SWDEV-555888 - Refactor Numa code (#1191)

1. Create a set of mini numa interface.
In Linux, the interface is based on system call rather than libnuma.
In Windows, the interface can also work, but the policy class is dummy.
Different from Linux, Windows doesn't provide numactl tool or numa lib to setup numa policy, thus
the default policy is followed in Windows, that is, using the closest host numa node to allocate
pinned host memory in hipHostMalloc().
To get the closest host numa node of a GPU device, you need query the new attribute
hipDeviceAttributeHostNumaId. Then you can create a thread with CPU affinity on the numa node.
For example, reference the test in hip-tests/catch/perftests/memory/hipPerfHostNumaAllocWin.cc.

2. Remove pfnSetThreadGroupAffinity and pfnGetNumaNodeProcessorMaskEx as the functions have been exposed since Win7 and Win server 2008.

3. Other minor fixes.

projects/clr/rocclr/device/rocm/rocdevice.cpp

@@ -54,10 +54,6 @@
 #include <iostream>
 #include <iomanip>
 #include <memory>
-#ifdef ROCCLR_SUPPORT_NUMA_POLICY
-#include <numa.h>
-#include <numaif.h>
-#endif  // ROCCLR_SUPPORT_NUMA_POLICY
 #include <sstream>
 #include <vector>
 
@@ -2039,30 +2035,17 @@ void* Device::hostAlloc(size_t size, size_t alignment, MemorySegment mem_seg,
 // ================================================================================================
 void* Device::hostNumaAlloc(size_t size, size_t alignment, MemorySegment mem_seg) const {
   void* ptr = nullptr;
-#ifndef ROCCLR_SUPPORT_NUMA_POLICY
-  ptr = hostAlloc(size, alignment, mem_seg, cpu_agent_info_);
-#else
-  int mode = MPOL_DEFAULT;
-  int maxNodes = numa_num_possible_nodes();
-  bitmask* nodeMask = numa_bitmask_alloc(maxNodes);
-  auto cpuCount = cpu_agents_.size();
-
-  long res = get_mempolicy(&mode, nodeMask->maskp, nodeMask->size, NULL, 0);
-  if (res) {
-    LogPrintfError("get_mempolicy failed with error %ld", res);
-    return ptr;
+  auto numa_node_count = cpu_agents_.size(); // count of host numa nodes
+  numa::NumaPolicy np(numa_node_count);
+  if (!np.GetMemPolicy()) {
+    return hostAlloc(size, alignment, mem_seg, cpu_agent_info_);
   }
-  ClPrint(amd::LOG_DETAIL_DEBUG, amd::LOG_RESOURCE,
-          "get_mempolicy() succeed with mode %d, nodeMask 0x%lx, cpuCount %zu", mode,
-          *nodeMask->maskp, cpuCount);
-
-  switch (mode) {
-    // For details, see "man get_mempolicy".
-    case MPOL_BIND:
-    case MPOL_PREFERRED:
+  switch (np.GetPolicy()) {
+    case numa::NumaPolicy::Policy::kPrefered:
+    case numa::NumaPolicy::Policy::kBind:
       // We only care about the first CPU node
-      for (unsigned int i = 0; i < cpuCount; i++) {
-        if ((1u << i) & *nodeMask->maskp) {
+      for (unsigned int i = 0; i < numa_node_count; i++) {
+        if (np.IsPolicySetAt(i)) {
           ptr = hostAlloc(size, alignment, mem_seg, &cpu_agents_[i]);
           break;
         }
@@ -2072,8 +2055,6 @@ void* Device::hostNumaAlloc(size_t size, size_t alignment, MemorySegment mem_seg
       //  All other modes fall back to default mode
       ptr = hostAlloc(size, alignment, mem_seg, cpu_agent_info_);
   }
-  numa_free_cpumask(nodeMask);
-#endif  // ROCCLR_SUPPORT_NUMA_POLICY
   return ptr;
 }