[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-Id: <20230531193231.58732-1-minhuadotchen@gmail.com>
Date: Thu, 1 Jun 2023 03:32:31 +0800
From: Min-Hua Chen <minhuadotchen@...il.com>
To: david.laight@...lab.com
Cc: jmorris@...ei.org, linux-kernel@...r.kernel.org,
linux-security-module@...r.kernel.org, minhuadotchen@...il.com,
paul@...l-moore.com, serge@...lyn.com
Subject: RE: [PATCH] capabilities: use logical OR
>From: Min-Hua Chen <minhuadotchen@...il.com>
>> Sent: 29 May 2023 23:55
>>
>> Use logical OR to fix the following sparse warnings:
>>
>> security/commoncap.c:1358:41: sparse: warning: dubious: !x | y
>>
>> No functional changes intended.
>
>Except it will run just that teeny, weeny bit slower.
>
> David
>
Thanks for the comment.
I wrote a test case on my x86 machine to see the difference
between '|' and '||' versions.
The '|' version does 2 OR operations and 1 test. (all cases)
The '||' (logical or) version does 3 cmpl (worst case)
and 1 cmpl and 1 jump if the first argument is TRUE (best case).
For the readers, I think the logical OR version is clearer.
thanks,
Min-Hua
my test case:
void foo(void)
{
}
void bar(void)
{
}
int main(void)
{
int a, b, c;
if (a || b || c)
foo();
else
bar();
}
if (a || b || c)
logical OR
Dump of assembler code for function main:
0x000000000000113f <+0>: endbr64
0x0000000000001143 <+4>: push %rbp
0x0000000000001144 <+5>: mov %rsp,%rbp
0x0000000000001147 <+8>: sub $0x10,%rsp
0x000000000000114b <+12>: cmpl $0x0,-0xc(%rbp)
0x000000000000114f <+16>: jne 0x115d <main+30>
0x0000000000001151 <+18>: cmpl $0x0,-0x8(%rbp)
0x0000000000001155 <+22>: jne 0x115d <main+30>
0x0000000000001157 <+24>: cmpl $0x0,-0x4(%rbp)
0x000000000000115b <+28>: je 0x1164 <main+37>
0x000000000000115d <+30>: call 0x1129 <foo>
0x0000000000001162 <+35>: jmp 0x1169 <main+42>
0x0000000000001164 <+37>: call 0x1134 <bar>
0x0000000000001169 <+42>: mov $0x0,%eax
0x000000000000116e <+47>: leave
0x000000000000116f <+48>: ret
End of assembler dump.
if (a | b | c)
Dump of assembler code for function main:
0x000000000000113f <+0>: endbr64
0x0000000000001143 <+4>: push %rbp
0x0000000000001144 <+5>: mov %rsp,%rbp
0x0000000000001147 <+8>: sub $0x10,%rsp
0x000000000000114b <+12>: mov -0xc(%rbp),%eax
0x000000000000114e <+15>: or -0x8(%rbp),%eax
0x0000000000001151 <+18>: or -0x4(%rbp),%eax
0x0000000000001154 <+21>: test %eax,%eax
0x0000000000001156 <+23>: je 0x115f <main+32>
0x0000000000001158 <+25>: call 0x1129 <foo>
0x000000000000115d <+30>: jmp 0x1164 <main+37>
0x000000000000115f <+32>: call 0x1134 <bar>
0x0000000000001164 <+37>: mov $0x0,%eax
0x0000000000001169 <+42>: leave
0x000000000000116a <+43>: ret
End of assembler dump.
Powered by blists - more mailing lists