Intel 253668-032US Manual De Usuario

8-62   Vol. 3

MULTIPLE-PROCESSOR MANAGEMENT

using OS specific APIs.
// Allocate per processor arrays to store the Package_ID, Core_ID and SMT_ID for every started
// processor.
 

ThreadAffinityMask = 1;

     ProcessorNum = 0;

while (ThreadAffinityMask != 0 && ThreadAffinityMask <= SystemAffinity) {

// Check to make sure we can utilize this processor first.
if (ThreadAffinityMask & SystemAffinity){

Set thread to run on the processor specified in ThreadAffinityMask
Wait if necessary and ensure thread is running on specified processor

APIC_ID = GetAPIC_ID(); // 32 bit ID in Example 8-19 or 8-bit ID in Example 

Extract the Package_ID, Core_ID and SMT_ID as explained in three level extraction  

algorithm of Example 8-21

PackageID[ProcessorNUM] = PACKAGE_ID;
CoreID[ProcessorNum] = CORE_ID;
SmtID[ProcessorNum] = SMT_ID;
ProcessorNum++;

}
ThreadAffinityMask <<= 1;

}
NumStartedLPs = ProcessorNum;

b) Using the list of PACKAGE_ID to count the number of physical packages in a MP system and 

construct, for each package, a multi-bit mask corresponding to those logical processors residing in 

the same package.

// Compute the number of packages by counting the number of processors  

// with unique PACKAGE_IDs in the PackageID array.  

// Compute the mask of processors in each package. 

PackageIDBucket is an array of unique PACKAGE_ID values. Allocate an array of 

NumStartedLPs count of entries in this array.
PackageProcessorMask is a corresponding array of the bit mask of processors belonging to 

the same package, these are processors with the same PACKAGE_ID 
The algorithm below assumes there is symmetry across package boundary if more than  

one socket is populated in an MP system.
// Bucket Package IDs and compute processor mask for every package.

PackageNum = 1;
PackageIDBucket[0] = PackageID[0];
ProcessorMask = 1;