This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] glibc vDSO bits for ppc/ppc64 update
- From: Steve Munroe <sjmunroe at us dot ibm dot com>
- To: Benjamin Herrenschmidt <benh at kernel dot crashing dot org>
- Cc: libc-alpha at sources dot redhat dot com, linuxppc-dev at ozlabs dot org, linuxppc64-dev at ozlabs dot org, munroesj at us dot ibm dot com, Tom Gall <tom_gall at vnet dot ibm dot com>
- Date: Fri, 11 Nov 2005 16:32:11 -0600
- Subject: Re: [PATCH] glibc vDSO bits for ppc/ppc64 update
Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote on 11/11/2005
04:13:28 PM:
>
> > Additional hp_timer_t is now declared as unsigned long long int
> > to match up with benh's get_tbfreq on 32 bit ppc. (ppc64 is already
> > 64 bit)
>
> I'm not sure having hp_timer_t be 64 bits on 32 bits libraries is the
> right thing to do. I suppose it's a matter of performances tradeoff. The
> fact that the vDSO __kernel_get_tbfreq() returns a long long doesn't
> force you to use a long long hp_timer_t, the vDSO doesn't have to return
> a value of type hp_timer_t :)
>
> I did that on the vDSO to avoid having to change the vDSO interface if
> we ever have to deal with >4Ghz time sources.
>
A 64-bit timebase is required to the use hp-timing for clock_gettime(
CLOCK_PROCESS_CPUTIME_ID, ) and clock_gettime(CLOCK_THREAD_CPUTIME_ID, ).
So 64-bit it is.
Steven J. Munroe
Linux on Power Toolchain Architect
IBM Corporation, Linux Technology Center