[Gems-users] Running bayes with high contention

[sjafri@xxxxxxxxxx]

Hi,

I tried running bayes on unmodified log-tm-SE with eager CD and eager VM with
perfect filters.


During the second phase of execution, the benchmark stopped making progress and
I saw the following output on screen:

 12750590 5 [5,0] TRAP TO HANDLER: TID: 4 TRAP_TYPE 5 TRAP ADDRESS 0x0
NUM_RETRIES 0 LOG_SIZE 0 XACT_LEVEL 0 XACT_LOWEST_CONFLICT_LEVEL 1 Handler
Address = [0x1aff0, line 0x1afc0] PC = [0x1007050, line 0x1007040]
 12761695 5 [5,0] CONTINUE EXECUTION CALLBACK New PC:[0x16234, line 0x16200]
 14258197 3 [3,0] TRAP TO HANDLER: TID: 2 TRAP_TYPE 5 TRAP ADDRESS 0x0
NUM_RETRIES 0 LOG_SIZE 0 XACT_LEVEL 0 XACT_LOWEST_CONFLICT_LEVEL 1 Handler
Address = [0x1aff0, line 0x1afc0] PC = [0x1007050, line 0x1007040]
 14263650 7 [7,0] TRAP TO HANDLER: TID: 6 TRAP_TYPE 5 TRAP ADDRESS 0x0
NUM_RETRIES 0 LOG_SIZE 2856 XACT_LEVEL 1 XACT_LOWEST_CONFLICT_LEVEL 1 Handler
Address = [0x1aff0, line 0x1afc0] PC = [0x1007050, line 0x1007040]
 14277934 7 [7,0] CONTINUE EXECUTION CALLBACK New PC:[0x18af8, line 0x18ac0]
 14293701 3 [3,0] CONTINUE EXECUTION CALLBACK New PC:[0x162f0, line 0x162c0]
 15013740 1 [1,0] TRAP TO HANDLER: TID: 7 TRAP_TYPE 5 TRAP ADDRESS 0x0
NUM_RETRIES 0 LOG_SIZE 0 XACT_LEVEL 0 XACT_LOWEST_CONFLICT_LEVEL 1 Handler
Address = [0x1aff0, line 0x1afc0] PC = [0x1007050, line 0x1007040]
 15021946 1 [1,0] CONTINUE EXECUTION CALLBACK New PC:[0x16234, line 0x16200]
[cpu2] v:0x000000000105b744 p:0x0000000002c5b744  be,pn %icc, 0x105b764


Reading the trace I saw that 7 out of 8 processors were executing OS code
(LWP_PARK) while proc 6 seemed to be looping forever in a user level code. Both
log-tm-SE and bayes were completely unmodified


Hope some one can help me

Ali