Timo, Harry, Alex, Thanks for the information. Since the integer-comparison Harry proposed is out of the question for performance reasons, I guess the fuzzy...
Hi Martijn, ... Why is that so bad? In fact, it was the next suggestion I was going to make: If you want to make the "equality" comparison as tight as...
Alex, I guess I'm just a bit anal about performance sometimes, sorry about that :) If it were the inner loop of a pixel rendering algo, performance may be an...
Hi All, My ISP (Verio) screwed up the DNS for the afh.com and filtermeister.com domains last night, and it may take 24 to 36 hours to resolve. In the...
Hi All, The DNS problems seem to be resolved (though it may take some time to propagate to the ends of the earth), so the afh.com and filtermeister.com domains...
Hi Richard, I managed to make the OnCtl handler multithreaded. I added triggerThread() function that works just like triggerEvent(), but runs the OnCtl handler...
Hi Harry, this sounds fantastic! Just what we need, and a fine addition to FM. I agree that having access to the new features is redundant unless we know how...
Hi Richard, The new countProcessors() function tells you how many processors are available. These "processors" can be CPUs, cores or hyperthreading pipes. You...
Hi, As I understand it, it might be an idea to base the number of threads not just on the number of cores but also on the size of the workload and generally go...
Hi Martijn, You are basically right. If a processor-intensive application runs in the background, it would be more effective to use more threads than the...
Hi All, In further support of multithreading in FM, I am exposing the Win32 Critical Sections api via the following API in FM: int createCriticalSection(void) ...
Hi Martijn, Thinking about it more I am not sure if using more threads than cores will make processing any faster if another processor-intensive process is...
Hi, I would would for the CS abreviation. You could also combine the enter and tryEnter calls with a second parameter. Or do we have to be absolutely compliant...
I'd vote to use the full name as that makes the Microsoft docs easy to find and FM Wiki/docs can just refer to the MS docs. Keep in mind with respect to...
I have a lot of experience with multiprocessor mainframes. Ron's remarks about overhead are very true. The rule of thumb that we used is - denote single...
Hi Bill, My tests one year ago with a multi-threaded gaussian blur indicated that the speed increase is 1.9 to 1.95 with a second core. So for my example it...
That's encouraging. My experience was multiprocessor mainframes which by design probably have slower communication paths between processors than multicore...
I think in Bill's case each CPU was running a separate OS while with Windows, one CPU drives all of the others. Bill's example would show more overhead. ...
A multicore architecture is different from a multiprocessor architecture, for instance you don't have a slow bus between the cores and all cores can have...
Extremely good points. Nothing comes free - n processors will not reduce wall clock time by 1/n but 1/(n+o) in all cases and the o is overhead related to...
I think in Bill's case each CPU was running a separate OS while with Windows, one CPU drives all of the others. Bill's example would show more overhead. ...
Thirdly, from Bill's description of it being a fault tolerant mainframe, I'm guessing it's a safe bet to say it was a Tandem ... Correct also. I was at...
If necessary, I can do some testing. I have two dual-cores and a quad-core at home. At the office, one of my employees has a personal dual quad-core Mac...
Hi Richard, Thanks for the offer of testing on your multi-core systems -- that will be very useful. The closest thing I have to a multi-core PC is a Toshiba...
Hi, I found what I think is a small bug in the display of scrollbars. I have a vertical scrollbar, anchored top and bottom to resize as the dialog height is...
Hello Duncan, Thanks for the bug report and excellent test case! I've verified that the bug has been present since at least FM 1.0 Beta 9.1, and it affects...
Due to an oversight in my FM build process, FM beta 9a, 9b, and 9c were built without setting the LARGE_ADDRESS_AWARE bit. This bit allows an application (and...
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.
