... The writing back zeros to every position was an an interesting method for avoiding memory fences. Wonder if a block based equivalent of this algorithm...
There was a little bit of discussion about this article a while ago: http://groups.yahoo.com/group/liblf-dev/messages/156?threaded=1&m=e&var=1&tidx=1...
I'm currently refactoring a library of mine and, among other things, working on improving thread-safety. Since the library is for working with audio, a common...
This pattern exists: it is an exponential-backoff spinlock with timeout. (Your high-priority thread is using a very low timeout, your low-priority thread not...
... If only we had a lock free library with a portable CAS :) If a system call penalty does not harm the performance of the audio thread that much, then I...
... Well we have a sourceforge site now: http://sourceforge.net/projects/liblfds Let me know your sf user names and I'll add you to the list... I'm sure we can...
... Hello all, I just joined the group today. I have created an extensive library of lockfree and other atomic inline primitives, lists, queues, locks,...
... Hi Sean Nice to hear from you. Other ABA possibilities I know of are SMR and using CAS64 (CAS128) to keep a change counter (or some equivalent mechanism...
Hi Josh I'd say there is interest in yours. Mine is written in x86 ASM and uses 64bit CAS to do refcounts instead of hazard ptrs so from a brief examanation of...
Hi I'm new to this group & haven't read every message yet, but I'll share my thoughts about this. Plz note that I'll be debating about Windows, since I have no...
... I think I saw a paper describing the problem with hazard pointers being that you need an awful lot of memory fences to make them work. Of course the...
Sorry to reply to my own post but... ... I got SMR and RCU mixed up. The paper I mentioned is "Comparative Performance of Memory Reclamation Strategies for ...
... Not currently. The functions are currently part of a proprietary commercial application for Mac OS X. I do post occasional code snippets and will gladly...
... Not currently. The functions are currently part of a proprietary commercial application for Mac OS X. I do post occasional code snippets and will gladly...
... Welcome to the list. However, what you propose is not a viable solution for what I think is the intended goal of people on this list. At least a few of the...
Yes, welcome :-) I agree with what James has said. There is certainly a lot of interest (in general) in lock-free techniques as an optimisation to increase...
Hi all, I've been profiling my code and found something pretty nasty. I'm not sure if I'm doing something wrong, but it seems that while my production thread...
Hi Aaron, 1) In C# you have no control when garbage collection hits you. This means you have a less deterministic behaviour than in other languages. This is...
t. bosh
liblf@...
Nov 3, 2006 9:08 am
195
Hi, t bosh. (Sorry, I couldn't find your name on the post). Thanks for the advice. I'll reply inline. First of all WRT .NET, there is some extremely...
... You may find, if your benchmark is hitting your semaphore very hard, that a lot of time is spent waiting for the cacheline involved to bounce between the...
Hi Chris, Well, I did get the spinlocks to perform. I'm not sure what I did since I haven't changed the two lines of code I posted before! :-S I'm doing this...
Howdy, ... I'm not sure whether it is hard (no experience in C#), but recently read this: http://www.boyet.com/Articles/LockfreeStack.html .. where the guy has...
Jan Wassenberg
urkt@...
Nov 6, 2006 6:56 am
199
hi, i tested this on a couple of machines (4x Xeon, 16x Itanium, 4x Opteron) and it seems to work. tomorrow i'll try it with some more real-life test cases. ...
... the audio thread, no matter if it is cheaper overall. But wait.. I wasn't claiming that blocking was cheap, it's not at all, I was claiming that...
hi, i tested this code on the same machines as before. seems to work... chris ... // IA64 // g++ -ofifo_thread_ia64 -lpthread -D__SMP__ fifo_thread.c //...
Hi, ... Ah, I think I see the problem. Sleep(0) only blocks if there are runnable threads of equal or higher priority. In general, using thread priority for...
Jan Wassenberg
urkt@...
Nov 6, 2006 6:59 am
203
good morning, i just tested bjorn roche's ring-buffer-2 code (with mem-barriers) on the following machines: Red Hat Ent Linux AS 4.0 on a HP Integrity rx8620...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
