Re: [PATCH] powerpc: merge include/asm/cputable.h

I get the idea now, how about we make CPU_FTR_ALWAYS &  
CPU_FTR_POSSIBLE just #defines and leave it to various sub-archs to  
define CPU_FTR_POSSIBLE if they want to.

I see the classes of for FTR_POSSIBLE: CLASSIC_PPC, 8xx, 4xx, FSL- 
BOOKE, PPC64 (maybe more subclasses here).

The hugh enum while useful, is just really ugly and I can't believe  
it's worth it for the ~60 cases we are using cpu_has_feature() in.

- kumar

On Sep 14, 2005, at 6:58 PM, Arnd Bergmann wrote:

On Middeweken 14 September 2005 21:11, Kumar Gala wrote:

quoted

I not sure I understand what the introduction of the enum's gets us.

It doesn't have to be an enum, it could just as well be a #define,
if we find that to be better in some way (maybe compile-time).

The general idea is to convert run-time checks into compile-time
checks in order to improve the running kernel. If you have

// start code
enum {
    FEATURE_1 = 1,
    FEATURE_2 = 2,
    PLATFORM_1 = FEATURE_1,
    PLATFORM_2 = FEATURE_2,
    PLATFORM_3 = FEATURE_1 | FEATURE_2,
    FEATURE_POSSIBLE =
#ifdef CONFIG_PLATFORM_1
        PLATFORM_1 |
#endif
#ifdef CONFIG_FEATURE_2
        PLATFORM_2 |
#endif
#ifdef CONFIG_FEATURE_3
        PLATFORM_3 |
#endif
        0,
    FEATURE_ALWAYS =
#ifdef CONFIG_PLATFORM_1
        PLATFORM_1 &
#endif
#ifdef CONFIG_PLATFORM_2
        PLATFORM_2 &
#endif
#ifdef CONFIG_PLATFORM_3
        PLATFORM_3 &
#endif
        FEATURE_POSSIBLE,
};

static inline int have_feature(unsigned long feature)
{
    return (FEATURE_ALWAYS & feature) ||
           (FEATURE_POSSIBLE & runtime_feature & feature);
}

int foo();
int bar();
int main(void)
{
    if (have_feature(FEATURE_1))
        return foo();
    if (have_feature(FEATURE_2))
        return bar();
    return 0;
}
// end code

Then gcc will produce optimal object code for any combination
of CONFIG_PLATFORM_{1,2,3}. Of course I have to admit that the
header file is not exactly elegant ;-).

    Arnd <><