My first program from OpenCL
Hello. This is my first program from OpenCL, but dosn't work. Help me pls. What's wrong with this?clBuildProgram(program1, 0, NULL, clcompileflags, NULL, NULL); return error code -11 PS: sorry for my...
View ArticleRe: opencl equivalent to cuda warp vote functions
Thanks, Tzachi. It looks like OpenCL 2.0 has these methods: http://www.khronos.org/registry/cl/specs/opencl-2.0-openclc.pdf#page=152 So, I will wait until the 2.0 release. On Sat, Jun 14, 2014 at 2:54...
View ArticleRe: clBuildProgram had an unhandled exception
Hi Daniel,Thanks for your information.We'll try to reproduce the error and keep you updated. Meanwhile, as you've mentioned, building your program with optimization option "-cl-opt-disable" is the best...
View ArticleRe: My first program from OpenCL
you don't have newline \n behind that comment in kernel source. your whole kernel code is on single line. best is don't include that comment into kernel source at all. remove quotation marks around it...
View ArticleRe: printf from kernel misbehaves again
Hi, Sorry for the delayed response. Does the problem persist? Can you try with the latest driver and report back? --Prasad
View ArticleOpenCL fails on restart after installing drivers
I've got this problem that my OpenCL doesn't work properly. I've done some investigating and I've noticed that whenever I install the drivers for my graphicscard OpenCL is activated untill I restart my...
View ArticleRe: My first program from OpenCL
Hi, Welcome to the OpenCL community. After calling the API clBuildProgram() [ i.e. [line] clerr = clBuildProgram(program1, 0, NULL, clcompileflags, NULL, NULL); ], please call clGetProgramBuildInfo()...
View ArticleMost efficient way of dividing by power of two
I have an array of signed ints, and I would like to divide by 2^n.I currently use this routine to do it: int out = in >= 0 ? int >> n : -((int & 0x7FFFFFFF) >> n); Is there a better...
View ArticleRe: Why does it make sense for catalyst to be closed source?
Your first section doesn't seem to contribute to an answer. The second however relies on the idea of AMD dev's out optimizing Intel or Nvidia -- I don't consider this realistic given the huge...
View ArticleRe: Bug in CPU implementation
Hi bubu I'm reviving this thread. Do you still see this issue? --Prasad
View ArticleStill confused about LDS
From the AMD App programming guide, from chapter 5.2, Local Memory (LDS) Optimizations, I read:Bank conflicts are determined by what addresses are accessed on each half wavefront boundary. Threads 0...
View ArticleRe: Most efficient way of dividing by power of two
Checking the OpenCL docs, it says that arithmetic shift is used when the left operand is signed, so that workaround is unnecessary.http://www.khronos.org/registry/cl/specs/opencl-1.x-latest.pdf#page=143
View ArticleRe: OpenCL fails on restart after installing drivers
Sorry, I don't have any solution at this point. But, I've few questions as given below:1.@"tried numerous drivers for the graphicscard". So, can I assume that you've tried with latest driver?...
View ArticleRe: Most efficient way of dividing by power of two
Thanks, realhet. So, negative numbers are rounded up towards zero? The spec mentiones setting empty bits, but this is a bit vague>
View ArticleRe: Bug in CPU implementation
Yep, Catalyst 14.4 still exposes the problem ... amplified The render differs from the GPU implementation and also from the C++/SIMD implementation. It only happens with the AMD's OpenCL CPU...
View ArticleRe: Can I force the compiler to reduce VGPR usage?
Would like to know the current status. Or at least to get feedback from i.e. CodeXL where those VGPRs counts come from. Simple counting in the code doesn't work.
View ArticleRe: Bug in CPU implementation
i still have same problem so i try to resolve some time but can not ca cuoc onlinevery bed I'm using W8.1 x64 ( also W7 SP1 x64 ), Cat.14.4, Radeon 7770-1Gb, Intel i7-2700K Sandy Bridge.
View ArticleHow can I easily switch between driver versions on Linux?
Using LKM functions (modprobe, rmmod, lsmod, etc) it is easy to install and change kernel-modules. I'd like to be able to easily switch between different versions of fglrx, so I can quickly test out a...
View ArticleRe: GaussianNoiseGL sample from APP fails with 4.x GL context on Linux
Hm, that doesn't solve it for me at least. I didn't try with the sample, but with a bigger application which has the same issue, and using glXCreateContext didn't change the behaviour there. I'll check...
View ArticleRe: Most efficient way of dividing by power of two
Hi, Indeed that -1 problem... A simlpe >> just fails on it. So I've checked how efficient are the different methods: int out = in >= 0 ? in >> n : -(-in >> n); //this is re OP's...
View Article