Hi all,
Greetings from me!
I want to use "perf stat" command to do a statistic of
"MEM_UOP_RETIRED.ALL_LOADS" and "MEM_UOP_RETIRED.ALL_STORES" of the
program:
#include <stdio.h>
#define MAX_ARRAY_SIZE (10000000)
double a[MAX_ARRAY_SIZE];
double b[MAX_ARRAY_SIZE];
double c[MAX_ARRAY_SIZE];
int main(void)
{
ssize_t i = 0;
for (i = 0; i < MAX_ARRAY_SIZE; i++)
{
c[i] = a[i] + b[i];
}
return 0;
}
The disassemble code of the program (no Optimization):
Dump of assembler code for function main:
0x00000000004004f0 <+0>: push %rbp
0x00000000004004f1 <+1>: mov %rsp,%rbp
0x00000000004004f4 <+4>: movq $0x0,-0x8(%rbp)
0x00000000004004fc <+12>: movq $0x0,-0x8(%rbp)
0x0000000000400504 <+20>: jmp 0x400536 <main+70>
0x0000000000400506 <+22>: mov -0x8(%rbp),%rax
0x000000000040050a <+26>: movsd 0x9e97860(,%rax,8),%xmm1
0x0000000000400513 <+35>: mov -0x8(%rbp),%rax
0x0000000000400517 <+39>: movsd 0x601060(,%rax,8),%xmm0
0x0000000000400520 <+48>: addsd %xmm1,%xmm0
0x0000000000400524 <+52>: mov -0x8(%rbp),%rax
0x0000000000400528 <+56>: movsd %xmm0,0x524c460(,%rax,8)
0x0000000000400531 <+65>: addq $0x1,-0x8(%rbp)
0x0000000000400536 <+70>: cmpq $0x98967f,-0x8(%rbp)
0x000000000040053e <+78>: jle 0x400506 <main+22>
0x0000000000400540 <+80>: mov $0x0,%eax
0x0000000000400545 <+85>: pop %rbp
0x0000000000400546 <+86>: retq
End of assembler dump.
The loop statement code is:
0x0000000000400506 <+22>: mov -0x8(%rbp),%rax
0x000000000040050a <+26>: movsd 0x9e97860(,%rax,8),%xmm1
0x0000000000400513 <+35>: mov -0x8(%rbp),%rax
0x0000000000400517 <+39>: movsd 0x601060(,%rax,8),%xmm0
0x0000000000400520 <+48>: addsd %xmm1,%xmm0
0x0000000000400524 <+52>: mov -0x8(%rbp),%rax
0x0000000000400528 <+56>: movsd %xmm0,0x524c460(,%rax,8)
0x0000000000400531 <+65>: addq $0x1,-0x8(%rbp)
0x0000000000400536 <+70>: cmpq $0x98967f,-0x8(%rbp)
The number of load operations is 7:
0x0000000000400506 <+22>: mov -0x8(%rbp),%rax
0x000000000040050a <+26>: movsd 0x9e97860(,%rax,8),%xmm1
0x0000000000400513 <+35>: mov -0x8(%rbp),%rax
0x0000000000400517 <+39>: movsd 0x601060(,%rax,8),%xmm0
0x0000000000400524 <+52>: mov -0x8(%rbp),%rax
0x0000000000400531 <+65>: addq $0x1,-0x8(%rbp)
0x0000000000400536 <+70>: cmpq $0x98967f,-0x8(%rbp)
The number of store operations is 2:
0x0000000000400528 <+56>: movsd %xmm0,0x524c460(,%rax,8)
0x0000000000400531 <+65>: addq $0x1,-0x8(%rbp)
So the number of total load operations is about 70,000,000, and the
number of total store operations is about 20,000,000.
After executing "perf stat" command, the output is:
[root@Linux test]# perf stat -a -e "r81d0","r82d0" ./a
Performance counter stats for './a':
71,779,954 r81d0
[100.00%]
26,601,675 r82d0
0.036929715 seconds time elapsed
I can see there are about 1,779,954(71,779,954 - 70,000,000) more load
operations and 6,601,675(26,601,675 - 20,000,000) more store
operations.
I can't understand why there are so many more load/store operations
from "perf stat" output. Could anyone can explain this? Thanks very
much in advance!
--
Best Regards
Nan Xiao
--
To unsubscribe from this list: send the line "unsubscribe linux-perf-users" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
