Re: testing ProcArrayLock patches

Robert Haas <robertmhaas@gmail.com> wrote:
> I was actually thinking it would be interesting to oprofile the
> read-only test; see if we can figure out where those slowdowns are
> coming from.
CPU: Intel Core/i7, speed 2262 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with
a unit mask of 0x00 (No unit mask) count 100000
samples  %        image name      symbol name
3124242   5.7137  postgres        s_lock
2555554   4.6737  postgres        AllocSetAlloc
2403412   4.3954  postgres        GetSnapshotData
1967132   3.5975  postgres        SearchCatCache
1872176   3.4239  postgres        base_yyparse
1327256   2.4273  postgres        hash_search_with_hash_value
1040131   1.9022  postgres        _bt_compare
1038976   1.9001  postgres        LWLockAcquire
817122    1.4944  postgres        MemoryContextAllocZeroAligned
738321    1.3503  postgres        core_yylex
622613    1.1386  postgres        MemoryContextAlloc
597054    1.0919  postgres        PinBuffer
556138    1.0171  postgres        ScanKeywordLookup
552318    1.0101  postgres        expression_tree_walker
494279    0.9039  postgres        LWLockRelease
488628    0.8936  postgres        hash_any
472906    0.8649  postgres        nocachegetattr
396482    0.7251  postgres        grouping_planner
382974    0.7004  postgres        LockAcquireExtended
375186    0.6861  postgres        AllocSetFree
375072    0.6859  postgres        ProcArrayLockRelease
373668    0.6834  postgres        new_list
365917    0.6692  postgres        fmgr_info_cxt_security
301398    0.5512  postgres        ProcArrayLockAcquire
300647    0.5498  postgres        LockReleaseAll
292073    0.5341  postgres        DirectFunctionCall1Coll
285745    0.5226  postgres        MemoryContextAllocZero
284684    0.5206  postgres        FunctionCall2Coll
282701    0.5170  postgres        SearchSysCache
max_connections = 100
max_pred_locks_per_transaction = 64
shared_buffers = 8GB
maintenance_work_mem = 1GB
checkpoint_segments = 300
checkpoint_timeout = 15min
checkpoint_completion_target = 0.9
wal_writer_delay = 20ms
seq_page_cost = 0.1
random_page_cost = 0.1
cpu_tuple_cost = 0.05
effective_cache_size = 40GB
default_transaction_isolation = '$iso'
pgbench -i -s 100
pgbench -S -M simple -T 300 -c 80 -j 80
transaction type: SELECT only
scaling factor: 100
query mode: simple
number of clients: 80
number of threads: 80
duration: 300 s
number of transactions actually processed: 104391011
tps = 347964.636256 (including connections establishing)
tps = 347976.389034 (excluding connections establishing)
vmstat 1 showed differently this time -- no clue why.
procs --------------memory------------- ---swap-- -----io----r  b   swpd    free      buff    cache   si   so    bi
bo---system---------cpu------   in      cs us sy id wa st
 
91  0   8196 4189436 203925700 52314492    0    0     0     032255 1522807 85 13  1  0  0
92  0   8196 4189404 203925700 52314492    0    0     0     032796 1525463 85 14  1  0  0
67  0   8196 4189404 203925700 52314488    0    0     0     032343 1527988 85 13  1  0  0
93  0   8196 4189404 203925700 52314488    0    0     0     032701 1535827 85 13  1  0  0
-Kevin