b319929f39
Make sure swapchain images are given TRANSFER_SOURCE usage
2016-02-07 18:43:45 +01:00
8f91342f33
Check memory requirements for buffers when preparing initial state
2016-02-07 18:43:44 +01:00
1068288a17
Remove some outdated VKTODOs
2016-02-07 18:43:43 +01:00
6a208e608c
Add vulkan pipeline state to replay proxy serialising
2016-02-07 18:43:42 +01:00
2e08ef8ef5
32-bit compile warning fix
2016-02-07 18:43:41 +01:00
ad59636cde
Clarify that there are leaks on capture and replay
2016-02-07 18:43:40 +01:00
b1b6e29162
Free memory for initial states, and our own command pool
2016-02-07 18:43:39 +01:00
a9bf0a2538
Destroy debug manager and serialiser on shutdown
2016-02-07 18:43:38 +01:00
347cd67354
Make sure to delete buffers after use
2016-02-07 18:43:37 +01:00
7a63e824cc
Make sure images always have SAMPLED and TRANSFER_SOURCE usage bits
2016-02-07 18:43:37 +01:00
963986b9e9
Do swapchain resizing properly and re-enable thumbnails (working now?!)
2016-02-07 18:43:36 +01:00
4336722e34
Tidy up FetchTexture data for vulkan, fill out all fields properly
2016-02-07 18:43:35 +01:00
313f53a820
Remove stored fake backbuffer data, display backbuffer on present calls
2016-02-07 18:43:34 +01:00
19a13eefee
Allow multiple frame capture
2016-02-07 18:43:33 +01:00
e841bd9969
Implement proper pixel picking via RenderTexture & F32 readback
2016-02-07 18:43:32 +01:00
4ebbf09364
Get replay to display textures in gamma-correct fashion
2016-02-07 18:43:31 +01:00
64be44bf8b
Fix mip display of textures
2016-02-07 18:43:30 +01:00
feec902280
Implement remaining API functions
...
* Some functions, particularly synchronisation related event/fence
functions, are mostly untested and likely don't work yet.
2016-02-07 18:43:29 +01:00
b3d06da6a0
Implement more Cmd functions. CmdExecuteCommands is stubbed out
2016-02-07 18:43:28 +01:00
f6b6104977
Implement a few more API entry points, some are stubbed out
2016-02-07 18:43:28 +01:00
5e624d0f58
Add current release notes
2016-02-07 18:43:27 +01:00
9cc52ebc7e
Windows compile fix - header declare Deserialise template functions
...
* MSVC generates multiply defined symbols otherwise.
2016-02-07 18:43:26 +01:00
2b5fe533e7
Remove VKTODO for Deserialise - completed by @greg-lunarg
2016-02-07 18:43:25 +01:00
38ebce536e
Finish move to Deserialise function; remove Deserialise class
2016-02-07 18:43:24 +01:00
c3553d26a9
Create Deserialize function to eventually replace Deserialise class
2016-02-07 18:43:23 +01:00
7cab194395
add Deserialise code for latest vulkan header/sdk
2016-02-07 18:43:22 +01:00
494af5a9f4
Add Deserialise support for object created with SerialiseBuffer.
2016-02-07 18:43:21 +01:00
cf1b3a2aa5
Add Deserialise support for objects created with SerialisePODArray.
2016-02-07 18:43:20 +01:00
ad77fc2c5d
fix vulkan header name changes from cherry-pick.
2016-02-07 18:43:20 +01:00
d1a80c6657
Add Deserialise support for all objects created with SerialiseComplexArray
...
and SerialiseOptionalObject.
2016-02-07 18:43:19 +01:00
17fb1deb84
Add Deserialise for VkDescriptorSetLayoutCreateInfo.
2016-02-07 18:43:18 +01:00
33ca020012
Add Deserialise class to deallocate nested Vulkan structs
...
Currently deallocating VkDeviceCreateInfo and VkPipelineLayoutCreateInfo.
Still more to be done.
2016-02-07 18:43:17 +01:00
8b7744e71c
Handle descriptor pools deleting their children implicitly on destroy
2016-02-07 18:43:16 +01:00
b28428cdf1
Don't try and index into a descriptor set or binding that doesn't exist
...
* This can happen with unused binds - I think it's valid for a bind in a
shader to index outside of the descriptor set as long as it's never
used.
2016-02-07 18:43:15 +01:00
2b9c3118a5
Fix null deref becaues of incorrect error print
2016-02-07 18:43:14 +01:00
2c0751c338
Change texdisplay.frag shader to be work with LunarG intel icd compiler
2016-02-07 18:43:13 +01:00
1fd3d4ba29
Handle buffer or image ending up as VK_NULL_HANDLE in memory barriers
2016-02-07 18:43:13 +01:00
2686ad698c
Mark queue as referenced when submitting through it
2016-02-07 18:43:12 +01:00
e958b9925c
Demote logs to debug prints
2016-02-07 18:43:11 +01:00
307052b467
For now, persistent maps are ones that have been going for many frames
2016-02-07 18:43:10 +01:00
ac0f1c6540
Only put memory binds into frame capture record if they're not the first
2016-02-07 18:43:09 +01:00
08a63f2c0b
Serialise image layout in CmdCopyBufferToImage properly
2016-02-07 18:43:08 +01:00
31274c7c1c
Make sure going between WRITING_IDLE and WRITING_CAPFRAME is atomic
...
* This fixes a bug we had before, where we'd go into CAPFRAME then do a
bunch of work before clearing the frame record.
* The threading around this probably needs more careful thought - e.g.
the snapshotting and chunk insertion on end capture.
2016-02-07 18:43:07 +01:00
18bcd75db7
Fix LogicalNot vs LogicalNotEqual
2016-02-07 18:43:06 +01:00
0c55322abf
Allow sType to be wrong for VkDeviceQueueCreateInfo for now
...
* Dota passes 0 (APPLICATION_INFO) - presumably it's init'd with {0}
and the code doesn't set it properly yet as it's new.
2016-02-07 18:43:05 +01:00
b72858b42f
Check image aspect mask by & not by ==
2016-02-07 18:43:05 +01:00
e645728300
Make sure to serialise extensibility members in VkDeviceQueueCreateInfo
2016-02-07 18:43:04 +01:00
9bbfa2a55f
Add todo about fetching dynamic state from either pipeline or dynamic
2016-02-07 18:43:03 +01:00
731c37b6c4
Check the right buffer resource object for different descriptor types
2016-02-07 18:43:02 +01:00
ca79cfecb6
Don't try and wait on semaphores on replay
...
* We don't track & restore semaphore signalled/unsignalled state on
replay, so it's not valid anyway. Just be conservative and wait for
idle.
2016-02-07 18:43:01 +01:00
baldurk