On Mon, Mar 12, 2007 at 02:09:18PM +0100, Andi Kleen wrote: > On Monday 12 March 2007 14:02, Joerg Roedel wrote: > > On Fri, Mar 09, 2007 at 08:10:03PM +0200, Avi Kivity wrote: > > > Joerg Roedel wrote: > > > >From: Joerg Roedel <[EMAIL PROTECTED]> > > > > > > > >This patch simplifies the get_cycles_sync() function by removing > > > >the #ifdefs from it. Further it introduces an optimization for AMD > > > >processors. There the RDTSCP instruction is used instead of CPUID;RDTSC > > > >which is helpfull if the kernel runs as a KVM guest. Running as a guest > > > >makes CPUID very expensive because it causes an intercept of the guest. > > > > > > > > +#define RDTSCP ".byte 0x0f, 0x01, 0xf9" > > > >+ alternative_io_two("cpuid\nrdtsc", > > > >+ "rdtsc", X86_FEATURE_SYNC_RDTSC, > > > >+ ".byte 0x0f, 0x01, 0xf9", X86_FEATURE_RDTSCP, > > > > > > > > > > why not use the RDTSCP macro here? > > > > Does this macro exist? I couldn't found it in the current git tree. And > > the rdtscp macros in msr.h use the plain opcode too. > > It doesn't exist. The rdtscp macros are also not used currently, that > is why nobody's binutils complained. > > Doing the .bytes is ok > > I still don't like the alternative() record complications though.
Do you think of another way to make use of RDTSCP in the get_cycles_sync function? Using CPUID in a function called such often is bad when running Linux as a virtualization guest... So using RDTSCP there might be a goog idea. Regards, Joerg -- Joerg Roedel Operating System Research Center AMD Saxony LLC & Co. KG - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/