Re: Re: [PATCH] newlib: libc: Improved the readability of strpbrk with minor optimization

2023-12-22 03:05  Jeff Johnston <[email protected]> wrote:
>
 
>Patch merged.  Thanks. 
Thank you, Jeff.
Perhaps the two emails are too similar, I would like to remind you that there is a similar optimization on
<https://sourceware.org/pipermail/newlib/2023/020771.html>.

>
>-- Jeff J.
>
>On Wed, Dec 20, 2023 at 8:16 PM Xiao Zeng <[email protected]>
>wrote:
>
>> 2023-12-21 06:48  Jeff Johnston <[email protected]> wrote:
>> >
>>
>> >Unless I am missing something, I think the code could be made even simpler
>> >to return the pointer when found and return NULL otherwise.
>> Yes, you have found a more efficient method.
>> After my testing, fewer assembly instructions were generated. This is
>> really exciting.
>>
>> Xiao's patch:
>> ---------------------------------------------------------
>> libc_a-strpbrk.o:     file format elf64-littleriscv
>>
>>
>> Disassembly of section .text:
>>
>> 0000000000000000 <strpbrk>:
>>    0:   00054683                lbu     a3,0(a0)
>>
>> 0000000000000004 <.LVL1>:
>>    4:   04068063                beqz    a3,44 <.L10>
>>    8:   0005c603                lbu     a2,0(a1)
>>    c:   04060063                beqz    a2,4c <.L14>
>>
>> 0000000000000010 <.L5>:
>>   10:   00058793                mv      a5,a1
>>   14:   00060713                mv      a4,a2
>>   18:   00c0006f                j       24 <.L4>
>>
>> 000000000000001c <.L17>:
>>   1c:   0007c703                lbu     a4,0(a5)
>>   20:   00070863                beqz    a4,30 <.L16>
>>
>> 0000000000000024 <.L4>:
>>   24:   00178793                addi    a5,a5,1
>>
>> 0000000000000028 <.LVL5>:
>>   28:   fee69ae3                bne     a3,a4,1c <.L17>
>>
>> 000000000000002c <.L1>:
>>   2c:   00008067                ret
>>
>> 0000000000000030 <.L16>:
>>   30:   00154683                lbu     a3,1(a0)
>>   34:   00150513                addi    a0,a0,1
>>
>> 0000000000000038 <.LVL8>:
>>   38:   fc069ce3                bnez    a3,10 <.L5>
>>
>> 000000000000003c <.L6>:
>>   3c:   0007c783                lbu     a5,0(a5)
>>
>> 0000000000000040 <.LVL9>:
>>   40:   fe0796e3                bnez    a5,2c <.L1>
>>
>> 0000000000000044 <.L10>:
>>   44:   00000513                li      a0,0
>>
>> 0000000000000048 <.LVL11>:
>>   48:   00008067                ret
>>
>> 000000000000004c <.L14>:
>>   4c:   00154683                lbu     a3,1(a0)
>>   50:   00150513                addi    a0,a0,1
>>   54:   fe069ce3                bnez    a3,4c <.L14>
>>   58:   00058793                mv      a5,a1
>>   5c:   fe1ff06f                j       3c <.L6>
>> ---------------------------------------------------------
>>
>> Jeff's patch
>> ---------------------------------------------------------
>> libc_a-strpbrk.o:     file format elf64-littleriscv
>>
>>
>> Disassembly of section .text:
>>
>> 0000000000000000 <strpbrk>:
>>    0:   00054683                lbu     a3,0(a0)
>>    4:   02068c63                beqz    a3,3c <.L8>
>>    8:   0005c603                lbu     a2,0(a1)
>>    c:   02060c63                beqz    a2,44 <.L12>
>>
>> 0000000000000010 <.L5>:
>>   10:   00058793                mv      a5,a1
>>   14:   00060713                mv      a4,a2
>>   18:   00c0006f                j       24 <.L4>
>>
>> 000000000000001c <.L15>:
>>   1c:   0007c703                lbu     a4,0(a5)
>>   20:   00070863                beqz    a4,30 <.L14>
>>
>> 0000000000000024 <.L4>:
>>   24:   00178793                addi    a5,a5,1
>>
>> 0000000000000028 <.LVL4>:
>>   28:   fed71ae3                bne     a4,a3,1c <.L15>
>>   2c:   00008067                ret
>>
>> 0000000000000030 <.L14>:
>>   30:   00154683                lbu     a3,1(a0)
>>   34:   00150513                addi    a0,a0,1
>>
>> 0000000000000038 <.LVL6>:
>>   38:   fc069ce3                bnez    a3,10 <.L5>
>>
>> 000000000000003c <.L8>:
>>   3c:   00000513                li      a0,0
>>
>> 0000000000000040 <.LVL7>:
>>   40:   00008067                ret
>>
>> 0000000000000044 <.L12>:
>>   44:   00154683                lbu     a3,1(a0)
>>   48:   00150513                addi    a0,a0,1
>>   4c:   fe069ce3                bnez    a3,44 <.L12>
>>   50:   fedff06f                j       3c <.L8>
>> ---------------------------------------------------------
>> After careful comparison, it was found that there are fewer assembly
>> instructions after the Jeff's patch.
>>
>> >
>> >index 774db1e..d984745 100644
>> >--- a/newlib/libc/string/strpbrk.c
>> >+++ b/newlib/libc/string/strpbrk.c
>> >@@ -29,23 +29,16 @@ strpbrk (const char *s1,
>> >        const char *s2)
>> > {
>> >   const char *c = s2;
>> >-  if (!*s1)
>> >-    return (char *) NULL;
>> >
>> >   while (*s1)
>> >     {
>> >       for (c = s2; *c; c++)
>> >        {
>> >          if (*s1 == *c)
>> >-           break;
>> >+           return (char *) s1;
>> >        }
>> >-      if (*c)
>> >-       break;
>> >       s1++;
>> >     }
>> >
>> >-  if (*c == '\0')
>> >-    s1 = NULL;
>> >-
>> >-  return (char *) s1;
>> >+  return (char *) NULL;
>> > }
>> >
>> >On Wed, Dec 20, 2023 at 1:12 AM Xiao Zeng <[email protected]>
>> >wrote:
>> >
>> >> Signed-off-by: Xiao Zeng <[email protected]>
>> >> ---
>> >>  newlib/libc/string/strpbrk.c | 5 ++---
>> >>  1 file changed, 2 insertions(+), 3 deletions(-)
>> >>
>> >> diff --git a/newlib/libc/string/strpbrk.c b/newlib/libc/string/strpbrk.c
>> >> index 774db1e6d..95e89c20c 100644
>> >> --- a/newlib/libc/string/strpbrk.c
>> >> +++ b/newlib/libc/string/strpbrk.c
>> >> @@ -37,15 +37,14 @@ strpbrk (const char *s1,
>> >>        for (c = s2; *c; c++)
>> >>         {
>> >>           if (*s1 == *c)
>> >> -           break;
>> >> +           goto end;
>> >>         }
>> >> -      if (*c)
>> >> -       break;
>> >>        s1++;
>> >>      }
>> >>
>> >>    if (*c == '\0')
>> >>      s1 = NULL;
>> >>
>> >> +end:
>> >>    return (char *) s1;
>> >>  }
>> >> --
>> >> 2.17.1
>> >>
>> >>
>>
>> Thanks
>> Xiao Zeng
>>
>>
 
Thanks
Xiao Zeng