Using Performance Monitoring Events B
B-45
1. A memory reference causing 64K aliasing conflict can be counted more than once in this stat. The resulting
performance penalty can vary from unnoticeable to considerable. Some implementations of the Pentium 4 processor
family can incur significant penalties from loads that alias to preceding stores.
2. Currently, bugs in this event can cause both overcounting and undercounting by as much as a factor of 2.
3. Most MMX technology instructions, Streaming SIMD Extensions and Streaming SIMD Extensions 2 decode into a
single
μop. There are some instructions that decode into several μops; in these limited cases, the metrics count the
number of
μops that are actually tagged.
Stalled Cycles
of Store Buffer
Resources
(non-standard
5
)
The duration of stalls
due to lack of store
buffers.
Resource_stall SBFULL
Stalls of Store
Buffer
Resources
(non-standard
5
)
The number of
allocation stalls due
to lack of store
buffers.
Resource_stall SBFULL
(Also Set the
following CCCR bits:
Compare=1;
Edge=1;
Threshold=0)
Machine Clear Metrics
Machine Clear
Count
The number of
cycles that the entire
pipeline of the
machine is cleared
for all causes.
Machine_clear CLEAR
(Also Set the
following CCCR bits:
Compare=1;
Edge=1;
Threshold=0)
Memory Order
Machine Clear
The number of times
that the entire
pipeline of the
machine is cleared
due to memory-
ordering issues.
Machine_clear MOCLEAR
Self-modifying
Code Clear
The number of times
the entire pipeline of
the machine is
cleared due to
self-modifying code
issues.
Machine_clear SMCCLEAR
Table B-1 Pentium 4 Processor Performance Metrics (continued)
Metric Description
Event Name or Metric
Expression
Event Mask Value
Required
To Next Page
Loading...