commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gary Lucas (JIRA)" <j...@apache.org>
Subject [jira] [Comment Edited] (IMAGING-95) Some tiff processing takes very long
Date Thu, 18 Oct 2012 20:20:02 GMT

    [ https://issues.apache.org/jira/browse/IMAGING-95?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13479190#comment-13479190
] 

Gary Lucas edited comment on IMAGING-95 at 10/18/12 8:18 PM:
-------------------------------------------------------------

Well, I took another look at this, and it appears that my first tests weren't accessing the
ByteSourceInputStream, but were using the ByteSourceFile.  So I tried it again, following
my own suggestion about using the ApacheImagingSpeedAndMemoryTest application.

The results loading a 3600 by 1800 pixel image

   using current version:   763.8 milliseconds
   using your version:      176.4 milliseconds

So it looks like you're on the right track.  Strictly as an aside,
I also tried testing with the ByteSourceFile and got a load time of
98 milliseconds.

{monospaced}
Using ByteSourceInputStream -------------------------------------
Processing file: BlueMarble.TIFF
 image size: 3600 by 1800

 time to load image    --         memory
 time ms      avg ms   --    used mb   total mb
  787.793     0.000    --     63.075    92.191 
  859.587     0.000    --     63.070    92.191 
  772.965   772.965    --     62.894    84.473 
  651.912   712.439    --     63.058    92.203 
  851.362   758.747    --     63.061    92.203 
  818.632   773.718    --     62.890    84.484 
  659.307   750.836    --     63.058    92.203 
  911.201   777.563    --     63.061    92.203 
  781.244   778.089    --     62.890    84.484 
  663.920   763.818    --     63.058    92.203 
 

Processing file: BlueMarble.TIFF
 image size: 3600 by 1800

 time to load image    --         memory
 time ms      avg ms   --    used mb   total mb
  295.915     0.000    --     63.073    92.188 
  193.905     0.000    --     63.068    92.188 
  191.630   191.630    --     62.892    84.473 
  214.639   203.134    --     63.056    92.199 
  211.340   205.870    --     63.059    92.199 
  182.537   200.036    --     62.888    84.480 
  175.037   195.037    --     63.056    92.199 
  166.024   190.201    --     63.059    92.199 
  149.707   184.416    --     62.888    84.480 
  176.374   183.411    --     63.057    92.199 



Using ByteSourceFile instead of ByteSourceInputStream  ------------------

Processing file: BlueMarble.TIFF
 image size: 3600 by 1800

 time to load image    --         memory
 time ms      avg ms   --    used mb   total mb
  184.074     0.000    --     26.417    40.285 
   90.738     0.000    --     29.509    51.410 
   88.949    88.949    --     30.485    47.816 
  104.790    96.869    --     26.166    40.289 
   99.904    97.881    --     26.084    40.289 
   98.153    97.949    --     26.084    40.289 
   98.774    98.114    --     26.084    40.289 
   98.111    98.113    --     26.084    40.289 
   97.828    98.073    --     26.084    40.289 
   98.742    98.156    --     26.084    40.289 

{monospaced}
                
      was (Author: gwlucas):
    Well, I took another look at this, and it appears that my first tests weren't accessing
the ByteSourceInputStream, but were using the ByteSourceFile.  So I tried it again, following
my own suggestion about using the ApacheImagingSpeedAndMemoryTest application.

The results loading a 3600 by 1800 pixel image

   using current version:   763.8 milliseconds
   using your version:      176.4 milliseconds

So it looks like you're on the right track.  Strictly as an aside,
I also tried testing with the ByteSourceFile and got a load time of
98 milliseconds.

{monospaced}
Using ByteSourceInputStream -------------------------------------
Processing file: BlueMarble.TIFF
 image size: 3600 by 1800

 time to load image    --         memory
 time ms      avg ms   --    used mb   total mb
  787.793     0.000    --     63.075    92.191 
  859.587     0.000    --     63.070    92.191 
  772.965   772.965    --     62.894    84.473 
  651.912   712.439    --     63.058    92.203 
  851.362   758.747    --     63.061    92.203 
  818.632   773.718    --     62.890    84.484 
  659.307   750.836    --     63.058    92.203 
  911.201   777.563    --     63.061    92.203 
  781.244   778.089    --     62.890    84.484 
  663.920   763.818    --     63.058    92.203 
 

Processing file: BlueMarble.TIFF
 image size: 3600 by 1800

 time to load image    --         memory
 time ms      avg ms   --    used mb   total mb
  295.915     0.000    --     63.073    92.188 
  193.905     0.000    --     63.068    92.188 
  191.630   191.630    --     62.892    84.473 
  214.639   203.134    --     63.056    92.199 
  211.340   205.870    --     63.059    92.199 
  182.537   200.036    --     62.888    84.480 
  175.037   195.037    --     63.056    92.199 
  166.024   190.201    --     63.059    92.199 
  149.707   184.416    --     62.888    84.480 
  176.374   183.411    --     63.057    92.199 



Using ByteSourceFile instead of ByteSourceInputStream  ------------------

Processing file: BlueMarble.TIFF
 image size: 3600 by 1800

 time to load image    --         memory
 time ms      avg ms   --    used mb   total mb
  184.074     0.000    --     26.417    40.285 
   90.738     0.000    --     29.509    51.410 
   88.949    88.949    --     30.485    47.816 
  104.790    96.869    --     26.166    40.289 
   99.904    97.881    --     26.084    40.289 
   98.153    97.949    --     26.084    40.289 
   98.774    98.114    --     26.084    40.289 
   98.111    98.113    --     26.084    40.289 
   97.828    98.073    --     26.084    40.289 
   98.742    98.156    --     26.084    40.289 

                  
> Some tiff processing takes very long
> ------------------------------------
>
>                 Key: IMAGING-95
>                 URL: https://issues.apache.org/jira/browse/IMAGING-95
>             Project: Commons Imaging
>          Issue Type: Bug
>          Components: Format: TIFF
>    Affects Versions: 1.0
>            Reporter: Amit Gupta
>         Attachments: tiff_perf_fix2.patch
>
>
> org.apache.commons.imaging.formats.tiff.TiffReader.getTiffRawImageData(ByteSource, TiffDirectory)
226635	1
> org.apache.commons.imaging.common.bytesource.ByteSourceInputStream.getBlock(int, int)
226588	5616
> org.apache.commons.imaging.common.BinaryFileFunctions.skipBytes(InputStream, int)	226526
5616
> org.apache.commons.imaging.common.BinaryFileFunctions.skipBytes(InputStream, int, String)
226526	5616
> org.apache.commons.imaging.common.bytesource.ByteSourceInputStream$CacheReadingInputStream.read(byte[],
int, int)	226526	188656860
> org.apache.commons.imaging.common.bytesource.ByteSourceInputStream$CacheBlock.getNext()
64581	188651244
> skip bytes is being called repeatedly again and again, last column is invocation count
in one call tree. Second column is total number of time taken by that method in that call
tree..
> and skip method is not overridden org.apache.commons.imaging.common.bytesource.ByteSourceInputStream.CacheReadingInputStream
and default implementation of InputStream tries to use read method (which is overridden in
CacheReadingInputStream) to skip. 
> In case of a tiff, which has large number of strips, skip is repeatedly called and use
of read is inefficient as it tried to do a System.arraycopy. array copy is not needed in case
of skip operation, as the bytes were already read in block/cached, we can simply jump the
pointer (block by block)

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message