Re: [PATCH 2/2] powerpc: Add optimized rawmemchr for POWER10

[Matheus Castanho <msc@linux.ibm.com>]

Raphael M Zinsly <rzinsly@linux.ibm.com> writes:

> Hi Matheus, adding the commentary bellow the patch LGTM, thanks!
>
> On 12/05/2021 11:27, Matheus Castanho via Libc-alpha wrote:
>> ...
>> @@ -50,19 +90,13 @@
>>   	li	  r6,offset;		    \
>>   	LXVP(v4+32,offset,addr);	    \
>>   	LXVP(v6+32,offset+32,addr);	    \
>> +	RAWMEMCHR_SUBTRACT_VECTORS;	    \
>>   	vminub	  v14,v4,v5;		    \
>>   	vminub	  v15,v6,v7;		    \
>>   	vminub	  v16,v14,v15;		    \
>> -	vcmpequb. v0,v16,v18;		    \
>> +	vcmpequb. v0,v16,VREG_ZERO;	    \
>>   	bne	  cr6,L(label)
>> -#define TAIL(vreg,increment)	   \
>> -	vctzlsbb  r4,vreg;	   \
>> -	subf	  r3,r3,r5;	   \
>> -	addi	  r4,r4,increment; \
>> -	add	  r3,r3,r4;	   \
>> -	blr
>> -
>>   /* Implements the function
>>      int [r3] strlen (const void *s [r3])
>
> Missing the rawmemchar "definition" here.
>
>
> Regards,

What about this?

    /* Implements the function

       int [r3] strlen (const void *s [r3])

       but when USE_AS_RAWMEMCHR is set, implements the function

       void* [r3] rawmemchr (const void *s [r3], int c [r4])

       The implementation can load bytes past a matching byte, but only
       up to the next 64B boundary, so it never crosses a page.  */

--
Matheus Castanho