Chapter 11. Performance Monitor
11-17
Event Selection
19 (01_0011)
L2SRAM read
cycles
Incremented for each beat of a read from the L2SRAM. A beat consists of 8 bytes
if L2CR[L2DSIZ] = 0, otherwise it is 16 bytes.
20 (01_0100)
BIU ARTRYed TS
assertions
Counts transactions initiated by this processor retried on the system interface.
21 (01_0101)
BIU 16-byte read
cycles
Counts assertions of TA for (16-byte) 128-bit reads
22 (01_0110)
dRLT SMMs
Counts data reload table store-miss merges
23 (01_0111)
L2 write-hit-on-SHD
occurrence
Counts store misses from the data reload table that hit on a shared cache block
in the L2 tag
24 (01_1000)
Cache-inhibited
stores
Counts cache-inhibited stores written to bus. If store gathering is occurring, this
counts stores after gathering
25 (01_1001)
iL1 reloads
Counts 32-byte cache blocks loaded into the L1 instruction cache
26 (01_1010)
L2 data side misses Counts times a load or store miss from the data reload table missed in the L2 tag.
Does not include hits due to L1 data cache cast outs, cache operations, or
write-through stores.
27 (01_1011)
L2 instruction side
hits
L1 instruction cache cacheable misses that hit in the L2 cache
28 (01_1100)
L2 dL1 castout
misses
L1 data cache castouts to L2 that miss in the L2 cache
29 (01_1101)
L2 sectors castout
Counts 0, 1, 2, 3, or 4 for a given L2 sector cast out
30 (01_1110)
dst
L1 cache block
misses
Counts
dst
and
dstst
instructions that miss in the L1 data cache and queue in
the data reload table. This event counts the times the
dst
cache block fetch
precedes the demand load or store by enough cycles to access the L1 data
cache Trst.
31 (01_1111)
dst
stream 1 cache
block fetches
Counts
dst
stream 1 cache block fetches
32 (10_0000)
dst
cache block
fetches
dst
cache block fetchesIncludes all streams
33 (10_0001)
dst
s prematurely
cancelled
Incremented when a stream is cancelled due to branch speculation cancel,
inappropriate translation protection, or WIMG. Includes whenever a
dst
stream is
cancelled for any reason other than reaching the end of the stream, refresh, or
dss
. Can count 0, 1, 2, 3, or 4 at a time.
34 (10_0010)
VFPU traps
Counts VFPU traps. These conditions are generated from certain operations
when in JAVA-mode.
35 (10_0011)
dRLT snoop hits
Counts serviced snoop requests that also hit on a data reload table entry
36 (10_0100)
Hit-style modiTed
intervention
data-only transfers
Increments when
modiTed data is provided to the system interface using data
intervention (enhanced mode is enabled) due to a snoop
37 (10_0101)
Snoop hits
Counts serviced snoop requests that hit on one or more caches or buffers
38 (10_0110)
First speculative
branch predicted
correctly
Counts correctly predicted branches in the Trst speculative stream
Table 11-9. PMC2 EventsMMCR0[26D31] Select Encodings (Continued)
Number
Event
Description