Hi list, I'm testing the last version of Gem from git and I'm really interested to context sharing. I tried to run 03.texture.sharing from examples but it crash pd. I'm using ubuntu linux 14.04, kernel 3.18, pd 0.46 from git, and glxwindow as render. my console give me this error after crash: intel_do_flush_locked failed: No such file or directory
Is this a bug or I'm missing something? What is the right procedure to share context inside same gem process? I would like to get a copy of the entire data in the context without using pix_snap (without pass for the cpu). Am I on the right path?
thanks
husk
On 2015-06-22 22:53, Husk 00 wrote:
intel_do_flush_locked failed: No such file or directory
afaik the problem only occurs on intel drivers, though with intel drivers it crashes consistently.
i'm tempted to believe that this is a bug in the drivers. and (worse), iirc the priority to fix that problem is rather low (who needs multiple openGL contexts in a single application?).
fgmasdr IOhannes
On Tue, Jun 23, 2015 at 3:12 PM, IOhannes m zmoelnig zmoelnig@iem.at wrote:
On 2015-06-22 22:53, Husk 00 wrote:
intel_do_flush_locked failed: No such file or directory
afaik the problem only occurs on intel drivers, though with intel drivers it crashes consistently.
i'm tempted to believe that this is a bug in the drivers.
Seems correct. I'll try with different kernel versions and report here if I find something
and (worse), iirc the priority to fix that problem is rather low (who needs multiple openGL contexts in a single application?).
I'm understand that priority is low. Still I believe that share context within same istance of Gem could be very handy for some situation like have a preview or PiP (in this case ContextInContext). Don't you think?
cheers husk
fgmasdr IOhannes
GEM-dev mailing list GEM-dev@lists.iem.at http://lists.puredata.info/listinfo/gem-dev
On 2015-06-24 08:53, Husk 00 wrote:
On Tue, Jun 23, 2015 at 3:12 PM, IOhannes m zmoelnig zmoelnig@iem.at wrote:
On 2015-06-22 22:53, Husk 00 wrote:
intel_do_flush_locked failed: No such file or directory
afaik the problem only occurs on intel drivers, though with intel drivers it crashes consistently.
i'm tempted to believe that this is a bug in the drivers.
Seems correct. I'll try with different kernel versions and report here if I find something
and (worse), iirc the priority to fix that problem is rather low (who needs multiple openGL contexts in a single application?).
I'm understand that priority is low. Still I believe that share context within same istance of Gem could be very handy for some situation like have a preview or PiP (in this case ContextInContext). Don't you think?
i think that's a misunderstanding: i think that the priority is low for the intel device driver developers. so the drivers are unlikely to be fixed. from the Gem side i would *love* to have multi-context support (that's one of the main changes in the last few years!), and consider the priority rather high.
so if it's fixable on the Gem side, it shall be done.
anyhow: there is a bug in current Gem that prevents multi-window support in non-context shared environments. if this bug was fixed, [gemglxwindow] could be compiled without context sharing support and still allow previewing in a separate window (though less performant).
finally: afair glfw3 has support for context-sharing, and it (the glfw3 demo showing this feature) doesn't seem to crash on intel cards. which hints that the bug is really with Gem (not the drivers). i can't remember the state of the [gemglfw3window] regarding context sharing.
gmasdr IOhannes
On Wed, Jun 24, 2015 at 9:27 AM, IOhannes m zmoelnig zmoelnig@iem.at wrote:
i think that's a misunderstanding: i think that the priority is low for the intel device driver developers. so the drivers are unlikely to be fixed. from the Gem side i would *love* to have multi-context support (that's one of the main changes in the last few years!), and consider the priority rather high.
so if it's fixable on the Gem side, it shall be done.
great, it would be amazing!
anyhow: there is a bug in current Gem that prevents multi-window support in non-context shared environments. if this bug was fixed, [gemglxwindow] could be compiled without context sharing support and still allow previewing in a separate window (though less performant).
Is this bug traced on the wiki? There is a way to know if/when it will be solved?
finally:
afair glfw3 has support for context-sharing, and it (the glfw3 demo showing this feature) doesn't seem to crash on intel cards. which hints that the bug is really with Gem (not the drivers). i can't remember the state of the [gemglfw3window] regarding context sharing.
I want to test gemglfw3window with my drivers but I had this error: /usr/lib/pd/extra/Gem/gemglfw3window.pd_linux: undefined symbol: glfwSetCursorPosCallback
GLFW3 was compiled from source (git repo). Gem was configured with "--with-glfw3-CFLAGS=/usr/include/GLFW/ --with-glfw3-LIBS=/usr/lib/" flags.
What's wrong?
thanks husk
On 2015-06-24 10:22, Husk 00 wrote:
Is this bug traced on the wiki? There is a way to know if/when it will be solved?
not in the wiki. in the issue-tracker: https://github.com/umlaeute/Gem/issues/87
finally:
afair glfw3 has support for context-sharing, and it (the glfw3 demo showing this feature) doesn't seem to crash on intel cards. which hints that the bug is really with Gem (not the drivers). i can't remember the state of the [gemglfw3window] regarding context sharing.
I want to test gemglfw3window with my drivers but I had this error: /usr/lib/pd/extra/Gem/gemglfw3window.pd_linux: undefined symbol: glfwSetCursorPosCallback
according to [1], the function is in glfw3. so either it's a transient bug in glfw3 or...
GLFW3 was compiled from source (git repo). Gem was configured with "--with-glfw3-CFLAGS=/usr/include/GLFW/ --with-glfw3-LIBS=/usr/lib/" flags.
...your build-flags told Gem not to link against libglfw.
if you must pass --with-glfw3-*FLAGS, then you should use: "--with-glfw3-CFLAGS=$(pkg-config --cflags glfw3) --with-glfw3-LIBS=$(pkg-config --libs glfw3)"
but really, Gem should detect glfw3 automatically (did you try that?)
fgmasdr IOhannes
-
Husk 00 -
IOhannes m zmoelnig