ATI released OpenCL SDK with hardware support

Discussions related to GPU Acceleration in LuxRender

Moderators: Dade, jromang, tomb, coordinators

Re: ATI released OpenCL SDK with hardware support

Postby Sheltem » Sat Jan 09, 2010 5:46 am

@ Dade thanks for cornell test scene. The mini is running at 63K samples per sec. in luxrender...Oh man that is quite a difference! Will be watching this tread very closely! :shock:

Keep up the great work guys!!!!

Sheltem
Sheltem
 
Posts: 122
Joined: Sun Jun 15, 2008 11:06 am

Re: ATI released OpenCL SDK with hardware support

Postby Eros » Sat Jan 09, 2010 9:06 am

Unfortunately i cant get the OSX version to work still - gives me a black screen still :( - No matter though i guess things will be more stable with the arrival of 10.6.3

In the meantime ill be getting my Linux box fixed up with Fedora rather than Ubuntu (can be anything really i guess but ill give that a go first) so i can give the GTX295 a spin.
User avatar
Eros
 
Posts: 418
Joined: Wed Jul 22, 2009 8:37 am

Re: ATI released OpenCL SDK with hardware support

Postby mitchde » Sat Jan 09, 2010 10:05 am

Hi, the SmallptGPU (1.6) works well on my 8800GTX (hackintosh, real macs didnt have GTX )
Question: Can on user run my compiled commandline app (plus that preprocessed kernel) also on an ATI GPU ?
Or does perporcessed kernel means no, only for that card found on compie time ?
Thanks

PS: Around 1000K/Sec, cornel szene
Attachments
Bildschirmfoto 2010-01-09 um 16.03.57.jpg
8800GTX Mac OS X
mitchde
 
Posts: 256
Joined: Fri Dec 25, 2009 2:13 am

Re: ATI released OpenCL SDK with hardware support

Postby jensverwiebe » Sat Jan 09, 2010 10:32 am

@ mitch, it only means the includes inside kernel are preprocessed cause the "-I." statement does not work in all compilers, so we "bake" that together first.
Has nothing todo with vendor-specific .

@eros, get the newest smallpt 1.6 version from Dade´s site, has all needed changes now in, just comment in/out the platform-flags. Runs for all OSX.
If you only have a 32bit mashine, add -m32 -arch i386 to your flags.

@ Dade, got smallLuxGPU to run with these __global changes, but runs painfully slow, must do some other changes as you did to smallpt also there ( cl_mem )
CPU works still fine.
Can i remover the CL_MEM_USE_HOST_PTR and point to NULL instead as done in smallpt ? Ah now c++, looks all different :)

BTW: perhaps found solution for the not working -I.
less error-prone method of accomplishing this is to open
the current directory (`.') and use the fchdir(2) function to return. ---> #include <unistd.h> needed it seems


Jens
Last edited by jensverwiebe on Sat Jan 09, 2010 2:31 pm, edited 5 times in total.
User avatar
jensverwiebe
Developer
 
Posts: 3378
Joined: Wed Apr 02, 2008 4:34 pm

Re: ATI released OpenCL SDK with hardware support

Postby Dade » Sat Jan 09, 2010 10:33 am

mitchde wrote:Hi, the SmallptGPU (1.6) works well on my 8800GTX (hackintosh, real macs didnt have GTX )
Question: Can on user run my compiled commandline app (plus that preprocessed kernel) also on an ATI GPU ?
Or does perporcessed kernel means no, only for that card found on compie time ?


Preprocessed kernel is just a workaround to another Apple bug (i.e. it looks like the compiler "-I." directive doesn't work). It is just the kernel and the headers files packed in a single text file. OpenCL works like OpenGL, one executable works with any driver so, yes, ATI users can run your executable compiled on a Mac with an NVIDIA card. No problem.

P.S. Jens, was faster than me to answer :D
User avatar
Dade
Developer
 
Posts: 8291
Joined: Sat Apr 19, 2008 6:04 pm
Location: Italy

Re: ATI released OpenCL SDK with hardware support

Postby Eros » Sat Jan 09, 2010 6:42 pm

Ok massive nub question - I finally have ubuntu up (had to abandon the Fedora...) with the 190 release drivers which *should* have the OpenCL support - it is the latest driver available, I installed that, i also tried to install the Cuda drivers, tool kit SDK along with the OpenCL toolkit and SDK. (i think its installed properly)

I am trying to compile the smallptGPU code now and get the following error

make
gcc -O3 -msse2 -mfpmath=sse -ftree-vectorize -funroll-loops -Wall -I/home/ward/NVIDIA_GPU_Computing_SDK/include -L/home/ward/NVIDIA_GPU_Computing_SDK/lib/ -lglut -lOpenCL -DSMALLPT_CPU -o smallptCPU smallptCPU.c displayfunc.c
geomfunc.h:348: warning: ‘RadianceDirectLighting’ defined but not used
/usr/bin/ld: cannot find -lOpenCL
collect2: ld returned 1 exit status
make: *** [smallptCPU] Error 1


Im missing something or i got my Include wrong? any ideas?
Ill keep firing away, with this kind of stuff i tend to just guess alot more than knowing what im missing
User avatar
Eros
 
Posts: 418
Joined: Wed Jul 22, 2009 8:37 am

Re: ATI released OpenCL SDK with hardware support

Postby jensverwiebe » Sun Jan 10, 2010 3:05 am

@ Dade

Hey, was nearly out-of-the-box-compile ;) Runs really nice here. Pls take a look at the results, if all is plausible and as intended.
Code: Select all
jens-macpro:SmallptGPU-v2.0alpha1_OSX jensverwiebe$ ./smallptgpuUsage: ./smallptgpu
Usage: ./smallptgpu <use CPU devices (0 or 1)> <use GPU devices (0 or 1)> <GPU workgroup size (0=default value or anything > 0 and power of 2)> <window width> <window height> <scene file>
Reading scene: scenes/cornell.scn
Scene size: 9
OpenCL Platform 0: Apple
OpenCL Device name 0: GeForce 8800 GT
OpenCL Device type 0: TYPE_GPU
OpenCL Device units 0: 14
OpenCL Device name 1: Intel(R) Xeon(R) CPU            5150  @ 2.66GHz
OpenCL Device type 1: TYPE_CPU
OpenCL Device units 1: 4
[Device::GeForce 8800 GT] Kernel file size 13351bytes
[Device::GeForce 8800 GT] Compilation result:
[Device::GeForce 8800 GT] Suggested work group size: 192
[Device::GeForce 8800 GT] Camera buffer size: 0Kb
[Device::GeForce 8800 GT] Scene buffer size: 0Kb
[Device::Intel(R) Xeon(R) CPU            5150  @ 2.66GHz] Kernel file size 13351bytes
[Device::Intel(R) Xeon(R) CPU            5150  @ 2.66GHz] Compilation result:
[Device::Intel(R) Xeon(R) CPU            5150  @ 2.66GHz] Suggested work group size: 1
[Device::Intel(R) Xeon(R) CPU            5150  @ 2.66GHz] Camera buffer size: 0Kb
[Device::Intel(R) Xeon(R) CPU            5150  @ 2.66GHz] Scene buffer size: 0Kb
OpenCL Device used: [GeForce 8800 GT][Intel(R) Xeon(R) CPU            5150  @ 2.66GHz]
Updating OpenCL Device workloads
Updating OpenCL Device workloads
Updating OpenCL Device workloads
Releasing resources
[Device::GeForce 8800 GT] Rendering thread halted
[Device::Intel(R) Xeon(R) CPU            5150  @ 2.66GHz] Rendering thread halted
Done.


smallptGPU2.png
visual output


Again, adapted source with OSX_diff and compiled 64bit binary: http://www.jensverwiebe.de/LuxRender/SmallptGPU-v2.0alpha1_OSX.zip
Note: for non LuxRender builders: you need libboost_thread-xgcc40-mt-1_39.a now too. ( naming can differ )
You can find my instructions for building boostlibs for OSX in "complation and portability" threads (viewtopic.php?f=22&t=472#p21831)

Some command-examples:

./smallptgpu 1 1 0 640 480 scenes/cornell.scn ( run on cpu/gpu with max available workgroupsize )
./smallptgpu 0 1 32 640 480 scenes/cornell.scn ( run on GPU only with workgroupsize = 32
./smallptgpu 1 1 0 800 600 scenes/simple.scn ( run on cpu/gpu with max available workgroupsize , widowsize 800x600 and simple scene )

..and..Dade, must repeat here: you are doing great work :) I love being on bleeding edge :D

Jens
User avatar
jensverwiebe
Developer
 
Posts: 3378
Joined: Wed Apr 02, 2008 4:34 pm

Re: ATI released OpenCL SDK with hardware support

Postby Dade » Sun Jan 10, 2010 4:18 am

jensverwiebe wrote:Hey, was nearly out-of-the-box-compile ;) Runs really nice here. Pls take a look at the results, if all is plausible and as intended.


Jens, to check if the results are plausible, just run "./smallptgpu 0 1 ..." than "./smallptgpu 1 0 ..." and finally "./smallptgpu 1 1 ...". The number of samples/sec should be, in the case of CPU+GPU, only slightly lower than the sum of results with CPU alone and GPU alone. It is useful to do a comparison with SmallptGPU 1.6 too.

Your numbers look a bit low to me but I guess we just need to tune the code a bit.

jensverwiebe wrote:..and..Dade, must repeat here: you are doing great work :) I love being on bleeding edge :D


Don't tell me, I'm not sleeping anymore :D

@Eros: can you post the result of a "ls -l /home/ward/NVIDIA_GPU_Computing_SDK/lib/" ? I have never used the NVIDIA SDK so it is hard for me to offer any help but just to check if the libraries are really there :idea:
User avatar
Dade
Developer
 
Posts: 8291
Joined: Sat Apr 19, 2008 6:04 pm
Location: Italy

Re: ATI released OpenCL SDK with hardware support

Postby jensverwiebe » Sun Jan 10, 2010 4:30 am

The number of samples/sec should be, in the case of CPU+GPU, only slightly lower than the sum of results with CPU alone and GPU alone.


Yeah, thats definitely not the case :
cpu.png
CPU alone, cpu @ 380% ( max, 400 )

gpu.png
GPU alone, cpu @ 99% ( max, 400 )

cpu_gpu.png
CPU + GPU, cpu @ 270% ( max, 400 )


Looks as it is just balancing the profiled max to both gpu and cpu.
In dualmode the cpu is not maxed out any longer , so is the gpu.( spending lot of time in sys, again a bloody threading thing? i guess we should use CL-atomics.
Edit: added a timeprofile
Code: Select all
# Report 1 - Session 1 - Time Profile (All Thread States) of smallptgpu
SharkProfileViewer
# Generated from the visible portion of the outline view
+ 31.5%, __workq_kernreturn, libSystem.B.dylib
| + 31.5%, _pthread_wqthread, libSystem.B.dylib
| |   31.5%, start_wqthread, libSystem.B.dylib
+ 18.5%, mach_msg_trap, libSystem.B.dylib
| + 18.5%, mach_msg, libSystem.B.dylib
| | + 11.1%, glcDebugListener, OpenGL
| | | + 11.1%, _pthread_start, libSystem.B.dylib
| | | |   11.1%, thread_start, libSystem.B.dylib
| | + 7.3%, io_connect_method, IOKit
| | | + 7.3%, IOConnectCallMethod, IOKit
| | | | + 7.3%, 0x113f01070 [216B], libclh.dylib
| | | | | + 7.3%, 0x113ed9c17 [unknown], libclh.dylib
| | | | | | + 7.3%, 0x113ed8e30 [1.0KB], libclh.dylib
| | | | | | | + 7.3%, glrCompWaitForEvent, GeForce8xxxGLDriver
| | | | | | | | + 7.3%, cldWaitForEvent, OpenCL
| | | | | | | | | + 7.3%, clWaitForEvent, OpenCL
| | | | | | | | | | + 7.3%, clWaitForEvents, OpenCL
| | | | | | | | | | | + 7.3%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
| | | | | | | | | | | | + 7.3%, thread_proxy, smallptgpu
| | | | | | | | | | | | | + 7.3%, _pthread_start, libSystem.B.dylib
| | | | | | | | | | | | | |   7.3%, thread_start, libSystem.B.dylib
| | | | + 0.0%, gldDestroyMemoryPlugin, GeForce8xxxGLDriver
| | | | | + 0.0%, gldReclaimBuffer, GeForce8xxxGLDriver
| | | | | | + 0.0%, gldDestroyBuffer, GeForce8xxxGLDriver
| | | | | | | + 0.0%, cleDestroyTransferBuffer, OpenCL
| | | | | | | | + 0.0%, clReleaseCommand, OpenCL
| | | | | | | | | + 0.0%, clReleaseCommandsUpTo, OpenCL
| | | | | | | | | | + 0.0%, clFinish, OpenCL
| | | | | | | | | | | + 0.0%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
| | | | | | | | | | | | + 0.0%, thread_proxy, smallptgpu
| | | | | | | | | | | | | + 0.0%, _pthread_start, libSystem.B.dylib
| | | | | | | | | | | | | |   0.0%, thread_start, libSystem.B.dylib
| | | | | | | + 0.0%, _CLMemDeallocate, OpenCL
| | | | | | | | + 0.0%, _CFRelease, CoreFoundation
| | | | | | | | | + 0.0%, clReleaseMemObject, OpenCL
| | | | | | | | | | + 0.0%, RenderDevice::SetWorkLoad(unsigned int, unsigned int, unsigned int, unsigned int, unsigned int*), smallptgpu
| | | | | | | | | | | + 0.0%, idleFunc(), smallptgpu
| | | | | | | | | | | | + 0.0%, -[GLUTApplication run], GLUT
| | | | | | | | | | | | | + 0.0%, glutMainLoop, GLUT
| | | | | | | | | | | | | | + 0.0%, main, smallptgpu
| | | | | | | | | | | | | | |   0.0%, start, smallptgpu
| | | | + 0.0%, gldCreateTexture, GeForce8xxxGLDriver
| | | | | + 0.0%, gldGetQueryInfo, GeForce8xxxGLDriver
| | | | | | + 0.0%, gldCreateTexture, GeForce8xxxGLDriver
| | | | | | | + 0.0%, gldGetQueryInfo, GeForce8xxxGLDriver
| | | | | | | | + 0.0%, gldUpdateDispatch, GeForce8xxxGLDriver
| | | | | | | | | + 0.0%, glBegin_Exec, GLEngine
| | | | | | | | | | + 0.0%, gleDrawPixelsFast, GLEngine
| | | | | | | | | | | + 0.0%, glDrawPixels_Exec, GLEngine
| | | | | | | | | | | | + 0.0%, displayFunc(), smallptgpu
| | | | | | | | | | | | | + 0.0%, -[GLUTView drawRect:], GLUT
| | | | | | | | | | | | | | + 0.0%, -[GLUTView handleWorkEvent:], GLUT
| | | | | | | | | | | | | | | + 0.0%, processWindowWorkList, GLUT
| | | | | | | | | | | | | | | | + 0.0%, __glutProcessWorkEvents, GLUT
| | | | | | | | | | | | | | | | | + 0.0%, -[GLUTApplication run], GLUT
| | | | | | | | | | | | | | | | | | + 0.0%, glutMainLoop, GLUT
| | | | | | | | | | | | | | | | | | | + 0.0%, main, smallptgpu
| | | | | | | | | | | | | | | | | | | |   0.0%, start, smallptgpu
| | | | + 0.0%, gldInitDispatch, GeForce8xxxGLDriver
| | | | | + 0.0%, gldPageoffBuffer, GeForce8xxxGLDriver
| | | | | | + 0.0%, gldUpdateDispatch, GeForce8xxxGLDriver
| | | | | | | + 0.0%, glBitmap_Exec, GLEngine
| | | | | | | | + 0.0%, glutBitmapCharacter, GLUT
| | | | | | | | | + 0.0%, displayFunc(), smallptgpu
| | | | | | | | | | + 0.0%, -[GLUTView drawRect:], GLUT
| | | | | | | | | | | + 0.0%, -[GLUTView handleWorkEvent:], GLUT
| | | | | | | | | | | | + 0.0%, processWindowWorkList, GLUT
| | | | | | | | | | | | | + 0.0%, __glutProcessWorkEvents, GLUT
| | | | | | | | | | | | | | + 0.0%, -[GLUTApplication run], GLUT
| | | | | | | | | | | | | | | + 0.0%, glutMainLoop, GLUT
| | | | | | | | | | | | | | | | + 0.0%, main, smallptgpu
| | | | | | | | | | | | | | | | |   0.0%, start, smallptgpu
| | - 0.0%, 0x113f00b40 [98B], libclh.dylib
| | - 0.0%, io_connect_map_memory_into_task, IOKit
+ 11.6%, __semwait_signal, libSystem.B.dylib
| + 11.6%, _pthread_cond_wait, libSystem.B.dylib
| | + 11.6%, boost::condition_variable::wait(boost::unique_lock<boost::mutex>&), smallptgpu
| | | + 10.7%, idleFunc(), smallptgpu
| | | | + 10.7%, -[GLUTApplication run], GLUT
| | | | | + 10.7%, glutMainLoop, GLUT
| | | | | | + 10.7%, main, smallptgpu
| | | | | | |   10.7%, start, smallptgpu
| | | + 0.9%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
| | | | + 0.9%, thread_proxy, smallptgpu
| | | | | + 0.9%, _pthread_start, libSystem.B.dylib
| | | | | |   0.9%, thread_start, libSystem.B.dylib
+ 11.1%, kevent, libSystem.B.dylib
| + 11.1%, _dispatch_mgr_invoke, libSystem.B.dylib
| | + 11.1%, _dispatch_queue_invoke, libSystem.B.dylib
| | | + 11.1%, _dispatch_worker_thread2, libSystem.B.dylib
| | | | + 11.1%, _pthread_wqthread, libSystem.B.dylib
| | | | |   11.1%, start_wqthread, libSystem.B.dylib
+ 11.0%, 0x1007fd5c0 [5.7KB], Unknown Library
| + 11.0%, cld_exec_workitem, libcldcpuengine.dylib
| | + 11.0%, _dispatch_apply2, libSystem.B.dylib
| | | + 8.2%, _dispatch_worker_thread2, libSystem.B.dylib
| | | | + 8.2%, _pthread_wqthread, libSystem.B.dylib
| | | | |   8.2%, start_wqthread, libSystem.B.dylib
| | | + 2.7%, dispatch_apply_f, libSystem.B.dylib
| | | | + 2.7%, cld_exec_kernel, libcldcpuengine.dylib
| | | | | + 2.7%, _dispatch_queue_drain, libSystem.B.dylib
| | | | | | + 2.7%, _dispatch_queue_invoke, libSystem.B.dylib
| | | | | | | + 2.7%, _dispatch_worker_thread2, libSystem.B.dylib
| | | | | | | | + 2.7%, _pthread_wqthread, libSystem.B.dylib
| | | | | | | | |   2.7%, start_wqthread, libSystem.B.dylib
+ 10.7%, semaphore_wait_signal_trap, libSystem.B.dylib
| + 10.7%, pthread_mutex_lock, libSystem.B.dylib
| | + 6.3%, _CLEventDeallocate, OpenCL
| | | + 6.3%, _CFRelease, CoreFoundation
| | | | + 6.3%, clReleaseCommand, OpenCL
| | | | | + 6.3%, clReleaseCommandsUpTo, OpenCL
| | | | | | + 6.3%, clFinish, OpenCL
| | | | | | | + 6.3%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
| | | | | | | | + 6.3%, thread_proxy, smallptgpu
| | | | | | | | | + 6.3%, _pthread_start, libSystem.B.dylib
| | | | | | | | | |   6.3%, thread_start, libSystem.B.dylib
| | + 2.6%, processWork, OpenCL
| | | + 2.6%, clCommandDispatch, OpenCL
| | | | + 2.6%, clEnqueueReadBuffer, OpenCL
| | | | | + 2.6%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
| | | | | | + 2.6%, thread_proxy, smallptgpu
| | | | | | | + 2.6%, _pthread_start, libSystem.B.dylib
| | | | | | | |   2.6%, thread_start, libSystem.B.dylib
| | + 1.1%, clFinish, OpenCL
| | | + 1.1%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
| | | | + 1.1%, thread_proxy, smallptgpu
| | | | | + 1.1%, _pthread_start, libSystem.B.dylib
| | | | | |   1.1%, thread_start, libSystem.B.dylib
| | + 0.6%, cldWaitForEvent, OpenCL
| | | + 0.6%, clWaitForEvent, OpenCL
| | | | + 0.6%, clWaitForEvents, OpenCL
| | | | | + 0.6%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
| | | | | | + 0.6%, thread_proxy, smallptgpu
| | | | | | | + 0.6%, _pthread_start, libSystem.B.dylib
| | | | | | | |   0.6%, thread_start, libSystem.B.dylib
| | + 0.0%, clEnqueueNDRangeKernel, OpenCL
| | | + 0.0%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
| | | | + 0.0%, thread_proxy, smallptgpu
| | | | | + 0.0%, _pthread_start, libSystem.B.dylib
| | | | | |   0.0%, thread_start, libSystem.B.dylib
- 3.3%, semaphore_wait_trap, libSystem.B.dylib
- 0.9%, 0x7fff86ea5b53 [725B], libSystem.B.dylib
- 0.5%, cosf, libSystem.B.dylib
- 0.3%, powf$fenv_access_off, libSystem.B.dylib
- 0.3%, gldGetTextureLevelImage, GeForce8xxxGLDriver
- 0.1%, io_connect_method, IOKit
- 0.1%, __memcpy, libSystem.B.dylib
- 0.1%, 0x1007efde0 [53.0KB], Unknown Library
- 0.1%, IOConnectCallMethod, IOKit
- 0.0%, void glgConvertTo_32<GLGConverter_ABGR8_ARGB8, (GLGMemory)2>(GLGOperation const*, GLDPixelMode const*), libGLImage.dylib
- 0.0%, fmaxf$fenv_access_off, libSystem.B.dylib
- 0.0%, mach_msg, libSystem.B.dylib
  0.0%, start_wqthread, libSystem.B.dylib
- 0.0%, memcpy, libSystem.B.dylib
- 0.0%, gldUpdateDispatch, GeForce8xxxGLDriver
- 0.0%, cld_exec_workitem, libcldcpuengine.dylib
- 0.0%, __memcpy_chk, libSystem.B.dylib
- 0.0%, semaphore_signal_trap, libSystem.B.dylib
- 0.0%, dyld_stub___memcpy_chk, IOKit
- 0.0%, 0x113f01070 [216B], libclh.dylib
- 0.0%, szone_size, libSystem.B.dylib
- 0.0%, semaphore_signal_all_trap, libSystem.B.dylib
- 0.0%, RenderDevice::RenderThread(RenderDevice*), smallptgpu
- 0.0%, pick_nextworkqueue_droplock, libSystem.B.dylib
- 0.0%, OSAtomicAdd32Barrier, libSystem.B.dylib
- 0.0%, glPixelStorei_Exec, GLEngine
- 0.0%, glGetIntegerv_Exec, GLEngine
- 0.0%, gleUpdateFragmentStateProgram, GLEngine
- 0.0%, gleGetState, GLEngine
- 0.0%, glBegin_Exec, GLEngine
- 0.0%, fminf$fenv_access_off, libSystem.B.dylib
- 0.0%, dyld_stub_memcpy, IOKit
- 0.0%, dyld_stub_mach_msg_trap, libSystem.B.dylib
- 0.0%, dyld_stub_glGetIntegerv, GLUT
- 0.0%, do_rand, libSystem.B.dylib
- 0.0%, __spin_lock, libSystem.B.dylib
- 0.0%, __longcopy, libSystem.B.dylib
- 0.0%, __cpu_number, libSystem.B.dylib
- 0.0%, 0x200367950 [214B], GeForce8xxxGLDriver
- 0.0%, 0x200350da0 [450B], GeForce8xxxGLDriver
- 0.0%, 0x113ed9c0d [unknown], libclh.dylib
- 0.0%, 0x113ed9c0b [unknown], libclh.dylib
- 0.0%, 0x113ed5617 [unknown], libclh.dylib
- 0.0%, 0x113ecfde8 [unknown], libclh.dylib


In solo-mode it is.
Note: smallptgpu-1.6 shows same results for both modes around 1100k on my sys, so cpu and gpu are around of the same power in my mashine.( for cornell )
BTW: in "Beyoind"-forums, results let mesuggest the same. Talonman´s results show a good spreading across multi-gpu indeed.

So could be interesting just to drop in a second card. Unfortunately my left empty pcie-slots are only v.1 and 8-lane.
Have another old NV 7300GT laying around, but that would not support openCL, mixing NV and ATI does too not sound good to me too ( in case of getting an 4870 )
( cries for a new mashine :P )

Don't tell me, I'm not sleeping anymore

..right, thats what we call enthusiasm :P You can count on me too, who needs sleep ? :o

Jens
Last edited by jensverwiebe on Sun Jan 10, 2010 5:39 am, edited 8 times in total.
User avatar
jensverwiebe
Developer
 
Posts: 3378
Joined: Wed Apr 02, 2008 4:34 pm

Re: ATI released OpenCL SDK with hardware support

Postby Eros » Sun Jan 10, 2010 4:34 am

Im not convinced things are installed correctly on my system, so im going to have a fish around, it does appear most of the folders of the SDK are empty... or at least they have examples in them but little in the way of libs or headers, and makefiles.

I will probably try and roll back and install the actual driver linked on the OpenCL webpage rather than just the latest, though i always just assumed that the latest would be fine. Inspection of pages via google say that the OpenCL files are only included in the 190.29 drivers, and NOT the 190.42 drivers which i have.

*shakes fist*

I 'should' get this up and running fairly quickly now i know this

[Edit] This was my problem, sorry for all the spam, Im pleased to report that everything compiled fine, and this is the result.

Two shots of SmallptGPU-v1.6, getting about 4Msamples/s on cornell and 20M-ish on the simple scene.
Attachments
Screenshot-1.png
Screenshot.png
User avatar
Eros
 
Posts: 418
Joined: Wed Jul 22, 2009 8:37 am

PreviousNext

Return to GPU Acceleration

Who is online

Users browsing this forum: No registered users and 1 guest