May 2006 - Darwin-dev - lists.apple.com

Re: gettimeofday cost on intel?
by Matt Watson 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 9, 2006, at 9:32 AM, Andrew Gallatin wrote: On May 9, 2006, at 8:11 AM, Andrew Gallatin wrote: I did not know about this. Nm reveals its symbol in libc (at least on ppc). Is it documented somewhere? Does it also use the commpage? Yes, it uses the commpage. Here's a technote on its usage: http://developer.apple.com/qa/qa2004/qa1398.html matt. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Matt Watson writes: I need a cheap way to measure events with at least a 10ms granularity that is monotonically increasing, regardless of the CPU my thread has been scheduled upon. I'd be surprised if mach_absolute_time() does not work for this. If you don't need the conversion to nanoseconds, you can just use the raw function. This email sent to site_archiver(a)lists.apple.com

1 0

Re: gettimeofday cost on intel?
by Andrew Gallatin 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Matt Watson writes: > > On May 9, 2006, at 8:11 AM, Andrew Gallatin wrote: > > > > > I need a cheap way to measure events with at least a 10ms granularity > > that is monotonically increasing, regardless of the CPU my thread has > > been scheduled upon. > > I'd be surprised if mach_absolute_time() does not work for this. I did not know about this. Nm reveals its symbol in libc (at least on ppc). Is it documented somewhere? Does it also use the commpage? > gettimeofday() is implemented in the commpage for both PowerPC and > Intel. Great! Thanks, Drew _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: gettimeofday cost on intel?
by Matt Watson 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 9, 2006, at 8:11 AM, Andrew Gallatin wrote: I need a cheap way to measure events with at least a 10ms granularity that is monotonically increasing, regardless of the CPU my thread has been scheduled upon. I'd be surprised if mach_absolute_time() does not work for this. Can I count on gettimeofday() using the commpage (and, hence, being very cheap) on x86 like it is on ppc? matt. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… gettimeofday() is implemented in the commpage for both PowerPC and Intel. This email sent to site_archiver(a)lists.apple.com

1 0

gettimeofday cost on intel?
by Andrew Gallatin 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com I need a cheap way to measure events with at least a 10ms granularity that is monotonically increasing, regardless of the CPU my thread has been scheduled upon. Can I count on gettimeofday() using the commpage (and, hence, being very cheap) on x86 like it is on ppc? On some OSes, gettimeofday() will always, or often result in an expensive syscall, so we mmap a counter out of our driver which is incremented every clock tick. If MacOSX's gettimeofday() can be counted upon to be cheap, I can avoid this effort on MacOSX. Thanks, Drew _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: How to use VARIANT Structure of C on MAC OSX and XCODE
by plumber Idraulico 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com *check variant implementation of oleaut32.dll from wine wine-X.X.X/dll/oleaut32/variant.h wine-X.X.X/dll/oleaut32/variant.c http://www.winehq.com/ http://darwine.opendarwin.org/ Best Regards _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: How to use VARIANT Structure of C on MAC OSX and XCODE
by Michael Smith 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com You can get therapy for that. I have implemented Variant structure on MAC but it gives error VARIANT V; V.vt = VAR_INT; = Mike _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… On May 8, 2006, at 12:03 PM, "Gaurav Kumar" <gaurav.kumar(a)spsoftindia.com> wrote I want Variant type structure as is in C which can accept any data type. Is there any replacement for Variant is on Mac or we can use them same as in MFC Classes. Use the language as it was intended. VARIANT is a bad attempt at handling opaque data. A slightly more mature approach is the use of (void *) in conjunction with out-of-band type information (since the singleton argument is just a special case of the arbitrary argument set). Better still is to take advantage of the language's type-safety and encapsulation features; if you must pass variant data, use a collection structure with a union of the full argument sets as a member. At least this gives you basic protection against mis-matched collections, and some assurance that you're looking at the right container in the first place. It gives error as VAR_INT is not declared and vt is of non type void *. That is because you do not appear to have "implemented" anything other than a compiler error message generator. It looks like you have attempted to *use* the (nonexistent on Mac OS X) VARIANT type. If I can use Variant on MAC, please provide one or two examples or some other thing that I can use to accept any primitive data type in a single data type. Happily, Mac OS X does not support anything quite this disgusting. You will need to either implement (note: implement, not use) an equivalent as per other suggestions in this thread, or (better but much less likely) fix the code that you are obviously porting so that it doesn't contain any more of this bogosity. This email sent to site_archiver(a)lists.apple.com

1 0

Re: pthread_cancel()
by André-John Mas 09 May '06

09 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thanks, I'll see if that works for me. On 8-May-06, at 16:08 , Matt Watson wrote: On May 8, 2006, at 12:59 PM, Andre-John Mas wrote: On May 8, 2006, at 12:45 PM, Andre-John Mas wrote: Could anyone tell me whether pthread_cancel() is planned to be implemented properly in a future version of Darwin? From what I understand the current implementation does not conform to the posix implementation. If you compile with _APPLE_C_SOURCE, _POSIX_C_SOURCE or _XOPEN_SOURCE, the implementation should behave as you want. This is when compiling user-level programs. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Thanks. Just one question are these for my applications or for compiling the kernel? matt. This email sent to site_archiver(a)lists.apple.com

1 0

RE: question for CFRunLoopStop()
by Philip Lukidis 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: AcZwk74Xolo3ir3gSZye46HpuYclIACCtz5Q Thread-topic: question for CFRunLoopStop() Thanks very much for your help and analysis. The times which CFRunLoopTimerInvalidate() worked for me without calling CFRunLoopStop/CFRunLoopWakeUp could be explained (as you say) by there having been a timer callback which was currently executing, and so the timer, being invalid, could never cause another timer callback. I'll remember your tip for CFRunLoopWakeUp(), which might be helpful to me in another context. thanks, Philip Lukidis -----Original Message----- From: Kevin Van Vechten [mailto:kevin@opendarwin.org] Sent: Friday, May 05, 2006 6:30 PM To: Philip Lukidis Cc: darwin-dev(a)lists.apple.com Subject: Re: question for CFRunLoopStop() On May 5, 2006, at 12:20 PM, Philip Lukidis wrote: > Perhaps I am missing something, but removing/invalidating the timer > from > the child runloop does not always provoke the child thread to > exit. So > after having read your reply, I suppose that the best way to solve my > problem would be to stop the runloop. But the documentation says that > removing all sources from a runloop causes CFRunLoopRun() to return: > > "The current thread's run loop runs in the default mode (see "Default > Run Loop Mode") until the run loop is stopped with CFRunLoopStop or > all > the sources and timers are removed from the default run loop mode." Looking at the CF sources, it appears that CFRunLoopTimerInvalidate() calls CFRunLoopRemoveTimer() to remove the timer from its runloop in all modes that it's registered in. > However, I do not consistently find that this is true in my case. > Perhaps there is a fault in my code, but invalidating the timer > does not > always cause the child thread to exit (and according to the docs, > invalidating a timer removes it from the runloop). However, so far > CFRunLoopStop() consistently works. Offhand could you explain this? While the runloop is waiting for events, it is blocked on a mach port read. The only way to unblock it is for a timer to fire, a source to be signaled, or to call CFRunLoopWakeUp() from another thread. The documentation suggests that removing or invalidating the last source would stop the runloop (ostensibly via a call to CFRunLoopWakeUp()). However, looking at the source verifies your findings, namely that CFRunLoopWakeUp() is not called implicitly as part of the CFRunLoopTimerInvalidate() or CFRunLoopRemoveTimer() calls. CFRunLoopStop(), however, does implicitly call CFRunLoopWakeUp(). Seems like you can either continue calling CFRunLoopStop(), or you can manually call CFRunLoopWakeUp() after removing/invalidating the timer. - Kevin _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: How to use VARIANT Structure of C on MAC OSX and XCODE 2.2...............
by Boyd Waters 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 8, 2006, at 3:04 AM, Jonas Maebe wrote: VARIANT V; V.vt = VAR_INT; Okay, you're hosed. This is not a feature of the language. - boyd Boyd Waters Socorro, New Mexico _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com It gives error as VAR_INT is not declared and vt is of non type void *. Indeed, it's a feature of some standard Windows library. It's possible to implement support for it in any language (also on non- Windows platforms), but that's quite a bit of work. I don't know of a pure C implementation. Sorry, I can't recall if you need interoperability with Microsoft implementations. Mozilla XPCOM might be an example implementation of this in C++ that works well with GCC. http://www.xulplanet.com/references/xpcomref/ifaces/nsIVariant.html Mozilla basically took the base Microsoft COM architecture and wrote a C++ implementation, which they later open-sourced. You can write a VARIANT implementation that is very inefficient, and I suppose you can write a better one if you pay attention to data- alignment issues. So it might be worthwhile to examine (or re-use) something like XPCOM if you must. smime.p7s

1 0

Re: pthread_cancel()
by Matt Watson 08 May '06

08 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 8, 2006, at 12:59 PM, Andre-John Mas wrote: On May 8, 2006, at 12:45 PM, Andre-John Mas wrote: Could anyone tell me whether pthread_cancel() is planned to be implemented properly in a future version of Darwin? From what I understand the current implementation does not conform to the posix implementation. If you compile with _APPLE_C_SOURCE, _POSIX_C_SOURCE or _XOPEN_SOURCE, the implementation should behave as you want. This is when compiling user-level programs. matt. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Thanks. Just one question are these for my applications or for compiling the kernel? This email sent to site_archiver(a)lists.apple.com

1 0