File sizes and linear offsets use a type MPI_Offset, which according
to the specification should be large enough to hold the size of any
file supported by the file system.
hmm, that may be but there are places in MPI I/O that limit to size of
signed int, here is cut direct from the mpi 3 standard:
int MPI_File_read_all(MPI_File fh, void *buf, int count,
MPI_Datatype datatype, MPI_Status *status)
the count argument is a signed int, so one can't read more than INT_MAX
in one go, and here is the VTK code where that is used
291 // Figure out how many bytes to read.
292 vtkIdType length = this->GetDataScalarTypeSize();
293 length *= this->NumberOfScalarComponents;
294 length *= extent[1]-extent[0]+1;
295 length *= extent[3]-extent[2]+1;
296 if (this->GetFileDimensionality() == 3) length *=
extent[5]-extent[4]+1;
297
298 if (length > VTK_INT_MAX)
299 {
300 vtkErrorMacro(<< "Cannot read more than " << VTK_INT_MAX
301 << " bytes at a time.");
302 }
303
304 // Do the read. This is a coordinated parallel operation for
efficiency.
305 MPICall(MPI_File_read_all(file.Handle, buffer,
static_cast<int>(length),
306 MPI_BYTE, MPI_STATUS_IGNORE));
the VTK code checks for over flow. however as that is not the error
reported by the user, I have to admit that I am wrong about Sam's
problem. Sorry about that.
Burlen
On 11/22/2016 02:11 PM, Moreland, Kenneth wrote:
Although we have certainly run into issues with communicating large
messages, I’m not sure this is the issue with MPI-I/O. File sizes and
linear offsets use a type MPI_Offset, which according to the
specification should be large enough to hold the size of any file
supported by the file system. Although it is true that creating the 3D
subarray uses 32-bit indices, those are for each of the separate
dimensions, and 2000 falls well below that limit.
Besides, it looks like that error happens when checking an ifstream
object, so it looks like the reader is bypassing the MPI-I/O path anyway.
I would follow Berk’s suggesting of looking into the filesystem before
diving into the guts of MPI-I/O.
-Ken
*From:*ParaView [mailto:[email protected]] *On Behalf Of
*Burlen Loring
*Sent:* Tuesday, November 22, 2016 12:55 PM
*To:* Geveci, Berk (External Contact) <[email protected]>; Keyes
S.D. <[email protected]>
*Cc:* [email protected]
*Subject:* [EXTERNAL] Re: [Paraview] RAW (binary) file import issues
I've hit that before. the RAW reader uses MPI-I/O, and MPI uses signed
int everywhere in its API, so there are limits that you end up
hitting. For instance to partition the data among processors one must
give MPI-I/O the starting offset of the data in the file as a signed
int, and that is limited to 2^31.
Burlen
On 11/22/2016 11:20 AM, Berk Geveci wrote:
Hmmm that happens to be around 4 GB which makes me wonder if there
is a limit somewhere in that reader... Any chance you can try this
on a Linux machine?
On Thu, Nov 17, 2016 at 10:30 AM, Keyes S.D.
<[email protected] <mailto:[email protected]>> wrote:
Dear all,
I am encountering issues when trying to import large 8 bit RAW
volumes to Paraview 5.1.2 (64 bit).
The file I wish to import has xyz dimensions of 2000x2000x1920
voxels (Unsigned char), 7.5GB. I am running a machine with
192 GB RAM and 12 CPU cores.
I can import up to a 2000x2000x1000 crop of this file, but I
encounter the following message when I attempt larger z
dimensions:
ERROR: In
C:\bbd\df0abce0\source-paraview\VTK\IO\Image\vtkImageReader2.cxx,
line 592
vtkPVImageReader (0000000007767F90): Initialize: Could not
open file
\\cseg_2\ERC\SLS_201606\Myco_2\CT2PpR3h1_\Hyphalseg_Raw_2000_2000_1920_8bit_SKELETON_FILT_PARAVIEW_TEST.raw
<file://cseg_2/ERC/SLS_201606/Myco_2/CT2PpR3h1_/Hyphalseg_Raw_2000_2000_1920_8bit_SKELETON_FILT_PARAVIEW_TEST.raw>
I cannot find an obvious memory ceiling within the paraview
settings - any ideas as to whether there is some effective
upper limit to import file size?
Sam
Samuel D Keyes, MEng, PhD
New Frontiers Fellow
Bioengineering / μVIS Centre for Computed Tomography
University of Southampton
E: [email protected]
<mailto:[email protected]><https://www.outlook.soton.ac.uk/owa/
<http://www.outlook.soton.ac.uk/owa/>redir.aspx?SURL=dhUg3da8XR77LeItLDzJQwhAwmH0Wx6PHrXeKfxx2tOD3LCr737TCG0AYQBpAGwAdABvADoAUwAuAEQALgBLAGUAeQBlAHMAQABzAG8AdABvAG4ALgBhAGMALgB1AGsA&URL=mailto%3aS.D.Keyes%40soton.ac.uk>
T: 07898720248
_______________________________________________
Powered by www.kitware.com <http://www.kitware.com>
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check the ParaView Wiki at:
http://paraview.org/Wiki/ParaView
<http://paraview.org/Wiki/ParaView>
Search the list archives at:
http://markmail.org/search/?q=ParaView
<http://markmail.org/search/?q=ParaView>
Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview
_______________________________________________
Powered bywww.kitware.com <http://www.kitware.com>
Visit other Kitware open-source projects
athttp://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check the ParaView Wiki
at:http://paraview.org/Wiki/ParaView
Search the list archives at:http://markmail.org/search/?q=ParaView
Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview
_______________________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check the ParaView Wiki at:
http://paraview.org/Wiki/ParaView
Search the list archives at: http://markmail.org/search/?q=ParaView
Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview
_______________________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check the ParaView Wiki at:
http://paraview.org/Wiki/ParaView
Search the list archives at: http://markmail.org/search/?q=ParaView
Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview
