Port-arm archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: NetBSD on ODROID-C1



I´ve checked the CPU frequency first. Both NetBSD and Raspian boot
with 700 MHz CPU clock. I changed that to 900 MHz for both and got the
following MD5 results:

NetBSD @ 900 MHz:   ~95.000
Linux @ 900 MHz:       ~440.000 (varying from 400.000 - 470.000 for
whatever reason)

I´ve also made a compile test - I compiled the very same software on
both platforms with 900 MHz CPU clock.

NetBSD @ 900 MHz:  6 min real (5m25sec. user, 27sec. sys)
Linux @ 900 MHz:      3m43s. real (3m32sec. user, 6.75sec. sys)

I then checked with bytebench (from pkgsrc) and got the following figures:

============================================================


  BYTE UNIX Benchmarks (Version 4.1.0)
  System -- rpi
  Start Benchmark Run: Wed Feb  4 11:54:35 JST 2015
   1 interactive users.
  11:54AM  up 9 mins, 1 user, load averages: 0.37, 0.37, 0.21
  -r-xr-xr-x  1 root  wheel  176971 Feb  2 02:21 /bin/sh
  /bin/sh: ELF 32-bit LSB executable, ARM, EABI5 version 1 (SYSV),
dynamically linked, interpreter /libexec/ld.elf_so, for NetBSD 7.99.4,
compiled for: earmv6hf, not stripped
  /dev/ld0a       7325933    2929537    4030100  42% /
Dhrystone 2 using register variables     965457.3 lps   (10.0 secs, 10 samples)
Double-Precision Whetstone                  326.3 MWIPS (10.0 secs, 10 samples)
System Call Overhead                      37616.7 lps   (10.0 secs, 10 samples)
Pipe Throughput                           72989.9 lps   (10.0 secs, 10 samples)
Pipe-based Context Switching              14421.9 lps   (10.0 secs, 10 samples)
Process Creation                            186.9 lps   (30.1 secs, 3 samples)
Execl Throughput                             74.3 lps   (29.8 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks     37496.0 KBps  (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks    63483.0 KBps  (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks     22027.0 KBps  (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks       15247.0 KBps  (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks      18122.0 KBps  (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks        7128.0 KBps  (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks     60878.0 KBps  (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks   163804.0 KBps  (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks     23597.0 KBps  (30.0 secs, 3 samples)
Shell Scripts (1 concurrent)                221.0 lpm   (60.2 secs, 3 samples)
Shell Scripts (8 concurrent)                 29.9 lpm   (60.2 secs, 3 samples)
Shell Scripts (16 concurrent)                15.0 lpm   (60.2 secs, 3 samples)
Arithmetic Test (type = short)                1.0 lps   (0.0 secs, 3 samples)
Arithmetic Test (type = int)                  1.0 lps   (0.0 secs, 3 samples)
Arithmetic Test (type = long)                 1.0 lps   (0.0 secs, 3 samples)
Arithmetic Test (type = float)                1.0 lps   (0.0 secs, 3 samples)
Arithmetic Test (type = double)               1.0 lps   (0.0 secs, 3 samples)
Arithoh                                       1.0 lps   (0.0 secs, 3 samples)
C Compiler Throughput                        48.9 lpm   (60.2 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places           3041.5 lpm   (30.1 secs, 3 samples)
Recursion Test--Tower of Hanoi            12167.7 lps   (20.0 secs, 3 samples)


                     INDEX VALUES
TEST                                        BASELINE     RESULT      INDEX

Dhrystone 2 using register variables        116700.0   965457.3       82.7
Double-Precision Whetstone                      55.0      326.3       59.3
Execl Throughput                                43.0       74.3       17.3
File Copy 1024 bufsize 2000 maxblocks         3960.0    22027.0       55.6
File Copy 256 bufsize 500 maxblocks           1655.0     7128.0       43.1
File Copy 4096 bufsize 8000 maxblocks         5800.0    23597.0       40.7
Pipe Throughput                              12440.0    72989.9       58.7
Pipe-based Context Switching                  4000.0    14421.9       36.1
Process Creation                               126.0      186.9       14.8
Shell Scripts (8 concurrent)                     6.0       29.9       49.8
System Call Overhead                         15000.0    37616.7       25.1
                                                                 =========
     FINAL SCORE                                                      39.1
============================================================

I´ve not run this benchmark on Linux yet, but I found these figures on
the internet:


========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)

   System: ExperiPi: GNU/Linux
   OS: GNU/Linux -- 3.12.35+ -- #730 PREEMPT Fri Dec 19 18:31:24 GMT 2014
   Machine: armv6l (unknown)
   Language: en_US.utf8 (charmap="ANSI_X3.4-1968", collate="ANSI_X3.4-1968")
   CPU 0: ARMv6-compatible processor rev 7 (v6l) (0.0 bogomips)

   18:28:40 up 1 min,  1 user,  load average: 1.27, 0.56, 0.21; runlevel 2

------------------------------------------------------------------------
Benchmark Run: Di Feb 10 2015 18:28:40 - 18:57:14
1 CPU in system; running 1 parallel copy of tests

Dhrystone 2 using register variables        1883098.1 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                      306.6 MWIPS (9.9 s, 7 samples)
Execl Throughput                                236.6 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks         40507.0 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           12736.3 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks         97998.2 KBps  (30.0 s, 2 samples)
Pipe Throughput                              167411.7 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                  20978.9 lps   (10.0 s, 7 samples)
Process Creation                                725.4 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                    443.0 lpm   (60.1 s, 2 samples)
Shell Scripts (8 concurrent)                     56.4 lpm   (60.7 s, 2 samples)
System Call Overhead                         431022.4 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0    1883098.1    161.4
Double-Precision Whetstone                       55.0        306.6     55.8
Execl Throughput                                 43.0        236.6     55.0
File Copy 1024 bufsize 2000 maxblocks          3960.0      40507.0    102.3
File Copy 256 bufsize 500 maxblocks            1655.0      12736.3     77.0
File Copy 4096 bufsize 8000 maxblocks          5800.0      97998.2    169.0
Pipe Throughput                               12440.0     167411.7    134.6
Pipe-based Context Switching                   4000.0      20978.9     52.4
Process Creation                                126.0        725.4     57.6
Shell Scripts (1 concurrent)                     42.4        443.0    104.5
Shell Scripts (8 concurrent)                      6.0         56.4     94.0
System Call Overhead                          15000.0     431022.4    287.3
                                                                   ========
System Benchmarks Index Score                                          97.7
====================================================

I think there are issues which are not just dedicated to libcrypt. I
think a good way to proceed is to test with the very same statically
linked binary on both OSes. Does the Linux emulation also work on arm?





2015-03-05 15:08 GMT+01:00 Greg Troxel <gdt%ir.bbn.com@localhost>:
>
> Stephan <stephanwib%googlemail.com@localhost> writes:
>
>> I wasn´t aware of this easy-at-hand OpenSSL benchmark. That motivated
>> me to do some comparing benchmarks. I noted the following numbers of
>> the 16 size blocks test:
>>
>> Raspberry Pi B NetBSD   ~70.000
>> Raspberry Pi B Raspian   ~300.000
>> Raspberry Pi 2 Raspian   ~800.000
>> NetBSD amd 64              ~4.400.000 (Intel Core 2 Extreme
>> (686-class), 2826.34 MHz, id 0x10676)
>
> NetBSD has opencrypto(4) that can use hardware acceleration, and openssl
> can be made to use it.  I have done this on Soekris net5501, but found
> that it was not 100% stable.  I am unclear on whether the RPI cpu has
> any crypto coprocessor or if Linux or NetBSD has support, but it's
> something to check.


Home | Main Index | Thread Index | Old Index