Hello Erik.
On Sat, 2013-05-18 at 21:50 -0400, Erik Schnetter wrote: > That's actually kind of good... it means that it should be easy to > find out what is going wrong. It could be something simple as a > missing entry in PATH or LD_LIBRARY_PATH. I hope so, but I'm a bit floating since my knowledge is limited :-( > > > (1) Are you running pocl from your build directory, or did you install > it? Both, on the machines I tested, it has already been installed since I tested it with my own little crapy program. The tests has been run from inside the build directory > (2) Is llvm (that will be used to compile the kernels) in your PATH > and LD_LIBRARY_PATH? On two boxes, I have already installed the port devel/llvm and devel/llvm-devel (latter one is on FreeBSD now llvm-3.4, first is still llvm-3.2: ohartmann@thor: [~] pkg info llvm-devel llvm-devel-3.4.r181598 Low Level Virtual Machine Compiling pocl-0.8 on that machine with llvm-3.4 eats up 31 GB of swap (and 8 GB of RAM) when llvm-linker is getting in action ;-) I guess a bug, llvm-3.2 doesn't show the problem) On both systems, the port system ensures that llvm and the appropriate libs are in the path. > (3) Can you look at the directory tests/testsuite.dir, and in one of > the subdirectories there (e.g. 001)? There should be a file > "testsuite.log" that should contain an error message. 001 looks like this: # -*- compilation -*- 1. testsuite.at:30: testing example1: dot product ... ./testsuite.at:39: $abs_top_builddir/examples/example1/example1 --- /dev/null 2013-05-19 10:38:10.000000000 +0200 +++ /usr/ports/devel/pocl/work/pocl-0.8/tests/testsuite.dir/at-groups/1/stderr 2013-05-19 10:38:12.295152942 +0200 @@ -0,0 +1,3 @@ +/usr/local/bin/llvm-link: /usr/ports/devel/pocl/work/pocl-0.8/lib/kernel/amd64-portbld-freebsd10.0/kernel-amd64-portbld-freebsd10.0.bc: error: Could not open input file: No such file or directory +/usr/local/bin/llvm-link: error loading file '/usr/ports/devel/pocl/work/pocl-0.8/lib/kernel/amd64-portbld-freebsd10.0/kernel-amd64-portbld-freebsd10.0.bc' +/usr/local/bin/opt: /tmp/poclLbh5aJ/pthread/dot_product/kernel_linked.bc: error: Could not open input file: No such file or directory --- expout 2013-05-19 10:38:10.557553142 +0200 +++ /usr/ports/devel/pocl/work/pocl-0.8/tests/testsuite.dir/at-groups/1/stdout 2013-05-19 10:38:12.298752817 +0200 @@ -1,5 +1,3 @@ -(0.000000, 0.000000, 0.000000, 0.000000) . (0.000000, 0.000000, 0.000000, 0.000000) = 0.000000 -(1.000000, 1.000000, 1.000000, 1.000000) . (1.000000, 1.000000, 1.000000, 1.000000) = 4.000000 -(2.000000, 2.000000, 2.000000, 2.000000) . (2.000000, 2.000000, 2.000000, 2.000000) = 16.000000 -(3.000000, 3.000000, 3.000000, 3.000000) . (3.000000, 3.000000, 3.000000, 3.000000) = 36.000000 -OK +ERROR +(0.000000, 0.000000, 0.000000, 0.000000) . (0.000000, 0.000000, 0.000000, 0.000000) = 1104145761423820062720.000000 +FAIL ./testsuite.at:39: exit code was 255, expected 0 1. testsuite.at:30: 1. example1: dot product (testsuite.at:30): FAILED (testsuite.at:39) On the console, when running (and failing) the tests, I always see the missing *.bc file error message (in all tests). ATTENTION: there is no folder amd64 in /usr/ports/devel/pocl/work/pocl-0.8/lib/kernel/ I reported this issue earlier since there was a simple build issue for pocl-0.7. The *BSD name their 64bit extentions AMD64, not X86_64, which seems to be the Intel/Linuxuism way. I never found that issue myself for repairing it - please enlighten me where to put my fingers on ;-) Regards, Oliver > > > -erik > > > On Sat, May 18, 2013 at 8:55 PM, O. Hartmann > <[email protected]> wrote: > On Sat, 2013-05-18 at 18:34 -0400, Erik Schnetter wrote: > > Oliver > > > > > > I am using pocl on OSX, and I am also building and using it > without > > ICD. I don't encounter any issues in this respect -- I > simply link > > against pocl instead of against Apple's OpenCL library, and > things > > work fine. > > > > > > Did you run "make check"? This should run a few self-tests, > including > > one called "Scalar wave equation" that I contributed. This > is a full > > physics application (albeit with a small grid), and if this > runs fine, > > things should be basically working for you. > > > > > > -erik > > [...] > > Hello, > > I tried the check (make check). They fail ALL. > > Oliver > >
signature.asc
Description: This is a digitally signed message part
------------------------------------------------------------------------------ AlienVault Unified Security Management (USM) platform delivers complete security visibility with the essential security capabilities. Easily and efficiently configure, manage, and operate all of your security controls from a single console and one unified framework. Download a free trial. http://p.sf.net/sfu/alienvault_d2d
_______________________________________________ pocl-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/pocl-devel
