Hi Victor, Sorry for the delay in responding. Like I told you the other night on gtalk, I am unable to get a log on Windows even using g_print as you suggested instead of g_message. On the Mac, I simply get that opencl is not being used since there is no shared object to load.
I think a lot of users will be using a single display (like I do using my laptop) and so, there would have to be some viable solution to problem of the nvidia driver killing the GPU job, don't you think? I didn't know about the GEGL_SWAP= RAM option and will add that to the environment while running Gimp. Thanks! Partha On Tue, Feb 19, 2013 at 1:53 PM, Victor Oliveira <victormath...@gmail.com>wrote: > Hi Partha, > > I couldn't really reproduce your error, so I assume it is because you're > trying to use the display GPU for processing and to avoid blocking the > screen, the nvidia driver just kills whatever is running in the GPU. > Also, I'm not sure if you're already using the GEGL_SWAP=RAM option to run > GIMP, it should improve performance and maybe you don't get this error. > > > On Sat, Feb 16, 2013 at 12:29 AM, Partha Bagchi <parth...@gmail.com>wrote: > >> Hi Victor, >> >> I have an NVIDIA GPU with 1G dedicated VRAM, i7 Core CPU, and 16G of >> RAM. In any case, 150x150 radius should be doable. In any case, I am sure >> that when Gimp goes production, it's bound to come up. >> >> Any further report on Mac? While I am finding that it's quite fast on an >> Mac with 8G RAM, I am not sure if the GPU is being used. >> >> Note that everything I have seen/ read about OpenCL says that you don't >> need to load the library at least on a Mac. >> >> Thanks, >> Partha >> >> >> >> >> >> On Fri, Feb 15, 2013 at 7:22 PM, Victor Oliveira <victormath...@gmail.com >> > wrote: >> >>> Hi Partha, >>> >>> Thanks for the bug report, I'll give a look when I can. >>> Notice that a 150x150 radius is very big and maybe yout GPU doesn't have >>> enough memory for that, but it was supposed to fallback to the CPU in this >>> case. >>> >>> Thanks >>> Victor >>> >>> >>> On Fri, Feb 15, 2013 at 8:47 AM, Tobias Ellinghaus <h...@gmx.de> wrote: >>> >>>> Am Freitag, 15. Februar 2013, 04:52:49 schrub Partha Bagchi: >>>> > Hi Victor, >>>> > >>>> > Latest git. I have an NVIDIA GeForce GT 230M card with 1G VRAM. The >>>> opengl >>>> > version is 1.1 CUDA 4.2.1 etc. Windows 7 64bit >>>> > >>>> > I think OpenCL is taking down my video every time. Here is a simple >>>> > repeatable test for me. >>>> > >>>> > 1. Open 16 bit tiff. >>>> > 2. Duplicate layer. >>>> > 3. Layer - desaturate -> invert >>>> > 4. GEGL gaussian blur x = y = 150. >>>> > >>>> > Takes the screen down (goes black and recovers) and leaves a dark >>>> tile on >>>> > the layer. >>>> > >>>> > Any ideas? >>>> >>>> That is normal when running OpenCL (or CUDA) kernels on a GPU that has a >>>> monitor connected. On Windows it will time out quite quickly, on Linux >>>> AFAIK >>>> not. However, that can be configured. Google is your friend, just >>>> search for >>>> "nvidia watchdog". >>>> >>>> [...] >>>> >>>> > Thanks, >>>> > Partha >>>> >>>> Tobias >>>> >>>> _______________________________________________ >>>> gimp-developer-list mailing list >>>> gimp-developer-list@gnome.org >>>> https://mail.gnome.org/mailman/listinfo/gimp-developer-list >>>> >>>> >>> >>> _______________________________________________ >>> gimp-developer-list mailing list >>> gimp-developer-list@gnome.org >>> https://mail.gnome.org/mailman/listinfo/gimp-developer-list >>> >>> >> >
_______________________________________________ gimp-developer-list mailing list gimp-developer-list@gnome.org https://mail.gnome.org/mailman/listinfo/gimp-developer-list
