amd APP for opencl on cpu?

[super gnu]

ok i downloaded the amd APP sdk v2.9 to try to test cpu based opencl. but i suspect either ive downloaded the wrong thing, or im missing a step.

on my renderbox with no gpu i get error 1001 (iirc) when i try to "select opencl devices"

on my main machine (with a gpu obvs.) i can open the selection dialogue, but i only see my gpu listed.

what am i missing?

[super gnu]

can i use this instead ?


OpenCL™ Runtime 14.2 for Intel® CPU and Intel® Xeon Phi™ coprocessors for Windows* (64-bit)

https://software.intel.com/en-us/art...drivers#phiwin

[savage309]

Both of them should work.
If you don't see the CPU device after installing the runtime (AMD or Intel), most likely you have to restart the machine.

[super gnu]

i restarted both after installing the amd one, no result.

i uninstalled it and tried the intel one.. works fine.

thanks! Robin.

[super gnu]

ok so my first test, using both gpu and cpu in first machine, and only cpu in second machine (using dr)...

i have an error on the second machine (first one is compiling trace program for gpu, will be a while!)

second machine (one with no gpu) says:

error:clcreatecontext() failed for device 0 (-6)!
warning: initdevices()failed


is this just because its stuck waiting for the master machine to start rendering, or an issue that needs fixing?

[super gnu]

ok so the building of the trace program failed on the main machine now:

ptxas fatal :memory allocation failure
error invalid binary (-42) at line 1514 , in file ./src/opencl_main.cpp !!!
warning:failed to compile opencl kernels,falling back to cpu mode.
error:buildprogram() failed for device 0 (-42)!
warning: initdevices() failed.



so thats both machines..

all fun eh?


im using vray 3.25 (build from vlado that fixed another RT issue i was having)

[super gnu]

i now tried disabling the gpu and using only the cpu using the "select opencl devices" tool


i disabled the gpu, but when i render RT states "found 2 opencl devices" and "using geforce 670 gtx for rendering " or something similar.. then starts building the trace program again.

in fact i noticed that wahtever i set in the "select opencl devices" dialogue, it only ever says "found 2 devices, using 1 device (gpu) "

[savage309]

This error is reported from the nVidia OpenCL, I guess ?
If so, try with driver 353.06 or 347.88.
We've tried to explain what is expected to work here http://docs.chaosgroup.com/display/VRAY3/GPU+Rendering.

[super gnu]

well it appears i have errors from both the cpus and the gpu, since as mentioned in the fifth post, the machine without a gpu (hence no nvidia driver) still errors out too.

plus the device selector isnt working.

i did install vray 3.25 to fix a crashing display driver when using rt cuda. it seemed to fix all my driver related problems. obviously not for opencl. ill try to roll back to an earlier geforce driver.

[savage309]

There might be a chance that the custom build you have has working CUDA, but not-working OpenCL.

[super gnu]

ok i shall also download the official build.. cheers!

[super gnu]

ok so quick update, after restarting my main machine again, opencl on cpu is working fine (using intel runtime)

so seems my build of vray has working opencl to some extent.

However, the second machine (via DR) which is identical spec to the first, with the same intel runtime installed, but without a gpu, still gives the following error:

error:clcreatecontext() failed for device 0 (-6)!
warning: initdevices()failed


the device (cpu) is detected properly in the "select devices" dialogue, and also shows up in the rt render server before the error.


only difference is lack of a gpu in that machine.. could that affect it? ive noticed there are actually some old nvidia drivers on there (331.something) since i sometimes plug in an old gpu to do bios stuff. normally i just leave it headless and remote into it.

[vlado]

Are you running the DR spawner as a service? What if you run it manually?

Best regards,
Vlado

[super gnu]

im running it manually...

[vlado]

Then try updating the nVidia drivers to a newer version - they should support OpenCL and CUDA through remote desktop too; maybe the two OpenCL implementations are connected somehow. Or else, plug a monitor and run the spawner from an actual service.

Best regards,
Vlado