[Bug 101572] glMemoryBarrier is backwards

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Tue Jun 27 16:46:18 UTC 2017


https://bugs.freedesktop.org/show_bug.cgi?id=101572

--- Comment #2 from Matias N. Goldberg <dark_sylinc at yahoo.com.ar> ---
That can't be right.

You're suggesting that in order to synchronize writes from FBO with a Compute
Shader I am going to dispatch (which btw the Compute Shader is accessing this
fbo as a regular sample fetch, not via imageLoad/imageStore) then I either need
to:

1. Switch to a dummy FBO, something that is mentioned nowhere: neither on
manuals or online documentation, wikis, or tutorials; also it's not mentioned
to be a guarantee in the spec either.

2. Use a function that was added in OpenGL 4.5; when Compute Shaders were added
in 4.3.

I may be misinterpreting the spec; but these solution don't make any sense.
Best case Mesa should detect that I am trying to read from an FBO in a compute
shader being dispatched and issue a barrier for me; worst case one of the
functionality already present in 4.3 (like glMemoryBarrier) that doesn't look
esoteric (like switching FBOs) should be enough to synchronize.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/dri-devel/attachments/20170627/7573f9b0/attachment.html>


More information about the dri-devel mailing list