Re: 2.8 release schedule - tentitively Wednesday 27 October.

2010-10-26 Thread Jan Lentfer
On Sat, 23 Oct 2010 11:06:05 -0700 (PDT), Matthew Dillon
dil...@apollo.backplane.com wrote:
 :Openssl hardware crypto support is not working yet, wich is a show
 stopper imo.
 :
 :Jan
 
 I don't have any crypto hardware to test that with, its up to you
 guys to figure out what is going on.  We know the kernel hw crypto
 works so it has to be the user library.
 
 If we have to MFC it after the release and roll a 2.8.2 we will, but
 it can't hold up the release any more.
 
   -Matt
   Matthew Dillon 
   dil...@backplane.com

My bad, according to http://www.a110wiki.de/wiki/VIA_Padlock you have to
use the (undocumented) commandline option -evp if you want to use hw
accellaration with openssl speed. If you do, it looks fine:

 openssl speed -evp aes-256-ecb
Doing aes-256-ecb for 3s on 16 size blocks: 895515 aes-256-ecb's in 2.65s
Doing aes-256-ecb for 3s on 64 size blocks: 221951 aes-256-ecb's in 2.55s
Doing aes-256-ecb for 3s on 256 size blocks: 57388 aes-256-ecb's in 2.63s
Doing aes-256-ecb for 3s on 1024 size blocks: 13941 aes-256-ecb's in 2.55s
Doing aes-256-ecb for 3s on 8192 size blocks: 1740 aes-256-ecb's in 2.55s
OpenSSL 1.0.0a 1 Jun 2010
built on: Sun Oct 10 17:54:52 CEST 2010
options:bn(64,32) rc4(idx,int) des(ptr,risc1,16,long) aes(partial)
blowfish(idx)
compiler: cc
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes256 bytes   1024 bytes   8192
bytes
aes-256-ecb   5410.07k 5577.37k 5596.70k 5588.00k
5596.69k

 openssl speed -evp aes-256-ecb -engine padlock
engine padlock set.
Doing aes-256-ecb for 3s on 16 size blocks: 9950030 aes-256-ecb's in 2.50s
Doing aes-256-ecb for 3s on 64 size blocks: 8237622 aes-256-ecb's in 2.34s
Doing aes-256-ecb for 3s on 256 size blocks: 5745112 aes-256-ecb's in
2.61s
Doing aes-256-ecb for 3s on 1024 size blocks: 2159485 aes-256-ecb's in
2.53s
Doing aes-256-ecb for 3s on 8192 size blocks: 337084 aes-256-ecb's in
2.68s
OpenSSL 1.0.0a 1 Jun 2010
built on: Sun Oct 10 17:54:52 CEST 2010
options:bn(64,32) rc4(idx,int) des(ptr,risc1,16,long) aes(partial)
blowfish(idx)
compiler: cc
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes256 bytes   1024 bytes   8192
bytes
aes-256-ecb  63680.19k   225694.31k   563640.21k   873604.99k 
1030490.36k

---

 openssl speed -evp aes-256-ofb
Doing aes-256-ofb for 3s on 16 size blocks: 748438 aes-256-ofb's in 2.28s
Doing aes-256-ofb for 3s on 64 size blocks: 215132 aes-256-ofb's in 2.57s
Doing aes-256-ofb for 3s on 256 size blocks: 50022 aes-256-ofb's in 2.35s
Doing aes-256-ofb for 3s on 1024 size blocks: 14054 aes-256-ofb's in 2.63s
Doing aes-256-ofb for 3s on 8192 size blocks: 1746 aes-256-ofb's in 2.58s
OpenSSL 1.0.0a 1 Jun 2010
built on: Sun Oct 10 17:54:52 CEST 2010
options:bn(64,32) rc4(idx,int) des(ptr,risc1,16,long) aes(partial)
blowfish(idx)
compiler: cc
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes256 bytes   1024 bytes   8192
bytes
aes-256-ofb   5249.32k 5356.72k 5445.58k 5466.13k
5547.92k

 openssl speed -evp aes-256-ofb -engine padlock
engine padlock set.
Doing aes-256-ofb for 3s on 16 size blocks: 7089446 aes-256-ofb's in 2.64s
Doing aes-256-ofb for 3s on 64 size blocks: 5129120 aes-256-ofb's in 2.52s
Doing aes-256-ofb for 3s on 256 size blocks: 2605388 aes-256-ofb's in
2.60s
Doing aes-256-ofb for 3s on 1024 size blocks: 775776 aes-256-ofb's in
2.36s
Doing aes-256-ofb for 3s on 8192 size blocks: 119139 aes-256-ofb's in
2.59s
OpenSSL 1.0.0a 1 Jun 2010
built on: Sun Oct 10 17:54:52 CEST 2010
options:bn(64,32) rc4(idx,int) des(ptr,risc1,16,long) aes(partial)
blowfish(idx)
compiler: cc
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes256 bytes   1024 bytes   8192
bytes
aes-256-ofb  42956.17k   130085.92k   256376.44k   336697.06k  
376284.02k

Jan

-- 
professional: http://www.oscar-consult.de
private: http://neslonek.homeunix.org/drupal/


Re: Is anyone going to benchmark 2.8?

2010-10-26 Thread Jan Lentfer
On Tue, 26 Oct 2010 09:20:37 +0530, Siju George sgeorge...@gmail.com
wrote:
 On Mon, Oct 25, 2010 at 9:11 AM,  elekktrett...@exemail.com.au wrote:
 This release has had substantial work on SMP scaling. It would be great
 to
 see the actual performance increase with MySQL and PostgreSQL in
 DragonFly
 2.8 versus FreeBSD 8/9 versus NetBSD 5.

 Is anyone planning to do this?

 
 I really don't know how to make a bench mark.
 If you can let me know I can do it :-)
 In fact I was wondering of using dfly as the mysql server for a new
 project instead of debian.

If I find the time I will try to continue my benchmarking with postgresql.

Jan

-- 
professional: http://www.oscar-consult.de
private: http://neslonek.homeunix.org/drupal/


Re: 2.8 release schedule - tentitively Wednesday 27 October.

2010-10-26 Thread Jan Lentfer

I installed the 2.8.1 iso on a VM and it seems that it boots into an SMP 
kernel, regardless of what option I choose.

Also I found it somewhat unusual that if I choose to boot the Install-CD with 
SMP Kernel, SMP kernel is not default on the installed system afterwards, but 
UP is and I still have to choose SMP in the loader. My expectation is different.

Jan

--
professional: http://www.oscar-consult.de
private: http://neslonek.homeunix.org/drupal/



Re: 2.8 release schedule - tentitively Wednesday 27 October.

2010-10-26 Thread Sascha Wildner

On 10/26/2010 21:52, Jan Lentfer wrote:

I installed the 2.8.1 iso on a VM and it seems that it boots into an SMP
kernel, regardless of what option I choose.


I've fixed the bug in my local tree. It seems to be a CD9660 issue 
(thanks to Samuel J. Greear for helping with bug hunting).


UFS (and hence our USB images) doesn't have it. I guess it has to do 
with the dots in the directory names (like, kernel.GENERIC) which CD9660 
doesn't like.



Also I found it somewhat unusual that if I choose to boot the Install-CD
with SMP Kernel, SMP kernel is not default on the installed system
afterwards, but UP is and I still have to choose SMP in the loader. My
expectation is different.


I have also added this feature to the installer (locally as well) and am 
testing it at the moment. Will report back later.


Sascha