[gdal-dev] GDAL/OGR 1.11.2 RC2 Available for Review

Even Rouault even.rouault at spatialys.com
Fri Feb 13 04:25:39 PST 2015


Le vendredi 13 février 2015 03:30:48, Kurt Schwehr a écrit :
> I'll have to look into this more.
> 
> gdal_grid  --debug on -txe 440720.0 441920.0 -tye 3751320.0 3750120.0 \
> 
> >                  -outsize 20 20 -ot Float64 -l grid \
> >                 
> >                 -a
> 
> invdist:power=2.0:smoothing=0.0:radius1=0.0:radius2=0.0:angle=0.0:max_point
> s=0:min_points=0:nodata=0.0 data/grid.vrt out.tif
> OGR: OGROpen(data/grid.vrt/0x7f8c9be00240) succeeded as VRT.
> GDAL: GDALDriver::Create(GTiff,out.tif,20,20,1,Float64,0x0)
> OGR: OGROpen(data/grid.csv/0x7f8c9bc24350) succeeded as CSV.
> Grid data type is "Float64"
> Grid size = (20 20).
> Corner coordinates = (440690.000000 3750090.000000)-(441950.000000
> 3751350.000000).
> Grid cell size = (60.000000 -60.000000).
> Source point count = 400.
> Algorithm name: "invdist".
> Options are
> "power=2.000000:smoothing=0.000000:radius1=0.000000:radius2=0.000000:angle=
> 0.000000:max_points=0:min_points=0:nodata=0.000000"
> 
> GDAL_GRID: Work buffer: 20 * 20
> GDAL_GRID: Using AVX optimized version
> GDAL_GRID: Using 8 threads
> Segmentation fault: 11

Interesting... In trunk, I've added more verbose messages regarding 
test_gdal_grid_3.

The AVX implementation works on one of my Linux boxes. On the Travis-CI Mac 
instance, the AVX instruction set must not be available at runtime according 
to below traces:
"""
  TEST: test_gdal_grid_3 ... Step 1: Disabling AVX/SSE optimized versions...
Step 1: 2 threads used
Step 2: Trying SSE optimized version...
...SSE optimized version used
Step 3: Trying AVX optimized version...
...AVX optimized version NOT used
"""

The gdalgridavx.cpp file hasn't changed in a significant way in the 1.11 branch 
since GDAL 1.11.0 so it doesn't look like a regression introduced since.

> 
> 
> On Thu, Feb 12, 2015 at 12:23 PM, Even Rouault <even.rouault at spatialys.com>
> 
> wrote:
> > Le jeudi 12 février 2015 20:54:42, Kurt Schwehr a écrit :
> > > Here is the full output:
> > > 
> > > https://gist.github.com/schwehr/c9e9fb595d0370e7eff7
> > > 
> > > fink list -i libjpeg webp
> > > 
> > >  i   libjpeg-bin                   9a-2                   Executables
> > >  for
> > > 
> > > libjpeg package
> > > 
> > >  i   libjpeg9                      9a-2                   JPEG image
> > 
> > format
> > 
> > > handling library
> > > 
> > >  i   libjpeg9-shlibs               9a-2                   Shared
> > 
> > libraries
> > 
> > > for libjpeg package
> > > 
> > >  i   libwebp5                      0.4.0-1                Library for
> > > 
> > > manipulating WebP format
> > > 
> > >  i   libwebp5-shlibs               0.4.0-1                Library for
> > > 
> > > manipulating WebP format
> > > 
> > > And the details of test_gdal_grid_3:
> > > 
> > > ./test_gdal_grid.py
> > > 
> > >   TEST: test_gdal_grid_1 ... success
> > >   TEST: test_gdal_grid_2 ... success
> > >   TEST: test_gdal_grid_3 ... ERROR 4: `tmp/grid_invdist.tif' not
> > 
> > recognised
> > 
> > > as a supported file format.
> > > 
> > > fail (blowup)
> > > 
> > > Traceback (most recent call last):
> > >   File "../pymod/gdaltest_python2.py", line 42, in run_func
> > >   
> > >     result = func()
> > >   
> > >   File "./test_gdal_grid.py", line 273, in test_gdal_grid_3
> > >   
> > >     maxdiff = gdaltest.compare_ds(ds, ds_ref, verbose = 0)
> > >   
> > >   File "../pymod/gdaltest.py", line 1089, in compare_ds
> > >   
> > >     width = ds1.RasterXSize
> > > 
> > > AttributeError: 'NoneType' object has no attribute 'RasterXSize'
> > > 
> > >   TEST: test_gdal_grid_4 ... success
> > >   TEST: test_gdal_grid_5 ... success
> > >   TEST: test_gdal_grid_6 ... success
> > >   TEST: test_gdal_grid_7 ... success
> > >   TEST: test_gdal_grid_8 ... success
> > >   TEST: test_gdal_grid_9 ... success
> > >   TEST: test_gdal_grid_10 ... success
> > >   TEST: test_gdal_grid_cleanup ... ERROR 4: `tmp/grid_invdist.tif' not
> > > 
> > > recognised as a supported file format.
> > > 
> > > success
> > 
> > Kurt,
> > 
> > I'm wondering if it wouldn't be an issue with SSE / AVX optimizations.
> > 
> > Could you go into autotest/utilities and run manually the following ?
> > 
> > $ gdal_grid  --debug on -txe 440720.0 441920.0 -tye 3751320.0 3750120.0 \
> > 
> >                  -outsize 20 20 -ot Float64 -l grid \
> >                 
> >                 -a
> > 
> > invdist:power=2.0:smoothing=0.0:radius1=0.0:radius2=0.0:angle=0.0:max_poi
> > nts=0:min_points=0:nodata=0.0 data/grid.vrt out.tif
> > 
> > Here's what I get :
> > 
> > OGR: OGROpen(data/grid.vrt/0x2482e50) succeeded as VRT.
> > GDAL: GDALDriver::Create(GTiff,out.tif,20,20,1,Float64,(nil))
> > OGR: OGROpen(data/grid.csv/0x24844c0) succeeded as CSV.
> > Grid data type is "Float64"
> > Grid size = (20 20).
> > Corner coordinates = (440690.000000 3750090.000000)-(441950.000000
> > 3751350.000000).
> > Grid cell size = (60.000000 -60.000000).
> > Source point count = 400.
> > Algorithm name: "invdist".
> > Options are
> > "power=2.000000:smoothing=0.000000:radius1=0.000000:radius2=0.000000:angl
> > e=0.000000:max_points=0:min_points=0:nodata=0.000000"
> > 
> > GDAL_GRID: Work buffer: 20 * 20
> > GDAL_GRID: Using SSE optimized version
> > GDAL_GRID: Using 4 threads
> > 0...10...20...30...40...50...60...70...80...90...100 - done.
> > VRT: 400 features read on layer 'grid'.
> > CSV: 400 features read on layer 'grid'.
> > GDAL: GDALClose(out.tif, this=0x24baac0)
> > 
> > If you have issues, you can try adding "--config GDAL_USE_AVX NO" and/or
> > "--config GDAL_USE_SSE NO"
> > You can also play with the GDAL_NUM_THREADS configuration option.
> > 
> > Does this work with trunk ? There have been some changes by moving SSE
> > specifc code into a specific file.
> > 
> > Even
> > 
> > > Test Script: test_gdal_grid
> > > Succeeded: 10
> > > Failed:    1 (1 blew exceptions)
> > > Skipped:   0
> > > Expected fail:0
> > > Duration:  1.09s
> > > 
> > > file tmp/grid_invdist.tif
> > > tmp/grid_invdist.tif: empty
> > > 
> > > 
> > > On Wed, Feb 11, 2015 at 12:28 AM, Even Rouault <
> > 
> > even.rouault at spatialys.com>
> > 
> > > wrote:
> > > > Le mardi 10 février 2015 20:05:52, Kurt Schwehr a écrit :
> > > > > I think it looks okay w.r.t. fink on mac osx 10.9.  autotest
> > > > > failures for anyone who is interested:
> > > > > https://gist.github.com/schwehr/757bae6bd28d9a7ffc4d
> > > > 
> > > > Kurt,
> > > > 
> > > > Some of the output traces that contain "fail" do not necessarily make
> > 
> > the
> > 
> > > > step
> > > > a failure, but should skip it. The error message and information
> > 
> > printed
> > 
> > > > and
> > > > line reported might be interesting.
> > > > Most must be due to a different version of the JPEG or webp lib that
> > 
> > must
> > 
> > > > produce slightly different results.
> > > > The exception stack associated with "TEST: test_gdal_grid_3 ... fail
> > > > (blowup)"
> > > > would be interesting
> > > > 
> > > > Even
> > > > 
> > > > > On Tue, Feb 10, 2015 at 4:30 AM, Even Rouault <
> > > > 
> > > > even.rouault at spatialys.com>
> > > > 
> > > > > wrote:
> > > > > > Hi,
> > > > > > 
> > > > > > I have prepared a second GDAL/OGR 1.11.2 release candidate, with
> > 
> > the
> > 
> > > > fix
> > > > 
> > > > > > for
> > > > > > the 1.11.0 regression in the PG driver identified yesterday
> > 
> > (#5837).
> > 
> > > > > > Please review and test.
> > > > > > 
> > > > > > Peek up an archive among the following ones (by ascending size):
> > > > > >   http://download.osgeo.org/gdal/1.11.2/gdal-1.11.2RC2.tar.xz
> > > > > >   http://download.osgeo.org/gdal/1.11.2/gdal-1.11.2RC2.tar.gz
> > > > > >   http://download.osgeo.org/gdal/1.11.2/gdal1112RC2.zip
> > > > > > 
> > > > > > The gdal-grass plugin archive hasn't been modified since RC1  :
> > > > > >   http://download.osgeo.org/gdal/1.11.2/gdal-grass-1.11.2.tar.gz
> > > > > > 
> > > > > > A snapshot of the gdalautotest suite is also available :
> > > > > >   http://download.osgeo.org/gdal/1.11.2/gdalautotest-1.11.2.tar.g
> > > > > >   z http://download.osgeo.org/gdal/1.11.2/gdalautotest-1.11.2.zip
> > > > > > 
> > > > > > The NEWS file is here :
> > > > > >   http://trac.osgeo.org/gdal/wiki/Release/1.11.2-News
> > > > > > 
> > > > > > I'll call for a vote promoting it soon if no serious problems are
> > > > > > reported before.
> > > > > > 
> > > > > > Best regards,
> > > > > > 
> > > > > > Even
> > > > > > --
> > > > > > Spatialys - Geospatial professional services
> > > > > > http://www.spatialys.com
> > > > > > _______________________________________________
> > > > > > gdal-dev mailing list
> > > > > > gdal-dev at lists.osgeo.org
> > > > > > http://lists.osgeo.org/mailman/listinfo/gdal-dev
> > > > 
> > > > --
> > > > Spatialys - Geospatial professional services
> > > > http://www.spatialys.com
> > 
> > --
> > Spatialys - Geospatial professional services
> > http://www.spatialys.com

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com


More information about the gdal-dev mailing list