Date: Mon, 24 May 2010 19:18:12 +0300 From: Andriy Gapon <avg@freebsd.org> To: Alexander Motin <mav@freebsd.org> Cc: svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org Subject: Re: svn commit: r208494 - in head/sys: amd64/amd64 amd64/include i386/i386 i386/include kern pc98/cbus sys x86/isa x86/x86 Message-ID: <4BFAA6C4.3040706@freebsd.org> In-Reply-To: <4BFAA409.4040008@FreeBSD.org> References: <201005241140.o4OBeova088506@svn.freebsd.org> <4BFA8F22.6040402@freebsd.org> <4BFAA409.4040008@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
on 24/05/2010 19:06 Alexander Motin said the following: > Andriy Gapon wrote: >> on 24/05/2010 14:40 Alexander Motin said the following: >>> Author: mav >>> Date: Mon May 24 11:40:49 2010 >>> New Revision: 208494 >>> URL: http://svn.freebsd.org/changeset/base/208494 >>> >>> Log: >>> - Implement MI helper functions, dividing one or two timer interrupts with >>> arbitrary frequencies into hardclock(), statclock() and profclock() calls. >>> Same code with minor variations duplicated several times over the tree for >>> different timer drivers and architectures. >>> - Switch all x86 archs to new functions, simplifying the code and removing >>> extra logic from timer drivers. Other archs are also welcome. >> could you please describe the new code/KPI in greater detail, perhaps on a more >> appropriate mailing list? > > Sent some summary to arch@. Thanks! >> For me it is not immediately obvious why IPI_PROFCLOCK is gone now. I haven't >> spent much time reverse engineering this change and perhaps it's easier for you >> to describe the change. > > IPI_PROFCLOCK is not used any more. Separation between statclock() and > profclock() now handled in common machine-independent code. Timer driver > generates interrupts, MD code manages timers configuration and events > redistribution to every CPU, using up to two IPIs if needed, MI code do > the rest of things. Thank you for the explanation. BTW, with e.g. HPET we may consider using a separate clock for profiling, but I am not sure. -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4BFAA6C4.3040706>