69 Replies
d3d11-cursed
its almost running first try lmao
the top parts a little fucked but its p good
👏
love to see it
25 API High Miscellaneous 2360752066 Validation Error: [ VUID-vkCmdDrawIndexed-renderPass-02684 ] Object 0: handle = Render Pass 458, type = VK_OBJECT_TYPE_RENDER_PASS; Object 1: handle = Render Pass 408, type = VK_OBJECT_TYPE_RENDER_PASS; | MessageID = 0x8cb637c2 | vkCmdDrawIndexed: RenderPasses incompatible between active render pass w/ VkRenderPass Render Pass 458 with a dependencyCount of 3 and pipeline state object w/ VkRenderPass Render Pass 408 with a dependencyCount of 1. The Vulkan spec states: The current render pass must be compatible with the renderPass member of the VkGraphicsPipelineCreateInfo structure specified when creating the VkPipeline bound to VK_PIPELINE_BIND_POINT_GRAPHICS (https://www.khronos.org/registry/vulkan/specs/1.3-extensions/html/vkspec.html#VUID-vkCmdDrawIndexed-renderPass-02684)
ah yes renderpasses my beloved
but why in #d3d11-cursed
dxvk-native
dxvk moment
yea
im p sure i am doing something spec related wrong
¯\_(ツ)_/¯
but renderdoc REALLY dislikes dxvk-native stuff
crashes with vulkan device init failed constantly
maybe attach it after startup?
https://github.com/Furball-Engine/Furball.Vixie/blob/d4a6b021194d97425ff15b1b741287e15d6c342b/Furball.Vixie.Backends.Direct3D11/Direct3D11Renderer.cs#L302
anything obvious you can see?
GitHub
Furball.Vixie/Direct3D11Renderer.cs at d4a6b021194d97425ff15b1b7412...
A Multi API Renderer specifically designed to be integrated into Furball, aswell as providing good Hardware support - Furball.Vixie/Direct3D11Renderer.cs at d4a6b021194d97425ff15b1b741287e15d6c342...
the problem is after the capture is already taken
if i select any event, after a min or two of being frozen, renderdoc will pop up an error and die
oh alright then
also shouldn't the nointerpolation in the shader be in the output struct, not the input struct?
oh
probably
fixing rn
1. please don't use Marshal.OffsetOf
2. don't you need semanticIndex: 1?
the semantic name should probably always be without a number
why not?
because it's not a simple "the layout of this in memory"
also i don't even think it's correct
i mean its worked fine so far
yeah just nitpickin
it should be OffsetOf + sizeof(int)
because you used a long
for whatever reason
2 before 1 beacuse im a baller
but yeah i already do + sizeof int
bindless handles
ugh fine
i just add 1 and 2 together in the shader
yeah i remember that
i think i get what you mean here, lemme recompile the shaders rq and test
(im manually compiling through FXC)
also Marshal.OffsetOf uses reflection so extra bleh
its only once at startup and it makes the code infinitely more readable so its a tradeoff, once nativeaot is in a better state ill be playing with reflectionless furball
but currently imgui no worky on nativeaot
segfault in cimgui
A semantic index is only needed in a case where there is more than one element with the same semantichttps://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/dx-graphics-hlsl-semantics hmm these docs are a bit confusing because a TEXCOORD can basically be anything yet they say
float4
hlsl do be quirkythat much is true technological pizza man
could you take a look at what i just pushed for d3d11? it seems that textures arent being applied properly and im not sure why, and since renderdoc shits itself for me i cant really check myself
alrighty then
i checked everything in c# land and it seems fine
thx dx, very cool
whats the stack trace
hmm
pop a breakpoint down in that function, see the params in the
desc
it works with
desc.CPUAccessFlags |= CpuAccessFlags.Write;
but then dies on the Map calloops fuck
ig vulkan doesnt have a problem if its not CPU writable :^)
¯\_(ツ)_/¯
weird that its failing to map
given it is Dynamic
alright fixed
MapMode.WriteDiscard 😎
ah
yeah that probably is what i want anyway
aaand it's bed time, bye 👋
a cya
i was able to get renderdoc working on my laptop
terrifyingly: the texture is 2.8*10^-45
the last int doesnt seem to be being picked up as one of the vertex attributes hm
hmmmmm
this is very odd i am in pain
@baebey how do i compile the hlsl
fxc.exe /Fo PixelShader.obj /T ps_5_0 /EPS_Main Shader.hlsl
fxc.exe /Fo VertexShader.obj /T vs_5_0 /EVS_Main Shader.hlsl
from the june 2010 dx sdkyou can also use
InputElementDescription.AppendAligned
to not need OffsetOfoh
neat
oh hey
hlsl very quirky
oh no what now
works
o h
i also did
nointerpolation uint TextureId : TEXID;
(changed to uint)also this warning
there was also a warning for the shader using
int
s but code using R32_UInt
it is odd im getting that warning
it seems to work though
code should still be
"TEXID", semanticIndex: 1
oh ok
¯\_(ツ)_/¯
huh yeah that seems to have fixed it
pog
okay well now i have to test a texture overflow
i need to draw >128 textures
i can't wait for my new spirv thing to fill most vertex layouts out for me 😩
creating 128 textures seems to still work pog
with that: all 3 backends seem to have a fully functional renderer now, so now i can start moving all the other tests over to it
then prob cleanup the code then create the PR
oh wait no i still have to create a VAO less version for GL<3.0
pain
and also have to figure out how to make ogl work consistently without the
layout
stuff
time to get it working with glsl 110
o wait i need to go to <#1010229292283728023>
okay im getting a weird crash now for no reaosn
when i end, draw, then begin a batch
in the middle of a frame
mainly for stuff like scissor
somehow the buffers are just
poofing
out of existance
they arent going into the queue
i checked and i def dont static anything rn
something about multiple renderers at once just this is NOT happy about
i dont know why its hitting this assert with it being false https://github.com/Furball-Engine/Furball.Vixie/blob/8a5dbfd7d3b49532b59ea0286f3c331d0cf5e5ab/Furball.Vixie.Backends.Direct3D11/Direct3D11Renderer.cs#L179
it should be impossible
this makes no sense PAIN
i genuinely cannot figure out where these buffers are going
unless actually wait
unless a frame doesnt allocate ANYTHING
thats when it might be null?
i think i'm leaking memory my computer froze
i think i'm leaking 20 buffers per frame
hm yeah im leaking memory somewhere and i have NO idea where lmao