Opened 17 years ago

Closed 9 years ago

#1346 closed defect (invalid)

Warping - doing work for nothing

Reported by: warmerdam Owned by: warmerdam
Priority: highest Milestone:
Component: GDAL_Raster Version: unspecified
Severity: blocker Keywords:
Cc: Markus Neteler, vali.system@…

Description (last modified by Markus Neteler)

In situations where the gdalwarp output chunk does not overlap input data
from the current file, a great deal of work is still done by the warper.

It is desirable:
 1) To short circuit chunks with no input data. 
 2) To shrink chunks for which there is not much coverage by input data. 

The SKIP_NOSOURCE warp option can be enabled to do (1), but it is not on
by default - presumably because there are some aspects of processing that 
we might need to do whether there is source data or not.  For instance
initializing the destination buffer to a background value.  Also, possibly 
producing the various output masks. 

This needs some thought, and a more holistic solution than hacks like
SKIP_NOSOURCE.

Attachments (1)

gdal-1.3.2.gdalwarp.patch (6.9 KB ) - added by vali.system@… 17 years ago.
First draft patch to reduce warping window

Download all attachments as: .zip

Change History (4)

by vali.system@…, 17 years ago

Attachment: gdal-1.3.2.gdalwarp.patch added

First draft patch to reduce warping window

comment:1 by Markus Neteler, 16 years ago

Cc: Markus Neteler added; neteler@… removed
Description: modified (diff)

comment:2 by sprice, 15 years ago

Could you take a look at the current trunk code and see if it addresses the issue? While fixing #2327 I made some optimizations which makes warping run much faster.

comment:3 by Even Rouault, 9 years ago

Resolution: invalid
Status: newclosed

Clsoing that ticket. If optimizations should be done, that should be in the warping algorithm itself, not in the utility. And the current patch implies SKIP_NOSOURCE. It also implies that the source dataset has a geotransform. There's another ticket with another approach for the same goal (although it should be verified how it works with respect to SKIP_NOSOURCE)

Note: See TracTickets for help on using tickets.