[Gems-users] Questions about "Other_GETS" in .sm file

[Jerry Lin <cosjerry@xxxxxxxxx>]

Hi all,
I am a little confused about the event "Other_GETS" in MOSI_SMP_bcast-cache.sm.
"Other_GETS" is triggered when other processors issue "GETS" request.
But does it check the address in the snooped message to see if the address is
now in the cache of snooping processor?
Or whenever a processor snoop a "GETS" request, it will trigger a "Other_GETS"
event no matter what the address is.

By the way, when I simulate application using MOSI_SMP_bcast protocol.
The "Transitions" part in ruby stats shows that only when the current
state is "NP", the events "Other_GETS", "Other_GETX" will be triggered.
Why other state, like "I", "S", won't trigger "Other_GETS" event?

Thanks for helping.

---------------- Selected parts of the stats file --------------------

Chip Stats
----------

 --- L1Cache ---
 - Event Counts -
Load  32123
Ifetch  40578
Store  18356
L1_to_L2  87103
L2_to_L1D  27518
L2_to_L1I  32080
L2_Replacement  1
Own_GETS  11898
Own_GET_INSTR  8498
Own_GETX  11202
Own_PUTX  1
Other_GETS  35694
Other_GET_INSTR  25494
Other_GETX  33606
Other_PUTX  0
Data  31598

 - Transitions -
NP  Load  11898
NP  Ifetch  8498
NP  Store  9156
NP  Other_GETS  35694
NP  Other_GET_INSTR  25494
NP  Other_GETX  33606
NP  Other_PUTX  0 <--

I  Load  0 <--
I  Ifetch  0 <--
I  Store  0 <--
I  L1_to_L2  0 <--
I  L2_to_L1D  0 <--
I  L2_to_L1I  0 <--
I  L2_Replacement  0 <--
I  Other_GETS  0 <--
I  Other_GET_INSTR  0 <--
I  Other_GETX  0 <--
I  Other_PUTX  0 <--

S  Load  10929
S  Ifetch  31946
S  Store  2046
S  L1_to_L2  59719
S  L2_to_L1D  11068
S  L2_to_L1I  31946
S  L2_Replacement  0 <--
S  Other_GETS  0 <--
S  Other_GET_INSTR  0 <--
S  Other_GETX  0 <--
S  Other_PUTX  0 <--