From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian.Campbell@citrix.com (Ian Campbell) Date: Tue, 28 Feb 2012 10:12:50 +0000 Subject: [PATCH-WIP 01/13] xen/arm: use r12 to pass the hypercall number to the hypervisor In-Reply-To: References: <1330019314-20865-1-git-send-email-stefano.stabellini@eu.citrix.com> <1330360043.8557.302.camel@zakaz.uk.xensource.com> Message-ID: <1330423970.31269.45.camel@zakaz.uk.xensource.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Mon, 2012-02-27 at 21:05 +0000, Peter Maydell wrote: > On 27 February 2012 16:27, Ian Campbell wrote: > > R12 is not accessible from the 16 bit "T1" Thumb encoding of mov > > immediate (which can only target r0..r7). > > > > Since we support only ARMv7+ there are "T2" and "T3" encodings available > > which do allow direct mov of an immediate into R12, but are 32 bit Thumb > > instructions. > > > > Should we use r7 instead to maximise instruction density for Thumb code? > > r7 is (used by gcc as) the Thumb frame pointer; I don't know if this > makes it worth avoiding in this context. I think it does. It actually sounds as if using r12 is fine here, the impact on code density should be pretty small -- there aren't really all that many call sites which involve hypercalls. By way of an example I measured an x86 kernel which should be using more hypercalls due to pv paging etc and found that 0.014% of the lines in "objdump -d" contained a call to the hypercall_page. (I know not all lines of objdump -d output are instructions but it's a reasonable approx IMHO). So I think using 3 16 bit instructions slots instead of 2 won't make much impact in practice. Thanks, Ian.