From: "Xiao Zeng" <zengxiao@eswincomputing.com>
To: "Jeff Johnston" <jjohnstn@redhat.com>
Cc: "newlib@sourceware.org" <newlib@sourceware.org>,
"palmer@rivosinc.com" <palmer@rivosinc.com>,
jeffreyalaw <jeffreyalaw@gmail.com>,
"Torbjorn SVENSSON" <torbjorn.svensson@foss.st.com>
Subject: Re: Re: [PATCH] newlib: libc: Improved the readability of strpbrk with minor optimization
Date: Thu, 21 Dec 2023 09:16:08 +0800 [thread overview]
Message-ID: <202312210916074865474@eswincomputing.com> (raw)
In-Reply-To: <CAOox84sw=FaVHqW9K=yoZnjiaDfoo4-BpOOncDiNWU0akT6yGg@mail.gmail.com>
2023-12-21 06:48 Jeff Johnston <jjohnstn@redhat.com> 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 <zengxiao@eswincomputing.com>
>wrote:
>
>> Signed-off-by: Xiao Zeng <zengxiao@eswincomputing.com>
>> ---
>> 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
next prev parent reply other threads:[~2023-12-21 1:16 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-20 6:10 Xiao Zeng
2023-12-20 6:25 ` Xiao Zeng
2023-12-20 22:48 ` Jeff Johnston
2023-12-21 1:16 ` Xiao Zeng [this message]
2023-12-21 19:05 ` Jeff Johnston
2023-12-22 1:21 ` Xiao Zeng
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=202312210916074865474@eswincomputing.com \
--to=zengxiao@eswincomputing.com \
--cc=jeffreyalaw@gmail.com \
--cc=jjohnstn@redhat.com \
--cc=newlib@sourceware.org \
--cc=palmer@rivosinc.com \
--cc=torbjorn.svensson@foss.st.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).