Энтузиаст протестировал новейший процессор Loongson 3C5000

Сравнение процессоров Loongson 3A5000 и 3C5000

Сравнение процессоров Loongson 3A5000 и 3C5000

Некоторое время назад с энтузиаст связался с разработчиками процессора Loongson и ему удалось протестировать сервер на процессоре Loongson 3С5000 (сервер содержит 2 процессора с 16 ядрами каждый).

Раннее энтузиаст уже проводил тесты десктопного процессора Loongson 3A5000, поэтому его будем сравнивать с процессором Loongson 3С5000.

Характеристики процессоров Loongson 3A5000 и Loongson 3С5000:

Loongson 3A5000

Loongson 3С5000

Семейство ISA

RISC

RISC

Архитектура

loongarch64

loongarch64

Микроархитектура

GS464V

GS464V

Частота (МГц)

2500

2200

Ядра

4

16

Тех процесс (нм)

12

12

TDP (Вт)

35

150

Кеш

64 Кб L1I, 64 Кб L1D,

1 Мб L2 (0,25×4)

16 Мб L3

64 Кб L1I, 64 Кб L1D,

1 Мб L2 (0,25×4)

64 Мб L3 (16×4)

Тип ОЗУ

DDR4–3200

DDR4–3200

GFLOPs

160

560

Год

2021

2022

Были проведены следующие тесты:

Сразу переходим к результатам, но детали тестов смотрите далее.

Результаты

Тест

Loongson 3A5000

Loongson 3С5000

Dhrystone [DMIPS]

18563

11806

Whetstone [MWIPS]

3990

3480

Whetstone MP [MWIPS]

15001

53408 (16 потоков)

106480 (32 потока)

Linpack 100 [MFLOPS]

2929

2600

Scimark 2 [Composite score]

1487

1464

Coremark (1T; MT)

14934; 58797

12521; 193271

MP MFLOPS

121167

719903

7zip (Comp; Decomp; Tot) (MT)

11360; 11638; 11081

37910; 38132; 38021

STREAM (Copy; Scale; Add; Triad) [MB/s]

14593; 14641; 13276; 13719

56452; 54762; 57358; 57051

SPEC 2006 INT

25

22

По результатам можно сделать вывод, что процессор Loongson 3C5000 примерно сравним с процессором Loongson 3A5000 на равных частотах.

Немного об архитектуре LoongArch

Loongson — процессор на основе сильно переработанной RISC архитектуре MIPS64, часть ненужных команд MIPS64 была удалена и добавлено большое число расширений (SIMD, шифрование, бинарная трансляция, виртуализация), всего порядка 2000+ инструкций.

lscpu 3A5000

Architecture:        loongarch64
Byte Order:          Little Endian
CPU(s):              4
On-line CPU(s) list: 0-3
Thread(s) per core:  1
Core(s) per socket:  4
Socket(s):           1
NUMA node(s):        1
Model name:          Loongson-3A5000
CPU max MHz:         2500.0000
CPU min MHz:         225.0000
BogoMIPS:            4992.00
L1d cache:           64K
L1i cache:           64K
L2 cache:            256K
L3 cache:            16384K
NUMA node0 CPU(s):   0-3

lscpu 3C5000

Architecture:        loongarch64
Byte Order:          Little Endian
CPU(s):              32
On-line CPU(s) list: 0-31
Thread(s) per core:  1
Core(s) per socket:  16
Socket(s):           2
NUMA node(s):        2
CPU family:          Loongson-64bit
Model name:          Loongson-3C5000
BogoMIPS:            4400.00
L1d cache:           64K
L1i cache:           64K
L2 cache:            256K
L3 cache:            16384K
NUMA node0 CPU(s):   0-15
NUMA node1 CPU(s):   16-31
Flags:               cpucfg lam ual fpu lsx lasx complex crypto lvz lbt_x86 lbt_arm lbt_mips

Dhrystone

Dhrystone достаточно древний тест 80х годов, написан на C. Тестирует целочисленную арифметику и работу со строками. Результаты измеряются в Dhrystone/s и DMIPS. (DMIPS = Dhrystone/s делить на 1757).

Вывод теста Dhrystone на 3С5000

Dhrystone Benchmark, Version 2.1 (Language: C or C++)

Optimisation    loongarch64 loongarch64 optimized
Register option not selected

       10000 runs   0.00 seconds 
      100000 runs   0.00 seconds 
     1000000 runs   0.05 seconds 
    10000000 runs   0.48 seconds 
    20000000 runs   0.96 seconds 
    40000000 runs   1.93 seconds 
    80000000 runs   3.86 seconds 

Final values (* implementation-dependent):

Int_Glob:      O.K.  5  Bool_Glob:     O.K.  1
Ch_1_Glob:     O.K.  A  Ch_2_Glob:     O.K.  B
Arr_1_Glob[8]: O.K.  7  Arr_2_Glob8/7: O.K.    80000010
Ptr_Glob->              Ptr_Comp:       *    0x12eac82a0
  Discr:       O.K.  0  Enum_Comp:     O.K.  2
  Int_Comp:    O.K.  17 Str_Comp:      O.K.  DHRYSTONE PROGRAM, SOME STRING
Next_Ptr_Glob->         Ptr_Comp:       *    0x12eac82a0 same as above
  Discr:       O.K.  0  Enum_Comp:     O.K.  1
  Int_Comp:    O.K.  18 Str_Comp:      O.K.  DHRYSTONE PROGRAM, SOME STRING
Int_1_Loc:     O.K.  5  Int_2_Loc:     O.K.  13
Int_3_Loc:     O.K.  7  Enum_Loc:      O.K.  1  
Str_1_Loc:                             O.K.  DHRYSTONE PROGRAM, 1'ST STRING
Str_2_Loc:                             O.K.  DHRYSTONE PROGRAM, 2'ND STRING

Nanoseconds one Dhrystone run:        48.21
Dhrystones per Second:             20744304
VAX  MIPS rating =                 11806.66

Вывод теста Dhrystone на 3A5000

Dhrystone Benchmark, Version 2.1 (Language: C or C++)

Optimisation    loongarch64 loongarch64 optimized
Register option not selected

       10000 runs   0.00 seconds 
      100000 runs   0.00 seconds 
     1000000 runs   0.03 seconds 
    10000000 runs   0.31 seconds 
    20000000 runs   0.61 seconds 
    40000000 runs   1.23 seconds 
    80000000 runs   2.45 seconds 

Final values (* implementation-dependent):

Int_Glob:      O.K.  5  Bool_Glob:     O.K.  1
Ch_1_Glob:     O.K.  A  Ch_2_Glob:     O.K.  B
Arr_1_Glob[8]: O.K.  7  Arr_2_Glob8/7: O.K.    80000010
Ptr_Glob->              Ptr_Comp:       *    0x1239782a0
  Discr:       O.K.  0  Enum_Comp:     O.K.  2
  Int_Comp:    O.K.  17 Str_Comp:      O.K.  DHRYSTONE PROGRAM, SOME STRING
Next_Ptr_Glob->         Ptr_Comp:       *    0x1239782a0 same as above
  Discr:       O.K.  0  Enum_Comp:     O.K.  1
  Int_Comp:    O.K.  18 Str_Comp:      O.K.  DHRYSTONE PROGRAM, SOME STRING
Int_1_Loc:     O.K.  5  Int_2_Loc:     O.K.  13
Int_3_Loc:     O.K.  7  Enum_Loc:      O.K.  1  
Str_1_Loc:                             O.K.  DHRYSTONE PROGRAM, 1'ST STRING
Str_2_Loc:                             O.K.  DHRYSTONE PROGRAM, 2'ND STRING

Nanoseconds one Dhrystone run:        30.66
Dhrystones per Second:             32616185
VAX  MIPS rating =                 18563.57

Whetstone

Тестирует арифметику с плавающей/фиксированной запятой, математические функции, ветвления, вызовов функций, присваиваний, работы с числами с фиксированной запятой, ветвлений. Результаты измеряются в MMIPS.

Вывод теста Whetstone на 3С5000

32 CPUs Available
##############################################

 Multithreading Single Precision Whetstones loongarch64 loongarch64 optimized

             Using 32 threads - Mon Dec 12 16:29:23 2022

 Calibrate
    0.00406 Seconds          1   Passes (x 100)
    0.01889 Seconds          5   Passes (x 100)
    0.09340 Seconds         25   Passes (x 100)
    0.46579 Seconds        125   Passes (x 100)

 Use 2683  passes (x 100)

 MFLOPS 1     596    596    596    596    596    596    596    596    596    596    596    596    596    596    596    596    596    596    596    595    595    595    595    595    595    595    595    590    574    574    554    554
 MFLOPS 2     596    596    596    596    596    596    596    596    596    596    596    596    596    596    596    596    596    595    595    595    595    595    595    595    595    595    595    595    594    586    552    540
 IFMOPS      6592   6591   6591   6590   6588   6587   6587   6586   6585   6583   6583   6582   6582   6581   6579   6579   6578   6576   6575   6574   6572   6571   6570   6569   6567   6566   6565   6563   5701   5650   5432   5387
 FIXPMOPS    2159   2158   2158   2158   2158   2158   2157   2157   2157   2156   2156   2156   2156   2155   2155   2155   2155   2155   2155   2155   2154   2154   2154   2154   2154   2153   2153   2153   2152   2151   1996   1988
 COSMOPS       99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     99     98     98     96     96
 MFLOPS 3     528    528    528    528    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    527    526    521    520    518    517
 EQUMOPS     4396   4395   4395   4395   4395   4395   4394   4394   4394   4394   4394   4393   4393   4393   4393   4393   4393   4393   4393   4392   4392   4392   4392   4392   4392   4392   4392   4391   4390   4365   4359   4335
 EXPMOPS       56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     56     55     55     55     54     54
 millisec    8133   8069   8035   8036   8046   8100   8056   8037   8040   8038   8042   8037   8039   8038   8039   8039   8039   8038   8040   8039   8041   8110   8047   8040   8044   8082   8058   8082   8045   8144   8113   8242
 MWIPS       3299   3325   3339   3339   3335   3312   3331   3338   3337   3338   3336   3338   3337   3338   3338   3338   3337   3338   3337   3338   3337   3308   3334   3337   3335   3320   3330   3320   3335   3294   3307   3255

         MWIPS MFLOPS MFLOPS MFLOPS    Cos    Exp  Fixpt     If  Equal
 Thread             1      2      3   MOPS   MOPS   MOPS   MOPS   MOPS

      1   3299    596    552    518     99     56   2159   6592   4359
      2   3325    596    596    528     99     56   1996   6590   4395
      3   3339    596    596    528     99     56   2158   6591   4396
      4   3339    596    596    528     99     56   2158   6591   4395
      5   3335    596    596    527     99     55   2158   6588   4394
      6   3312    596    596    528     96     56   2156   6587   4395
      7   3331    596    596    526     98     56   2158   6587   4393
      8   3338    596    596    527     99     56   2156   6585   4395
      9   3337    596    596    527     99     56   2158   6586   4394
     10   3338    596    596    527     99     56   2154   6583   4395
     11   3336    596    596    527     99     56   2157   6582   4393
     12   3338    596    596    527     99     56   2155   6582   4390
     13   3337    596    596    527     99     56   2157   6583   4393
     14   3338    596    595    527     99     56   2155   6579   4394
     15   3338    596    596    527     99     56   2157   6581   4393
     16   3338    590    596    527     99     56   2155   6578   4394
     17   3337    596    596    527     99     56   2156   6579   4393
     18   3338    595    595    527     99     56   2155   6575   4394
     19   3337    596    596    527     99     56   2156   6576   4392
     20   3338    596    595    527     99     56   2154   6574   4393
     21   3337    596    595    527     99     56   2155   6571   4392
     22   3308    595    595    527     99     54   2154   6572   4393
     23   3334    595    596    527     99     56   2155   5650   4392
     24   3337    595    595    527     99     56   2154   6570   4393
     25   3335    574    595    527     99     56   2154   6569   4392
     26   3320    595    595    520     99     56   2153   6566   4365
     27   3330    554    586    527     99     56   2153   6567   4392
     28   3320    595    595    527     98     55   2155   5432   4392
     29   3335    574    595    527     99     56   2153   6565   4392
     30   3294    595    594    521     99     54   2151   5701   4392
     31   3307    554    540    527     99     56   2152   6563   4391
     32   3255    595    595    517     96     55   1988   5387   4335
 Total  106480  18929  18949  16840   3152   1774  68644 206382 140462

 MWIPS  102841 Based on time for last thread to finish


 Results Of Calculations Thread 1

 MFLOPS 1   -1.12253940105438232      MFLOPS 2   -1.13133072853088379
 IFMOPS      1.00000000000000000      FIXPMOPS   12.00000000000000000
 COSMOPS     0.49911013245582581      MFLOPS 3    0.99999982118606567
 EQUMOPS     3.00000000000000000      EXPMOPS     0.93536460399627686

      Numeric results of the other 31 threads were same as above

               End of test Mon Dec 12 16:29:34 2022
##########################################
Single Precision C Whetstone Benchmark loongarch64 loongarch64 optimized, Mon Dec 12 16:28:37 2022

Calibrate
       0.00 Seconds          1   Passes (x 100)
       0.01 Seconds          5   Passes (x 100)
       0.07 Seconds         25   Passes (x 100)
       0.36 Seconds        125   Passes (x 100)
       1.80 Seconds        625   Passes (x 100)
       8.98 Seconds       3125   Passes (x 100)

Use 3480  passes (x 100)

          Single Precision C/C++ Whetstone Benchmark

Loop content                  Result              MFLOPS      MOPS   Seconds

N1 floating point     -1.12475013732910156       596.310              0.112
N2 floating point     -1.12274742126464844       730.378              0.640
N3 if then else        1.00000000000000000                4398.026    0.082
N4 fixed point        12.00000000000000000               65767.359    0.017
N5 sin,cos etc.        0.49911010265350342                  94.350    3.069
N6 floating point      0.99999982118606567       527.722              3.557
N7 assignments         3.00000000000000000                3297.946    0.195
N8 exp,sqrt etc.       0.75110864639282227                  55.667    2.326

MWIPS                                           3480.951              9.997

Вывод теста Whetstone на 3A5000

4 CPUs Available
##############################################

 Multithreading Single Precision Whetstones loongarch64 loongarch64 optimized

             Using 4 threads - Sun Dec 11 15:28:38 2022

 Calibrate
    0.00353 Seconds          1   Passes (x 100)
    0.01669 Seconds          5   Passes (x 100)
    0.08278 Seconds         25   Passes (x 100)
    0.41330 Seconds        125   Passes (x 100)

 Use 3024  passes (x 100)

 MFLOPS 1     678    678    677    676
 MFLOPS 2     678    676    676    670
 IFMOPS      7495   7494   7493   7489
 FIXPMOPS    2082   2082   2081   2080
 COSMOPS      112    112    112    111
 MFLOPS 3     600    599    599    597
 EQUMOPS     4997   4991   4991   4967
 EXPMOPS       63     63     63     63
 millisec    8055   8068   8047   8083
 MWIPS       3754   3748   3758   3741

         MWIPS MFLOPS MFLOPS MFLOPS    Cos    Exp  Fixpt     If  Equal
 Thread             1      2      3   MOPS   MOPS   MOPS   MOPS   MOPS

      1   3754    676    678    599    112     63   2082   7489   4991
      2   3748    678    676    599    111     63   2080   7494   4991
      3   3758    678    676    600    112     63   2081   7493   4997
      4   3741    677    670    597    112     63   2082   7495   4967
 Total   15001   2708   2700   2395    446    252   8325  29971  19946

 MWIPS   14939 Based on time for last thread to finish


 Results Of Calculations Thread 1

 MFLOPS 1   -1.13214290142059326      MFLOPS 2   -1.13333344459533691
 IFMOPS      1.00000000000000000      FIXPMOPS   12.00000000000000000
 COSMOPS     0.50000011920928955      MFLOPS 3    0.99999982118606567
 EQUMOPS     3.00000000000000000      EXPMOPS     0.93536460399627686

      Numeric results of the other 3 threads were same as above

               End of test Sun Dec 11 15:28:48 2022
##########################################
Single Precision C Whetstone Benchmark loongarch64 loongarch64 optimized, Sun Dec 11 15:27:54 2022

Calibrate
       0.00 Seconds          1   Passes (x 100)
       0.01 Seconds          5   Passes (x 100)
       0.06 Seconds         25   Passes (x 100)
       0.32 Seconds        125   Passes (x 100)
       1.60 Seconds        625   Passes (x 100)
       8.01 Seconds       3125   Passes (x 100)

Use 3900  passes (x 100)

          Single Precision C/C++ Whetstone Benchmark

Loop content                  Result              MFLOPS      MOPS   Seconds

N1 floating point     -1.12475013732910156       676.608              0.111
N2 floating point     -1.12274742126464844       828.798              0.632
N3 if then else        1.00000000000000000                   0.000    0.000
N4 fixed point        12.00000000000000000               74584.969    0.016
N5 sin,cos etc.        0.49911010265350342                 100.223    3.238
N6 floating point      0.99999982118606567       598.946              3.512
N7 assignments         3.00000000000000000                3742.728    0.193
N8 exp,sqrt etc.       0.75110864639282227                  63.179    2.296

MWIPS                                           3900.650              9.998

Coremark

Современный тест, который должен заменить Dhrystone и Whetstone. Написан на C. Считает различные массивы, матрицы, сортировка массивов и т. д. Предназначался для запуска на всём: от микроконтроллеров до мощных процессоров.

Вывод однопоточного теста Coremark на 3С5000

2K performance run parameters for coremark.
CoreMark Size    : 666
Total ticks      : 16048
Total time (secs): 16.048000
Iterations/Sec   : 12462.612164
Iterations       : 200000
Compiler version : GCC8.3.0 20190222 (Loongson 8.3.0-31 vec)
Compiler flags   : -Ofast   -DPERFORMANCE_RUN=1 -DUSE_FORK=1 -lrt
Memory location  : Please put data memory location here
			(e.g. code in flash, data on heap etc)
seedcrc          : 0xe9f5
[0]crclist       : 0xe714
[0]crcmatrix     : 0x1fd7
[0]crcstate      : 0x8e3a
[0]crcfinal      : 0x4983
Correct operation validated. See README.md for run and reporting rules.
CoreMark 1.0 : 12462.612164 / GCC8.3.0 20190222 (Loongson 8.3.0-31 vec) -Ofast   -DPERFORMANCE_RUN=1 -DUSE_FORK=1 -lrt / Heap

Вывод многопоточного теста Coremark на 3С5000

2K performance run parameters for coremark.
CoreMark Size    : 666
Total ticks      : 16557
Total time (secs): 16.557000
Iterations/Sec   : 193271.727970
Iterations       : 3200000
Compiler version : GCC8.3.0 20190222 (Loongson 8.3.0-31 vec)
Compiler flags   : -Ofast   -DPERFORMANCE_RUN=1 -DUSE_FORK=1 -lrt
Parallel Fork : 16
Memory location  : Please put data memory location here
			(e.g. code in flash, data on heap etc)
seedcrc          : 0xe9f5
[0]crclist       : 0xe714
[1]crclist       : 0xe714
[2]crclist       : 0xe714
[3]crclist       : 0xe714
[4]crclist       : 0xe714
[5]crclist       : 0xe714
[6]crclist       : 0xe714
[7]crclist       : 0xe714
[8]crclist       : 0xe714
[9]crclist       : 0xe714
[10]crclist       : 0xe714
[11]crclist       : 0xe714
[12]crclist       : 0xe714
[13]crclist       : 0xe714
[14]crclist       : 0xe714
[15]crclist       : 0xe714
[0]crcmatrix     : 0x1fd7
[1]crcmatrix     : 0x1fd7
[2]crcmatrix     : 0x1fd7
[3]crcmatrix     : 0x1fd7
[4]crcmatrix     : 0x1fd7
[5]crcmatrix     : 0x1fd7
[6]crcmatrix     : 0x1fd7
[7]crcmatrix     : 0x1fd7
[8]crcmatrix     : 0x1fd7
[9]crcmatrix     : 0x1fd7
[10]crcmatrix     : 0x1fd7
[11]crcmatrix     : 0x1fd7
[12]crcmatrix     : 0x1fd7
[13]crcmatrix     : 0x1fd7
[14]crcmatrix     : 0x1fd7
[15]crcmatrix     : 0x1fd7
[0]crcstate      : 0x8e3a
[1]crcstate      : 0x8e3a
[2]crcstate      : 0x8e3a
[3]crcstate      : 0x8e3a
[4]crcstate      : 0x8e3a
[5]crcstate      : 0x8e3a
[6]crcstate      : 0x8e3a
[7]crcstate      : 0x8e3a
[8]crcstate      : 0x8e3a
[9]crcstate      : 0x8e3a
[10]crcstate      : 0x8e3a
[11]crcstate      : 0x8e3a
[12]crcstate      : 0x8e3a
[13]crcstate      : 0x8e3a
[14]crcstate      : 0x8e3a
[15]crcstate      : 0x8e3a
[0]crcfinal      : 0x4983
[1]crcfinal      : 0x4983
[2]crcfinal      : 0x4983
[3]crcfinal      : 0x4983
[4]crcfinal      : 0x4983
[5]crcfinal      : 0x4983
[6]crcfinal      : 0x4983
[7]crcfinal      : 0x4983
[8]crcfinal      : 0x4983
[9]crcfinal      : 0x4983
[10]crcfinal      : 0x4983
[11]crcfinal      : 0x4983
[12]crcfinal      : 0x4983
[13]crcfinal      : 0x4983
[14]crcfinal      : 0x4983
[15]crcfinal      : 0x4983
Correct operation validated. See README.md for run and reporting rules.
CoreMark 1.0 : 193271.727970 / GCC8.3.0 20190222 (Loongson 8.3.0-31 vec) -Ofast   -DPERFORMANCE_RUN=1 -DUSE_FORK=1 -lrt / Heap / 16:Fork

Вывод однопоточного теста Coremark на 3A5000

2K performance run parameters for coremark.
CoreMark Size    : 666
Total ticks      : 13392
Total time (secs): 13.392000
Iterations/Sec   : 14934.289128
Iterations       : 200000
Compiler version : GCC8.3.0
Compiler flags   : -Ofast   -DPERFORMANCE_RUN=1 -DUSE_FORK=1 -lrt
Memory location  : Please put data memory location here
			(e.g. code in flash, data on heap etc)
seedcrc          : 0xe9f5
[0]crclist       : 0xe714
[0]crcmatrix     : 0x1fd7
[0]crcstate      : 0x8e3a
[0]crcfinal      : 0x4983
Correct operation validated. See README.md for run and reporting rules.
CoreMark 1.0 : 14934.289128 / GCC8.3.0 -Ofast   -DPERFORMANCE_RUN=1 -DUSE_FORK=1 -lrt / Heap

Вывод многопоточного теста Coremark на 3A5000

2K performance run parameters for coremark.
CoreMark Size    : 666
Total ticks      : 13646
Total time (secs): 13.646000
Iterations/Sec   : 58625.238165
Iterations       : 800000
Compiler version : GCC8.3.0
Compiler flags   : -Ofast   -DPERFORMANCE_RUN=1 -DUSE_FORK=1 -lrt
Parallel Fork : 4
Memory location  : Please put data memory location here
			(e.g. code in flash, data on heap etc)
seedcrc          : 0xe9f5
[0]crclist       : 0xe714
[1]crclist       : 0xe714
[2]crclist       : 0xe714
[3]crclist       : 0xe714
[0]crcmatrix     : 0x1fd7
[1]crcmatrix     : 0x1fd7
[2]crcmatrix     : 0x1fd7
[3]crcmatrix     : 0x1fd7
[0]crcstate      : 0x8e3a
[1]crcstate      : 0x8e3a
[2]crcstate      : 0x8e3a
[3]crcstate      : 0x8e3a
[0]crcfinal      : 0x4983
[1]crcfinal      : 0x4983
[2]crcfinal      : 0x4983
[3]crcfinal      : 0x4983
Correct operation validated. See README.md for run and reporting rules.
CoreMark 1.0 : 58625.238165 / GCC8.3.0 -Ofast   -DPERFORMANCE_RUN=1 -DUSE_FORK=1 -lrt / Heap / 4:Fork

MP MFLOPS

Выполняет операции с плавающей запятой на векторах.

Вывод теста MP-MFLOPS на 3С5000

32 CPUs Available

##############################################

  64 Bit MP SSE MFLOPS Benchmark 1, 32 Threads, Mon Dec 12 16:22:54 2022

  Test             4 Byte  Ops/   Repeat    Seconds   MFLOPS       First   All
                    Words  Word   Passes                         Results  Same

 Data in & out     102400     2    80000   0.070316   233004    0.410898   Yes
 Data in & out    1024000     2     8000   0.065234   251158    0.802311   Yes
 Data in & out   10240000     2      800   0.213911    76593    0.976486   Yes

 Data in & out     102400     8    80000   0.151069   433816    0.558756   Yes
 Data in & out    1024000     8     8000   0.116899   560618    0.876281   Yes
 Data in & out   10240000     8      800   0.199468   328554    0.985828   Yes

 Data in & out     102400    32    80000   0.372965   702865    0.353359   Yes
 Data in & out    1024000    32     8000   0.431096   608087    0.709574   Yes
 Data in & out   10240000    32      800   0.364138   719903    0.962629   Yes

               End of test Mon Dec 12 16:22:56 2022

Вывод теста MP-MFLOPS на 3A5000

4 CPUs Available

##############################################

  64 Bit MP SSE MFLOPS Benchmark 1, 4 Threads, Sun Dec 11 15:22:37 2022

  Test             4 Byte  Ops/   Repeat    Seconds   MFLOPS       First   All
                    Words  Word   Passes                         Results  Same

 Data in & out     102400     2    10000   0.086061    23797    0.764063   Yes
 Data in & out    1024000     2     1000   0.096334    21259    0.970753   Yes
 Data in & out   10240000     2      100   0.597981     3425    0.997008   Yes

 Data in & out     102400     8    10000   0.068533   119533    0.850919   Yes
 Data in & out    1024000     8     1000   0.069136   118490    0.982347   Yes
 Data in & out   10240000     8      100   0.593080    13813    0.998205   Yes

 Data in & out     102400    32    10000   0.304694   107544    0.660093   Yes
 Data in & out    1024000    32     1000   0.304863   107484    0.953624   Yes
 Data in & out   10240000    32      100   0.620384    52819    0.995219   Yes

               End of test Sun Dec 11 15:22:40 2022

7zip

Встроенный тест архиватора 7z.

Запускается так:

7z b

Вывод теста 7z на 3A5000

7-Zip (a) 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=zh_CN.UTF-8,Utf16=on,HugeFiles=on,64 bits,4 CPUs LE)

LE
CPU Freq: 64000000 - - - - - - - -

RAM size:   16165 MB,  # CPU hardware threads:   4
RAM usage:    882 MB,  # Benchmark threads:      4

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:      11279   327   3352  10973  |     130124   396   2803  11102
23:      10681   325   3346  10883  |     127682   393   2810  11048
24:      11183   353   3406  12025  |     126626   395   2815  11116
25:      11099   360   3525  12673  |     124248   394   2805  11058
----------------------------------  | ------------------------------
Avr:             341   3407  11638  |              395   2808  11081
Tot:             368   3108  11360

Вывод теста 7z на 3С5000

# 7z b

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,32 CPUs LE)

LE
CPU Freq: 64000000 - - - - - - - -

RAM size:  129704 MB,  # CPU hardware threads:  32
RAM usage:   7060 MB,  # Benchmark threads:     32

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:      62370  2796   2170  60674  |     878738  3138   2388  74938
23:      60911  2793   2222  62062  |     861051  3149   2366  74508
24:      59913  2867   2247  64419  |     832658  3112   2349  73085
25:      61317  2941   2380  70010  |     815385  3087   2351  72564
----------------------------------  | ------------------------------
Avr:            2849   2255  64291  |             3122   2363  73774
Tot:            2985   2309  69032



# 7z b -mmt16                                            
                                                                                 
7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21                   
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,32 CPUs LE)
                                                                                 
LE                                                                               
CPU Freq: - - - - - - - - -                                                      
                                                                                 
RAM size:  129704 MB,  # CPU hardware threads:  32                               
RAM usage:   3530 MB,  # Benchmark threads:     16                               
                                                                                 
                       Compressing  |                  Decompressing             
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating             
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS             
                                                                                 
22:      36575  1439   2473  35580  |     450711  1594   2412  38441             
23:      36638  1469   2540  37330  |     443639  1589   2416  38384             
24:      35836  1519   2537  38532  |     434708  1591   2399  38157             
25:      35206  1523   2640  40198  |     421884  1573   2387  37546             
----------------------------------  | ------------------------------             
Avr:            1487   2548  37910  |             1587   2403  38132             
Tot:            1537   2476  38021

STREAM

Тест производительности ОЗУ.

Вывод теста STREAM на 3A5000

Function    Best Rate MB/s  Avg time     Min time     Max time
Copy:           16748.3     0.009969     0.009553     0.010576
Scale:          18735.0     0.009437     0.008540     0.010660
Add:            16333.2     0.015492     0.014694     0.016449
Triad:          17796.0     0.014789     0.013486     0.015294

Вывод теста STREAM на 3C5000

Function    Best Rate MB/s  Avg time     Min time     Max time
Copy:           53623.8     0.024431     0.023870     0.025548
Scale:          49980.5     0.026965     0.025610     0.028581
Add:            48869.8     0.041190     0.039288     0.052376
Triad:          49102.3     0.040250     0.039102     0.043391

Версия компилятора, ядра Linux

Ядро Linux на 3A5000:

Linux 4.19.167-rc5.lnd.1-loongson-3 #1 SMP Sat Apr 17 07:32:32 UTC 2021 loongarch64 loongarch64 loongarch64 GNU/Linux

Версия компилятора на 3A5000:  

gcc version 8.3.0 (Debian 8.3.0-6.lnd.vec.20).

Ядро Linux на 3C5000:

Linux 4.19.0-19-loongson-3 #1 SMP pkg_lnd10_4.19.190-7.7 Fri Nov 25 10:17:52 UTC 2022 loongarch64

Версия компилятора на 3C5000:  

cc (Loongnix 8.3.0-6.lnd.vec.34) 8.3.0

Заключение

Процессоры Loongson очень активно развиваются и догоняют по производительности процессоры мировых производителей на других архитектурах, имеют отличную документацию и набор портированного софта.

Другие результаты для сравнения:

a2fc10b93b9c6dd30113fd7200358b8f.png

CСсылки

© Habrahabr.ru