Page 1 of 1

The efficient matrix multiplication using opencl

PostPosted: Thu Jan 14, 2016 11:20 am
by Tiago
Hey guys! Did you already run the efficient matrix multiplication using opencl (http://www.adapteva.com/white-papers/ef ... ng-opencl/)? If so, could you help me on this? (I have a parellella with epiphany III)

Re: The efficient matrix multiplication using opencl

PostPosted: Thu Jan 14, 2016 5:39 pm
by jar
Do you have any specific questions that you need answered?

Re: The efficient matrix multiplication using opencl

PostPosted: Thu Jan 14, 2016 7:25 pm
by Tiago
jar wrote:Do you have any specific questions that you need answered?


Hey jar,
My question is: how to build and run that algorithm?

Re: The efficient matrix multiplication using opencl

PostPosted: Fri Jan 15, 2016 1:59 am
by jar
Look at some of the Makefiles in the browndeer/examples directory and follow the same compilation process

Re: The efficient matrix multiplication using opencl

PostPosted: Wed Jan 20, 2016 8:19 pm
by Tiago
hey guys, I build the example with (gcc matrixMult.c -o matrixMult.c.x -I/usr/local/browndeer_new/include -L/usr/local/browndeer_new/lib -locl -lm -fopenmp -lstdcl), but when I tried to execute I received this message:

Code: Select all
coprthr-1.6.0 (Freewill)
[12701] clmesg info: cmdsched.c(86): cmdqx1: run
[12701] clmesg WARNING: command_queue.c(39): __do_create_command_queue_1: cmdq exists
[12701] clmesg WARNING: command_queue.c(39): __do_create_command_queue_1: cmdq exists
dmalloc(0,65536): 0x8e104000 0x8e104000 0x90004000
dmalloc(0,65536): 0x8e104000 0x8e114000 0x90004000
dmalloc(0,65536): 0x8e104000 0x8e124000 0x90004000
[12701] clmesg info: cmdsched.c(86): cmdqx1: run
Segmentation fault


could you help with this?

Re: The efficient matrix multiplication using opencl

PostPosted: Thu Jan 21, 2016 12:28 am
by nickoppen
A segmentation fault usually indicates a problem with memory management on the host side.

I suggest you sharpen your debugging skills (gdb by itself or via eclipse or code::blocks) and that will take you to where the segmentation fault occurs and that will tell you where to start looking.

No one here on the forum can do that for you. These things are "can opener" jobs. You've got to get in there and figure out what is going on.

(Also be aware that the memcopy at the end of the kernel is not supported in the current realease.)