Hi Igor, I applied the patch onto SF_RELEASE_2_1_0 and it compiles fine. When I run the sample programs, CUDATopology and CUDAKernelTest work fine whilst others fail like this:
$ runx10 KMeansCUDA points: 100000 clusters: 8 dim: 4 Running using 1 GPUs. GPU known as (Place 1) gets role 0 offset 0 len 100000 100000 8 4 2.703 kernel: 1.689 dma: 0.447 cpu: 0.463 reduce: 0.101 Segmentation fault $ runx10 CUDABlackScholes Using the GPU at place (Place 1) This program only supports a single GPU. CUDA_ERROR_OUT_OF_MEMORY (At common/x10rt_cuda.cc:452) Aborted Are these failures considered 'expected' or my graphics card is not very intelligent for such tasks? I have a GeForce 8300 GS Device 0: "GeForce 8300 GS" > CUDA Driver Version: 3.20 > CUDA Runtime Version: 3.20 > CUDA Capability Major/Minor version number: 1.1 > Total amount of global memory: 133496832 bytes > Multiprocessors x Cores/MP = Cores: 1 (MP) x 8 (Cores/MP) > = 8 (Cores) > Total amount of constant memory: 65536 bytes > Total amount of shared memory per block: 16384 bytes > Total number of registers available per block: 8192 > Warp size: 32 > Maximum number of threads per block: 512 > Maximum sizes of each dimension of a block: 512 x 512 x 64 > Maximum sizes of each dimension of a grid: 65535 x 65535 x 1 > Maximum memory pitch: 2147483647 bytes > Texture alignment: 256 bytes > Clock rate: 0.92 GHz > Concurrent copy and execution: No > Run time limit on kernels: Yes > Integrated: No > Support host page-locked memory mapping: Yes > Compute mode: Default (multiple host Thanks Richard Gomes M: +44(77)9955-6813 http://tinyurl.com/frgomes twitter: frgomes JQuantLib is a library for Quantitative Finance written in Java. http://www.jquantlib.org/ twitter: jquantlib On 09/11/10 22:22, Igor Peshansky wrote: > Richard, > > "svn diff -r18088:18092 x10.runtime/x10rt/common/x10rt_cuda.cc" in our > repo should generate that patch. > Igor > > Richard Gomes<rgomes1...@yahoo.co.uk> wrote on 11/09/2010 05:11:43 PM: > >> Hi Dave, >> >> Could you please send the patch again? >> >> Thanks >> >> Richard Gomes >> M: +44(77)9955-6813 >> http://tinyurl.com/frgomes >> twitter: frgomes >> >> JQuantLib is a library for Quantitative Finance written in Java. >> http://www.jquantlib.org/ >> twitter: jquantlib >> >> On 09/11/10 03:45, Dave Cunningham wrote: >>> Thanks for trying out X10/CUDA >>> >>> Your initial problem with CUDATopology is due to the fact that > X10RT_ACCELS >>> is ineffective if X10 was built without -DX10RT_CUDA=true, this means > the >>> X10 application was unable to 'see' the accelerators. You correctly >>> surmised that building X10 from the source release was necessary. >>> >>> The build errors are due to nvidia adding more error codes and making >>> backwards incompatible changes in the CUDA API. This is now fixed in > SVN. >>> I checked the build with the following CUDA versions: >>> >>> cuda-2.2 cuda-2.3 cuda-3.0 cuda-3.1 cuda-3.2.12 >>> >>> If you decide to use SVN, there are some changes to the way kernels > should >>> be written in X10 that are currently undocumented (except via the code > in >>> the samples dir). Also, we can't guarantee there won't be more > changes (and >>> breakages) before the next release. However you will be able to try > new >>> features like clocks and constant memory on the GPU. >>> >>> If using SVN does not appeal, you can also patch the source release to > fix >>> this problem. Apply the attached patch from the root of the source > release >>> as follows, and rebuild: >>> >>> patch -p0< cuda_3.2.patch >>> >>> hope this helps ------------------------------------------------------------------------------ Centralized Desktop Delivery: Dell and VMware Reference Architecture Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View: A highly scalable, end-to-end client virtualization framework. Read more! http://p.sf.net/sfu/dell-eql-dev2dev _______________________________________________ X10-users mailing list X10-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/x10-users
