zdoom/qzdoom
0
0
Fork 0
mirror of https://github.com/ZDoom/qzdoom.git synced 2024-12-14 14:31:53 +00:00
Code Issues Projects Releases Packages Wiki Activity
17617 commits 86 branches 139 tags 124 MiB
Commit graph

348 commits

Author SHA1 Message Date
Christoph Oelckers
5e4de10f97 - fixed survey data retrieval for Vulkan. 2019-08-11 21:26:08 +02:00
Christoph Oelckers
1e4a32004b - fixed bad texture canvas checks when in truecolor software rendering. 2019-08-09 13:59:54 +02:00
Magnus Norddahl
d1026e4b5b - call vkDeviceWaitIdle before RAII tears down buffers 2019-08-08 20:56:11 +02:00
alexey.lysiuk
ff4bc8bab9 - use EXT_metal_surface extension for Vulkan surface creation 
Prefer EXT_metal_surface over deprecated MVK_macos_surface extension, and use it only when modern one is not available
 2019-08-03 13:16:12 +03:00
alexey.lysiuk
cf81ab465e - updated Vulkan headers to version 1.1.114 
This is the latest version released with Vulkan SDK
New headers are needed for EXT_metal_surface
 2019-08-03 10:52:04 +03:00
alexey.lysiuk
5870cb7ea1 - recreate Vulkan swapchain in case of surface lost error 
Do not abort with fatal error when VK_ERROR_SURFACE_LOST_KHR was returned from vkAcquireNextImageKHR() or vkQueuePresentKHR()
So far, only gfx-portability implementation is reporting this error from time to time, usually on startup, entering the game, or task switching
 2019-08-02 10:21:06 +03:00
alexey.lysiuk
370f5ef234 - fixed writing of texture matrices in Vulkan render backend 
https://forum.zdoom.org/viewtopic.php?t=65479
 2019-07-30 11:46:14 +03:00
Christoph Oelckers
afbd7f7a24 - increased StreamBuffer size because I got it to crash with comatose.wad once due to an overflow here. 2019-07-28 19:05:39 +02:00
Magnus Norddahl
ad8c5bb96e - add VkRenderPassManager::GetVertexFormat so that all VkRenderState access is done through functions that can then be made thread safe 2019-07-28 17:20:49 +02:00
Magnus Norddahl
76675a4d61 - add VkMatrixBufferWriter so that both buffers are used same way in VkRenderState 2019-07-28 17:04:56 +02:00
Magnus Norddahl
4adac3fe59 - move streaming uniform buffers out of VkRenderState 2019-07-28 16:28:43 +02:00
Christoph Oelckers
7346288bf5 - moved some more files. 2019-07-14 21:09:49 +02:00
Marisa Kirisame
bdb083f457 Shape2D drawing will gracefully abort the VM on any out of bounds access. 2019-07-12 17:44:54 +02:00
Christoph Oelckers
5df8919b90 Merge branch 'master' into userlights 2019-07-07 08:11:09 +02:00
Dzmitry Malyshau
3467e0edcf Switch command buffers to be one-time-submittable (#885) 2019-07-05 02:15:59 +02:00
Dzmitry Malyshau
e7e46f8c7c vk: Check for the time query support on the graphics queue (#884) 2019-07-03 16:49:06 +02:00
Christoph Oelckers
5ef43143f8 - removed unused CVAR_GUID type 2019-06-27 09:16:34 +02:00
Christoph Oelckers
3d3f364874 - consolidated postprocessing code a bit. 
The code that determines the order of effects should not be duplicated between renderers.
 2019-06-20 11:55:45 +02:00
Magnus Norddahl
774955dbc4 - write into stream buffers before opening the next render pass 2019-06-16 13:08:25 +02:00
Magnus Norddahl
d31dbf14e4 - wait for space if the stream buffers are exhausted 2019-06-16 13:05:12 +02:00
Magnus Norddahl
240a32f4c8 - also reset framebuffers when the image is reset 2019-06-16 10:31:12 +02:00
alexey.lysiuk
7e901055ea - fixed: light mode from mapinfo had no effect 
https://forum.zdoom.org/viewtopic.php?t=64997
 2019-06-12 09:49:40 +03:00
Magnus Norddahl
a093f686a7 - store the vulkan framebuffer with the image (this ensures the framebuffer object is destroyed along with the image) 2019-06-10 22:14:02 +02:00
Christoph Oelckers
2766303cfc - consolidated the 3 atterm implementations. 
Each platform had its own copy. Why?
 2019-06-10 12:01:01 +02:00
Christoph Oelckers
9f2fbc1294 - renamed GLViewpointBuffer to HWViewpointBuffer. 2019-06-10 10:50:45 +02:00
Magnus Norddahl
13cab1c0b6 - delay postprocess texture destruction until the end of the frame 2019-06-10 09:13:56 +02:00
Magnus Norddahl
8853872bb4 - fix wipe screen when using -loadgame 2019-06-10 08:40:49 +02:00
Christoph Oelckers
037b69c8a7 - reworked buffer binding logic. 
This shouldn't be in the hardware independent interface because the semantics on OpenGL and Vulkan are too different, so a common implementation is not possible.
Most bind calls were in the GL interface anyway, so these no longer pass through hardware independent code.

This also moves the bind calls in the shadowmap code into the GL interface - these never did anything useful in Vulkan and aren't needed there.

Last but not least, this moves the legacy buffer binding handling into FGLRenderState and performs the initial binding for the light buffer in a more suitable place so that this doesn't have to pollute the render state.
 2019-06-09 20:37:11 +02:00
Gutawer
324a720d56 - added a linear transformation system to Shape2D so that things like rotations and scales can be changed without having to clear and push a large amount of data 2019-06-08 14:05:32 +02:00
Magnus Norddahl
c9ee880196 - fix loadgame crash 2019-06-06 18:56:34 +02:00
Magnus Norddahl
ae69abc049 - FXAA_DISCARD should never be enabled. It only works when the output framebuffer is the same as the original. 2019-06-05 19:06:33 +02:00
alexey.lysiuk
4d29b5b5e7 - limited disabling of FXAA discard to Vulkan on AMD hardware 2019-06-04 12:43:08 +03:00
alexey.lysiuk
dc8a944382 - implemented GPU vendor string assingment in Vulkan backend 2019-06-04 12:42:01 +03:00
alexey.lysiuk
addcad8ac0 - disabled discard in FXAA shader with Vulkan backend 
Discard caused graphical corruptions on AMD hardware with Vulkan backend enabled

https://forum.zdoom.org/viewtopic.php?t=64230
 2019-06-04 11:32:00 +03:00
Edoardo Prezioso
589f9b5c24 - Fixed uninitialized variable in case GL_MAX_VERTEX_SHADER_STORAGE_BLOCKS does not exist. 
Old graphics, like mine (Intel Sandybridge Mobile, GL 3.0 Mesa) do not support this, therefore most of the time RFL_SHADER_STORAGE_BUFFER wasn't unset (I found no consequence of this on my machine, but better safe than sorry).

Found out by Valgrind.
 2019-06-02 21:20:53 +02:00
Christoph Oelckers
6ace482266 - fixed fillSimplePoly for indexed data. 
On OpenGL this merely caused some visual glitches on the cheat automap - but the bad data in the buffer can make Vulkan abort entirely.
 2019-05-30 13:32:42 +02:00
alexey.lysiuk
da2d0e47e6 - removed array length() function from shadowmap shader 
Array's length() function is not yet supported by SPIRV-cross and MoltenVK
Its usage was replaced by explicit nodes count value passed as uniform
 2019-05-24 14:29:37 +03:00
Magnus Norddahl
756c593e96 - create a new error class for vulkan errors as they are only recoverable during initialization (unlike CRecoverableError which is recoverable during normal processing) 
- improve vulkan errors by including the status code returned by vulkan if they fail
 2019-05-23 14:09:05 +02:00
Magnus Norddahl
e4dab707e4 - don't run the custom shaders twice on screenshots 2019-05-20 12:46:05 +02:00
Magnus Norddahl
2a15f65065 - apply present shader on screenshots 2019-05-20 12:36:45 +02:00
alexey.lysiuk
bbb75a40ca - fixed erroneous texture mipmapping in Vulkan render 
https://forum.zdoom.org/viewtopic.php?t=64257
 2019-05-19 14:24:59 +02:00
Magnus Norddahl
680a6f348b - use one render pass for the entire scene or until postprocess or command buffer flushing forces it to end 2019-05-18 08:12:46 +02:00
Magnus Norddahl
7e37d640dc - avoid creating a new render pass if a pipeline bind will suffice 2019-05-18 06:54:35 +02:00
Magnus Norddahl
9ab19d057d - centralize how image transitions are done in the vulkan backend 2019-05-16 18:26:47 +02:00
Magnus Norddahl
99f58855ad - transition scene images used for the screenshot to color attachment optimal before rendering to them 2019-05-15 16:11:43 +02:00
Magnus Norddahl
3a112c2a47 - fix stencil test not being active when applying SSAO to portals 2019-05-13 22:06:17 +02:00
Magnus Norddahl
f2e37d506e - fix model chasecam crash 2019-05-12 23:05:31 +02:00
Christoph Oelckers
cf58666708 - fixed: The software renderer was accessing the global viewpoint in several places. 
Aside from using incorrect values this could crash the renderer if used in the very first rendering frame because the software renderer always operates on a copy.
 2019-05-12 09:16:17 +02:00
Magnus Norddahl
7957c423aa - fix wrong screenshot gamma 2019-05-10 18:02:27 +02:00
Magnus Norddahl
7c3e99a6f1 - Intel on Linux used the footnote in the standard saying a descriptor pool can become fragmented (thanks guys!) 2019-05-10 02:16:26 +02:00