RFCL cpuset APIs

Gedare Bloom gedare at rtems.org
Sat Oct 12 01:06:19 UTC 2013

I think it would be better to have a clean implementation than to
carry around a bunch of CPP and have to deal with them anytime we want
to synchronize with upstream.

On Fri, Oct 11, 2013 at 8:46 PM, Joel Sherrill
<Joel.Sherrill at oarcorp.com> wrote:
> No more than a couple of hundred lines in a .h file for cpu set.. I don't think the FreeBSD version is much more than that and it has kernel prototypes we don't need.
> Gedare Bloom <gedare at rtems.org> wrote:
> How much code are we talking about here? I would lean toward a clean
> RTEMS implementation. Is this going to be implemented in the score?
> -Gedare
> On Fri, Oct 11, 2013 at 5:06 PM, Joel Sherrill
> <joel.sherrill at oarcorp.com> wrote:
>> Hi
>> We are finishing up a proposal for pthread affinity APIs. We
>> think we should follow the Linux API but have our own
>> implementation.  I am trying to use the FreeBSD cpu set code
>> with minimal ifdefs, but there are differences which are
>> adding up to the point where I have a problem. Do I put in
>> lots of ifdef's? Or just create our own version using
>> the BSD version as inspiration and code fragements?
>> The first difference is Linux uses cpu_set_t while
>> FreeBSD uses cpuset_t. This can be accounted for with
>> a typedef. A global search and replace or using ifdefs
>> will result in a file with lots of changes. Assume a
>> typedef is OK since it minimizes impact of that.
>> Then you have supporting macros.  The following summarizes
>> the similarities and differences.
>> These macros appear to be the same on Linux and FreeBSD.
>>   void CPU_ZERO(cpu_set_t *set);
>>   void CPU_SET(int cpu, cpu_set_t *set);
>>   void CPU_CLR(int cpu, cpu_set_t *set);
>>   int  CPU_ISSET(int cpu, cpu_set_t *set);
>> For comparing sets, Linux has CPU_EQUAL while BSD has CPU_CMP.
>> The signature is the same so a renaming macro would address this.
>>   int  CPU_EQUAL(cpu_set_t *set1, cpu_set_t *set2);
>> FreeBSD has these extra:
>>   void CPU_FILL(cpu_set_t *set)
>>   bool CPU_EMPTY(cpu_set_t *set) /* is empty? */
>>   bool CPU_SUBSET(cpu_set_t *p,cpu_set_t *c) /* is c a subset of p? */
>>   bool CPU_OVERLAP(cpu_set_t *p,cpu_set_t *c) /* do c and p overlap? */
>> And these differ between FreeBSD and Linux. The Linux API has three
>> parameters and is performs bitwise operations on the two sources
>> and puts the result in destination. FreeBSD has an "|=" or "&="
>> and takes destination and the set to operate with. This is the
>> Linux API. Just drop a parameter to get FreeBSD but that makes
>> them incompatible.
>>   void CPU_AND(cpu_set_t *destset,
>>      cpu_set_t *srcset1, cpu_set_t *srcset2);
>>   void CPU_OR(cpu_set_t *destset,
>>      cpu_set_t *srcset1, cpu_set_t *srcset2);
>>   void CPU_XOR(cpu_set_t *destset,
>>      cpu_set_t *srcset1, cpu_set_t *srcset2);
>> FreeBSD also has NAND.
>> Comments, suggestions, thoughts welcomed.
>> --
>> Joel Sherrill, Ph.D.             Director of Research & Development
>> joel.sherrill at OARcorp.com        On-Line Applications Research
>> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>> Support Available                (256) 722-9985
>> _______________________________________________
>> rtems-devel mailing list
>> rtems-devel at rtems.org
>> http://www.rtems.org/mailman/listinfo/rtems-devel

More information about the devel mailing list