Penguin XEN: Benchmarks

Last update : Monday the 1st of March 2006
maintained by: Guillaume Thouvenin

Table of Contents [Back to main menu]

  1. Kernel compilation
  2. Kernbench
  3. LMBench
  4. SysBench
  5. Volano
  6. Glucas



Kernel compilation

Synopsis

We use the command /usr/bin/time sh -c "make oldconfig && make". All operating systems are running with 1Go of memory. The test is launched five times and we take the average.

The xen-unstable (changeset 9021) comes with a 2.6.16-rc5 kernel. The Linux-2.6.16-rc5 'vanilla' kernel is booted with mem=1G parameter.

The domU (called capablanca) has the following configuration [/etc/xen/capablanca]:

# cat /etc/xen/capablanca
kernel="/boot/vmlinuz-2.6.16-rc5-xenU"
memory=1024 
name="capablanca"
disk=['phy:sdc1,sdc1,w','phy:sdc5,sdc5,w','phy:sdc6,sdc6,w']
vif = [ '' ]
vcpus = 2
root="/dev/sdc1 ro"

The machine is an Intel Xeon 3.00Ghz processor with 4G of RAM. The domain 0 is running a Debian amd64 stable distribution. The two unprivileged domains used in the test are also running a Debian amd64 stable distribution. Hyper-threading is enabled.

Results

Following graphics are showing that domain 0 must be used only for managing other domains. If you stress it, performance can go down dramatically. They are also showing that time taken to compile a kernel under a XenLinux kernel is not so close than a vanilla Linux kernel especially if we look the system time. The performance loss is around -32.1% between system time on dom0 and vanilla Linux and -34.3% between domU and vanilla Linux.

We ran new test to see if the difference is not due to sEDF CPU scheduler which is not necessarily the best for every workload as suggested by Matt Ayres and BVT scheduler is really better for this kind of work.

One compilation by one.

The first test is the compilation of a 2.6.12 kernel by a 2.6.16-rc5 Linux kernel, a 2.6.16-rc5-xen0 XenLinux kernel and a 2.6.16-rc5-xenU XenLinux kernel. On dom0, compilation was done without other active domain. The test under domU was launched whereas there was no extra activity in dom0.

test1

Concurrent compilations between dom0 and domU

The second test is the compilation of a 2.6.12 kernel in parallel between a dom0 and domU. We run a 2.6.16-xen XenLinux. We don't use the 2.6.16-rc5 because we ran this test after the others and when we ran them, the new xen kernel was available.

test2

To see if the problem is linked to the scheduler's choice, we try the BVT scheduler and here is the result:

bvt_test

Concurrent compilations between domU and domU

The last test is always a compilation of a 2.6.12 kernel in parallel but this time we use two domU with a 2.6.16-rc5-xenU. The XenLinux is launched with default options that means, for example, that sEDF (a CPU scheduler) is well balanced between both domains.

test3

Influence of the sEDF (Earliest Deadline First) CPU scheduler

We test the influence of the sEDF scheduler. This time we use a vmlinuz-2.6.16-xen kernel with a xen-3.0-unstable hypervisor. We are running the following tests:

  1. We change the weight:
    guill@bazar:~$ sudo xm list
    Name                              ID Mem(MiB) VCPUs State  Time(s)
    Domain-0                           0      960     4 r-----   208.8
    capablanca                         2     1024     4 -b----   116.7
    morphy                             3     1024     4 -b----    67.0
    steinitz                           1     1024     4 -b----  1799.3
    guill@bazar:~$ sudo xm sched-sedf 1 -w 1
    guill@bazar:~$ sudo xm sched-sedf 2 -w 10
    guill@bazar:~$ sudo xm sched-sedf 3 -w 100
    guill@bazar:~$ sudo xm sched-sedf
    Name                              ID Period(ms) Slice(ms) Lat(ms) Extra Weight
    Domain-0                           0      20.0      15.0     0.0      1      0
    capablanca                         2     100.0       1.8     0.0      1     10
    morphy                             3     100.0      18.0     0.0      1    100
    steinitz                           1     100.0       0.2     0.0      1      1
    

sedf_test1

kernbench

Synopsis

http://freshmeat.net/projects/kernbench/

kernbench is a CPU throughput benchmark. It is designed to compare kernels on the same machine, or to compare hardware.

Tests are done on a bi-Xeon 3GHz with 1G of ram. Hyper-threading is enabled thus the machine has 4 CPU.

We use the following command line:

guill@bazar:~/src/linux-2.6.12$ ~/work/benchs/kernbench-0.30/kernbench -M
4 cpus found
Cleaning source tree...
Making defconfig...
Making dep if needed...
Caching kernel source in ram...
Half load is 2 jobs, changing to 3 as a kernel compile won't guarantee 2 jobs

Performing 5 runs of
make -j 3
make -j 16

Results

If we're looking at the compilation time, this test is showing the same thing than the previous kernel compilation test. The differences between a vanilla Linux and a XenLinux is not very big. This test offers something more that are data about percent CPU usage, context switches and sleeps. We can see that context switches are more important in XenLinux and this explains why it takes a little more time in XenLinux to compile a kernel compared to Linux.

Compilation of a 2.6.12 Linux kernel
2.6.16 2.6.16-xen0 2.6.16-xenU 2.6.16-xenU // 2.6.16-xenU
Average Half Load -j 3 Run:
Elapsed time 179.73 213.23 207.994 505.778 - 378.564
User Time 460.804 492.372 479.73 837.988 - 741.45
System Time 66.632 96.766 93.364 151.296 - 136.606
Percent CPU 292.8 275.8 275.2 205.2 - 238.2
Context Switches 9514.4 14251 14199.4 13598.6 - 14585
Sleeps 28326.2 23860.6 23737 25515 - 24842.4
Average Optimal -j 16 Load Run:
Elapsed time 164.604 190.616 186.448 314.206 - 353.596
User Time 562.824 609.74 595.486 922.934 - 1020.13
System Time 77.462 127.41 126.226 185.85 - 198.172
Percent CPU 388.2 386.2 386.6 360 - 346
Context Switches 25030 26886 26003.2 32273.6 - 33654.4
Sleeps 28903.6 27792.8 27442.2 28242 - 28356.6

During the execution of the test we executed a 'top' command and here is the output:

xentop - 11:32:10   Xen 3.0-unstable
3 domains: 3 running, 0 blocked, 0 paused, 0 crashed, 0 dying, 0 shutdown
Mem: 4193324k total, 3215268k used, 978056k free    CPUs: 4 @ 2992MHz
      NAME  STATE   CPU(sec) CPU(%)     MEM(k) MEM(%)  MAXMEM(k) MAXMEM(%) VCPUS NETS NETTX(k) NETRX(k) SSID
capablanca -----r       2719  107.1    1048320   25.0    1048576      25.0     4    1       25      673    0
  Domain-0 -----r      14115    1.4    1043640   24.9   no limit       n/a     4    8       94      744    0
    morphy -----r       1179  201.5    1048404   25.0    1048576      25.0     4    1       33      147    0

LMbench

Synopsis

http://lmbench.sourceforge.net/

It's a suite of simple, portable benchmarks in order to compare different systems performances. It incorporates bandwidth and latency benchmarks.

LMbench was used under linux-2.6.15-rc5 vanilla, linux-2.6.15-rc5-xen0 and linux-2.6.15-rc5-xenU. For XenLinux, we used the xen-unstable [ChangeSet: Mon Mar 6 23:50:30 2006 +0100 9156:583d01868d17]

LMbench was launched five times and we've made the average of those five runs.

Results

Work in progress...

                 L M B E N C H  3 . 0   S U M M A R Y
                 ------------------------------------
		 (Alpha software, do not distribute)

Basic system parameters
------------------------------------------------------------------------------
Host                 OS Description              Mhz  tlb  cache  mem   scal
                                                     pages line   par   load
                                                           bytes  
--------- ------------- ----------------------- ---- ----- ----- ------ ----
bazar      2.6.16-rc5          x86_64-linux-gnu 2991                       1
bazar        rc5-xen0          x86_64-linux-gnu 2933                       1
capablanca   rc5-xenU          x86_64-linux-gnu 2979                       1

Processor, Processes - times in microseconds - smaller is better
------------------------------------------------------------------------------
Host                 OS  Mhz null null      open slct sig  sig  fork exec sh  
                             call  I/O stat clos TCP  inst hndl proc proc proc
--------- ------------- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ----
bazar      2.6.16-rc5   2991 0.22 0.30 1.98 3.03 8.75 0.33 3.33 184. 603. 5934
bazar        rc5-xen0   2933 1.57 1.89 4.29 7.57 11.1 1.98 5.85 533. 1365 8687
capablanca   rc5-xenU   2979 1.56 1.91 4.08 7.20 10.7 1.94 5.63 517. 1384 4568

Basic integer operations - times in nanoseconds - smaller is better
-------------------------------------------------------------------
Host                 OS  intgr intgr  intgr  intgr  intgr  
                          bit   add    mul    span    mod   
--------- ------------- ------ ------ ------ ------ ------ 
bazar      2.6.16-rc5   0.3400 0.3400 3.3400   20.8   28.0
bazar        rc5-xen0   0.3400 0.3400 3.4400   21.3   28.9
capablanca   rc5-xenU   0.3400 0.3400 3.3600   20.8   28.2

Basic float operations - times in nanoseconds - smaller is better
-----------------------------------------------------------------
Host                 OS  float  float  float  float
                         add    mul    span    bogo
--------- ------------- ------ ------ ------ ------ 
bazar      2.6.16-rc5   1.6700 2.3400   10.8   10.7
bazar        rc5-xen0   1.7200 2.4200   11.1   11.0
capablanca   rc5-xenU   1.6800 2.3500   10.8   10.8

Basic double operations - times in nanoseconds - smaller is better
------------------------------------------------------------------
Host                 OS  double double double double
                         add    mul    div    bogo
--------- ------------- ------  ------ ------ ------ 
bazar      2.6.16-rc5   1.6750 2.3400   17.9   13.4
bazar        rc5-xen0   1.7200 2.4100   18.3   13.8
capablanca   rc5-xenU   1.6800 2.3500   17.9   13.4

Context switching - times in microseconds - smaller is better
-------------------------------------------------------------------------
Host                 OS  2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
                         ctxsw  ctxsw  ctxsw ctxsw  ctxsw   ctxsw   ctxsw
--------- ------------- ------ ------ ------ ------ ------ ------- -------
bazar      2.6.16-rc5   5.1500 5.2400 5.5900 5.4500    9.6 6.50000    20.4
bazar        rc5-xen0   4.2000 4.3025 4.3150 7.1000   12.3 6.87750    23.2
capablanca   rc5-xenU   5.5325 4.2425 4.3025 5.1575   10.3 6.55250    21.1

*Local* Communication latencies in microseconds - smaller is better
---------------------------------------------------------------------
Host                 OS 2p/0K  Pipe AF     UDP  RPC/   TCP  RPC/ TCP
                        ctxsw       UNIX         UDP         TCP conn
--------- ------------- ----- ----- ---- ----- ----- ----- ----- ----
bazar      2.6.16-rc5   5.150  12.6 24.8  19.0  28.3  24.3  36.8  50.
bazar        rc5-xen0   4.200  29.3 31.2  42.5  54.9  47.7  65.9  69.

capablanca   rc5-xenU   4.090  18.4 29.0  26.0  38.4  43.6  62.9  66.
capablanca   rc5-xenU   9.960  30.2 29.4  39.5  53.3  45.0  56.1  66.
capablanca   rc5-xenU   4.050  30.1 29.7  39.4        45.2        66.
capablanca   rc5-xenU   4.030  18.2 29.5  39.5  53.0  44.8        69.

File & VM system latencies in microseconds - smaller is better
-------------------------------------------------------------------------------
Host                 OS   0K File      10K File     Mmap    Prot   Page   100fd
                        Create Delete Create Delete Latency Fault  Fault  selct
--------- ------------- ------ ------ ------ ------ ------- ----- ------- -----
bazar      2.6.16-rc5     14.9   12.3   48.9   26.8  6345.0 0.878 1.94970 8.058
bazar        rc5-xen0     23.3   17.8   68.5   33.6   18.0K 2.146 5.05430 10.00
capablanca   rc5-xenU     21.1   16.7   58.1   31.8   17.7K 2.170 4.96762 9.666

*Local* Communication bandwidths in MB/s - bigger is better
-----------------------------------------------------------------------------
Host                OS  Pipe AF    TCP  File   Mmap  Bcopy  Bcopy  Mem   Mem
                             UNIX      reread reread (libc) (hand) read write
--------- ------------- ---- ---- ---- ------ ------ ------ ------ ---- -----
bazar      2.6.16-rc5   1276 2575 1031 2332.9 3554.9  970.8  902.0 3339 1521.
bazar        rc5-xen0   1177 2139 820. 1893.4 3270.4  951.4  884.5 3069 1454.
capablanca   rc5-xenU   1219 2203 939. 2021.2 3503.7  962.3  896.0 3294 1499.

Memory latencies in nanoseconds - smaller is better
    (WARNING - may not be correct, check graphs)
------------------------------------------------------------------------------
Host                 OS   Mhz   L1 $   L2 $    Main mem    Rand mem    Guesses
--------- -------------   ---   ----   ----    --------    --------    -------
bazar      2.6.16-rc5    2991 1.3400 9.2167   55.8       234.6
bazar        rc5-xen0    2933 1.3657 9.5557   61.9       259.0
capablanca   rc5-xenU    2979 1.3470 9.4300   56.5       240.3

We notice that during the test of 2.6.16-rc5-xen0, some messages are displayed on the console. They look like:

...
Timer ISR/0: Time went backwards: delta=20003146 cpu_delta=-459996854 
shadow=389000341715 off=480933330 processed=389460053726 
cpu_processed=389940053726
 0: 389940053726
 1: 388940053726
printk: 147 messages suppressed.
Timer ISR/0: Time went backwards: delta=50003080 cpu_delta=-419996920 
shadow=394000366851 off=510930137 processed=394460053726 
cpu_processed=394930053726
 0: 394930053726
 1: 393960053726
printk: 205 messages suppressed.
...

SysBench

Synopsis

http://sysbench.sourceforge.net/

SysBench is a modular, cross-platform and multi-threaded benchmark tool for evaluating OS parameters that are important for a system running a database under intensive load.

The machine is a single Intel Xeon 3.00Ghz processor with hyper-threading enabled.

CPU

The cpu is one of the most simple benchmarks in SysBench. In this mode each request consists in calculation of prime numbers up to a value specified by the --cpu-max-primes option. All calculations are performed using 64-bit integers.

Each thread executes the requests concurrently until either the total number of requests or the total execution time exceeds the limits specified with the common command line options.

Test bed:

Kernel is a 2.6.16-rc5
MemTotal:      1028800 kB

processor       : 0
vendor_id       : GenuineIntel
cpu family      : 15
model           : 4
model name      :                   Intel(R) Xeon(TM) CPU 3.00GHz
cpu MHz         : 2992.610
cache size      : 1024 KB
bogomips        : 5994.15

processor       : 1
vendor_id       : GenuineIntel
cpu family      : 15
model           : 4
model name      :                   Intel(R) Xeon(TM) CPU 3.00GHz
cpu MHz         : 2992.610
cache size      : 1024 KB
bogomips        : 5985.35
Kernel is a 2.6.16-rc5-xen0

MemTotal:      1043456 kB

processor       : 0
vendor_id       : GenuineIntel
cpu family      : 15
model           : 4
model name      :                   Intel(R) Xeon(TM) CPU 3.00GHz
cpu MHz         : 2992.496
cache size      : 1024 KB
bogomips        : 5987.86

processor       : 1
vendor_id       : GenuineIntel
cpu family      : 15
model           : 4
model name      :                   Intel(R) Xeon(TM) CPU 3.00GHz
stepping        : 1
cpu MHz         : 2992.496
cache size      : 1024 KB
bogomips        : 5987.86
Kernel is a 2.6.16-rc5-xenU

MemTotal:      1048576 kB

processor       : 0
vendor_id       : GenuineIntel
cpu family      : 15
model           : 4
model name      :                   Intel(R) Xeon(TM) CPU 3.00GHz
cpu MHz         : 2992.492
cache size      : 1024 KB
bogomips        : 5987.46

processor       : 1
vendor_id       : GenuineIntel
cpu family      : 15
model           : 4
model name      :                   Intel(R) Xeon(TM) CPU 3.00GHz
cpu MHz         : 2992.492
bogomips        : 5987.46

We notice that during the test of 2.6.16-rc5-xen0 and 2.6.16-rc5-xenU, some messages are displayed on the console. They look like:

...
Timer ISR/0: Time went backwards: delta=2102697 cpu_delta=-7897303
shadow=19152011999583 off=41398996 processed=19152050064532 
cpu_processed=19152060064532
 0: 19152060064532
 1: 19152050064532

Results

Work in progress...

# sysbench --test=cpu --cpu-max-prime=20000 run
sysbench v0.4.5: multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 20000

2.6.16-rc5 2.6.16-rc5-xen0 2.6.16-rc5-xenU
Test execution summary:
total time 66.2819s 68.0954s 66.9405s
total number of events: 10000 10000 10000
total time taken by event execution: 66.2756 68.0725 66.9191
per-request statistics:
min:0.0066s
avg:0.0066s
max:0.0071s
approx. 95 percentile:0.0067s
min:0.0066s
avg:0.0068s
max:0.0187s
approx. 95 percentile:0.0071s
min:0.0066s
avg:0.0067s
max:0.0097s
approx. 95 percentile:0.0071s
Threads fairness:
events (avg/stddev): 1000.0000/0.00 1000.0000/0.00 1000.0000/0.00
execution time (avg/stddev): 66.2756/0.80 68.0725/0.00 66.9191/0.00

Threads

This test mode was written to benchmark scheduler performance, more specifically the cases when a scheduler has a large number of threads competing for some set of mutexes.

We test a Xen version 3.0-unstable (guill@frec.bull.fr) (gcc version 3.3.5 (Debian 1:3.3.5-13)) Fri Mar 17 09:11:15 CET 2006 Latest ChangeSet: Thu Mar 16 12:31:16 2006 +0100 9255:00aba64e034e.

The command used is:

guill@capablanca:~$ sysbench --num-threads=64 --test=threads --thread-yields=100  --thread-locks=2 run
sysbench v0.4.5:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 64

Doing thread subsystem performance test
Thread yields per test: 100 Locks used: 2
Threads started!
Done.
1 VCPU 2 VCPUS 3 VCPUS 4 VCPUS 5 VCPUS 6 VCPUS 7 VCPUS 8 VCPUS
Test execution summary:
total time 10.2517s 9.4455s 6.2093s 5.7713s 3.2753s 3.5179s 3.2911s 4.0511s
total number of events: 10000 10000 10000 10000 10000 10000 10000 10000
total time taken by event execution: 654.1123 602.9263 396.0599 368.2627 208.2764 224.2287 209.6522 258.1759
per-request statistics:
min:0.0020s
avg:0.0654s
max:4.7625s
approx. 95 percentile: 0.3888s
min:0.0007s
avg:0.0603s
max:3.9564s
approx. 95 percentile: 0.3924s
min:0.0002s
avg:0.0396s
max:1.8567s
approx. 95 percentile: 0.2565s
min:0.0002s
avg:0.0368s
max:2.1267s
approx. 95 percentile: 0.1650s
min:0.0002s
avg:0.0208s
max:0.9850s
approx. 95 percentile: 0.1185s
min:0.0002s
avg:0.0224s
max:0.6630s
approx. 95 percentile: 0.1139s
min:0.0002s
avg:0.0210s
max:1.0656s
approx. 95 percentile: 0.0952s
min:0.0002s
avg:0.0258s
max:1.3062s
approx. 95 percentile: 0.1190s
Threads fairness:
events (avg/stddev): 156.2500/89.77 156.2500/65.80 156.2500/48.90 156.2500/49.28 156.2500/41.98 156.2500/36.97 156.2500/43.28 156.2500/41.40
execution time (avg/stddev): 10.2205/0.02 9.4207/0.01 6.1884/0.01 5.7541/0.01 3.2543/0.01 3.5036/0.01 3.2758/0.00 4.0340/0.01

Here is a graph extracted from the previous results that is showing there is a optimum number of virtual CPU.

sysbench_threads

We also extracted per-request statistics. We used a Y-axis logarithm scale because minimum is very low and we wanted to show the global form of the graph.

per-request-stat

Volano

Synopsis

http://www.volano.com/report/index.html

VolanoMark is a pure Java server benchmark characterized by long-lasting network connections and high thread counts.

First we test the domain0 alone which runs a 2.6.16-xen0 kernel. The machine is a bi-Xeon. We allow 1G of RAM to the domain0.

Results

Work in progress...

guill@bazar:~/work/benchs/volano/bench$ java -Xms8m -Xmx64m COM.volano.Main &
[1] 3728
guill@bazar:~/work/benchs/volano/bench$ VolanoChat(TM) Server Version 2.1.2
Copyright (C) 1996-1999 Volano LLC.  All rights reserved.
Loading server properties from /home/guill/work/benchs/volano/bench/properties.txt.
bazar.frec.bull.fr (172.16.109.122) VolanoChatPro - unlimited connections.

guill@bazar:~/work/benchs/volano/bench$ java -Xms8m -Xmx64m COM.volano.Mark -count 100

VolanoMark version = 2.1.2
Messages sent      = 20000
Messages received  = 380000
Total messages     = 400000
Elapsed time       = 19.624 seconds
Average throughput = 20383 messages per second

guill@bazar:~/work/benchs/volano/bench$ java -Xms8m -Xmx64m COM.volano.Mark -count 500

VolanoMark version = 2.1.2
Messages sent      = 100000
Messages received  = 1900000
Total messages     = 2000000
Elapsed time       = 111.62501 seconds
Average throughput = 17917 messages per second

guill@bazar:~/work/benchs/volano/bench$ java -Xms8m -Xmx64m COM.volano.Mark -count 1000

VolanoMark version = 2.1.2
Messages sent      = 200000
Messages received  = 3800000
Total messages     = 4000000
Elapsed time       = 235.97601 seconds
Average throughput = 16951 messages per second

Glucas

Synopsis

http://glucas.sourceforge.net/glucas/index.html

SELFTEST INFORMATION
Host: bazar.
OS: Linux. Release: 2.6.16. Version: #1 SMP Wed Mar 22 14:10:08 CET 2006
Machine: x86_64
Glucas 2.9.2-20051226 (experimental)
-DY_AVAL=4 -DY_MEM_THRESHOLD=8192 -DY_BLOCKSIZE=128 -DY_SHIFT=6
-DY_TARGET=0 -DY_USE_SSE2 -DY_VECTORIZE
Selftest 21 (2048 K FFT-runlength). 100 iterations for M36700159...
[Fri Mar 24 08:21:21 2006]
Going to work with exponent 36700159
Starting from iteration 1. Exponent 36700159. Initial shift 18825244.
Iter.  10 ( 10.00%), Err= 0.000, 1.35 user 100% CPU (0.1350 sec/iter).
Iter.  20 ( 20.00%), Err= 0.005, 1.34 user 100% CPU (0.1345 sec/iter).
Iter.  30 ( 30.00%), Err= 0.086, 1.34 user 100% CPU (0.1345 sec/iter).
Iter.  40 ( 40.00%), Err= 0.094, 1.34 user 100% CPU (0.1346 sec/iter).
Iter.  50 ( 50.00%), Err= 0.086, 1.34 user 100% CPU (0.1345 sec/iter).
Iter.  60 ( 60.00%), Err= 0.000, 1.29 user 100% CPU (0.1286 sec/iter).
Iter.  70 ( 70.00%), Err= 0.000, 1.28 user 100% CPU (0.1286 sec/iter).
Iter.  80 ( 80.00%), Err= 0.000, 1.29 user 100% CPU (0.1287 sec/iter).
Iter.  90 ( 90.00%), Err= 0.000, 1.29 user 100% CPU (0.1289 sec/iter).
Iter. 100 (100.00%), Err= 0.000, 1.37 user 100% CPU (0.1370 sec/iter).
[Fri Mar 24 08:21:34 2006]
M36700159. Iteration 100. Res64: 3A368A8683127C76. Glucas 2.9.2 selftest
[Fri Mar 24 08:21:34 2006]
SELFTEST INFORMATION
Host: bazar.
OS: Linux. Release: 2.6.16-xen0. Version: #2 SMP Wed Mar 22 09:42:26 CET 2006
Machine: x86_64
Glucas 2.9.2-20051226 (experimental)
-DY_AVAL=4 -DY_MEM_THRESHOLD=8192 -DY_BLOCKSIZE=128 -DY_SHIFT=6
-DY_TARGET=0 -DY_USE_SSE2 -DY_VECTORIZE
Selftest 21 (2048 K FFT-runlength). 100 iterations for M36700159...
[Fri Mar 24 08:07:37 2006]
Going to work with exponent 36700159
Starting from iteration 1. Exponent 36700159. Initial shift 7363321.
Iter.  10 ( 10.00%), Err= 0.000, 1.39 user 100% CPU (0.1392 sec/iter).
Iter.  20 ( 20.00%), Err= 0.006, 1.39 user 100% CPU (0.1389 sec/iter).
Iter.  30 ( 30.00%), Err= 0.094, 1.39 user 100% CPU (0.1387 sec/iter).
Iter.  40 ( 40.00%), Err= 0.102, 1.38 user  99% CPU (0.1388 sec/iter).
Iter.  50 ( 50.00%), Err= 0.102, 1.39 user 100% CPU (0.1389 sec/iter).
Iter.  60 ( 60.00%), Err= 0.000, 1.33 user 100% CPU (0.1330 sec/iter).
Iter.  70 ( 70.00%), Err= 0.000, 1.33 user 100% CPU (0.1331 sec/iter).
Iter.  80 ( 80.00%), Err= 0.000, 1.34 user 100% CPU (0.1331 sec/iter).
Iter.  90 ( 90.00%), Err= 0.000, 1.33 user 100% CPU (0.1330 sec/iter).
Iter. 100 (100.00%), Err= 0.000, 1.41 user  99% CPU (0.1418 sec/iter).
[Fri Mar 24 08:07:51 2006]
M36700159. Iteration 100. Res64: 3A368A8683127C76. Glucas 2.9.2 selftest
[Fri Mar 24 08:07:51 2006]
SELFTEST INFORMATION
Host: capablanca.
OS: Linux. Release: 2.6.16-xenU. Version: #1 SMP Wed Mar 22 09:25:43 CET 2006
Machine: x86_64
Glucas 2.9.2-20051226 (experimental)
-DY_AVAL=4 -DY_MEM_THRESHOLD=8192 -DY_BLOCKSIZE=128 -DY_SHIFT=6
-DY_TARGET=0 -DY_USE_SSE2 -DY_VECTORIZE
Selftest 21 (2048 K FFT-runlength). 100 iterations for M36700159...
[Fri Mar 24 08:09:07 2006]
Going to work with exponent 36700159
Starting from iteration 1. Exponent 36700159. Initial shift 15795012.
Iter.  10 ( 10.00%), Err= 0.000, 1.36 user  99% CPU (0.1367 sec/iter).
Iter.  20 ( 20.00%), Err= 0.005, 1.37 user 100% CPU (0.1364 sec/iter).
Iter.  30 ( 30.00%), Err= 0.094, 1.36 user 100% CPU (0.1365 sec/iter).
Iter.  40 ( 40.00%), Err= 0.102, 1.37 user 100% CPU (0.1365 sec/iter).
Iter.  50 ( 50.00%), Err= 0.086, 1.36 user 100% CPU (0.1361 sec/iter).
Iter.  60 ( 60.00%), Err= 0.000, 1.31 user 100% CPU (0.1307 sec/iter).
Iter.  70 ( 70.00%), Err= 0.000, 1.30 user 100% CPU (0.1306 sec/iter).
Iter.  80 ( 80.00%), Err= 0.000, 1.31 user 100% CPU (0.1304 sec/iter).
Iter.  90 ( 90.00%), Err= 0.000, 1.30 user 100% CPU (0.1306 sec/iter).
Iter. 100 (100.00%), Err= 0.000, 1.39 user 100% CPU (0.1389 sec/iter).
[Fri Mar 24 08:09:20 2006]
M36700159. Iteration 100. Res64: 3A368A8683127C76. Glucas 2.9.2 selftest
[Fri Mar 24 08:09:20 2006]

Valid XHTML 1.0 Strict