I know how multisampled images work. I was wondering how you got the multisampled data, since there are no image formats for storing such data and there is no mechanism (in Vulkan or OpenGL) to directly read data from a multisampled image.
Multisampled images mean something more than just having multiple samples per texel. Those samples are expected to correspond to particular locations within each pixel. This is information the multisample resolve can use as part of its resolve process. So dumping data into a multisample image like you’re trying to do could cause problems.
I said “storage images”. As in the Vulkan descriptor type. In OpenGL, it would be called using image load/store.
Also, be advised that implementations don’t have to allow you to create multisampled storage images. You have to check to see if the implementation offers that as a feature (