[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20171013053005.cj5vhqehkwlnhyla@gmail.com>
Date: Fri, 13 Oct 2017 07:30:05 +0200
From: Ingo Molnar <mingo@...nel.org>
To: Thomas Gleixner <tglx@...utronix.de>
Cc: Andi Kleen <andi@...stfloor.org>, x86@...nel.org,
linux-kernel@...r.kernel.org, Andi Kleen <ak@...ux.intel.com>,
Jonathan McDowell <noodles@...th.li>
Subject: Re: [PATCH v9 2/5] x86/cpuid: Add generic table for cpuid
dependencies
* Thomas Gleixner <tglx@...utronix.de> wrote:
> On Thu, 12 Oct 2017, Ingo Molnar wrote:
> >
> > * Andi Kleen <andi@...stfloor.org> wrote:
> >
> > > --- /dev/null
> > > +++ b/arch/x86/kernel/cpu/cpuid-deps.c
> > > @@ -0,0 +1,109 @@
> > > +/* Declare dependencies between CPUIDs */
> > > +#include <linux/kernel.h>
> > > +#include <linux/init.h>
> > > +#include <linux/module.h>
> > > +#include <asm/cpufeature.h>
> > > +
> > > +struct cpuid_dep {
> > > + unsigned short feature;
> > > + unsigned short depends;
> > > +};
> >
> > Why are these 16-bit fields? 16-bit data types should be avoided as much as
> > possible, as they generate suboptimal code.
>
> I was looking at that as well and decided that we preferrably have a
> compressed data structure. The code which walks the table is hardly
> performance critical and the difference in text size is marginal.
So the code should all be __init (once that is fixed), hence data and text size
literally does not matter - it gets freed.
So the only effect the 16-bit variables have is (marginally) worse boot times.
Thanks,
Ingo
Powered by blists - more mailing lists