Changes between Version 2 and Version 3 of Notes/MiriadScalingLimitations

Timestamp:

06/07/07 18:11:36 (17 years ago)

Author:

mci156

Comment:

--

Notes/MiriadScalingLimitations

@@ -9 +9 @@
 There is also a {{{MAXBUF}}} parameter for the C parts of MIRIAD, defined in {{{inc/maxdimc.h}}}, but used only by tasks that use routines from {{{xyzio.c}}}.
 
-The maximum image size that can be handled is to some extent defined by the {{{MAXDIM}}} parameter in {{{inc/maxdim.h}}}, which is often used to indicate the maximum size of maps (the longest side will be less than {{{MAXDIM}}}).
-However the actual maximum may vary from that suggested by MAXDIM, depending on how big {{{MAXBUF}}} is, whether the task processes the entire image or in a raster fashion, and whether any subregions have been allocated out of the space defined by {{{MAXBUF}}}. This last item adds some unpredictability to the whole issue; different ordering of clean windows may change the efficiency of allocation of chunks of {{{MAXBUF}}}.
+The parameter {{{MAXDIM}}} parameter in {{{inc/maxdim.h}}}, is often used to indicate the maximum size of maps (the longest side will be less than {{{MAXDIM}}}).
+However the actual maximum that can successfully be allocated may vary from that suggested by MAXDIM, depending on how big {{{MAXBUF}}} is, whether the task processes the entire image or in a raster fashion, and whether any subregions have been allocated out of the space defined by {{{MAXBUF}}}. This last item adds some unpredictability to the whole issue; for example different ordering of clean windows may change the way in which the memory reserved by {{{MAXBUF}}} is fragmented, with the result that the packing is less efficient on some runs, and the maximum image size that can be accommodated may be different.
 
-There are (now) comments in the include files explaining the upper limits in some detail.
+There are (now) comments in the include files noted above, that explain the upper limits in some detail.
 
 For [http://www.atnf.csiro.au/cgi-bin/pm/manage/040 CABB], the biggest change is along the spectral (channel) axis; since MIRIAD usually handles each channel separately we do not expect that scaling will present much of a problem for single fields.