[SYCL] Usm allocator refactor (#7785)

This commit makes USMAllocator unaware of the memory type on which it is
operating. The user of USMAllocator is responsible
for setting appropriate options for each instance of USMAllocator.

This is the first step to making USMAllocator generic. In the next step,
I'd like to start integration with Unified Memory Allocation
(oneMemory). USMAllocator would be used as a heap manager by UMA.

Author: igchor

sycl/plugins/level_zero/CMakeLists.txt

@@ -90,6 +90,8 @@ add_sycl_plugin(level_zero
     "../unified_runtime/ur/ur.cpp"
     "../unified_runtime/ur/usm_allocator.cpp"
     "../unified_runtime/ur/usm_allocator.hpp"
+    "../unified_runtime/ur/usm_allocator_config.cpp"
+    "../unified_runtime/ur/usm_allocator_config.hpp"
     "../unified_runtime/ur/adapters/level_zero/ur_level_zero.hpp"
     "../unified_runtime/ur/adapters/level_zero/ur_level_zero.cpp"
     # Following are the PI Level-Zero Plugin only codes.

sycl/plugins/level_zero/pi_level_zero.cpp

@@ -26,6 +26,7 @@
 
 #include <zet_api.h>
 
+#include "ur/usm_allocator_config.hpp"
 #include "ur_bindings.hpp"
 
 extern "C" {
@@ -96,6 +97,8 @@ static const bool IndirectAccessTrackingEnabled = [] {
          nullptr;
 }();
 
+static usm_settings::USMAllocatorConfig USMAllocatorConfigInstance;
+
 // Map from L0 to PI result.
 static inline pi_result mapError(ze_result_t Result) {
   return ur2piResult(ze2urResult(Result));
@@ -775,18 +778,24 @@ pi_result _pi_context::initialize() {
   auto createUSMAllocators = [this](pi_device Device) {
     SharedMemAllocContexts.emplace(
         std::piecewise_construct, std::make_tuple(Device->ZeDevice),
-        std::make_tuple(std::unique_ptr<SystemMemory>(
-            new USMSharedMemoryAlloc(this, Device))));
+        std::make_tuple(
+            std::unique_ptr<SystemMemory>(
+                new USMSharedMemoryAlloc(this, Device)),
+            USMAllocatorConfigInstance.Configs[usm_settings::MemType::Shared]));
 
     SharedReadOnlyMemAllocContexts.emplace(
         std::piecewise_construct, std::make_tuple(Device->ZeDevice),
         std::make_tuple(std::unique_ptr<SystemMemory>(
-            new USMSharedReadOnlyMemoryAlloc(this, Device))));
+                            new USMSharedReadOnlyMemoryAlloc(this, Device)),
+                        USMAllocatorConfigInstance
+                            .Configs[usm_settings::MemType::SharedReadOnly]));
 
     DeviceMemAllocContexts.emplace(
         std::piecewise_construct, std::make_tuple(Device->ZeDevice),
-        std::make_tuple(std::unique_ptr<SystemMemory>(
-            new USMDeviceMemoryAlloc(this, Device))));
+        std::make_tuple(
+            std::unique_ptr<SystemMemory>(
+                new USMDeviceMemoryAlloc(this, Device)),
+            USMAllocatorConfigInstance.Configs[usm_settings::MemType::Device]));
   };
 
   // Recursive helper to call createUSMAllocators for all sub-devices
@@ -808,7 +817,8 @@ pi_result _pi_context::initialize() {
   // are device-specific. Host allocations are not device-dependent therefore
   // we don't need a map with device as key.
   HostMemAllocContext = std::make_unique<USMAllocContext>(
-      std::unique_ptr<SystemMemory>(new USMHostMemoryAlloc(this)));
+      std::unique_ptr<SystemMemory>(new USMHostMemoryAlloc(this)),
+      USMAllocatorConfigInstance.Configs[usm_settings::MemType::Host]);
 
   // We may allocate memory to this root device so create allocators.
   if (SingleRootDevice &&
@@ -8116,12 +8126,6 @@ pi_result USMHostMemoryAlloc::allocateImpl(void **ResultPtr, size_t Size,
   return USMHostAllocImpl(ResultPtr, Context, nullptr, Size, Alignment);
 }
 
-MemType USMSharedMemoryAlloc::getMemTypeImpl() { return MemType::Shared; }
-
-MemType USMDeviceMemoryAlloc::getMemTypeImpl() { return MemType::Device; }
-
-MemType USMHostMemoryAlloc::getMemTypeImpl() { return MemType::Host; }
-
 void *USMMemoryAllocBase::allocate(size_t Size) {
   void *Ptr = nullptr;
 
@@ -8150,8 +8154,6 @@ void USMMemoryAllocBase::deallocate(void *Ptr, bool OwnZeMemHandle) {
   }
 }
 
-MemType USMMemoryAllocBase::getMemType() { return getMemTypeImpl(); }
-
 pi_result piextUSMDeviceAlloc(void **ResultPtr, pi_context Context,
                               pi_device Device,
                               pi_usm_mem_properties *Properties, size_t Size,
@@ -9227,7 +9229,7 @@ pi_result _pi_buffer::getZeHandle(char *&ZeHandle, access_mode_t AccessMode,
     // The host allocation may already exists, e.g. with imported
     // host ptr, or in case of interop buffer.
     if (!HostAllocation.ZeHandle) {
-      if (enableBufferPooling()) {
+      if (USMAllocatorConfigInstance.EnableBuffers) {
         HostAllocation.ReleaseAction = allocation_t::free;
         PI_CALL(piextUSMHostAlloc(pi_cast<void **>(&ZeHandle), Context, nullptr,
                                   Size, getAlignment()));
@@ -9280,7 +9282,7 @@ pi_result _pi_buffer::getZeHandle(char *&ZeHandle, access_mode_t AccessMode,
       Allocation.Valid = true;
       return PI_SUCCESS;
     } else { // Create device allocation
-      if (enableBufferPooling()) {
+      if (USMAllocatorConfigInstance.EnableBuffers) {
         Allocation.ReleaseAction = allocation_t::free;
         PI_CALL(piextUSMDeviceAlloc(pi_cast<void **>(&ZeHandle), Context,
                                     Device, nullptr, Size, getAlignment()));
@@ -9341,7 +9343,7 @@ pi_result _pi_buffer::getZeHandle(char *&ZeHandle, access_mode_t AccessMode,
         // host ptr, or in case of interop buffer.
         if (!HostAllocation.ZeHandle) {
           void *ZeHandleHost;
-          if (enableBufferPooling()) {
+          if (USMAllocatorConfigInstance.EnableBuffers) {
             HostAllocation.ReleaseAction = allocation_t::free;
             PI_CALL(piextUSMHostAlloc(&ZeHandleHost, Context, nullptr, Size,
                                       getAlignment()));

sycl/plugins/level_zero/pi_level_zero.hpp

@@ -118,23 +118,20 @@ class USMMemoryAllocBase : public SystemMemory {
   // type
   virtual pi_result allocateImpl(void **ResultPtr, size_t Size,
                                  pi_uint32 Alignment) = 0;
-  virtual MemType getMemTypeImpl() = 0;
 
 public:
   USMMemoryAllocBase(pi_context Ctx, pi_device Dev)
       : Context{Ctx}, Device{Dev} {}
   void *allocate(size_t Size) override final;
   void *allocate(size_t Size, size_t Alignment) override final;
   void deallocate(void *Ptr, bool OwnZeMemHandle) override final;
-  MemType getMemType() override final;
 };
 
 // Allocation routines for shared memory type
 class USMSharedMemoryAlloc : public USMMemoryAllocBase {
 protected:
   pi_result allocateImpl(void **ResultPtr, size_t Size,
                          pi_uint32 Alignment) override;
-  MemType getMemTypeImpl() override;
 
 public:
   USMSharedMemoryAlloc(pi_context Ctx, pi_device Dev)
@@ -146,7 +143,6 @@ class USMSharedReadOnlyMemoryAlloc : public USMMemoryAllocBase {
 protected:
   pi_result allocateImpl(void **ResultPtr, size_t Size,
                          pi_uint32 Alignment) override;
-  MemType getMemTypeImpl() override { return MemType::SharedReadOnly; }
 
 public:
   USMSharedReadOnlyMemoryAlloc(pi_context Ctx, pi_device Dev)
@@ -158,7 +154,6 @@ class USMDeviceMemoryAlloc : public USMMemoryAllocBase {
 protected:
   pi_result allocateImpl(void **ResultPtr, size_t Size,
                          pi_uint32 Alignment) override;
-  MemType getMemTypeImpl() override;
 
 public:
   USMDeviceMemoryAlloc(pi_context Ctx, pi_device Dev)
@@ -170,7 +165,6 @@ class USMHostMemoryAlloc : public USMMemoryAllocBase {
 protected:
   pi_result allocateImpl(void **ResultPtr, size_t Size,
                          pi_uint32 Alignment) override;
-  MemType getMemTypeImpl() override;
 
 public:
   USMHostMemoryAlloc(pi_context Ctx) : USMMemoryAllocBase(Ctx, nullptr) {}