Forum Linux.général Probleme de performance.

Posté par .
Tags : aucun
1
25
nov.
2010
Bonjour tout le monde,

je demande votre aide car j'ai un probleme de performance sur linux.

J'ai deux systemes : un ubuntu et un slitaz (c'est une mini distro, je l'utilise car j'ai besoin d'un systeme léger)

Le probleme c'est que slitaz est bien plus lent que ubuntu.

J'ai fais des benchmarks en utilisant geekbench, je vous met les resultats en dessous.

Si vous avez la moindre idée d'ou peut venir le probleme je suis preneur.

slitaz :

System Information
Platform: Linux x86 (32-bit)
Compiler: GCC 4.1.2 20070925 (Red Hat 4.1.2-33)
Operating System: Linux 2.6.36-vigilant i686
Model: Linux PC (Intel Atom N270)
Motherboard: Unknown Motherboard
Processor: Intel Atom N270
Processor ID: GenuineIntel Family 6 Model 28 Stepping 2
Logical Processors: 2
Physical Processors: 1
Processor Frequency: 1.60 GHz
L1 Instruction Cache: 0.00 B
L1 Data Cache: 0.00 B
L2 Cache: 512 KB
L3 Cache: 0.00 B
Bus Frequency: 0.00 Hz
Memory: 996 MB
Memory Type: N/A
SIMD: 1
BIOS: N/A
Processor Model: Intel(R) Atom(TM) CPU N270 @ 1.60GHz
Processor Cores: 2

Integer
Blowfish
single-threaded scalar 536 ||
multi-threaded scalar 707 ||
Text Compress
single-threaded scalar 503 ||
multi-threaded scalar 652 ||
Text Decompress
single-threaded scalar 510 ||
multi-threaded scalar 697 ||
Image Compress
single-threaded scalar 439 |
multi-threaded scalar 565 ||
Image Decompress
single-threaded scalar 340 |
multi-threaded scalar 447 |
Lua
single-threaded scalar 529 ||
multi-threaded scalar 758 |||

Floating Point
Mandelbrot
single-threaded scalar 403 |
multi-threaded scalar 556 ||
Dot Product
single-threaded scalar 730 ||
multi-threaded scalar 1077 ||||
single-threaded vector 685 ||
multi-threaded vector 1088 ||||
LU Decomposition
single-threaded scalar 295 |
multi-threaded scalar 262 |
Primality Test
single-threaded scalar 517 ||
multi-threaded scalar 589 ||
Sharpen Image
single-threaded scalar 838 |||
multi-threaded scalar 1216 ||||
Blur Image
single-threaded scalar 934 |||
multi-threaded scalar 1211 ||||

Memory
Read Sequential
single-threaded scalar 1329 |||||
Write Sequential
single-threaded scalar 1268 |||||
Stdlib Allocate
single-threaded scalar 578 ||
Stdlib Write
single-threaded scalar 442 |
Stdlib Copy
single-threaded scalar 842 |||

Stream
Stream Copy
single-threaded scalar 878 |||
single-threaded vector 1031 ||||
Stream Scale
single-threaded scalar 520 ||
single-threaded vector 1000 ||||
Stream Add
single-threaded scalar 739 ||
single-threaded vector 1073 ||||
Stream Triad
single-threaded scalar 567 ||
single-threaded vector 892 |||

Integer Score: 556 ||
Floating Point Score: 742 ||
Memory Score: 891 |||
Stream Score: 837 |||

Overall Geekbench Score: 716 ||

______________________________________________________________________________
ubuntu :

System Information
Platform: Linux x86 (32-bit)
Compiler: GCC 4.1.2 20070925 (Red Hat 4.1.2-33)
Operating System: Ubuntu 10.04 LTS 2.6.32-21-generic i686
Model: Linux PC (Intel Atom N270)
Motherboard: Unknown Motherboard
Processor: Intel Atom N270
Processor ID: GenuineIntel Family 6 Model 28 Stepping 2
Logical Processors: 2
Physical Processors: 1
Processor Frequency: 1.60 GHz
L1 Instruction Cache: 0.00 B
L1 Data Cache: 0.00 B
L2 Cache: 512 KB
L3 Cache: 0.00 B
Bus Frequency: 0.00 Hz
Memory: 993 MB
Memory Type: N/A
SIMD: 1
BIOS: N/A
Processor Model: Intel(R) Atom(TM) CPU N270 @ 1.60GHz
Processor Cores: 2

Integer
Blowfish
single-threaded scalar 716 ||
multi-threaded scalar 1192 ||||
Text Compress
single-threaded scalar 735 ||
multi-threaded scalar 1049 ||||
Text Decompress
single-threaded scalar 760 |||
multi-threaded scalar 1132 ||||
Image Compress
single-threaded scalar 626 ||
multi-threaded scalar 934 |||
Image Decompress
single-threaded scalar 500 ||
multi-threaded scalar 717 ||
Lua
single-threaded scalar 755 |||
multi-threaded scalar 1118 ||||

Floating Point
Mandelbrot
single-threaded scalar 484 |
multi-threaded scalar 913 |||
Dot Product
single-threaded scalar 938 |||
multi-threaded scalar 1709 ||||||
single-threaded vector 878 |||
multi-threaded vector 1869 |||||||
LU Decomposition
single-threaded scalar 336 |
multi-threaded scalar 507 ||
Primality Test
single-threaded scalar 699 ||
multi-threaded scalar 953 |||
Sharpen Image
single-threaded scalar 1031 ||||
multi-threaded scalar 1793 |||||||
Blur Image
single-threaded scalar 1085 ||||
multi-threaded scalar 1933 |||||||

Memory
Read Sequential
single-threaded scalar 1829 |||||||
Write Sequential
single-threaded scalar 1580 ||||||
Stdlib Allocate
single-threaded scalar 882 |||
Stdlib Write
single-threaded scalar 1266 |||||
Stdlib Copy
single-threaded scalar 2548 ||||||||||

Stream
Stream Copy
single-threaded scalar 1155 ||||
single-threaded vector 1245 ||||
Stream Scale
single-threaded scalar 719 ||
single-threaded vector 1217 ||||
Stream Add
single-threaded scalar 891 |||
single-threaded vector 1407 |||||
Stream Triad
single-threaded scalar 697 ||
single-threaded vector 1043 ||||

Integer Score: 852 |||
Floating Point Score: 1080 ||||
Memory Score: 1621 ||||||
Stream Score: 1046 ||||

Overall Geekbench Score: 1104 ||||
  • # Il y a peut-être autre chose qui utilise des ressources

    Posté par (page perso) . Évalué à 1.

    Il serait intéressant de jeter un coup d'oeil aux processus qui tournent sur chacun des systèmes quand tu lances ton bench (ps, top...)
    • [^] # Re: Il y a peut-être autre chose qui utilise des ressources

      Posté par . Évalué à 2.

      Bien vu,

      j'avais oublié quelques processus lourds...

      j'ai refais le benchmark avec 0 charge.

      Je le met en dessous.

      Il montre que y'a un probleme au niveau de la mémoire. Quelques sont les optimisations possible pour les acces RAM ?

      System Information
      Platform: Linux x86 (32-bit)
      Compiler: GCC 4.1.2 20070925 (Red Hat 4.1.2-33)
      Operating System: Linux 2.6.36-vigilant i686
      Model: Linux PC (Intel Atom N270)
      Motherboard: Unknown Motherboard
      Processor: Intel Atom N270
      Processor ID: GenuineIntel Family 6 Model 28 Stepping 2
      Logical Processors: 2
      Physical Processors: 1
      Processor Frequency: 1.60 GHz
      L1 Instruction Cache: 0.00 B
      L1 Data Cache: 0.00 B
      L2 Cache: 512 KB
      L3 Cache: 0.00 B
      Bus Frequency: 0.00 Hz
      Memory: 996 MB
      Memory Type: N/A
      SIMD: 1
      BIOS: N/A
      Processor Model: Intel(R) Atom(TM) CPU N270 @ 1.60GHz
      Processor Cores: 2

      Integer
      Blowfish
      single-threaded scalar 717 ||
      multi-threaded scalar 1194 ||||
      Text Compress
      single-threaded scalar 725 ||
      multi-threaded scalar 1049 ||||
      Text Decompress
      single-threaded scalar 784 |||
      multi-threaded scalar 1162 ||||
      Image Compress
      single-threaded scalar 626 ||
      multi-threaded scalar 917 |||
      Image Decompress
      single-threaded scalar 499 |
      multi-threaded scalar 704 ||
      Lua
      single-threaded scalar 771 |||
      multi-threaded scalar 1130 ||||

      Floating Point
      Mandelbrot
      single-threaded scalar 485 |
      multi-threaded scalar 914 |||
      Dot Product
      single-threaded scalar 939 |||
      multi-threaded scalar 1703 ||||||
      single-threaded vector 879 |||
      multi-threaded vector 1850 |||||||
      LU Decomposition
      single-threaded scalar 338 |
      multi-threaded scalar 341 |
      Primality Test
      single-threaded scalar 716 ||
      multi-threaded scalar 971 |||
      Sharpen Image
      single-threaded scalar 1033 ||||
      multi-threaded scalar 1801 |||||||
      Blur Image
      single-threaded scalar 1087 ||||
      multi-threaded scalar 1929 |||||||

      Memory
      Read Sequential
      single-threaded scalar 1767 |||||||
      Write Sequential
      single-threaded scalar 1526 ||||||
      Stdlib Allocate
      single-threaded scalar 808 |||
      Stdlib Write
      single-threaded scalar 503 ||
      Stdlib Copy
      single-threaded scalar 1007 ||||

      Stream
      Stream Copy
      single-threaded scalar 1121 ||||
      single-threaded vector 1214 ||||
      Stream Scale
      single-threaded scalar 720 ||
      single-threaded vector 1191 ||||
      Stream Add
      single-threaded scalar 887 |||
      single-threaded vector 1367 |||||
      Stream Triad
      single-threaded scalar 699 ||
      single-threaded vector 1014 ||||

      Integer Score: 856 |||
      Floating Point Score: 1070 ||||
      Memory Score: 1122 ||||
      Stream Score: 1026 ||||

      Overall Geekbench Score: 1001 ||||
      • [^] # Re: Il y a peut-être autre chose qui utilise des ressources

        Posté par . Évalué à 2.

        un diff des .config des kernel ?
      • [^] # Re: Il y a peut-être autre chose qui utilise des ressources

        Posté par (page perso) . Évalué à 2.

        Apparemment, il y a de grosses différences dans les stdlib allocate/write/copy.
        Jette un coup d'oeil aux versions des librairies C.
      • [^] # Re: Il y a peut-être autre chose qui utilise des ressources

        Posté par . Évalué à 1.

        Un top indique aussi que le programme que je fais tourner prend 5.5% de mémoire sur ubuntu et 29% sur slitaz.

        j'ai fait un cat /proc/meminfo et voila ce que ça donne :

        Slitaz :

        MemTotal: 1020288 kB
        MemFree: 897312 kB
        Buffers: 1668 kB
        Cached: 56240 kB
        SwapCached: 0 kB
        Active: 58664 kB
        Inactive: 51092 kB
        Active(anon): 51848 kB
        Inactive(anon): 0 kB
        Active(file): 6816 kB
        Inactive(file): 51092 kB
        Unevictable: 0 kB
        Mlocked: 0 kB
        HighTotal: 129608 kB
        HighFree: 18280 kB
        LowTotal: 890680 kB
        LowFree: 879032 kB
        SwapTotal: 0 kB
        SwapFree: 0 kB
        Dirty: 4 kB
        Writeback: 0 kB
        AnonPages: 51924 kB
        Mapped: 23320 kB
        Shmem: 0 kB
        Slab: 5124 kB
        SReclaimable: 1484 kB
        SUnreclaim: 3640 kB
        KernelStack: 632 kB
        PageTables: 660 kB
        NFS_Unstable: 0 kB
        Bounce: 0 kB
        WritebackTmp: 0 kB
        CommitLimit: 510144 kB
        Committed_AS: 133060 kB
        VmallocTotal: 122880 kB
        VmallocUsed: 2788 kB
        VmallocChunk: 117428 kB
        DirectMap4k: 909304 kB
        DirectMap4M: 0 kB


        Ubuntu :

        MemTotal: 1016908 kB
        MemFree: 791340 kB
        Buffers: 48940 kB
        Cached: 97500 kB
        SwapCached: 0 kB
        Active: 111408 kB
        Inactive: 90356 kB
        Active(anon): 55656 kB
        Inactive(anon): 0 kB
        Active(file): 55752 kB
        Inactive(file): 90356 kB
        Unevictable: 0 kB
        Mlocked: 0 kB
        HighTotal: 129608 kB
        HighFree: 252 kB
        LowTotal: 887300 kB
        LowFree: 791088 kB
        SwapTotal: 230392 kB
        SwapFree: 230392 kB
        Dirty: 4 kB
        Writeback: 0 kB
        AnonPages: 55320 kB
        Mapped: 28504 kB
        Shmem: 336 kB
        Slab: 10184 kB
        SReclaimable: 5232 kB
        SUnreclaim: 4952 kB
        KernelStack: 1288 kB
        PageTables: 1032 kB
        NFS_Unstable: 0 kB
        Bounce: 0 kB
        WritebackTmp: 0 kB
        CommitLimit: 738844 kB
        Committed_AS: 155868 kB
        VmallocTotal: 122880 kB
        VmallocUsed: 12896 kB
        VmallocChunk: 99356 kB
        HardwareCorrupted: 0 kB
        HugePages_Total: 0
        HugePages_Free: 0
        HugePages_Rsvd: 0
        HugePages_Surp: 0
        Hugepagesize: 4096 kB
        DirectMap4k: 16376 kB
        DirectMap4M: 892928 kB
  • # Faire remonter

    Posté par (page perso) . Évalué à 3.

    Tu as pensé à faire remonter aux gens qui font SliTaz? Ils seront sûrement intéressés...
    • [^] # Re: Faire remonter

      Posté par . Évalué à 1.

      Oui j'ai un thread chez eux.

      Mais rien n'indique que ce soit un probleme slitaz en fait... Etant donné que j'utilise un kernel custom, ça vient probablement de la.
      • [^] # Re: Faire remonter

        Posté par . Évalué à 2.

        Etant donné que j'utilise un kernel custom, ça vient probablement de la

        kernel custom des deux cotés (slitaz et ubuntu ?)

        si oui, meme version, meme options à la compilation ?
        • [^] # Re: Faire remonter

          Posté par . Évalué à 1.

          Kernel custom uniquement sur slitaz, kernel ubuntu sur ubuntu.

          Apparement ce serai un probleme de cache.

          Quelqu'un connait quelles options du kernel peuvent influencer la taille du cache, le nombre d'objets slab disponibles ?
          • [^] # Re: Faire remonter

            Posté par . Évalué à 3.

            prend le .config du ubuntu et met le sur ton slitaz avant de recompiler
            • [^] # Re: Faire remonter

              Posté par . Évalué à 1.

              Oui j'ai fais ça, ça ne résoud pas le pb.

              Du coup si ça ne vient pas du kernel, je sais pas d'ou ça peut venir.
              • [^] # Re: Faire remonter

                Posté par (page perso) . Évalué à 2.

                La libc? Tu avais l'air de dire que les deux distributions n'utilisent pas la même implémentation.
                • [^] # Re: Faire remonter

                  Posté par . Évalué à 2.

                  J'avais changé la libc, j'ai essayé une autre glibc et une eglibc, sans amélioration.
  • # config des kernels

    Posté par . Évalué à 1.

    un diff des .config des kernels : slitaz a gauche, ubuntu a droite.

    http://www.fileshost.com/download.php?id=80ED64E91


    --------------------------------------
    http://Proxy1gratuit.blogspot.com

Suivre le flux des commentaires

Note : les commentaires appartiennent à ceux qui les ont postés. Nous n'en sommes pas responsables.