Re: [PATCH 3/3] x86: enlightenment for ticket spinlocks - remove NOPs from unlock path

["H. Peter Anvin" <hpa@zytor.com>]

On 01/29/2010 12:02 AM, Jan Beulich wrote:
> Under the assumption that the nop-s added by the base ticket spinlock
> enlightenment patch might be considered undesirable (or worse), here
> is an optional patch to eliminate these nop-s again. This is done
> through extending the memory operands of the inc instructions used for
> unlocking ticket locks to the necessary size, using assembler and
> linker features.

> --- 2.6.33-rc5-virt-spinlocks.orig/arch/x86/include/asm/alternative-asm.h
> +++ 2.6.33-rc5-virt-spinlocks/arch/x86/include/asm/alternative-asm.h
> @@ -1,3 +1,7 @@
> +#if 0 /* Hide this from compiler. */
> +	.if 0 # Hide assembly source stuff when assembling compiler output.
> +#endif
> +
>  #ifdef __ASSEMBLY__
>  
>  #include <asm/asm.h>
> @@ -16,3 +20,58 @@
>  #endif
>  
>  #endif  /*  __ASSEMBLY__  */
> +
> +#if 0 /* Hide this from compiler. */
> +	.else # Code to be used in compiler output:
> +
> +	.weak _$.zero
> +
> +	.macro unary opc arg1 arg2 arg3
> +	 .Lempty=2
> +	 .irpc c,"\arg2"
> +	  .Lempty=3
> +	 .endr
> +	 .irpc c,"\arg3"
> +	  .Lempty=0
> +	 .endr
> +	 .Lsym=1
> +	 .Lnum=0
> +	 .irpc c,"\arg1"
> +	  .irpc m,"(123456789-0"
> +	   .ifeqs "\c","\m"
> +	    .Lsym=0
> +	    .exitm
> +	   .endif
> +	   .Lnum=1
> +	  .endr
> +	  .exitm
> +	 .endr
> +	 .if .Lempty == 2
> +	  .if .Lsym
> +	   \opc \arg1
> +	  .elseif .Lnum
> +	   \opc _$.zero+\arg1
> +	  .else
> +	   \opc _$.zero\arg1
> +	  .endif
> +	 .elseif .Lempty == 3
> +	  .if .Lsym
> +	   \opc \arg1,\arg2
> +	  .elseif .Lnum
> +	   \opc _$.zero+\arg1,\arg2
> +	  .else
> +	   \opc _$.zero\arg1,\arg2
> +	  .endif
> +	 .else
> +	  .if .Lsym
> +	   \opc \arg1,\arg2,\arg3
> +	  .elseif .Lnum
> +	   \opc _$.zero+\arg1,\arg2,\arg3
> +	  .else
> +	   \opc _$.zero\arg1,\arg2,\arg3
> +	  .endif
> +	 .endif
> +	.endm
> +
> +	.endif
> +#endif

Okay, I have absolutely no idea what this macro either *does* or what
it's *supposed to do* nor if it matches... you kind of forgot to
describe that.  The other bit is that this whole handling with .if and
#if is just too ugly to live.  Create two include files at the very minimum.

I'd like to figure out if there isn't a better idea to do what you're
trying to do here, though...
	
	-hpa