HEADS UP - SITE OFFLINE FOR MAJOR UPGRADE THURSDAY MAY 14
Everyone: DevGurus is scheduled to go offline about 10:30 AM Central time on May 14 as we build the new AMD Community. We expect to be back up at some point late in the afternoon of the 16th, or...
View ArticleDig out memory usage by using CodeXL
Hi there, I am new CodeXL, and I was wondering is there anyway by using it to dig out the memory usage of the kernel.The memory usage section in the debugger here only indicate the memory usage by the...
View ArticleMistake in AMD GCN 3 ISA Reference (VOP_SDWA, VOP_DPP)
Hi, I am amateur. I found mistake in AMD GCN3 ISA Reference in VOP2, VOP1 and VOPC encoding format. In these places manual describes values in SRC0 field anddescribes 249 and 250 values as reserved....
View ArticleInstalling catalyst driver on Kaveri APU
I have been trying to install the catalyst driver for last few days, but its harder than I was expecting:-( I am giving you the detailed info thinking that it might be helpful to resolve the issue and...
View ArticleA Survey of CPU-GPU Heterogeneous Computing Techniques
https://www.academia.edu/12355899/A_Survey_of_CPU-GPU_Heterogeneous_Computing_Techniques Accepted in ACM Computing Surveys 2015 This paper clearly shows how research on heterogeneous computing has...
View ArticleHEADS UP: Forum Upgrade Coming
Folks, We are upgrading the forum software to a new version. This should happen the weekend of May 16, if all goes according to plan. So, about three weeks from now. I'll alert you of the precise dates...
View ArticleRe: clCreateKernel SIGSEGVs if used with binary image in openCL 1.2
Its good that you've identified the problem. As per the OpenCL spec, kernel program needs to be built/compiled irrespective of its created from source or binary. That's why, clBuildProgram() is a...
View ArticleRe: Device queues
Hi Jason,Creating device-side queue requires OpenCL 2.0 supported devices. If you check the clinfo output, it should report the "Device OpenCL C version" as "OpenCL C 2.0" for that particular device....
View ArticleIs global synchronization in OpenCL possible?
Hello everyone! As well known OpenCL barrier() function works only for single workgroup, and there is no direct possibility to synchronize workgroups. If it possible what's best approach for global...
View ArticleRe: clCreateKernel SIGSEGVs if used with binary image in openCL 1.2
The bitnes of kernels depends on bitness of device. AFAIK you need set GPU_FORCE_64BIT_PTR=1 enviroment variable to build 64 bit code for device.
View ArticleRe: Question about HSA-OpenMP-GCC-AMD
rtgllc, Thanks for testing gcc hsa branch. This is in active development and you got burned by the recent update to HSA 1.0F and the elimimation of Okra. This new update needs HSA 1.0F kernel and...
View ArticleRe: 2D array using clSVMAlloc ?
HI dipak ,Thanks for your answer.The sample code as follow ,"typedef struct { int a ; double c; double d; double d; ......}test;int szie = 25600 ;test *temp[5000] ;temp[0] = (temp...
View ArticleWhat attributes are correlated with "Could not debug kernel".
I have a Windows 8.1 Pro machine (Xeon E5-2650 v2 @ 2.6GHz (16 processors), 32GB RAM, 256GB SSD, 1TB drive, OpenCL2.0 AMD-APP 1642.5) with the latest AMD releases of Catalyst (Driver Packaging Version:...
View ArticleRe: Deep Learning on OpenCL
I absolutely agree on the benefits of OpenCL acceleration in these frameworks. I can't talk about where we're headed, but I can say this is a bleeding edge area that we are really interested in. In the...
View ArticleRe: What is right order to queue multiple kernels in loop
Thanks all to advice/answers, I got stable code, but problem was outside. Current version is: cl_event ndrEvt = 0; for(int i=0;i<numIter-1;i++){ // numIter = 1000 status =...
View ArticleRe: clCreateKernel SIGSEGVs if used with binary image in openCL 1.2
I'm not sure what you mean by --load and --dump command lines. I just use the FFT_Kernels.cl from the samples to do my fft.My program doesn't take these args. I only have 1 GPU device and I save binary...
View ArticleRe: What is right order to queue multiple kernels in loop
If you want to enqueue all the kernel commands into the same queue, you can even avoid using any event object at all. The reason is, host queues are in-order by default. You don't need any explicit...
View ArticleRe: Question about HSA-OpenMP-GCC-AMD
Some clarification: apparently it is a linker problem. Gcc couldn't find some .so files. So we modified the makefile and produced what follows: GCCHSAIL=$(GCC_HSA)/bin/gccLDFLAGS=-lhsa-runtime64...
View ArticleRe: clCreateKernel SIGSEGVs if used with binary image in openCL 1.2
Did you try the command line arguments "--dump" and "--load" to generate and consume the device binary for any of SDK sample? Please make sure that you are providing the correct device binary during...
View Article