perf mem 不记录堆上的 mem 负载

Question

我在我的测试 C 程序中使用了 perf mem，因为我想在同一页面上获取 mem 加载（和存储）的次数。

一个。我的 C 程序在这里：

“测试.c”

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

const int TESTSIZE = 1024*10;

int main(){
    int* ptr_obj = (int*)malloc(sizeof(int)*TESTSIZE);
    int count = 0;
    while(count < 1000){
            for (int i=0; i< TESTSIZE; i++) {
                ptr_obj[i] = 12;
            }

            for (int i=1; i < TESTSIZE-1; i++) {
                ptr_obj[i] = ptr_obj[i-1] + ptr_obj[i+1] - ptr_obj[i] + 1;
            }
            
            count ++;
            sleep(1);
    }
return 0;
}

湾。然后我用这个命令构建它：

gcc -O0 -o stest stest.c

C。接下来由 perf mem 记录使用此命令：

perf 内存记录 ./test

大约 5 秒后，我通过 CTRL + C 中断了该过程。

d。我使用此命令获取报告：

perf mem 报告 --stdio -d 测试

这是输出。（仅内存加载）

# To display the perf.data header info, please use --header/--header-only options.
#
# dso: stest
#
# Total Lost Samples: 0
#
# Samples: 94  of event 'cpu/mem-loads,ldlat=30/P'
# Total weight : 5792
# Sort order   : local_weight,mem,sym,dso,symbol_daddr,dso_daddr,snoop,tlb,locked
#
# Overhead       Samples  Local Weight  Memory access             Symbol    Data Symbol             Data Object  Snoop         TLB access              Locked
# ........  ............  ............  ........................  ........  ......................  ...........  ............  ......................  ......
#
     6.22%             8  45            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     5.92%             7  49            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     5.28%             6  51            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     4.87%             6  47            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     4.45%             6  43            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     3.97%             5  46            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     3.45%             4  50            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     2.75%             3  53            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     2.69%             3  52            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     2.49%             3  48            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     2.21%             4  32            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     2.07%             3  40            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.93%             2  56            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.90%             2  55            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.86%             2  54            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.81%             3  35            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.76%             3  34            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.71%             3  33            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.52%             2  44            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.45%             2  42            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.42%             2  41            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.35%             2  39            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     1.00%             1  58            LFB or LFB hit            [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     0.66%             1  38            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     0.64%             1  37            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No
     0.62%             1  36            L1 or L1 hit              [.] main  [.] 0x00007ffd193dbd50  [stack]      None          L1 or L2 hit            No

显然，这里没有堆记录。

我的问题是

    a. why no heap record?
    b. how to get the record.
    c. or maybe my test C program has something wrong

Answer 1

您正在分配非常小的内存（10k）。将其增加到更大的值。

$ head ./stest.c
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

const int TESTSIZE = 1024*1024*1024;

int main(){

在perf record之后，

$ perf mem report --stdio -d stest
# Sort order   : local_weight,mem,sym,dso,symbol_daddr,dso_daddr,snoop,tlb,locked,blocked,local_ins_lat,p_stage_cyc
...
  0.16%             1  4583          Local RAM or RAM hit      [.] main  [.] 0x00007ff5b2808580  anon         Hit           L1 or L2 hit            No       N/A        0        
...