Re: [uml-devel] [patch 02/12] uml: cpu_relax fix

[Bodo S. <bst...@fu...>]

bla...@ya... wrote:
> Use rep_nop instead of barrier for cpu_relax, following $(SUBARCH)'s doing
> that (i.e. i386 and x86_64).

IIRC, Jeff had the idea, to use sched_yield() for this (from a discussion on #uml).
S390 does something similar using a special DIAG-opcode that gives permission to zVM,
that another Guest might run.

On a host running many UMLs, this might improve performance.

So, I would like to have the small patch below (it's not tested, just an idea).

		Bodo


> diff -puN include/asm-um/processor-generic.h~uml-cpu_relax include/asm-um/processor-generic.h
> --- linux-2.6.11/include/asm-um/processor-generic.h~uml-cpu_relax	2005-03-22 16:52:25.000000000 +0100
> +++ linux-2.6.11-paolo/include/asm-um/processor-generic.h	2005-03-22 16:54:41.000000000 +0100
> @@ -16,7 +16,8 @@ struct task_struct;
>  
>  struct mm_struct;
>  
> -#define cpu_relax()   barrier()
> +#include "kern.h"
> +#define cpu_relax()   sched_yield()
>  
>  struct thread_struct {
>  	int forking;