Intel 253668-032US Benutzerhandbuch

Seite von 806
Vol. 3   8-43
MULTIPLE-PROCESSOR MANAGEMENT
8.7.11 
MICROCODE UPDATE Resources
In an Intel processor supporting Intel Hyper-Threading Technology, the microcode 
update facilities are shared between the logical processors; either logical processor 
can initiate an update. Each logical processor has its own BIOS signature MSR 
(IA32_BIOS_SIGN_ID at MSR address 8BH). When a logical processor performs an 
update for the physical processor, the IA32_BIOS_SIGN_ID MSRs for resident logical 
processors are updated with identical information. If logical processors initiate an 
update simultaneously, the processor core provides the necessary synchronization 
needed to insure that only one update is performed at a time. 
Operating system microcode update drivers that adhere to Intel’s guidelines do not 
need to be modified to run on processors supporting Intel Hyper-Threading Tech-
nology.
8.7.12 
Self Modifying Code
Intel processors supporting Intel Hyper-Threading Technology support self-modifying 
code, where data writes modify instructions cached or currently in flight. They also 
support cross-modifying code, where on an MP system writes generated by one 
processor modify instructions cached or currently in flight on another. See Section 
8.1.3, “Handling Self- and Cross-Modifying Code,”
 for a description of the require-
ments for self- and cross-modifying code in an IA-32 processor.
8.7.13 Implementation-Specific 
Intel HT Technology Facilities
The following non-architectural facilities are implementation-specific in IA-32 proces-
sors supporting Intel Hyper-Threading Technology:
Caches
Translation lookaside buffers (TLBs)
Thermal monitoring facilities
The Intel Xeon processor MP implementation is described in the following sections.
8.7.13.1   Processor Caches
For processors supporting Intel Hyper-Threading Technology, the caches are shared. 
Any cache manipulation instruction that is executed on one logical processor has a 
global effect on the cache hierarchy of the physical processor. Note the following:
WBINVD instruction — The entire cache hierarchy is invalidated after modified 
data is written back to memory. All logical processors are stopped from executing 
until after the write-back and invalidate operation is completed. A special bus 
cycle is sent to all caching agents. The amount of time or cycles for WBINVD to 
complete will vary due to the size of different cache hierarchies and other factors.