Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

TurboBench: Maximum Compression Benchmark - Skylake i7-6700 3.4GHz #17

Open
powturbo opened this issue Aug 9, 2019 · 2 comments
Open

Comments

@powturbo
Copy link
Owner

powturbo commented Aug 9, 2019

Latest update: 2019.08.09

Binary:mingw-w64_x86_64-6.3.0-posix-seh-rt_v5-rev2.tar

     C Size   ratio%     C MB/s     D MB/s   Name       (bold = pareto) MB=1.000.0000
    42179292     9.1       0.91     189.42   lzturbo 49      
    42457543     9.1       2.07     186.33   lzma 9d29:fb273:mf=bt4
    42988097     9.2       0.57     857.10   brotli 11d29    
    45871347     9.9       1.50    2206.52   lzturbo 39      
    46458069    10.0       2.66    1491.27   zstd 22d29      
    56248534    12.1       1.93    2966.68   lzturbo 29      
    58474096    12.6      16.25      27.38   bsc 0         
    58654705    12.6     125.48    2100.32   lzturbo 32      
    60760645    13.1      13.12      60.33   lzturbo 59      
    60765065    13.1      14.08      90.42   lzturbo 59t2    
    73060389    15.7       2.24    5717.94   lzturbo 19
    87416420    18.8      11.75    1397.49   zstd 15         
    88006994    18.9      45.49     514.92   brotli 5        
    88252775    19.0       1.78    2232.29   lizard 49       
    94672486    20.3       1.91    3073.42   lizard 29 
   111853281    24.0      12.56      39.18   bzip2           
   126614955    27.2       9.25     340.34   zlib 9          
   133921751    28.8       2.97    2230.60   lizard 39       
   143616390    30.9      38.77    3233.08   lz4 9           
   144311136    31.0       3.34    3519.31   lizard 19       
   145898393    31.3       1.51    3828.81   lzsse8 17       
   174569457    37.5     871.54    4503.78   lzturbo 10
   179181598    38.5     657.00    2866.63   lz4 1 
   465457156   100.0   13824.50   14242.87   memcpy          

English Text:Encyclopaedia_Judaica_(in_22_volumes)_TXT.tar

     C Size   ratio%     C MB/s     D MB/s   Name        
    21905524    20.3       9.45      15.92   bsc 0         
    22279381    20.7       9.46      49.89   lzturbo 59      
    22280917    20.7      10.04      71.61   lzturbo 59t2    
    27070097    25.1       0.91      82.32   lzturbo 49      
    27526894    25.5       1.23      86.14   lzma 9d29:fb273:mf=bt4
    27600942    25.6       0.46     305.70   brotli 11d29    
    27945445    25.9       1.06     672.33   lzturbo 39      
    28008216    26.0       1.43     640.32   zstd 22d29      
    31570314    29.3       1.10    1124.56   lzturbo 29      
    31734097    29.4      12.91      29.08   bzip2           
    32696227    30.3      46.86     781.41   lzturbo 32      
    33249008    30.8       4.94     851.66   zstd 15         
    35077489    32.5       1.85     694.91   lizard 49       
    36277020    33.7      26.67     367.32   brotli 5        
    41420278    38.4       1.92     973.00   lizard 29       
    42104125    39.1       8.15    1145.53   lizard 39       
    42236588    39.2      12.32     243.79   zlib 9          
    43696364    40.5       9.52    3085.19   lzsse8 17       
    45341158    42.1       1.28    2958.75   lzturbo 19      
    47834366    44.4       8.36    1555.15   lizard 19       
    48214128    44.7      22.56    2825.87   lz4 9 
    67850583    63.0     438.16    3994.67   lzturbo 10
    68340992    63.4     358.85    2790.46   lz4 1          
   107784196   100.0   14225.18   14694.50   memcpy          

PDF: Learn-Hot-English_magazine_(18-issues).tar

      C Size  ratio%     C MB/s     D MB/s   Name 
   244285827    54.0       0.47     212.50   brotli 11d30    
   245265022    54.2       2.17      33.91   lzturbo 49      
   246117453    54.4       3.22    3616.72   lzturbo 39      
   246415683    54.5       3.48      34.90   lzma 9d30:fb273 
   246786277    54.6       3.84    2182.46   zstd 22d30      
   250400776    55.4      81.47    3913.83   lzturbo 32
   253265114    56.0       3.62    7728.73   lzturbo 19      
   253761470    56.1       3.53    6090.31   lzturbo 29      
   289049146    63.9       5.62    6516.19   lizard 29       
   289084982    63.9       5.30    5445.98   lizard 49       
   299881226    66.3      17.58    4348.64   zstd 15
   330124786    73.0      47.57    6174.21   lz4 9          
   330179011    73.0       8.09    6278.61   lizard 39       
   330769147    73.1      28.04     351.13   zlib 9
   330921620    73.2       9.18    6710.84   lizard 19       
   334380724    73.9    1940.57    7960.84   lzturbo 10
   349951390    77.4       0.69    4303.37   lzsse8 17       
   368898163    81.6       0.69    2528.31   lzsse4 17       
   403005235    89.1       0.69    1329.14   lzsse2 17  
   452223492   100.0   13956.22   13771.35   memcpy 
@Sanmayce
Copy link

Nothing short of amazing, as usual.

Hamid, I bend a knee before you, your work is ecstatic and inspirational, straight up.

The original "logo" for this corpus is:
[img]http://www.sanmayce.com/Nakamichi/LearnHotEnglish_18_half.png[/img]
My browser on Windows XP prevents me to use GitHub in its full functionality.

These three corpora are among the "telling" ones, Nakamichi is so inferior with PDFs, ugh, my Kaby Lake i5-7200U @3.1GHz decompresses the super important corpus [b]Learn-Hot-English_magazine_(18-issues)_(True-PDFs).tar[/b] at:

   249569351                       4366      lzturbo 29
   252162586                       4980      oodle 99 'Mermaid'
   289049146                       5316      lizard 29
   289944694                       2264      Nakamichi

Speaking of importance and the ability to "open" PDFs FAST, please consider running 'Hot-English' corpus on ARM, there is a trend ARM to overshadow x86 in supercomputers also. My word here is for an ebook-reader or smartphone, how fast the user can open the needed True-PDF magazine/ebook when compressed with nowadays schemes.

With incoming Gen4 SSDs, with 4GB/s read/write rates your LzTurbo 10 fits beautifully - those encoding speeds - all Pareto!

Excellent source material for learning English these magazines are, they represent nicely all the TRUE-PDF "openings", wonder how all PDF users would react if they knew there was much faster way to open their collections, e.g. 3616.72/351.13= 10.3x than the "standard" way, while being much smaller!

   246117453    54.4       3.22    3616.72   lzturbo 39      
   246415683    54.5       3.48      34.90   lzma 9d30:fb273 
   253265114    56.0       3.62    7728.73   lzturbo 19      
   330769147    73.1      28.04     351.13   zlib 9

I guess the 3616.72/34.90= 103.6x boost is not enough for powerusers to state the obvious - a DIAMANTINE performance.
Perhaps, "opening" True-PDFs at 7728.73/351.13= 22.0x higher speeds is 'meh' for those who got no sense of code being lyrics, of each line being a verse.

"What power, what a shine / Welche Kraft, was für ein Schein
Beautiful as a diamond / Wunderschön wie ein Diamant"
/DIAMANT lyrics - Rammstein/

@Sanmayce
Copy link

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants