A handful of interface changes to support OpenXR future rebasing #1846
Conversation
|
CI gfxreconstruct build queued with queue ID 290663. |
jzulauf-lunarg
requested review from
MarkY-LunarG,
bradgrantham-lunarg and
panos-lunarg
|
CI gfxreconstruct build # 5232 running. |
|
CI gfxreconstruct build # 5232 passed. |
|
CI gfxreconstruct build queued with queue ID 291019. |
|
CI gfxreconstruct build # 5237 running. |
|
CI gfxreconstruct build # 5237 passed. |
|
|
||
| GFXRECON_BEGIN_NAMESPACE(gfxrecon) | ||
| GFXRECON_BEGIN_NAMESPACE(decode) | ||
| template <typename T, typename = void> |
There was a problem hiding this comment.
Nit: Put a blank line above this.
|
CI gfxreconstruct build queued with queue ID 291462. |
|
CI gfxreconstruct build # 5244 running. |
|
CI gfxreconstruct build # 5244 passed. |
There was a problem hiding this comment.
@davidd-lunarg , @bradgrantham-lunarg -- this change, particularly, I'd like your review of
| const StructPointerDecoder<Decoded_VkInstanceCreateInfo>* pCreateInfo, | ||
| const StructPointerDecoder<Decoded_VkAllocationCallbacks>* pAllocator, | ||
| HandlePointerDecoder<VkInstance>* pInstance) | ||
| void VulkanReplayConsumerBase::ModifyCreateInstanceInfo( |
There was a problem hiding this comment.
@davidd-lunarg -- @bradgrantham-lunarg ... starting here.
There was a problem hiding this comment.
This looks like it keeps the original functionality but split into separate pieces. Looks OK to me.
|
CI gfxreconstruct build queued with queue ID 291493. |
|
CI gfxreconstruct build # 5245 running. |
|
CI gfxreconstruct build # 5245 passed. |
|
CI gfxreconstruct build queued with queue ID 291531. |
|
CI gfxreconstruct build # 5246 running. |
|
That's all folks. |
|
CI gfxreconstruct build # 5246 passed. |
| return output_data_; | ||
| } | ||
|
|
||
| typename T::struct_type* AllocateOutputData(size_t len, const typename T::struct_type& init) | ||
| { | ||
| assert(!DecoderHasOutputAllocator<T>::value); |
There was a problem hiding this comment.
Prefer GFXRECON_ASSERT to assert. The CMake option GFXRECON_ENABLE_RELEASE_ASSERTS can be enabled to log failed GFXRECON_ASSERTs as errors in release builds.
There was a problem hiding this comment.
The rest of this file is all assert(...)
There was a problem hiding this comment.
That's fine. For new code please prefer GFXRECON_ASSERT. There may come a day when we change as many assert as we can to GFXRECON_ASSERT
| const StructPointerDecoder<Decoded_VkInstanceCreateInfo>* pCreateInfo, | ||
| const StructPointerDecoder<Decoded_VkAllocationCallbacks>* pAllocator, | ||
| HandlePointerDecoder<VkInstance>* pInstance) | ||
| void VulkanReplayConsumerBase::ModifyCreateInstanceInfo( |
There was a problem hiding this comment.
This looks like it keeps the original functionality but split into separate pieces. Looks OK to me.
| @@ -68,12 +92,13 @@ class StructPointerDecoder : public PointerDecoderBase | |||
| typename T::struct_type* AllocateOutputData(size_t len) | |||
| { | |||
| output_len_ = len; | |||
| output_data_ = DecodeAllocator::Allocate<typename T::struct_type>(len); | |||
| output_data_ = StructPointerOutputDataAllocator(decoded_structs_, len); | |||
There was a problem hiding this comment.
Is this change for a new use of StructPointerDecoder? Or is there an existing use that wasn't being handled correclty?
|
If @davidd-lunarg and @MarkY-LunarG approve the PR I trust their approval. |
For layered API support.
For layered API support.
Layered API support
Prevents uninitialized next/pNext. The lower level decoders aren't always setting values. Set a good state at initialization (zeros) Layered API support.
Prevent collisions with other API consumers and decoders Layered API support.
Split vulkan instance and device creation override into reusable pieces suitable for layered API support (specifically OpenXR) Make neeed information public for other replay consumers Layered API support.
Vulkan captures unwrapped physical device handles, Layered API (like OpenXR) captures wrapped handles. During replay two HandleId's will reference the same VkPhysical device. The vulkan_alias is the handleId as known by the vulkan_consumer, which will be created/updated, etc, by all Vulkan replay calls.
Move replay options around that are shared between multiple replay paths. For layered API support.
jzulauf-lunarg
force-pushed
the
zulauf_oxr_infrastructure
branch
from
6b7e2a1
to
a08a7ca
Compare
|
CI gfxreconstruct build queued with queue ID 292063. |
|
CI gfxreconstruct build # 5252 running. |
|
CI gfxreconstruct build # 5252 passed. |
Not sure if these will interfere with @panos-lunarg pending changes, will wait if needed