Intel 253668-032US User Manual

Page of 806
8-44   Vol. 3
MULTIPLE-PROCESSOR MANAGEMENT
As a consequence, the use of the WBINVD instruction can have an impact on 
interrupt/event response time.
INVD instruction — The entire cache hierarchy is invalidated without writing 
back modified data to memory. All logical processors are stopped from executing 
until after the invalidate operation is completed. A special bus cycle is sent to all 
caching agents.
CLFLUSH instruction — The specified cache line is invalidated from the cache 
hierarchy after any modified data is written back to memory and a bus cycle is 
sent to all caching agents, regardless of which logical processor caused the cache 
line to be filled.
CD flag in control register CR0 — Each logical processor has its own CR0 
control register, and thus its own CD flag in CR0. The CD flags for the two logical 
processors are ORed together, such that when any logical processor sets its CD 
flag, the entire cache is nominally disabled. 
8.7.13.2   Processor Translation Lookaside Buffers (TLBs)
In processors supporting Intel Hyper-Threading Technology, data cache TLBs are 
shared. The instruction cache TLB may be duplicated or shared in each logical 
processor, depending on implementation specifics of different processor families.
Entries in the TLBs are tagged with an ID that indicates the logical processor that 
initiated the translation. This tag applies even for translations that are marked global 
using the page-global feature for memory paging. See Section 4.10, “Caching Trans-
lation Information,” for inform
ation about global translations.
When a logical processor performs a TLB invalidation operation, only the TLB entries 
that are tagged for that logical processor are guaranteed to be flushed. This protocol 
applies to all TLB invalidation operations, including writes to control registers CR3 
and CR4 and uses of the INVLPG instruction.
8.7.13.3   Thermal Monitor
In a processor that supports Intel Hyper-Threading Technology, logical processors 
share the catastrophic shutdown detector and the automatic thermal monitoring 
mechanism (see Section 14.5, “Thermal Monitoring and Protection”). Sharing results 
in the following behavior:
If the processor’s core temperature rises above the preset catastrophic shutdown 
temperature, the processor core halts execution, which causes both logical 
processors to stop execution.
When the processor’s core temperature rises above the preset automatic thermal 
monitor trip temperature, the clock speed of the processor core is automatically 
modulated, which effects the execution speed of both logical processors.
For software controlled clock modulation, each logical processor has its own 
IA32_CLOCK_MODULATION MSR, allowing clock modulation to be enabled or