3e332567d6
Pipelines should have created renderpass as a parent
2016-02-07 18:45:19 +01:00
af8addf20e
Display if blend is enabled properly
2016-02-07 18:45:18 +01:00
c504cc456b
Improve printing of begin/end render pass load/store operations
2016-02-07 18:45:17 +01:00
b363dc1fb5
Vulkan scissor regions are x,y,width,height
2016-02-07 18:45:16 +01:00
ba95d887f8
Don't remap memory indices for vkAllocMemory while in replay
2016-02-07 18:45:15 +01:00
36d2ffd988
Handle super small buffers in buffer viewer without displaying nothing
2016-02-07 18:45:14 +01:00
484d70ddbd
Display buffers properly in resource list
2016-02-07 18:45:13 +01:00
354e64a844
Pass back buffers and buffer information
2016-02-07 18:45:12 +01:00
4ece1ed5b3
Handle vtx binds referencing unset vbuffers (valid if binds are unused)
2016-02-07 18:45:12 +01:00
b2676c2861
Store last swapchain presented and use it instead if none is specified
2016-02-07 18:45:11 +01:00
d198569313
Make swapchains parents of im and put GetSwapchainImages in swap record
2016-02-07 18:45:10 +01:00
58703dd348
Don't try and create our own layer on replay if it was in createInfo
2016-02-07 18:45:09 +01:00
2d8a6640eb
Update VKTODO with note on blocker
2016-02-07 18:45:08 +01:00
8696b2133c
Remove VKTODO that is handled
2016-02-07 18:45:07 +01:00
4fe74824b1
Switch buffer offsets and lengths to 64-bit, downcast only where needed
2016-02-07 18:45:06 +01:00
b569932a71
Update pipeline stage names to Vulkan nomenclature
2016-02-07 18:45:05 +01:00
f1f0d45113
Sum up release notes entry - not GPU portable at all
2016-02-07 18:45:05 +01:00
6f351c05e1
Fixes for storing & displaying dynamic offsets in pipeline state
2016-02-07 18:45:04 +01:00
bb4522c2b7
Minor shader cleanup
2016-02-07 18:45:03 +01:00
bd237aea83
Remove release note about memory maps
2016-02-07 18:45:02 +01:00
fb3ccc9994
Finish map-intercept handling for now.
...
* Currently I've decided to stick with the scheme of reading back from
mapped pointers where necessary to either compare or serialise. This
is because the bookkeeping for intercepting memory maps is quite
complex and right now it seems hard to justify as the overhead of the
readback is not significant enough to worry about. Unless we encounter
a case where the readback is hugely slower and frame captures begin to
take 10s of seconds longer, it's not justifiable.
2016-02-07 18:45:01 +01:00
34e37257bd
Add explanation of current plan for memory map handling
2016-02-07 18:45:00 +01:00
79e7785718
Move function to more appropriate file
2016-02-07 18:44:59 +01:00
156c75b26d
Modify memory properties to discourage coherent maps & remap memory idxs
2016-02-07 18:44:58 +01:00
6bdf285166
Use consistent serialiser pointer
2016-02-07 18:44:58 +01:00
7769e1d672
Change API events view to show { } sections in sub-nodes
2016-02-07 18:44:57 +01:00
685032e78c
Store renderpass load/store ops and display in drawcall names
2016-02-07 18:44:56 +01:00
dd57110036
Make a drawcall for render pass end
2016-02-07 18:44:55 +01:00
1d092f6b87
Remove some VKTODOs that are judged not relevant anymore
2016-02-07 18:44:54 +01:00
7b952c53ef
Move loadmodule/getprocaddress code into os_specific
2016-02-07 18:44:53 +01:00
55914f99ae
GL hooks no longer need to be disabled - just don't hook nv VK driver
2016-02-07 18:44:52 +01:00
6da044ae0a
Remove hardcoded hack to clear backbuffer to black
...
* Now handled as a runtime created init state, the same as if we started
skipping initial states for large images that are write-only
2016-02-07 18:44:51 +01:00
da15b25789
Handle image dirty & initial state tracking more correctly
...
* Still not doing buffer<->image copy for proper decode/encode.
* Images are dirtied in preference to their memory, and their initial
state consists of the memory backing behind them, which is copied
and restored in the same way as dirty memory.
2016-02-07 18:44:50 +01:00
97527d273d
Add memory creation info
2016-02-07 18:44:50 +01:00
a87587e91f
Memory no longer needs to be dirty from creation, tracking correctly now
2016-02-07 18:44:49 +01:00
518a86b4a4
Remove hack to propogate frame references to memory, track them properly
...
* This takes advantage of immutable non-sparse object memory bindings.
* Currently this will treat images the same as buffers, as read-only
containers with memory still being the onlything that is dirtied and
needs initial contents, but has some of the code needed for proper
separate image initial state/dirty tracking.
2016-02-07 18:44:48 +01:00
abe9696973
Reduce debug-print spam for dirty resources
2016-02-07 18:44:47 +01:00
da447ba5d9
Fixes for coherent persistent maps and multiple frame capture
2016-02-07 18:44:46 +01:00
0d66a78105
only flush pending dirty resources while idle
2016-02-07 18:44:45 +01:00
198ab2b714
Check descriptor sets are in range when iterating layout, warn otherwise
2016-02-07 18:44:44 +01:00
24ad3356dd
Deserialise fix - make sure scoping doesn't delete data we need
2016-02-07 18:44:43 +01:00
25c5de8c85
Make sure resource records are deleted as most specific type
...
* Don't need to use a virtual destructor as the information is already
in the templated resource manager
2016-02-07 18:44:42 +01:00
63e0cc3f39
Use correct uniform buffer alignment instead of hardcoded 256
2016-02-07 18:44:42 +01:00
c54b17284c
Remove some VKTODOs
2016-02-07 18:44:41 +01:00
4211cd7359
Remove VKTODOs that aren't needed anymore
2016-02-07 18:44:40 +01:00
900d65b770
Refactor out start/end frame capture from present code
2016-02-07 18:44:39 +01:00
c1959102c8
Rename GenericPipeline to HighlightBoxPipeline as it uses line lists
2016-02-07 18:44:38 +01:00
67d97f4a10
Change text rendering back to instanced triangle strips
2016-02-07 18:44:37 +01:00
ed70b4b58e
Remove VKTODO - image layout no longer part of framebuffer attachment
2016-02-07 18:44:36 +01:00
fa5bc912de
Tweak handling of sType/pNext. Currently not much to do
2016-02-07 18:44:35 +01:00
baldurk