Intel® SDK for OpenCL* Applications

Virtualbox and __ocl_svml_u8.so

OpenCL application on Fedora 20 on host machine works, and strace says that it loads __ocl_svml_e9.so. On vmachine inside Virtualbox the program fails at clCreateContext; strace says it doesn't find __ocl_svml_u8.so.

/proc/cpuinfo at host machine says that model name is Intel(R) Core(TM) i5-3470 CPU @ 3.20GHz, stepping is 9 and cpuid level is 13. On virtual machine cpuid level is 5. Is this related?

Is there some deep problem with OpenCL and virtualization? If not, known problem  with Virtualbox?

 

Coupling of OpenCL runtime versions between CPU and GPU drivers

I am trying to compare OpenCL kernel performance on some different CPUs, and I would like to use the same device version for this. For the machines that don't have an Intel GPU, I am using the CPU only runtime (3.0.1.15216), which makes the CPU Device report "OpenCL 1.2 (Build 80752)" as the version. That's all fine and good. But one of my machines is an i7-4700EQ with 4600 Graphics. I understand that for this machine, the CPU device support is also provided through the Graphics driver.

Barrier error, when using __local qualified memory

 

"aftrer using barrier function the value in memory, which is qualified as __local, is changed."

 

 

I could narrow down the range.

The problem comes from using barrier when I read and write some data in memory(array), which is qualified as __local.

I didn't see there is some limitation  the memory area must be used as only reading or writing.

But it seems to be operated that there is the cache and when local memory is read, that is saved in cache and flush when I use barrier, which argument is  "CLK_LOCAL_MEM_FENCE"

Dynamic parallelism in OpenCL drivers

We are evaluating GPU solutions to accelerate our algorithms, which have unavoidable sequential parts at the core.
Because of data dependencies we would need to send up-to 300000 commands per second to the GPU which introduce considerable slowdown.

It seems NVIDIA exposes "dynamic parallelism" through CUDA which would seems helpful to lessen this command submission stress. OpenCL 2.0 will offer something similar under the name "device enqueing" which will allow to launch child kernels and such.

About "Intel® SDK for OpenCL™ Applications 2014 Beta"

I found references about "Intel® SDK for OpenCL™ Applications 2014 Beta"  on Intel INDE page..

I wanted to know if it's coming soon as it seems to support OpenCL Android development on Intel GPUs?!..

This is good news! hope this motivates Intel to support also Intel GPUs in OpenCL on Linux as Android it's similar to Linux (not?)

Bug in OpenCL kernel builder on Intel HD Graphics

Hello. 

I found a curious bug in kernel builder of Intel OpenCL SDK (Version 2013 R3 64-bit) when it executes on Intel HD Graphics 4600 (Core i5 4570).

I write following kernel:

#define cn 1

__local uchar val1; val1 = 19;
__local uchar val2; val2 = 18;
__local uchar val3; val3 = 9;
__local uchar res; res = 255;
for (int c = 0; c < cn; ++c)
    if ( val2 > val1 || val3 < val1 )
        {
            res = 0;
            break;
        }

This is a simplified version of a piece of code of InRange function from OpenCV library. 

No luck with debugging OpenCL kernels using VC2013

Hello,

I would like to debug my opencl kernel using the opencl debugger plugin within VC2012 (+ Intel SDK for OpenCL Applications 2013 3.0.0.81147; Intel SDK for OpenCL - CPU Only Runtime Package 3.0.1.15216).

Here is what I did:

1. Called clBuildProgram with additional options, as mentioned in the documentation:
    clBuildProgram(program, 1, &device, "-g -s C:\\PathToMy\\File.cl", NULL, NULL)
2. Enabled the OpenCL debugger, told it to trigger at Dim 0/0/0
3. Started debugging. VC2012 seems to halt, but I can't do any actual debugging.

"internal error" when building for Iris pro 5200 on Mac OS X

Hi,

I'm trying to get some kernels running on the Iris pro 5200 in my new late 2013 macbook pro 15". The kernel works fine on Windows with my 3rd gen i7 CPU as well as my GeForce GTX 760. However, on Mac OS X 10.9/Xcode5 I keep getting an "internal error" when building my kernel from source:
The exact debug output from within clBuildProgram is

[CL_DEVICE_NOT_AVAILABLE] : OpenCL Error : Error: Build Program driver returned (10015)

OpenCL Warning: clBuildProgram failed: could not build program for 0x1224500 (Iris Pro) (err: -2)

Iscriversi a Intel® SDK for OpenCL* Applications