[Bug 105496] Using a single-plane imageview from a multi-plane image is broken
bugzilla-daemon at freedesktop.org
bugzilla-daemon at freedesktop.org
Wed Mar 14 01:34:57 UTC 2018
https://bugs.freedesktop.org/show_bug.cgi?id=105496
Bug ID: 105496
Summary: Using a single-plane imageview from a multi-plane
image is broken
Product: Mesa
Version: git
Hardware: x86-64 (AMD64)
OS: Linux (All)
Status: NEW
Severity: normal
Priority: medium
Component: Drivers/Vulkan/intel
Assignee: intel-3d-bugs at lists.freedesktop.org
Reporter: atomnuker at gmail.com
QA Contact: intel-3d-bugs at lists.freedesktop.org
CC: jason at jlekstrand.net
Hi,
As the title suggests, creating a VkImageView with a format of
VK_FORMAT_R8_UNORM out of a VkImage with format
VK_FORMAT_G8_B8_R8_3PLANE_420_UNORM (and the mutable flag enabled too) and then
using the image as a storage image (e.g. layout (set = 0, binding = 0, rgba8)
uniform readonly image2D input_img; inside a GLSL shader) will result in
vec4(0.0f) actually being read instead of the pixel values inside the image.
Writing to such images however seems to be successful.
I've uploaded some code to replicate this, though I'm afraid its not a small
example. The repo is https://github.com/atomnuker/FFmpeg/tree/exp_vulkan. To
compile run ./configure with --enable-vulkan. The shader is in
libavfilter/vulkan/unsharp.comp and the code which creates the imageviews is in
libavfilter/vf_unsharp_vulkan. To run, "./ffmpeg_g -init_hw_device
"vulkan=vk:0" -i <some sane video or image> -filter_hw_device vk -vf
format=yuv420p,scale=1280x720,hwupload,unsharp_vulkan,hwdownload,format=yuv420p
-y out.mkv". I don't have glslang hooked up to the build system yet so you need
to regenerate the .h file with "glslangValidator -V unsharp.comp
--variable-name "unsharp_vulkan_comp_spv" -o unsharp.comp.h".
--
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20180314/110c20e3/attachment.html>
More information about the intel-3d-bugs
mailing list