-    } else if (CounterBasedEventEnabled && HostVisible &&
-               UsingImmediateCmdList) {
-      CacheType = HostVisibleCounterBasedImmediateCacheType;
-    } else if (CounterBasedEventEnabled && !HostVisible &&
-               UsingImmediateCmdList) {
-      CacheType = HostInvisibleCounterBasedImmediateCacheType;
-    } else if (!CounterBasedEventEnabled && HostVisible) {
-      CacheType = HostVisibleCacheType;
-    } else {
-      CacheType = HostInvisibleCacheType;
+      return WithProfiling ? &ZeEventPoolCache[index * 2]
+                           : &ZeEventPoolCache[index * 2 + 1];
     }
-    return UR_RESULT_SUCCESS;
   }
 
   // Decrement number of events living in the pool upon event destroy
@@ -314,33 +288,68 @@ struct ur_context_handle_t_ : _ur_object {
 
 private:
   // Get the cache of events for a provided scope and profiling mode.
-  auto getEventCache(bool HostVisible, bool WithProfiling,
-                     ur_device_handle_t Device) {
-    if (HostVisible) {
+  auto getEventCache(ur_event_flags_t Flags, ur_device_handle_t Device) {
+    if (Flags & HOST_VISIBLE) {
       if (Device) {
         auto EventCachesMap =
-            WithProfiling ? &EventCachesDeviceMap[0] : &EventCachesDeviceMap[1];
+            (Flags & ENABLE_PROFILER)
+                ? &EventCachesDeviceMap[HostVisibleProfilingCacheType]
+                : &EventCachesDeviceMap[HostVisibleRegularCacheType];
         if (EventCachesMap->find(Device) == EventCachesMap->end()) {
           EventCaches.emplace_back();
           EventCachesMap->insert(
               std::make_pair(Device, EventCaches.size() - 1));
         }
         return &EventCaches[(*EventCachesMap)[Device]];
       } else {
-        return WithProfiling ? &EventCaches[0] : &EventCaches[1];
+        return (Flags & ENABLE_PROFILER)
+                   ? &EventCaches[HostVisibleProfilingCacheType]
+                   : &EventCaches[HostVisibleRegularCacheType];
       }
     } else {
       if (Device) {
         auto EventCachesMap =
-            WithProfiling ? &EventCachesDeviceMap[2] : &EventCachesDeviceMap[3];
+            (Flags & ENABLE_PROFILER)
+                ? &EventCachesDeviceMap[HostInvisibleProfilingCacheType]
+                : &EventCachesDeviceMap[HostInvisibleRegularCacheType];
         if (EventCachesMap->find(Device) == EventCachesMap->end()) {
           EventCaches.emplace_back();
           EventCachesMap->insert(
               std::make_pair(Device, EventCaches.size() - 1));
         }
         return &EventCaches[(*EventCachesMap)[Device]];
       } else {
-        return WithProfiling ? &EventCaches[2] : &EventCaches[3];
+        return (Flags & ENABLE_PROFILER)
+                   ? &EventCaches[HostInvisibleProfilingCacheType]
+                   : &EventCaches[HostInvisibleRegularCacheType];
+      }
+    }
+  };
+  auto getCounterBasedEventCache(ur_event_flags_t Flags,
+                                 ur_device_handle_t Device) {
+    if (Flags & USING_IMM_CMDLIST) {
+      if (Device) {
+        auto EventCachesMap =
+            (Flags & ENABLE_PROFILER)
+                ? &EventCachesDeviceMap[CounterBasedImmediateProfilingCacheType]
+                : &EventCachesDeviceMap[CounterBasedImmediateCacheType];
+        return &EventCaches[(*EventCachesMap)[Device]];
+      } else {
+        return (Flags & ENABLE_PROFILER)
+                   ? &EventCaches[CounterBasedImmediateProfilingCacheType]
+                   : &EventCaches[CounterBasedImmediateCacheType];
+      }
+    } else {
+      if (Device) {
+        auto EventCachesMap =
+            (Flags & ENABLE_PROFILER)
+                ? &EventCachesDeviceMap[CounterBasedRegularProfilingCacheType]
+                : &EventCachesDeviceMap[CounterBasedRegularCacheType];
+        return &EventCaches[(*EventCachesMap)[Device]];
+      } else {
+        return (Flags & ENABLE_PROFILER)
+                   ? &EventCaches[CounterBasedRegularProfilingCacheType]
+                   : &EventCaches[CounterBasedRegularCacheType];
       }
     }
   }
diff --git a/source/adapters/level_zero/event.cpp b/source/adapters/level_zero/event.cpp
@@ -1277,15 +1277,24 @@ ur_result_t EventCreate(ur_context_handle_t Context, ur_queue_handle_t Queue,
   bool ProfilingEnabled =
       ForceDisableProfiling ? false : (!Queue || Queue->isProfilingEnabled());
   bool UsingImmediateCommandlists = !Queue || Queue->UsingImmCmdLists;
+  ur_event_flags_t Flags = 0;
+  if (ProfilingEnabled)
+    Flags |= ENABLE_PROFILER;
+  if (UsingImmediateCommandlists)
+    Flags |= USING_IMM_CMDLIST;
+  if (HostVisible)
+    Flags |= HOST_VISIBLE;
+  if (IsMultiDevice)
+    Flags |= MULTIDEVICE;
+  if (CounterBasedEventEnabled)
+    Flags |= COUNTER_BASED;
 
   ur_device_handle_t Device = nullptr;
-
-  if (!IsMultiDevice && Queue) {
+  if (!(Flags & MULTIDEVICE) && Queue) {
     Device = Queue->Device;
   }
 
-  if (auto CachedEvent = Context->getEventFromContextCache(
-          HostVisible, ProfilingEnabled, Device, CounterBasedEventEnabled)) {
+  if (auto CachedEvent = Context->getEventFromContextCache(Flags, Device)) {
     *RetEvent = CachedEvent;
     return UR_RESULT_SUCCESS;
   }
@@ -1295,16 +1304,15 @@ ur_result_t EventCreate(ur_context_handle_t Context, ur_queue_handle_t Queue,
 
   size_t Index = 0;
 
-  if (auto Res = Context->getFreeSlotInExistingOrNewPool(
-          ZeEventPool, Index, HostVisible, ProfilingEnabled, Device,
-          CounterBasedEventEnabled, UsingImmediateCommandlists))
+  if (auto Res = Context->getFreeSlotInExistingOrNewPool(ZeEventPool, Index,
+                                                         Flags, Device))
     return Res;
 
   ZeStruct<ze_event_desc_t> ZeEventDesc;
   ZeEventDesc.index = Index;
   ZeEventDesc.wait = 0;
 
-  if (HostVisible || CounterBasedEventEnabled) {
+  if ((Flags & USING_IMM_CMDLIST) || (Flags & COUNTER_BASED)) {
     ZeEventDesc.signal = ZE_EVENT_SCOPE_FLAG_HOST;
   } else {
     //
@@ -1329,8 +1337,8 @@ ur_result_t EventCreate(ur_context_handle_t Context, ur_queue_handle_t Queue,
   } catch (...) {
     return UR_RESULT_ERROR_UNKNOWN;
   }
-  (*RetEvent)->CounterBasedEventsEnabled = CounterBasedEventEnabled;
-  if (HostVisible)
+  (*RetEvent)->CounterBasedEventsEnabled = (Flags & COUNTER_BASED);
+  if (Flags & HOST_VISIBLE)
     (*RetEvent)->HostVisibleEvent =
         reinterpret_cast<ur_event_handle_t>(*RetEvent);
 
diff --git a/source/adapters/level_zero/event.hpp b/source/adapters/level_zero/event.hpp
@@ -132,6 +132,7 @@ struct ur_event_handle_t_ : _ur_object {
     OwnNativeHandle = OwnZeEvent;
   }
 
+  ur_event_flags_t flags;
   // Level Zero event handle.
   ze_event_handle_t ZeEvent;
 

Original file line number	Diff line number	Diff line change
`@@ -132,6 +132,7 @@ struct ur_event_handle_t_ : _ur_object {`
`132`	`132`	`OwnNativeHandle = OwnZeEvent;`
`133`	`133`	`}`
`134`	`134`
	`135`	`+ ur_event_flags_t flags;`
`135`	`136`	`// Level Zero event handle.`
`136`	`137`	`ze_event_handle_t ZeEvent;`
`137`	`138`