yuikns / byte-unixbench Goto Github PK

What steps will reproduce the problem?
Run with the default command line ('./Run') on a machine with > 16 CPUs.

What is the expected output? What do you see instead?
Expected would be a run with a single serial run, followed by an N-way parallel 
run. Instead, it skips all of the tests because of a hardcoded limit of '16'.

What version of the product are you using? On what operating system?
v5.1.3 on Linux 2.6.32.

Please provide any additional information below.
Is there a reason for the hardcoded limit of 16-way parallelism? It seems 
reasonable to be able to match the number of processors detected in the system.

Q. What steps will reproduce the problem?
./Run

Q. What is the expected output? What do you see instead?
Usually compilers cannot find functions like sin cos etc without this

Q. What version of the product are you using? On what operating system?
Ubuntu 12.04

You can see the diff file as an attachment.

What steps will reproduce the problem?
1. configure 1 cpu to the system
2. run 1 copy, 2, 4, copies of whetstone
3. the throughput should be the same as there is only one cpu. The throughput 
scales linearly.

What is the expected output? What do you see instead?
Expected to use wall clock time for calculations instead it uses process time.

What version of the product are you using? On what operating system?
ubuntu 15.04 - latest unixbench

Please provide any additional information below.
I modified the compilation flag from UNIX to GTODay and it fixed the issue. 
GTODay uses wall clock time to calculate the results which is what is needed as 
using process time may result in undue advantages to processors having higher 
single threaded performance while lack in throughput mode ( multiple SMT copies 
). This does not also differentiate a configuration with higher number of cpus.

What steps will reproduce the problem?
1. when we are in the unixbench directory ,use ./Run arg ===>that's right
But,when we are not there ,such as: USAGE: /home/unixbench/Run arg

HERE the error message

even we change the Run file on 《system("cat 
\"${BINDIR}/unixbench.logo\"");》 
that's also can't run either.

maybe it is the <BINDIR> variable problem,but i am not used to perl

PLEASE  

3Q

unixbench tries to execute '3dinfo' command to get info about graphic system (I 
guess).
I found this program only in OpenSuse package 3ddiag and it looks that right 
now dumping expected info is commented out.
So, on most (all?) systems trying to run 3dinfo is pointless.
I think that 3dinfo should be replaced with something newer - maybe some part 
of 'glxinfo' output or 'hwinfo'.

What steps will reproduce the problem?
1. ./Run -i 1 int
Also reproducible here for the tests arithoh, short, int, long, float, double 

What is the expected output? What do you see instead?
I am expecting the normal result listing.


What version of the product are you using? On what operating system?
Version 5.1.3


Please provide any additional information below.

Console output:
make all
make[1]: Entering directory 
`/home/scosu/ptestsuite/suite/tests/unixbench/UnixBench'
Checking distribution of files
./pgms  exists
./src  exists
./testdir  exists
./tmp  exists
./results  exists
make[1]: Leaving directory 
`/home/scosu/ptestsuite/suite/tests/unixbench/UnixBench'
sh: 3dinfo: command not found
sh: runlevel: command not found

   #    #  #    #  #  #    #          #####   ######  #    #   ####   #    #
   #    #  ##   #  #   #  #           #    #  #       ##   #  #    #  #    #
   #    #  # #  #  #    ##            #####   #####   # #  #  #       ######
   #    #  #  # #  #    ##            #    #  #       #  # #  #       #    #
   #    #  #   ##  #   #  #           #    #  #       #   ##  #    #  #    #
    ####   #    #  #  #    #          #####   ######  #    #   ####   #    #

   Version 5.1.3                      Based on the Byte Magazine Unix Benchmark

   Multi-CPU version                  Version 5 revisions by Ian Smith,
                                      Sunnyvale, CA, USA
   January 13, 2011                   johantheghost at yahoo period com


1 x Arithmetic Test (int)  1Can't take log of 0 at ./Run line 935.

Hi,

I'm trying to get UnixBench running on AIX.
As make does not work on AIX, I used gmake.

Now, I'm getting the following error.
Checking distribution of files
./pgms  exists
./src  exists
./testdir  exists
./tmp  exists
./results  exists
gcc -o ./pgms/syscall -DTIME -Wall -pedantic -ansi -O2 -fomit-frame-pointer 
-fforce-addr -ffast-math -Wall ./src/syscall.c
In file included from ./src/syscall.c:31:
./src/timeit.c: In function 'wake_me':
./src/timeit.c:39: warning: implicit declaration of function 'alarm'
./src/syscall.c: In function 'main':
./src/syscall.c:67: warning: implicit declaration of function 'close'
./src/syscall.c:67: warning: implicit declaration of function 'dup'
./src/syscall.c:68: warning: implicit declaration of function 'getpid'
./src/syscall.c:69: warning: implicit declaration of function 'getuid'
./src/syscall.c:70: warning: implicit declaration of function 'umask'
./src/syscall.c:88: error: 'pid_t' undeclared (first use in this function)
./src/syscall.c:88: error: (Each undeclared identifier is reported only once
./src/syscall.c:88: error: for each function it appears in.)
./src/syscall.c:88: error: expected ';' before 'pid'
./src/syscall.c:89: error: 'pid' undeclared (first use in this function)
./src/syscall.c:93: warning: implicit declaration of function 'execl'
./src/syscall.c:97: warning: implicit declaration of function 'waitpid'
gmake: *** [pgms/syscall] Error 1

The system is AIX.

If someone could help, that'd be great.

Jean-Louis

This is really outdated and even detrimental to performance: -O2 
-fomit-frame-pointer -fforce-addr -ffast-math -Wall

A better combination will be:  -O2 -ffast-math -Wall -march=native

What steps will reproduce the problem?
1. Running the dc benchmark

The command dc < dc.dat is computed very fast on modern computers, less than a 
millisecond. So the dc benchmark is benchmarking fork and exec instead 
(looper.c).

It is easy to change, by increasing the precision of sqrt(2) calculation.

Thanks

What steps will reproduce the problem?
1. Run "perl Run" on OSX 10.9

What is the expected output? What do you see instead?

I expect UnixBench results.

Here's the output instead:

perl Run
make all
Checking distribution of files
./pgms  exists
./src  exists
./testdir  exists
./tmp  exists
./results  exists
Use of uninitialized value $coll in substitution (s///) at Run line 741.
Use of uninitialized value $coll in sprintf at Run line 742.
sh: 3dinfo: command not found
sh: runlevel: command not found

   #    #  #    #  #  #    #          #####   ######  #    #   ####   #    #
   #    #  ##   #  #   #  #           #    #  #       ##   #  #    #  #    #
   #    #  # #  #  #    ##            #####   #####   # #  #  #       ######
   #    #  #  # #  #    ##            #    #  #       #  # #  #       #    #
   #    #  #   ##  #   #  #           #    #  #       #   ##  #    #  #    #
    ####   #    #  #  #    #          #####   ######  #    #   ####   #    #

   Version 5.1.3                      Based on the Byte Magazine Unix Benchmark

   Multi-CPU version                  Version 5 revisions by Ian Smith,
                                      Sunnyvale, CA, USA
   January 13, 2011                   johantheghost at yahoo period com

Use of uninitialized value in printf at Run line 1369.
Use of uninitialized value in printf at Run line 1567.

1 x Dhrystone 2 using register variables  1 2 3 4 5 6 7 8 9 10

1 x Double-Precision Whetstone  1 2 3 4 5 6 7 8 9 10

1 x Execl Throughput  1 2 3

1 x File Copy 1024 bufsize 2000 maxblocks  1 2 3

1 x File Copy 256 bufsize 500 maxblocks  1 2 3

1 x File Copy 4096 bufsize 8000 maxblocks  1 2 3

1 x Pipe Throughput  1 2 3 4 5 6 7 8 9 10

1 x Pipe-based Context Switching  1
**********************************************
Run: "Pipe-based Context Switching": slave read failed: Invalid argument; 
aborting

What steps will reproduce the problem?
1. Run make -j 4 (anything > 3 will cause failure)
2. Observe errors about dhry_1.o or dhry_2.o files.

What is the expected output? What do you see instead?

Project should compile without error using parallel make. Instead, project 
fails to compile due to race conditions in the construction and deletion of  
the dhry_1.o and dhry_2.o files.

What version of the product are you using? On what operating system?

All versions as far back as 4.0.1 (at least, probably always). Linux.

Please provide any additional information below.

Attached is a patch file to change the dhry2reg target to compile its versions 
of the dhry_1.o and dhry_2.o files to dhry_1_reg.o and dhry_2_reg.o so that the 
dhry2 target doesn't overwrite them.

What steps will reproduce the problem?
1. Execute the test harness with ./Run
2. Notice that while the Pipe-based Context Switching test is running, it
fails with the error message: "slave write failed: Broken pipe; aborting"
3. This only happens on certain platforms, as I don't see this occur on all
hardware/software configurations. But I have seen it happen on certain
Ubuntu releases.

I'm currently seeing this with Ubuntu Karmic derivative with the 2.6.30
kernel, i686, running on an ASUS EEE PC with INtel Atom CPU N280 @ 1.66ghz.

What steps will reproduce the problem?
1. ./Run arthrimetic
2. ./Run -c 4 arthrimetic
3. ./Run -c 8 arthrimetic
4. ./Run -c 12 arthrimetic
5. ./Run -c 16 arthrimetic

What is the expected output? What do you see instead?

the system CPU has only 4 cores.

-c 4 ~ 16 should have result slightly less than 4x of single process test 
result.
but, the results are:
4: 3380.9       
8: 6756.2   
12: 10130.1 
16: 13504.2

What version of the product are you using? On what operating system?
unixbench 5.1.3
3.2.0-4-amd64 #1 SMP Debian 3.2.63-2 x86_64 GNU/Linux
one Intel XEON E3-1225

Please provide any additional information below.
also got same kind of result on a Apple MBP, with Core i7 3615, 4C8T

1: 959.8    
4: 3902.9   
8: 7139.1
12: 10673.6
16: 14252.6

What steps will reproduce the problem?
1. Download and run on a recent distro (Ubuntu 12, CentOS 6)

What is the expected output? What do you see instead?
Any tests that utilize arith.c (arithoh, long, short, etc) will return 0 for 
iter instead of the expected value.

What version of the product are you using? On what operating system?
5.1.3 on Ubuntu 12.04 and CentOS 6.

Please provide any additional information below.
This is trivially fixed by changing the definition of the iter variable in 
arith.c to volatile.

The existing code attempts to prevent the loop from being optimized away by 
feeding the result variable back in to the dumb_stuff function. That should be 
completely unnecessary if iter is defined as volatile.

What steps will reproduce the problem?
1. Download and decompress
2. Comment out graphics test per instructions in README
3. execute make
4. execute Run

What is the expected output? What do you see instead?
Quad-core G5 has 4 CPUs, but 0 CPUs are reported. Please see additional info 
below.

What version of the product are you using? On what operating system?
v5.1.2 on YDL 6.2 (2.6.29-3.ydl61.4)

Please provide any additional information below.

Output snippet:
========================================================================
   BYTE UNIX Benchmarks (Version 5.1.2)

   System: sorrel: GNU/Linux
   OS: GNU/Linux -- 2.6.29-3.ydl61.4 -- #1 SMP Mon Sep 7 14:50:27 PDT 2009
   Machine: ppc64 (ppc64)
   Language: en_US.utf8 (charmap="UTF-8", collate="UTF-8")
   09:51:37 up 18:57,  2 users,  load average: 0.22, 0.70, 0.70; runlevel 3

------------------------------------------------------------------------
Benchmark Run: Tue Nov 09 2010 09:51:37 - 10:19:32
0 CPUs in system; running 1 parallel copy of tests

Dhrystone 2 using register variables        5749777.3 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     1246.8 MWIPS (9.7 s, 7 samples)
Execl Throughput                               2021.4 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        190662.5 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           57739.1 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        474500.7 KBps  (30.0 s, 2 samples)
Pipe Throughput                              401805.7 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 120129.4 lps   (10.0 s, 7 samples)
Process Creation                               6058.6 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   3675.4 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   1057.4 lpm   (60.0 s, 2 samples)
System Call Overhead                         567026.7 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0    5749777.3    492.7
Double-Precision Whetstone                       55.0       1246.8    226.7
Execl Throughput                                 43.0       2021.4    470.1
File Copy 1024 bufsize 2000 maxblocks          3960.0     190662.5    481.5
File Copy 256 bufsize 500 maxblocks            1655.0      57739.1    348.9
File Copy 4096 bufsize 8000 maxblocks          5800.0     474500.7    818.1
Pipe Throughput                               12440.0     401805.7    323.0
Pipe-based Context Switching                   4000.0     120129.4    300.3
Process Creation                                126.0       6058.6    480.8
Shell Scripts (1 concurrent)                     42.4       3675.4    866.8
Shell Scripts (8 concurrent)                      6.0       1057.4   1762.3
System Call Overhead                          15000.0     567026.7    378.0
                                                                   ========
System Benchmarks Index Score                                         490.6

What steps will reproduce the problem?
1. fscopy test cases

What is the expected output? What do you see instead?

1. The test cases with a biggest file of 4096 buffer and 8000 blocks is ~ 32MB 
file size. When I run this on a Haswell server (Intel(R) Xeon(R) CPU E5-2690 
v3), this fits entirely in the L3 memory when I run only one guest. Even though 
the test case might be valid the enviroment or scenario where only one guest is 
running will give undue advantage as L3 is shared. This requires that the size 
of file be increased

2. I tried a cp command on ubuntu 15.04 on a file which contains only "0" as 
contents

read(3, "0\n", 131072)                  = 2
write(4, "0\n", 2)                      = 2

and I see that it uses a buffer size of 128k. I also noticed the same on ppc64 
version of ubuntu15.04

 So, the buffer size also needs to be increased.

What version of the product are you using? On what operating system?

ubuntu 5.1.3 / ubuntu 15.04 (ext4)

Please provide any additional information below.

In RHEL 7.0 (xfs) the cp command uses 64k buffer size.
So, it is necessary that the buffer and block size be increased.

I looked at 5.1.3 tarbal and was sad to see that it broke with the convenient 
name-version.tar.gz scheme (and a tarball containing name-version/ toplevel 
directory) but being some funky camelcase homegrown thing with a 
non-predictable directory inside.

Please read http://www.gnu.org/prep/standards/standards.html#Releases and don't 
change what you don't know, it might have been done properly already and 
distribution packagers like me won't become this grumpy.

Thanks for efforts though!

What steps will reproduce the problem?
1. Put some "middle" CPUs offline (example: cpus 1 and 2 on a 4 cpus system). 
This can be done using "echo 0 > /sys/devices/system/cpu/cpuX/online".
2. Run test


What is the expected output? What do you see instead?

Unixbench should count 2 CPUs but counts 4, this is due to getCpuInfo routine 
which uses CPU ID as array index.


What version of the product are you using? On what operating system?

UnixBench 5.1.3, will behave like this on all systems


Please provide any additional information below.

The routine is also not written properly, the "processor : 0" line is not 
parsed because of the "next if (!$prop || !$val);" line. Thankfully the $cpu is 
initialized to 0 so there is no functionnal issue there.

I ran into this issue by trying to run the test with less cores than the system 
really has, due to hyperthreading I had to disable core 4-7 and 12-15 (as 
opposed to disabling core 8-15)

Trying to run the UnixBench 5 on my ubuntu system, but i get the error 
[pgms/ubgears] error 1, i know this error is related to setting up the 
variables for Graphics test, but i really dont know how to go about it