In regards to bug 157084, I can answer how much performance the MEMSET
macro gains you:
[EMAIL PROTECTED]:test-memset$ cat test-memset.c
#include <string.h>
int set(char *x) {
memset(x, 0, 10);
}
int main() {
char buf[10];
set(buf);
}
[EMAIL PROTECTED]:test-memset$ gcc -O2 -c test-memset.c
[EMAIL PROTECTED]:test-memset$ objdump -d test-memset.o
test-memset.o: file format elf64-x86-64
Disassembly of section .text:
0000000000000000 <set>:
0: 48 c7 07 00 00 00 00 movq $0x0,(%rdi)
7: 66 c7 47 08 00 00 movw $0x0,0x8(%rdi)
d: c3 retq
e: 66 data16
f: 90 nop
0000000000000010 <main>:
10: 48 83 ec 18 sub $0x18,%rsp
14: 48 89 e7 mov %rsp,%rdi
17: e8 00 00 00 00 callq 1c <main+0xc>
1c: 48 83 c4 18 add $0x18,%rsp
20: c3 retq
That's right. GCC knows what memset does, and doesn't even generate a
function call. It generates optimal assembly based upon the amount of
memory being set.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
