Track Buffer-Device-Addresses #1712
Conversation
- codegen adapted to forward device-addresses to overrides - store both capture/replay addresses
- do some sanity-checks with new buffer-tracker and - check sbt-properties, add warning
- trimming: covers case when no vkGetBufferDeviceAddress was captured
|
CI gfxreconstruct build queued with queue ID 249835. |
|
CI gfxreconstruct build # 4767 running. |
|
CI gfxreconstruct build queued with queue ID 249858. |
|
CI gfxreconstruct build # 4768 running. |
|
CI gfxreconstruct build # 4768 passed. |
|
belongs to portable raytracing issue |
|
CI gfxreconstruct build queued with queue ID 250067. |
|
CI gfxreconstruct build # 4773 running. |
|
CI gfxreconstruct build # 4773 passed. |
|
I suspect that in the end we won't actually need the replayer address lookup functionality, just for the capture address. We have a patch in review that removes the last uses of this in our raytracing branch. |
good point, it could be omitted. I thought the range-based lookup 'might' be useful, also for replay-addresses, in other places but can't give an actual example. |
- we'll prob. only require lookup of capture-time addresses
|
CI gfxreconstruct build queued with queue ID 253263. |
|
CI gfxreconstruct build # 4805 running. |
|
CI gfxreconstruct build # 4805 passed. |
|
CI gfxreconstruct build queued with queue ID 253933. |
|
CI gfxreconstruct build # 4808 running. |
|
CI gfxreconstruct build # 4808 passed. |
|
CI gfxreconstruct build queued with queue ID 257812. |
|
CI gfxreconstruct build # 4839 running. |
|
many thanks @charles-lunarg & @MarkY-LunarG! |
|
CI gfxreconstruct build # 4839 passed. |
VulkanBufferTrackerclass -> map device-addresses (ranges) to BufferInfo*, O(log n)OverrideGetBufferDeviceAddress, track capture/replay buffer-addresses during replayOverrideCmdTraceRaysKHR, sprinkle asserts using buffer-tracker queriesOverrideBindBufferMemorywith tracking logicBuilding up necessary foundation as demonstrated by @bartosz-muszarski-arm's portable-raytracing branch