From mboxrd@z Thu Jan 1 00:00:00 1970 From: linux@arm.linux.org.uk (Russell King - ARM Linux) Date: Tue, 1 Jul 2014 23:44:38 +0100 Subject: [PATCH v7 0/9] ARM: VDSO In-Reply-To: References: <1403493118-7597-1-git-send-email-nathan_lynch@mentor.com> <53B2E311.6030609@mentor.com> Message-ID: <20140701224438.GD32514@n2100.arm.linux.org.uk> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Tue, Jul 01, 2014 at 10:08:07PM +0200, Ard Biesheuvel wrote: > We (at Linaro) had independently identified VDSO for ARM as a topic of > interest after seeing gettimeofday() turn up in a couple of benchmarks > (which btw seem to be the primary users of gettimeofday() at such a > high rate), so of course, when someone else steps up and does the > work, i am happy to at least spend some time testing it. Therefore, I > would be happy to test your new version either before or after you > (re)post it, whichever you prefer. As has been covered by Nathan in a previous reply - the VDSO is only of real benefit on CPUs with the architected timer. Those without this do not have access from userspace to an appropriate timer to allow the VDSO to compute the current time. Hence, the VDSO will just call into the kernel via the standard (and very same) syscalls that glibc would otherwise have used. This is illustrated nicely by the results posted for iMX6, which reflects how the VDSO is going to affect platforms without the architected timer. -- FTTC broadband for 0.8mile line: now at 9.7Mbps down 460kbps up... slowly improving, and getting towards what was expected from it.