Intel® SDK for OpenCL* Applications

Will you open source LLVM MIC (xeon phi) backend.. (seen in ocl 2013 beta binaries)

Hi seeing in binaries I found references to llvm.mic instructions so is clear xeon Phi (MIC) support in OpenCL will be trough a LLVM MIC backend..
Hope in time for release you contribute back this backend to llvm community as only right now MIC is supported through Intel compilers and a simple GCC port (with no new sopport for 512bit vector instructions)..

Same OpenCL kernel gets different results on CPU and IGP

I wrote a very simple code for testing share resouces. My envorinment: i7-3770k win7 64-bit Intel OpenCL SDK 2013. The attached is my project. When OpenCL kernel ran on CPU device, the results are right. However, when it ran  on GPU device, the results are wrong. Anyone can help me explain this problem?

Thanks in advance.

is the instruction set architecture for HD graphics 4000 publicly available?

Hi,  We are doing a performance comparison of Ivy Bridge, Tesla, and APU for SGEMM. We found that the Ivy Brdige CPU and (integrated) GPU only reaches the 13% (29 GFLOPS) and 33% (49 GFLOPS) of the theoretical peak performance respectively, and we would like to figure out the sources of inefficiencies by looking at the assembly code. Is there a way to view the assembly code for Intel HD graphics 4000 in Ivy Bridge? We have tried Intel offline compiler, but it only gives us the CPU assembly and the intermediate LLVM code. Any advice is greatly appreciated. Yao

OpenCL compiler generating mad instruction

I have a compute instensive kernel performing lots of multiply and adds.
Am using opencl mad() function and float16 for the compiler to generate avx mad instruction.
But when I see the ASM dump from intel offline compiler, it shows mul and add instructions on YMM registers but no mad at all.

Problems with Compiler Flags in 2013 Beta


I seeing serveral issues with 2013 Beta when compiling OpenCL kernels:

When compiling for CPU target (installed CPU i7-3770K) with the "-cl-strict-aliasing" option, I'm getting a build error (-43).

When compiling for iGPU target with "-cl-fast-relaxed-math" performance drops by 30%. This issue also occurs when using the 2012 SDK.

I can provide a reproducer if needed.


Compiling SHOC benchmark suite for HD4000


I'm currently trying to compile SHOC (, version 1.1.4) for GT2. As there is no OpenCL in Linux I'm trying to build it in Windows using cygwin.

I did the following:

./configure -with-cuda=no CPPFLAGS="-I/cygdrive/c/Users/heinecke/Documents/spaffy-shoc-7380bc4/include -Dalloca=__builtin_alloca" LDFLAGS="-L/cygdrive/c/Users/heinecke/Documents/spaffy-shoc-7380bc4/lib/x64"

Include and lib dir are links into the SDK folder in C:\\Program Files (x86)\\Intel\\OpenCL SDK\\2.0\\

Intel® SDK for OpenCL* Applications abonnieren