Use on-demand paging for CpuSaveStates read/write. It was measured about 200us for either read or write the PI CpuSaveStates to framework, ~400us in total on a platform with 80 CPUs with original for loop implementation. So with on-demand paging, if the framework SMI handler doesn’t read/write CpuSaveStates, ~400us will be saved. If the handler happens to use CpuSaveStates, there will be about 20us overhead for either read or write a page which contains 5 continuous CpuSaveStates.

git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@10328 6f19259b-4bc3-4df7-8a09-765794883524

EdkCompatibilityPkg/Compatibility/SmmBaseHelper/SmmBaseHelper.inf

@@ -30,6 +30,16 @@
 
 [Sources]
   SmmBaseHelper.c
+
+[Sources.Ia32]
+  PageFaultHandler.c
+  
+[Sources.X64]
+  X64/PageFaultHandler.asm | MSFT
+
+  X64/PageFaultHandler.asm | INTEL
+  
+  X64/PageFaultHandler.S   | GCC
  
 [Packages]
   MdePkg/MdePkg.dec
@@ -46,6 +56,8 @@
   DevicePathLib
   CacheMaintenanceLib
   MemoryAllocationLib
+  SynchronizationLib
+  CpuLib
 
 [Guids]
   gEfiSmmBaseThunkCommunicationGuid