Re: [PATCH 01/10] x86: add support for multiple choice alternatives

From: Avi Kivity <avi@redhat.com>
To: Luca Barbieri <luca@luca-barbieri.com>
Cc: mingo@elte.hu, hpa@zytor.com, a.p.zijlstra@chello.nl,
	akpm@linux-foundation.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 01/10] x86: add support for multiple choice alternatives
Date: Wed, 17 Feb 2010 14:47:29 +0200	[thread overview]
Message-ID: <4B7BE561.3070403@redhat.com> (raw)
In-Reply-To: <1266406962-17463-2-git-send-email-luca@luca-barbieri.com>

On 02/17/2010 01:42 PM, Luca Barbieri wrote:
> This patch modifies the x86 alternative macros to allow more than one
> alternative code sequence.
>
> This is done by simply adding multiple alternative patches, which are
> applied in sequence, overwriting previous ones.
>
> +/* alternative assembly primitive: */
> +#define ALTERNATIVE(oldinstr, newinstr, feature)			\
> +      "661:\n\t" oldinstr "\n662:\n"					\
> +      ALTERNATIVE_PATCH("661b", "662b", newinstr, feature)
> +
> +#define ALTERNATIVE3(oldinstr, newinstr1, feature1, newinstr2, feature2)			\
> +      "661:\n\t" oldinstr "\n662:\n"					\
> +      ALTERNATIVE_PATCH("661b", "662b", newinstr1, feature1) "\n" \
> +      ALTERNATIVE_PATCH("661b", "662b", newinstr2, feature2)
> +
> +#define ALTERNATIVE4(oldinstr, newinstr1, feature1, newinstr2, feature2, newinstr3, feature3)			\
> +      "661:\n\t" oldinstr "\n662:\n"					\
> +      ALTERNATIVE_PATCH("661b", "662b", newinstr1, feature1) "\n" \
> +      ALTERNATIVE_PATCH("661b", "662b", newinstr2, feature2) "\n" \
> +      ALTERNATIVE_PATCH("661b", "662b", newinstr3, feature3)
> +
>    

Suggest documenting the precedence of alternatives: if both feature1 and 
feature2 are present, which newinstr is patched in?

-- 
error compiling committee.c: too many arguments to function