Comment # 4
on bug 104492
from florian.will@googlemail.com
Maybe the bug is in compiler/glsl/lower_buffer_access.cpp? https://cgit.freedesktop.org/mesa/mesa/tree/src/compiler/glsl/lower_buffer_access.cpp?id=396c006d907b023f9b187db618ee2a6e4e1b8a85#n51 I'm not sure about the control flow for radeonsi, but if that pass is used in my setup, and this emit_access() code is used to break down struct derefs into multiple scalar/array/vec derefs, then it seems like the check for packing == GLSL_INTERFACE_STD430 is missing in lines 77/78 and 85, and the std140 layout is assumed instead. I think that might explain the observed incorrect behaviour. In that same file, the check for std430 was added in a few places, like in line 338. I'd love to give this a try and add checks to use std430_base_alignment() and std430_size() if appropriate, but I'm not really prepared to compile Mesa myself right now. So, if someone who knows this code feels like my suggested change is correct and required, I'd be more than happy if they could take care of this. Otherwise I might give it a try myself in some time.
You are receiving this mail because:
- You are the assignee for the bug.
_______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel
