This fixes a performance regression when loading captures from <=v1.6
(before the new image state code went in). The code to import the old
image state information was incorrectly discarding the frame ref
information, forcing the images to be reinitialized on every replay.
In old captures `ImageRefs` chunk is processed, creating `ImageState`
objects that only contain `refType`; then `BeginCaptureFrame` processes
the `ImageLayouts` info, to construct a new `ImageState` object,
containing the layout information. This `ImageState` was overwriting the
`ImageState` from the `ImageRefs` chunk, discarding the `refType` info.
This change merges the `ImageState`s from `ImageRefs`, where each field
of the merged image state comes from whichever `ImageState` had a
non-undefined value for that field.
For new captures, this change should be safe, since the image state
should be undefined when the `BeginCaptureFrame` chunk is processed, so
the `MergCaptureBeginState` call will just take all the state fields
from the `ImageState` stored in the `BeginCaptureFrame` chunk.
Change-Id: Id2252edb6382bd02bc0236b77c7cca667e97bf29
This is needed for Vulkan Pixel history, where we need to know render
state at current point of time for a command buffer during callback
replay.
Modifications:
- Removed m_pDriver and m_CreationInfo members from VulkanRenderState,
instead these are accepted as arguments to member functions.
- BakedCmdBufferInfo state is now full VulkanRenderState
* Enter and shift-enter in the find text go forwarwd and backwards respectively.
* This replaces the previous system of selecting a direction explicitly and only
ever going in that direction when finding.
* The builtin is more directly useful but the signature index allows us to match
up inputs and outputs that don't have a builtin meaning (pure interpolators).
Added fetching of texture data for UAVs, to allow stores to occur.
Fixed issues with sample/gather when debugging a compute shader.
Added a demo for dispatch in D3D12.
This fixes an ambiguity between old captures (where missing MemRefs
indicates a lack of information, and we need to act pessimistically), vs
new captures where the memory contents were not accessed (e.g. only
bound to images). Now the latter case will have an explicit MemRefs
entry indicating that the memory was not referenced.
Change-Id: Ib76d56ef6f2e250952412b4cdbe0b2432a4ca0ec
* When we need a SPIR-V pointer we box up a real pointer to a ShaderVariable,
and optionally column/scalar selectors if it's a pointer to a scalar within a
vector or matrix.
* For globals in opaque storage classes we'll allocate our own ShaderVariables
so that we can safely keep pointers to them. Otherwise IDs of temporary
pointers (from OpAccessChain) should always be a strict subset of the lifetime
of the ID it's pointing to, so we can keep the pointer around. The variables
don't move and are persistently allocated so they won't be relocated from
under us (we pre-allocate space for the IDs).
* Before passing back a ShaderVariable to the outside world, we evaluate its
current value. Thus pointers are presented as if they are just variables that
happen to update when their parent does.
Benson Joeris