Re: rthreads vs pthreads performance test
yeap. mysql compiled with libpthread.so.10.0(maybe 4.5 version) so. i did recompiled and just move libpthread.so.x.y -> librthread.so.x.y here result. (super-smack error and sysbench slow performance. hackbench error) * super-smack # ./super-smack -d mysql select-key.smack 10 100 Query Barrel Report for client smacker1 connect: max=4ms min=0ms avg= 2ms from 10 clients Query_type num_queries max_timemin_timeq_per_s select_index20000 0 9448.89 # ./super-smack -d mysql select-key.smack 10 1000 Error running query select * from http_auth where username = 'kwscxbmlsfgq13920':MySQL server has gone away super-smack: aborting on failed query Error running query select * from http_auth where username = 'ebxcipdukaud8400':MySQL server has gone away super-smack: aborting on failed query super-smack: Could not lock the pipe semaphore 1048577 errno = 22, error is Invalid argument super-smack: Could not lock the pipe semaphore 1048577 errno = 22, error is Invalid argument super-smack: Could not lock the pipe semaphore 1048577 errno = 22, error is Invalid argument super-smack: Could not lock the pipe semaphore 1048577 errno = 22, error is Invalid argument super-smack: Could not lock the pipe semaphore 1048577 errno = 22, error is Invalid argument super-smack: Could not lock the pipe semaphore 1048577 errno = 22, error is Invalid argument super-smack: Could not lock the pipe semaphore 1048577 errno = 22, error is Invalid argument Query Barrel Report for client smacker1 connect: max=0ms min=-1ms avg= 6ms from 1 clients Query_type num_queries max_timemin_timeq_per_s select_index20000 0 711.86 * sysbench (threads test is very slow perfomance.) # sysbench --num-threads=1024 --test=threads --thread-yields=100 --thread-locks=2 run sysbench v0.4.8: multi-threaded system evaluation benchmark Running the test with following options: Number of threads: 1024 Doing thread subsystem performance test Thread yields per test: 100 Locks used: 2 Threads started! Done. Test execution summary: total time: 63.4129s total number of events: 1 total time taken by event execution: 64381.3383 per-request statistics: min:0.0005s avg:6.4381s max:63.3723s approx. 95 percentile: 40.1140s Threads fairness: events (avg/stddev): 9.7656/95.47 execution time (avg/stddev): 62.8724/0.38 * hackbench (Segmentation Fault) - http://people.redhat.com/mingo/cfs-scheduler/tools/hackbench.c # gdb hackbench GNU gdb 6.3 Copyright 2004 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions. There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "i386-unknown-openbsd4.7"... (gdb) set args 2 thread 2 (gdb) r Starting program: /tmp/hackbench 2 thread 2 Running with 2*40 (== 80) tasks. Program received signal SIGSEGV, Segmentation fault. 0x04fee097 in _rthread_free_stack (stack=0x0) at /usr/src/lib/librthread/rthread_stack.c:85 85 munmap(stack->base, stack->len); (gdb) bt #0 0x04fee097 in _rthread_free_stack (stack=0x0) at /usr/src/lib/librthread/rthread_stack.c:85 #1 0x04fed8a0 in _rthread_reaper () at /usr/src/lib/librthread/rthread.c:203 #2 0x04fedac2 in pthread_create (threadp=0x31fbf321, attr=0xcfbe5a38, start_routine=0x31fbf321, arg=0x31fbf321) at /usr/src/lib/librthread/rthread.c:303 #3 0x1c000f44 in create_worker (ctx=0x850a60c0, func=0x1c000de0 ) at hackbench.c:160 #4 0x1c00106c in group (pth=0x87fc8c00, num_fds=20, ready_out=7, wakefd=8) at hackbench.c:210 #5 0x1c001319 in main (argc=0, argv=0xcfbe5b5c) at hackbench.c:280 (gdb) info reg eax0x31fbf321 838595361 ecx0x3c0032e0 1006645984 edx0x24feed10 620686608 ebx0x24fec29c 620675740 esp0xcfbe59b8 0xcfbe59b8 ebp0xcfbe59c8 0xcfbe59c8 esi0x0 0 edi0x24feb1dc 620671452 eip0x4fee0970x4fee097 eflags 0x10296 66198 cs 0x2b 43 ss 0x33 51 ds 0x33 51 es 0x33 51 fs 0x33 51 gs 0x33 51 2010/4/17 Chris Cappuccio : > You don't need to recompile mysql or other tools. Just move the librthreads shared object library (librthreads.so.x.y) in place of libpthreads.so.x.y. It's designed to work this way. > > Jung [moor...@gmail.com] wrote: >> oops. >> sorry. i did not re-compile mysql and other tool. (pthreads vs >> rthreads result are same) >> so, previous e-mail's performance test is fail. >> i'll update test tools and i'll send t
Re: rthreads vs pthreads performance test
You don't need to recompile mysql or other tools. Just move the librthreads shared object library (librthreads.so.x.y) in place of libpthreads.so.x.y. It's designed to work this way. Jung [moor...@gmail.com] wrote: > oops. > sorry. i did not re-compile mysql and other tool. (pthreads vs > rthreads result are same) > so, previous e-mail's performance test is fail. > i'll update test tools and i'll send that results again. > > thanks.
Re: rthreads vs pthreads performance test
oops. sorry. i did not re-compile mysql and other tool. (pthreads vs rthreads result are same) so, previous e-mail's performance test is fail. i'll update test tools and i'll send that results again. thanks. 2010/4/13 Ted Unangst : > On Tue, Apr 13, 2010 at 4:33 AM, Jung wrote: >> rthreads vs pthreads performance test >> >> at today, i did test recent rthreads performance. >> previous e-mail http://marc.info/?l=openbsd-tech&m=120634514721997 >> rthreads performance was poor. but recent rthreads performance good like below. >> >> some test result on Core 2 Duo + 4.7-Current(GENERIC.MP)+RTHREADS option > > It's unfortunately still early to be testing performance, but thanks > for the effort. If rthreads is really performing the same as > pthreads, that's quite good for now.
Re: rthreads vs pthreads performance test
On Tue, Apr 13, 2010 at 4:33 AM, Jung wrote: > rthreads vs pthreads performance test > > at today, i did test recent rthreads performance. > previous e-mail http://marc.info/?l=openbsd-tech&m=120634514721997 > rthreads performance was poor. but recent rthreads performance good like > below. > > some test result on Core 2 Duo + 4.7-Current(GENERIC.MP)+RTHREADS option It's unfortunately still early to be testing performance, but thanks for the effort. If rthreads is really performing the same as pthreads, that's quite good for now.
rthreads vs pthreads performance test
rthreads vs pthreads performance test at today, i did test recent rthreads performance. previous e-mail http://marc.info/?l=openbsd-tech&m=120634514721997 rthreads performance was poor. but recent rthreads performance good like below. some test result on Core 2 Duo + 4.7-Current(GENERIC.MP)+RTHREADS option * super-smack test - rthreads) # ./super-smack -d mysql select-key.smack 10 1 Query Barrel Report for client smacker1 connect: max=2ms min=0ms avg= 1ms from 10 clients Query_type num_queries max_timemin_timeq_per_s select_index20 0 0 13294.37 # ./super-smack -d mysql select-key.smack 20 1 Query Barrel Report for client smacker1 connect: max=5ms min=0ms avg= 3ms from 20 clients Query_type num_queries max_timemin_timeq_per_s select_index40 3 0 13150.13 # ./super-smack -d mysql update-select.smack 10 1 Query Barrel Report for client smacker connect: max=2ms min=0ms avg= 1ms from 10 clients Query_type num_queries max_timemin_timeq_per_s select_index10 0 0 5856.31 update_index10 1 0 5856.31 # ./super-smack -d mysql update-select.smack 20 1 Query Barrel Report for client smacker connect: max=6ms min=0ms avg= 4ms from 20 clients Query_type num_queries max_timemin_timeq_per_s select_index20 1 0 5880.56 update_index20 1 0 5880.56 - pthreads) # ./super-smack -d mysql select-key.smack 10 1 Query Barrel Report for client smacker1 connect: max=2ms min=0ms avg= 1ms from 10 clients Query_type num_queries max_timemin_timeq_per_s select_index20 0 0 13492.67 # ./super-smack -d mysql select-key.smack 20 1 Query Barrel Report for client smacker1 connect: max=5ms min=0ms avg= 3ms from 20 clients Query_type num_queries max_timemin_timeq_per_s select_index40 4 0 12328.78 # ./super-smack -d mysql update-select.smack 10 1 Query Barrel Report for client smacker connect: max=3ms min=0ms avg= 1ms from 10 clients Query_type num_queries max_timemin_timeq_per_s select_index10 0 0 5887.26 update_index10 0 0 5887.26 # ./super-smack -d mysql update-select.smack 20 1 Query Barrel Report for client smacker connect: max=5ms min=0ms avg= 3ms from 20 clients Query_type num_queries max_timemin_timeq_per_s select_index20 1 0 5769.90 update_index20 2 0 5769.90 --- * sysbench test sysbench --test=oltp --num-threads=2 --mysql-user=root --mysql-socket=/var/run/mysql/mysql.sock --max-time=60 --max-requests=0 --oltp-read-only=on run rthreads) - client threads(Transactions/sec) 2 (412.58 per sec.) 5 (419.72 per sec.) 10 (421.68 per sec.) 15 (419.20 per sec.) 20 (412.36 per sec.) pthreads) - client threads(Transactions/sec) 2 (410.84 per sec.) 5 (408.39 per sec.) 10 (420.40 per sec.) 15 (416.54 per sec.) 20 (413.13 per sec.) --- - Jung
