Sistemi di Calcolo

Corso di Laurea in Ingegneria Informatica e Automatica - A.A. 2017-2018

HomePage | Avvisi | Diario lezioni | Programma | Materiale didattico | Esami | Forum | Login

Lunedì 30 ottobre e venerdì 3 novembre 2017 (Lezione 8 - 150 min) - Demetrescu/Coppa


Profilazione delle prestazioni con gprof

Esempio di programma da profilare con gprof:

prova.c
#include <string.h>
#include <stdlib.h>

#define M 1000000000
#define N 10000

void init(int *v, int n) {
    memset(v, n*sizeof(int), 0xFF);
}

void A(int *v, int n) {
    int i = 0;
    for (i=0; i<n; i++) v[i] = v[i] / 2; // provare anche con v[i]/291
}

void B(int *v, int n) {
    int i = 0;
    for (i=0; i<n; i++) v[i] = v[i] >> 1;
}

int main() {
    int* v = malloc(N*sizeof(int)), i;
    init(v, N);
    for (i=0; i<M/N; i++) {
        A(v, N);
        B(v, N);
    }
    free(v);
    return 0;
}


Sessione di profilazione:

> gcc -pg prova.c -o prova -O1 -m32
> ./prova
> gprof ./prova > report.txt


Contenuto del file report.txt generato:

report.txt
Flat profile:

Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total          
 time   seconds   seconds    calls  us/call  us/call  name    
 50.00      0.45     0.45   100000     4.50     4.50  A
 50.00      0.90     0.45   100000     4.50     4.50  B
  0.00      0.90     0.00        1     0.00     0.00  init

 %         the percentage of the total running time of the
time       program used by this function.

cumulative a running sum of the number of seconds accounted
 seconds   for by this function and those listed above it.

 self      the number of seconds accounted for by this
seconds    function alone.  This is the major sort for this
           listing.

calls      the number of times this function was invoked, if
           this function is profiled, else blank.
 
 self      the average number of milliseconds spent in this
ms/call    function per call, if this function is profiled,
       else blank.

 total     the average number of milliseconds spent in this
ms/call    function and its descendents per call, if this
       function is profiled, else blank.

name       the name of the function.  This is the minor sort
           for this listing. The index shows the location of
       the function in the gprof listing. If the index is
       in parenthesis it shows where it would appear in
       the gprof listing if it were to be printed.

Copyright (C) 2012 Free Software Foundation, Inc.

Copying and distribution of this file, with or without modification,
are permitted in any medium without royalty provided the copyright
notice and this notice are preserved.

             Call graph (explanation follows)


granularity: each sample hit covers 4 byte(s) for 1.11% of 0.90 seconds

index % time    self  children    called     name
                                                 <spontaneous>
[1]    100.0    0.00    0.90                 main [1]
                0.45    0.00  100000/100000      A [2]
                0.45    0.00  100000/100000      B [3]
                0.00    0.00       1/1           init [4]
-----------------------------------------------
                0.45    0.00  100000/100000      main [1]
[2]     50.0    0.45    0.00  100000         A [2]
-----------------------------------------------
                0.45    0.00  100000/100000      main [1]
[3]     50.0    0.45    0.00  100000         B [3]
-----------------------------------------------
                0.00    0.00       1/1           main [1]
[4]      0.0    0.00    0.00       1         init [4]
-----------------------------------------------

 This table describes the call tree of the program, and was sorted by
 the total amount of time spent in each function and its children.

 Each entry in this table consists of several lines.  The line with the
 index number at the left hand margin lists the current function.
 The lines above it list the functions that called this function,
 and the lines below it list the functions this one called.
 This line lists:
     index  A unique number given to each element of the table.
        Index numbers are sorted numerically.
        The index number is printed next to every function name so
        it is easier to look up where the function is in the table.

     % time This is the percentage of the `total' time that was spent
        in this function and its children.  Note that due to
        different viewpoints, functions excluded by options, etc,
        these numbers will NOT add up to 100%.

     self   This is the total amount of time spent in this function.

     children   This is the total amount of time propagated into this
        function by its children.

     called This is the number of times the function was called.
        If the function called itself recursively, the number
        only includes non-recursive calls, and is followed by
        a `+' and the number of recursive calls.

     name   The name of the current function.  The index number is
        printed after it.  If the function is a member of a
        cycle, the cycle number is printed between the
        function's name and the index number.


 For the function's parents, the fields have the following meanings:

     self   This is the amount of time that was propagated directly
        from the function into this parent.

     children   This is the amount of time that was propagated from
        the function's children into this parent.

     called This is the number of times this parent called the
        function `/' the total number of times the function
        was called.  Recursive calls to the function are not
        included in the number after the `/'.

     name   This is the name of the parent.  The parent's index
        number is printed after it.  If the parent is a
        member of a cycle, the cycle number is printed between
        the name and the index number.

 If the parents of the function cannot be determined, the word
 `<spontaneous>' is printed in the `name' field, and all the other
 fields are blank.

 For the function's children, the fields have the following meanings:

     self   This is the amount of time that was propagated directly
        from the child into the function.

     children   This is the amount of time that was propagated from the
        child's children to the function.

     called This is the number of times the function called
        this child `/' the total number of times the child
        was called.  Recursive calls by the child are not
        listed in the number after the `/'.

     name   This is the name of the child.  The child's index
        number is printed after it.  If the child is a
        member of a cycle, the cycle number is printed
        between the name and the index number.

 If there are any cycles (circles) in the call graph, there is an
 entry for the cycle-as-a-whole.  This entry shows who called the
 cycle (as parents) and the members of the cycle (as children.)
 The `+' recursive calls entry shows the number of function calls that
 were internal to the cycle, and the calls entry for each member shows,
 for that member, how many times it was called from other members of
 the cycle.

Copyright (C) 2012 Free Software Foundation, Inc.

Copying and distribution of this file, with or without modification,
are permitted in any medium without royalty provided the copyright
notice and this notice are preserved.

Index by function name

   [2] A                       [3] B                       [4] init


Esempio di ottimizzazioni del compilatore

test.c
int f(int x) {
    int z = x*3;
    return z;
}


Generazione codice macchina non ottimizzato:

> gcc -m32 test.c -c
> objdump -d test.o
Disassembly of section .text:

00000000 <f>:
   0:   55                      push   %ebp
   1:   89 e5                   mov    %esp,%ebp
   3:   83 ec 10                sub    $0x10,%esp
   6:   8b 55 08                mov    0x8(%ebp),%edx
   9:   89 d0                   mov    %edx,%eax
   b:   01 c0                   add    %eax,%eax
   d:   01 d0                   add    %edx,%eax
   f:   89 45 fc                mov    %eax,-0x4(%ebp)
  12:   8b 45 fc                mov    -0x4(%ebp),%eax
  15:   c9                      leave  
  16:   c3                      ret    


Generazione codice macchina ottimizzato a livello 1:

> gcc -m32 test.c -c -O1
> objdump -d test.o

test.o:     file format elf32-i386


Disassembly of section .text:

00000000 <f>:
   0:   8b 44 24 04             mov    0x4(%esp),%eax
   4:   8d 04 40                lea    (%eax,%eax,2),%eax
   7:   c3                      ret    


Si noti l'uso astuto della lea per effettuare l'operazione eax <- 3*eax. L'istruzione, pensata in generale per calcolare indirizzi, viene qui usata dal compilatore per calcolare espressioni su interi. In fondo gli indirizzi sono numeri, no? Bel trucco.
Valid XHTML 1.0 Transitional :: Valid CSS :: Powered by WikkaWiki
Page was generated in 0.0798 seconds