On Tue, Sep 29, 2015 at 2:48 AM, Timothy Arceri <t_arc...@yahoo.com.au> wrote: > On Tue, 2015-09-29 at 02:33 -0400, Ilia Mirkin wrote: >> On Tue, Sep 29, 2015 at 2:26 AM, Timothy Arceri < >> t_arc...@yahoo.com.au> wrote: >> > On Tue, 2015-09-29 at 02:08 -0400, Ilia Mirkin wrote: >> > > On Tue, Sep 29, 2015 at 2:05 AM, Timothy Arceri < >> > > t_arc...@yahoo.com.au> wrote: >> > > > On Tue, 2015-09-29 at 01:05 -0400, Ilia Mirkin wrote: >> > > No. I mean the line: >> > > >> > > brw->vtbl.emit_buffer_surface_state(brw, &surf_offsets[i], >> > > bo, >> > > >> > > Shouldn't that look up the atomic buffer index in prog->Uniforms >> > > instead of using 'i'? >> > >> > No that looks fine. The problem in the patch was we didn't have the >> > atomic buffex index and we used the binding as the offset which >> > doesn't >> > have to be the same as the buffer index. >> > >> > In brw_upload_abo_surfaces its the opposite we have the atomic >> > buffer >> > index which is i and we are looking up the binding to get the >> > buffer >> > object. >> > >> > Does that make sense? >> >> i is the binding index though... > > i is the buffer index > > See find_active_atomic_counters() for the code the counts the active > buffers. > > http://cgit.freedesktop.org/mesa/mesa/tree/src/glsl/link_atomics.cpp#n9 > 9 > > This count is then used to set NumAtomicBuffers and to create the prog > ->AtomicBuffers array so the array is no bigger than it needs to be. > This is the reason the binding and atomic buffer index can differ. > > http://cgit.freedesktop.org/mesa/mesa/tree/src/glsl/link_atomics.cpp#n1 > 73
Do you agree that NumAtomicBuffers can go up to MaxCombinedAtomicBuffers? If so, that's problematic, when it's > MaxAtomicBuffers, right? -ilia _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
