- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
pardiso claims to not have enough memory for a very small system with 54 unknown (and with many more small systems, but for this, I created an example code).
System is Windows and I use the MS Visual Studio 11 compiler using the threaded library with 32 bit indices.
I also had tested the 64 bit indices, with the same effect (maybe not with exactly this data).
I get the error code -2 (not enough memory) very often, preferably with smaller equation systems, but that's maybe because the smaller systems are used more often.
Any suggestion how to solve this, is welcome,
Harald
The small example code and test data that fails are attached.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Which version of MKL did you use? Your program built and gave the correct results using MSC 18.0 and MKL 11.2.2 as well as with the current version of Intel C and MKL:
Intel(R) Math Kernel Library Version 11.2.2 Product Build 20150120 for Intel(R) 64 architecture applications open file 'Nxb-38.txt' x[0] = 1.439951e-010 x[53] = -2.173159e-003
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I'm using MKL 11.3.0.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I see the similar result with the latest vesion of MKL
open file 'Nxb-38.txt'
Major version: 2017
Minor version: 0
Update version: 0
Product status: Product
Build: 20160801
Processor optimization: Intel(R) Advanced Vector Extensions 2 (Intel(R) AVX2) enabled processors
================================================================
x[0] == 1.43995e-010
x[53] == -0.00217316
passed
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Does PARDISO respects the number of threads set with the environment variable MKL_NUM_THREADS?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
yes, in the case if you linked with threaded version of MKL.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Maybe something with my compiler or linker calls is wrong:
Visual Studio is installed in: C:/Program Files (x86)/Microsoft Visual Studio 11.0/
set VC = C:/Program Files (x86)/Microsoft Visual Studio 11.0/VC/bin
%VC%/x86_amd64/cl /c /O2 /EHcs /I"%MKLROOT%/include" test_mkl.cpp /Fotest_mkl.obj
%VC%/x86_amd64/link.exe /libpath:"%MKLROOT%/lib/intel64_win" /machine:x64 mkl_intel_lp64.lib mkl_core.lib mkl_intel_thread.lib libiomp5md.lib /out:test_mkl.exe test_mkl.obj
I tried setting MKL_NUM_THREADS=1 and also to link against the sequential libs (see below) to exclude an influence of threading, but it leads to the same result.
%VC%/x86_amd64/link.exe /libpath:"%MKLROOT%/lib/intel64_win" /machine:x64 mkl_intel_lp64.lib mkl_core.lib mkl_sequential.lib /out:test_mkl.exe test_mkl.obj
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
at the first glance, all is OK with your settings. I have installed 11.3 update 3 version of MKL on my system, and I also used exactly the same linking line. Please see below the output i have.
...\_Forums\u684920>test.exe
open file 'Nxb-38.txt'
Major version: 11
Minor version: 3
Update version: 3
Product status: Product
Build: 20160413
Processor optimization: Intel(R) Advanced Vector Extensions (Intel(R) AVX) enabled processors
================================================================
x[0] == 1.43995e-010
x[53] == -0.00217316
passed
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page