Language:
switch to room list switch to menu My folders
Go to page: First ... 10 11 12 13 [14] 15 16 17 18 ... Last
[#] Mon Feb 16 2015 12:52:46 EST from dothebart

[Reply] [ReplyQuoted] [Headers] [Print]

So my performance evaluation is online:

https://www.arangodb.com/2015/02/16/comparing-atomic-mutex-rwlocks

afaik atomics are also available in c; so maybe some of the citserver code should be adopted ;-)

interestingly atomics are half as fast on windows on the same machine...



[#] Mon Feb 16 2015 18:16:18 EST from TaMeR to room_Programming@uncensored.citadel.org

Subject: Re: (no subject)

[Reply] [ReplyQuoted] [Headers] [Print]

On 02/16/2015 07:52 PM, dothebart wrote:
So my performance evaluation is online:

https://www.arangodb.com/2015/02/16/comparing-atomic-mutex-rwlocks

afaik atomics are also available in c; so maybe some of the citserver
code should be adopted ;-)

interestingly atomics are half as fast on windows on the same machine...



Yes you are right but it still looks to me like you get the best bang
for the bug on a Linux ARM (Allwinner A20 SOC on Cubie Truck).

What do those cost like $40?

[#] Mon Feb 16 2015 22:08:59 EST from fleeb

[Reply] [ReplyQuoted] [Headers] [Print]


That's a very interesting read, dothebart.

[#] Tue Feb 17 2015 05:40:49 EST from dothebart

Subject: Re: (no subject)

[Reply] [ReplyQuoted] [Headers] [Print]

well, the cubie truck isn't that cheap; However you can get Allwinner A20 boards for around that price.

It seems even MSVC-Devs feel urged to reply to something like that:

http://www.reddit.com/r/cpp/comments/2w3abx/comparing_lockless_programming_with_atomics_in_c/



[#] Wed Feb 18 2015 06:20:52 EST from fleeb

[Reply] [ReplyQuoted] [Headers] [Print]


You're right (in reddit) about it being a shame that Microsoft invests so much time only to have something like that screw it all up.

But it's interesting to see them aware of the issue, and trying to come up with a way to improve the performance.

[#] Wed Feb 18 2015 09:58:31 EST from dothebart

[Reply] [ReplyQuoted] [Headers] [Print]

need nine more github starts to get the 1k followers filled

https://github.com/arangodb/arangodb

;-)



[#] Sun Feb 22 2015 00:33:57 EST from ax25

[Reply] [ReplyQuoted] [Headers] [Print]

Yet another way for me to avoid Javascript!  Ho-ray :-)

http://jeanphix.me/Ghost.py/



[#] Sun Feb 22 2015 13:48:59 EST from TaMeR to room_Programming@uncensored.citadel.org

Subject: Re: (no subject)

[Reply] [ReplyQuoted] [Headers] [Print]

On 02/22/2015 07:33 AM, ax25 wrote:
Yet another way for me to avoid Javascript! Ho-ray :-)

http://jeanphix.me/Ghost.py/

You are still programming in js with that!!!
All you are doing is, you are shipping it with py.
Why would you try to avoid js so hard anyway?
That is all we have on client side.

[#] Sun Feb 22 2015 14:08:22 EST from dothebart

Subject: Re: (no subject)

[Reply] [ReplyQuoted] [Headers] [Print]

or if you use arangodb/FOXX, iojs or nodejs on the server

it hast its challenges, but I think its realy ok.

running the lint should be mandatorry however.



[#] Sun Feb 22 2015 23:13:21 EST from ax25

Subject: Re: (no subject)

[Reply] [ReplyQuoted] [Headers] [Print]

 

Sun Feb 22 2015 01:48:59 PM EST from TaMeR @ Uncensored Subject: Re: (no subject)
On 02/22/2015 07:33 AM, ax25 wrote:
Yet another way for me to avoid Javascript! Ho-ray :-)

http://jeanphix.me/Ghost.py/
You are still programming in js with that!!!
All you are doing is, you are shipping it with py.
Why would you try to avoid js so hard anyway?
That is all we have on client side.

Because someone left out the batteries :-)



[#] Thu Feb 26 2015 05:30:46 EST from dothebart

[Reply] [ReplyQuoted] [Headers] [Print]

WHOA! openssl 1.0.2 brings down the valgrind suppressions neccessary to:

{
   Ignore OpenSSL malloc
   Memcheck:Leak
   ...
   fun:SSL_library_init
   ...
   obj:*
}
{
   Ignore OpenSSL malloc
   Memcheck:Leak
   ...
   fun:SSL_load_error_strings
   ...
   obj:*
}
{
   Ignore OpenSSL malloc
   Memcheck:Leak
   ...
   fun:OpenSSL_add_all_algorithms
   ...
   obj:*
}

However, the current debian package simply removed SSLv3, which gave me random broken stuff all over the system like python, ruby, php...

fixed that by reverting debian rules to

CONFARGS  = --prefix=/usr --openssldir=/usr/lib/ssl --libdir=lib/$(DEB_HOST_MULTIARCH) no-idea no-mdc2 no-rc5 no-zlib  enable-tlsext no-ssl2 enable-unit-test



[#] Thu Feb 26 2015 08:04:45 EST from LoanShark

[Reply] [ReplyQuoted] [Headers] [Print]


Compiling it out is a non-starter anyway. Enterprise guys, especially those who use SSL/TLS in a backend scenario or over a VPN, are moving slowly on disabling SSLv3. You can't always control how fast your business partners move.


Durrr....

[#] Thu Mar 05 2015 22:09:57 EST from IGnatius T Foobar

[Reply] [ReplyQuoted] [Headers] [Print]

Yeah, pretty much the only people actively working on phasing out SSLv2 are folks legally obligated to do so (finance and health care sectors etc)

[#] Fri Mar 06 2015 04:23:33 EST from dothebart

[Reply] [ReplyQuoted] [Headers] [Print]

 

Thu Feb 26 2015 12:11:16 EST from dothebart @ Uncensored

Ok, details:

if you have static const objects with initializers, they are defined to be thread safe in linux and windows.

However, if they live in function/method scope, its still true for linux, but not for windows.so they basicaly need to be moved outside of the functions.



[#] Fri Mar 06 2015 14:07:39 EST from fleeb

[Reply] [ReplyQuoted] [Headers] [Print]


Oh, that's rather subtle. Grrr...

[#] Fri Mar 06 2015 20:03:39 EST from TaMeR to room_Programming@uncensored.citadel.org

Subject: Re: (no subject)

[Reply] [ReplyQuoted] [Headers] [Print]

On 03/06/2015 11:23 AM, dothebart wrote:


Thu Feb 26 2015 12:11:16 EST from dothebart @ Uncensored

lolwindows.
https://github.com/arangodb/arangodb/commit/3e85160b04fd86329bea63e5ab86ad75c39dc62e

fails... sometimes.



Ok, details:

if you have static const objects with initializers, they are defined to
be thread safe in linux and windows.

However, if they live in function/method scope, its still true for
linux, but not for windows.so they basicaly need to be moved outside of
the functions.


No wonder I couldn't figure out what the lol was all about.
Who still cares about windows?

[#] Sat Mar 07 2015 17:24:24 EST from fleeb

[Reply] [ReplyQuoted] [Headers] [Print]


Er, I have to care about Windows programming, at least.

And, this is fairly awful. One of the common idioms used in C++ for creating a singleton can involve a static const object in function/method scope.

[#] Mon Mar 09 2015 14:47:10 EDT from LoanShark

[Reply] [ReplyQuoted] [Headers] [Print]

Yeah, pretty much the only people actively working on phasing out SSLv2

are folks legally obligated to do so (finance and health care sectors

etc)

Or contractually; i.e. anyone who has sworn to uphold PCI Compliance.

[#] Tue Mar 10 2015 08:30:17 EDT from dothebart

Subject: Re: (no subject)

[Reply] [ReplyQuoted] [Headers] [Print]

While I personaly wouldn't care about windows its still many developers using it as their workstation of choice.

So if you want to reach them as users without clunky VMs they should install a native port is neccessary.



Go to page: First ... 10 11 12 13 [14] 15 16 17 18 ... Last