As of today, the latest Intel OpenCL drivers supports very few different image formats (only 32 formats for latest 188.8.131.5232 GPU driver and 43 formats for latest "beta" CPU driver). These are very few compared to the 80-100 formats supported by Nvidia and AMD. The lack of "full" image format support leads to undesirable introduction of Intel-specific work-arounds that both complexifies code and reduces performance. In particular, it had been a great advantage if "SNORM"-types could been supported for CL_R/CL_RG/CL_RGBA x INT8/INT16.
There seems to be an issue with CL context reference-counting on Intel OpenCL GPU devices. Calling clRetainContext followed by clReleaseContext on a context object results in a silent failure that invalidates the context and triggers a crash on subsequent OpenCL calls on my computer.
A minimal VS2010 project to reproduce is attached. The project has no dependencies besides an OpenCL SDK, and should built & run out of the box.
Hi, I have a simple test program using an image3d with CL_UNSIGNED_INT8 data in it. When I use the following sampler my test program works fine.
sampler_t trilinear_sampler = CLK_FILTER_LINEAR | CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAMP_TO_EDGE;
But if I switch to :
sampler_t trilinear_sampler = CLK_FILTER_LINEAR | CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAMP;
then the sampling seems to be done using Nearest and not Linear.
I have attached the driver Info for my configuration below.
The attached kernel causes a segfault when clBuildProgram is called. Steps to reproduce:
- On an Ubuntu 64-bit 12.04.1 system, install onlineclc (http://gitorious.org/onlineclc).
- Install the 2013 beta SDK as the only (or first) OpenCL platform on the system.
- Run onlineclc crash.cl. It will segfault.
The kernel is a minimal example created by deleting as much code as possible while still segfaulting, so it doesn't do anything useful.
Well, I think that this is my first post on the forum so hi! to everyone.
I have an issue when debugging an OpenCl kernel, when I try to enter a function declared outside it and invoqued inside the kernel I have the following error:
First-chance exception at 0x763cc41f in OpenCL_EGS.exe: Microsoft C++ exception: OclSocketError at memory location 0x02dff9f0..
The thread 'Win32 Thread' (0x11b4) has exited with code -1 (0xffffffff).
I’m happy to announce the availability of new Intel SDK for OpenCL Applications XE 2013 for Linux* Operating Systems with OpenCL 1.2 features on both Intel Xeon processor and Intel Xeon Phi coprocessor.
This new SDK helps preserve the OpenCL investment made on Intel® Xeon® processors, Intel® Core™ Processor, and Intel® HD Graphics. OpenCL code developed for these processors will run on Intel Xeon Phi coprocessor with minimum changes.
When I was using the image2D objects to optimize my application, there was a strange phenomenon happened.
I have set the width and the height of the image2D buffer and the performance would increase as the height increase while the width decrease.
However, after I modified my opencl code to try another implementation and then restore the code to its status before modification, the performance was slower than before. And the performance would increase as the height decrease while the width increase.
I'm using /opt/intel/opencl-1.2-3.0.56860/lib64/libintelocl.so
All the programs I have function using the Nvidia OpenCL driver for CUDA 5.0
All my programs seg-fault with the Intel device. I'm running .Red Hat Enterprise Linux Server release 6.2 (Santiago)
On a dual-socket 4 core Intel(R) Xeon(R) CPU E5620 @ 2.40GH
Heres a picture of the stack:
I'd like to know what the interoperability status is between Intel OpenCL and TBB. In specific, can I look forward to problems (or not) if I make calls to Intel OpenCL from within TBB tasks? Do I have to worry about oversubscription (will they use the same thread queue)? Do I have to worry about which version of TBB I use with which version of OpenCL. Sorry if this is answered somewhere in the docs already, I couldn't find it.
Build log claims:
fc1 build 1 succeded.
fc1 build 2 succeeded.
Error: Internal Error.
The same program works fine on the CPU; it only fails in the GPU. Any pointers on where to start? Thanks much (I am using OpenCL 2012 on a Windows 7 Pro with Visual Studio 2010 Express). I cannot install 2013, it fails for both the 32 bit and 64 bit versions)