I understand how (or why) dependencies exist, and I'm just trying to layout-transition an Image.
Why are there so many accessMasks and stageMasks and all that?
also does the synchronisation2 make anything better?
#How does the pipelineBarrier work
upper latch
jaunty sage
sync2 adds some missing things
and no, it doesn't make the mental model fundamentally easier
sync in Vulkan is delibaretly explicit and something you need to get used to
upper latch
sure, but what do the thingfs even mean
whats an access mask and stage mask?
why are there multiple of them, and why are they specified in many places?
jaunty sage
it's all described in the spec
maybe get comfortable with how GPUs work first
that's why there are so many stage and access masks
they mimic how current GPUs work
if you want to fully leverage the advantages you get from Vulkan, knowing how these make your GPU work is crucial knowledge
trail escarp
whats an access mask and stage mask?
My mental model is that a stage mask defines a bit of hardware in the GPU, and access masks define various caches in the GPU that need invalidating or flushing to vram.
But that's only a mental model for ease of programming. Ultimately those masks are abstract concepts defined in the memory model in the specification.
upper latch
My mental model is that a stage mask defines a bit of hardware in the GPU, and a...
ok that helps a lot
cinder swan
if you want to fully leverage the advantages you get from Vulkan, knowing how th...
Are there any good resources for learning how GPUs work in general?
Like do any vendors talk about how their devices are implemented? Because that seems like something they'd want to keep secret.
trail escarp
AMD is the most open here. You can get their ISA docs for free.
soft herald
