<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <font face="Times New Roman, Times, serif">No, this is unaaceptable,
      but I cannot reproduce it. The only problem I had was with control
      points that were too close together. Are yours evenly distributed?
      Can you do tests with subsets of control points?<br>
      <br>
    </font>
    <div class="moz-cite-prefix">On 12/12/2012 01:27 PM, Knut-Frode
      Dagestad wrote:<br>
    </div>
    <blockquote cite="mid:ka9t7u$lpq$1@ger.gmane.org" type="cite">Hi
      Jan,
      <br>
      <br>
      That sounds interesting and promising.
      <br>
      <br>
      For the mentioned file it takes about 2 minutes with "-et 5" (low
      accuracy), and 12 minutes without this switch. Without -tps it
      takes less than 0.5 seconds.
      <br>
      The machine is quite decent with 16 GB memory and SSD disk.
      <br>
      <br>
      Best regards from Knut-Frode
      <br>
      <br>
      <br>
      <br>
      On 12. des. 2012 13:24, Jan Hartmann wrote:
      <br>
      <blockquote type="cite">Hi Knut,
        <br>
        <br>
        What do you mean by "very slow"? I regularly use gdalwarp -tps
        on much
        <br>
        larger rasters with a few thousand gcp-s, and never noticed
        unacceptable
        <br>
        delays. Do you have very little physical memory, or many
        parallel
        <br>
        processes running?
        <br>
        <br>
        Jan
        <br>
        <br>
        On 12/12/2012 01:12 PM, Knut-Frode Dagestad wrote:
        <br>
        <blockquote type="cite">Hi list,
          <br>
          <br>
          When warping images with many GCPs, the -tps switch (Thin
          Plate
          <br>
          Spline) is found to be necessary to get decent accuracy. This
          makes
          <br>
          however warping very slow. The only method I found to increase
          speed
          <br>
          is the -et switch, but at cost of spatial accuracy.
          <br>
          <br>
          <br>
          Below some comments about the other tries which did not help:
          <br>
          <br>
          - compiling GDAL with armadillo support had no effect on
          speed.
          <br>
          Btw, to compile with armadillo I had to manually insert
          -llapack in
          <br>
          the following line in configure:
          <br>
          if test -z "`${CXX} testarmadillo.cpp -o testarmadillo
          -larmadillo
          <br>
          2>&1`" ; then .....
          <br>
          Perhaps ${LIBS} should be added permanently to this line in
          trunk?
          <br>
          <br>
          - compiling GDAL with OpenCL had also no effect. At first
          surprising,
          <br>
          but looking at the opencl warp kernel it seems that it only
          makes a
          <br>
          difference for other resampling algorithms than nearest
          neighbour?
          <br>
          <br>
          - Increasing  memory with -wm had no effect
          <br>
          <br>
          - Using several threads (-multi -wo NUM_THREADS=ALL_CPUS)
          actually
          <br>
          increased computing time significantly (my proj version is
          4.7.1).
          <br>
          <br>
          From the debug output, a lot of the time is apparently spent
          on:
          <br>
          "WARP: Copying metadata from first source to destination
          dataset"
          <br>
          Is this an indication that much time is simply spent on
          reading and
          <br>
          writing the GCPs to/from file? If so, we could perhaps expect
          improved
          <br>
          performance if Geolocation Arrays could be used instead (not
          possible
          <br>
          due to <a class="moz-txt-link-freetext" href="http://trac.osgeo.org/gdal/ticket/4907">http://trac.osgeo.org/gdal/ticket/4907</a>).
          <br>
          <br>
          <br>
          The tests are made with the following file and command on
          Ubuntu:
          <br>
          <a class="moz-txt-link-freetext" href="http://dl.dropbox.com/u/15885758/testgcp.tif">http://dl.dropbox.com/u/15885758/testgcp.tif</a>  (3212 GCPs and
          2048x2511
          <br>
          pixels covering Southern Europe)
          <br>
          <br>
          time gdalwarp --debug on -et 5 -tps -t_srs '+proj=merc'
          testgcp.tif
          <br>
          out.tif  (+ other swithces mentioned above)
          <br>
          <br>
          <br>
          Best regards from Knut-Frode
          <br>
          <br>
          _______________________________________________
          <br>
          gdal-dev mailing list
          <br>
          <a class="moz-txt-link-abbreviated" href="mailto:gdal-dev@lists.osgeo.org">gdal-dev@lists.osgeo.org</a>
          <br>
          <a class="moz-txt-link-freetext" href="http://lists.osgeo.org/mailman/listinfo/gdal-dev">http://lists.osgeo.org/mailman/listinfo/gdal-dev</a>
          <br>
        </blockquote>
        <br>
        <br>
        <br>
        _______________________________________________
        <br>
        gdal-dev mailing list
        <br>
        <a class="moz-txt-link-abbreviated" href="mailto:gdal-dev@lists.osgeo.org">gdal-dev@lists.osgeo.org</a>
        <br>
        <a class="moz-txt-link-freetext" href="http://lists.osgeo.org/mailman/listinfo/gdal-dev">http://lists.osgeo.org/mailman/listinfo/gdal-dev</a>
        <br>
        <br>
      </blockquote>
      <br>
      <br>
      _______________________________________________
      <br>
      gdal-dev mailing list
      <br>
      <a class="moz-txt-link-abbreviated" href="mailto:gdal-dev@lists.osgeo.org">gdal-dev@lists.osgeo.org</a>
      <br>
      <a class="moz-txt-link-freetext" href="http://lists.osgeo.org/mailman/listinfo/gdal-dev">http://lists.osgeo.org/mailman/listinfo/gdal-dev</a>
      <br>
    </blockquote>
    <br>
  </body>
</html>