[SYCL][L0] Use compute engine for memory fill command (#6802)

E2E test in intel/llvm-test-suite#1273
Signed-off-by: Sergey V Maslov <[email protected]>

Author: smaslov-intel

sycl/doc/EnvironmentVariables.md

@@ -187,6 +187,7 @@ variables in production code.</span>
 | `SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS` | Any(\*) | Enable support of device-scope events whose state is not visible to the host. If enabled mode is SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS=1 the Level Zero plugin would create all events having device-scope only and create proxy host-visible events for them when their status is needed (wait/query) on the host. If enabled mode is SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS=2 the Level Zero plugin would create all events having device-scope and add proxy host-visible event at the end of each command-list submission. The default is 2, meaning only the last event in a batch is host-visible. |
 | `SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS` | Integer | When set to a positive value enables use of Level Zero immediate commandlists, which means there is no batching and all commands are immediately submitted for execution. Default is 0. Note: When immediate commandlist usage is enabled it is necessary to also set SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS to either 0 or 1.  |
 | `SYCL_PI_LEVEL_ZERO_USE_MULTIPLE_COMMANDLIST_BARRIERS` | Integer | When set to a positive value enables use of multiple Level Zero commandlists when submitting barriers. Default is 0. |
+| `SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE_FOR_FILL` | Integer | When set to a positive value enables use of a copy engine for memory fill operations. Default is 0. |
 
 ## Debugging variables for CUDA Plugin
 

sycl/plugins/level_zero/pi_level_zero.cpp

@@ -6660,7 +6660,7 @@ enqueueMemCopyHelper(pi_command_type CommandType, pi_queue Queue, void *Dst,
   printZeEventList(WaitList);
 
   ZE_CALL(zeCommandListAppendMemoryCopy,
-    (ZeCommandList, Dst, Src, Size, ZeEvent, 0, nullptr));
+          (ZeCommandList, Dst, Src, Size, ZeEvent, 0, nullptr));
 
   if (auto Res =
           Queue->executeCommandList(CommandList, BlockingWrite, OkToBatch))
@@ -6913,15 +6913,17 @@ enqueueMemFillHelper(pi_command_type CommandType, pi_queue Queue, void *Ptr,
 
   auto &Device = Queue->Device;
 
-  // Performance analysis on a simple SYCL data "fill" test shows copy engine
-  // is faster than compute engine for such operations.
-  //
-  bool PreferCopyEngine = true;
+  // Default to using compute engine for fill operation, but allow to
+  // override this with an environment variable.
+  const char *PreferCopyEngineEnv =
+      std::getenv("SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE_FOR_FILL");
+  bool PreferCopyEngine =
+      PreferCopyEngineEnv ? std::stoi(PreferCopyEngineEnv) != 0 : false;
 
   // Make sure that pattern size matches the capability of the copy queues.
   // Check both main and link groups as we don't known which one will be used.
   //
-  if (Device->hasCopyEngine()) {
+  if (PreferCopyEngine && Device->hasCopyEngine()) {
     if (Device->hasMainCopyEngine() &&
         Device->QueueGroup[_pi_device::queue_group_info_t::MainCopy]
                 .ZeProperties.maxMemoryFillPatternSize < PatternSize) {