vanattab / slimdx

Automatically exported from code.google.com/p/slimdx
MIT License
0 stars 0 forks source link

fx_5_0-Effect with disabled hull/domain shaders causes an exception #633

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago
Hi, I have a problem. I have many effects, which describes shaders. But not
all of them need all the shaders. For example my gui effect needs only the
vertex and pixel shaders. So I have to disable the others, because if not,
they remain active.

A way to disable the shaders is to passing a "0" or "NULL" to the
SetHullShader(0); function in the technique of an effect.

I did this, and the shader seems to compile successfully using the fxc.exe
compiler. The problem occurs only, if I try to set the hull/domain shader
to "0" or "NULL". The vertex/geometry/pixel shaders accepts this values and
SlimDX has no problems with them. But for the hull/domain shaders, SlimDX
comes with the following problem:
[code]
SlimDX.Direct3D11.Direct3D11Exception: E_FAIL: An undetermined error
occurred (-2147467259)
   bei SlimDX.Result.Throw[T](Object dataKey, Object dataValue) in
dev\slimdx\source\result.cpp:Zeile 119.
   bei SlimDX.Result.Record[T](Int32 hr, Boolean failed, Object dataKey,
Object dataValue) in dev\slimdx\source\result.cpp:Zeile 150.
   bei SlimDX.Direct3D11.Effect..ctor(Device device, ShaderBytecode data,
EffectFlags effectFlags) in dev\slimdx\source\direct3d11\effect11.cpp:Zeile 56.
[/code]

I'm using SlimDX, C#, x86 target, Win7 64bit, 5870 with the newest radeon
driver.

Original issue reported on code.google.com by remi2...@googlemail.com on 10 Mar 2010 at 9:03

GoogleCodeExporter commented 9 years ago
Is there any output from the debug runtime layer?

Original comment by josh.petrie on 11 Mar 2010 at 2:41

GoogleCodeExporter commented 9 years ago
I have enabled the debug layer on DX11 (pointed to my application first, then 
set the
"Force on" radio button. No other checkboxes are on.) I also tried to enable the
debug unmanaged code option in my C# project. Nothing changes, the same problem
occur. Maybe there is something more I have to change?

I also tried to debug step by step the SlimDX code: The effect compiles well, 
there
are no compilation errors. But then, the call to D3DX11CreateEffectFromMemory 
in the
Effect-Constructor returns nothing for the effect and a negative hr.

Maybe this is a bug in DX? If you can explain step by step, how I can provide 
you
more info, for example, if I missed something to enable the debug layer, please 
tell me.

I think, you can try it yourself. Just create a simple effect and in the 
technique
set the hull and domain shader to "0". (Or use an existing DX11 example 
effect.) This
effect should compile fine using fxc.exe. But calling to:

new Effect(device, ShaderBytecode.CompileFromFile("effect.fx", "main", "fx_5_0",
ShaderFlags.None, EffectFlags.None), 0);

should throw an error.

Original comment by remi2...@googlemail.com on 11 Mar 2010 at 9:10

GoogleCodeExporter commented 9 years ago
LOL, now I have with 70% chance on starting my application this error:

Unbehandelte Ausnahme: System.ArgumentException: Ein Element mit dem gleichen
Schlüssel wurde bereits hinzugefügt. (This means: Tere is already an element 
with the
given key.)
   bei System.ThrowHelper.ThrowArgumentException(ExceptionResource resource)
   bei System.Collections.Generic.Dictionary`2.Insert(TKey key, TValue value, Boolean
add)
   bei SlimDX.ObjectTable.Add(ComObject comObject, ComObject owner) in
dev\slimdx\source\objecttable.cpp:Zeile 105.
   bei SlimDX.ComObject.Construct(IUnknown* pointer, ComObject owner) in
dev\slimdx\source\comobject.cpp:Zeile 116.
   bei SlimDX.Direct3D11.DeviceContext..ctor(ID3D11DeviceContext* pointer, ComObject
owner) in dev\slimdx\source\direct3d11\devicecontext11.cpp:Zeile 73.
   bei
SlimDX.ComObject.ConstructFromPointer<SlimDX::Direct3D11::DeviceContext,ID3D11De
viceContext>(ID3D11DeviceContext*
pointer, ComObject owner, ComObjectFlags flags) in
dev\slimdx\source\comobject.h:Zeile 94.
   bei SlimDX.Direct3D11.Device.get_ImmediateContext() in
dev\slimdx\source\direct3d11\device11.cpp:Zeile 113.

What I have done? Oo Why this now?

Original comment by remi2...@googlemail.com on 11 Mar 2010 at 9:17

GoogleCodeExporter commented 9 years ago
The last exception may be an unrelated bug. 

What feature levels are you using to create the device, and what feature level 
are
you actually getting selected at runtime? That will be neccessary to try to 
reproduce
this.

Original comment by josh.petrie on 11 Mar 2010 at 4:48

GoogleCodeExporter commented 9 years ago
Ok, I create my device this way:

factory = new Factory();
device = new Device(factory.GetAdapter(0));

For some reason, this is the one and only way I can create a device supporting 
the
Level_11_0. After that, if I ask the device, it has the proper Level_11_0. With 
a
lower level the "fx_5_0" target for the effect compiler would fail, I think. ^^

For the other bug: I posted this as an other isue, I hope, this is ok for you.

Original comment by remi2...@googlemail.com on 11 Mar 2010 at 6:47

GoogleCodeExporter commented 9 years ago
I tried to debug this today. All I know, is, that the shader compiles well, 
because
there are all values inside like technique etc. Then the 
D3DX11CreateEffectFromMemory
function fails, without touching the shader stream (but I don't know this for 
sure).
It's pointer seems to point still to index 0 after the 
D3DX11CreateEffectFromMemory
function returns.

If I remove the SetHullShader(0); and SetDomainShader(0); functions from my 
effect
file, all is running fine.

I have an other idea to explain this behaivour:

Maaaaaaaaaybeeee the HLSL team allow as to set this functions to 0, BUT the DX 
team
suppose, that IF we use those two shaders, then we always define both of them
together AND this will replace the old one. And after that, IF we want to use an
effect without those two shaders, THEN we can leave them out from the pass and 
then
we can't use control points as input anymore. But if we don't use control 
points,
then the old hull/domain shader will be automatically ignored -> implicitly 
disabled
for that pass ^^. And then the DX team leader said: "Guys, we do not have to 
support
SetHull/DomainShader(0); explicitly!"

Original comment by remi2...@googlemail.com on 13 Mar 2010 at 12:19

GoogleCodeExporter commented 9 years ago
I just ran into this exact same issue.

Setting the Hull and Domain shaders to 0 or NULL causes the Effect constructor 
to 
throw an exception.

The DirectX debug says this:
"Effects11: Internal loading error: invalid shader index."

Which leads me to believe it is on the DirectX side and that I'm just probably 
doing 
something wrong.  Is there another way to disable the Hull/Domain Shaders? 

I am using feature level 11_0.

Original comment by mccarthp on 17 Mar 2010 at 7:26

GoogleCodeExporter commented 9 years ago
I did some more searching online, and this is definitely on the DirectX side, 
but it 
is (apparently) the way DirectX works by design.

The solution is to disable them using the Device, like so:

device.ImmediateContext.HullShader.Set(null);
device.ImmediateContext.DomainShader.Set(null);

This solves the problem, so it is no longer an issue.

Original comment by mccarthp on 17 Mar 2010 at 7:43

GoogleCodeExporter commented 9 years ago
AS I said, if you disable them, you can't use control points anymore, you have 
to use
normal vertices instead. But if you use normal vertices, then you don't have to
disable the hull/domain shaders, because they are automatically ignored. So
specifing, that you want to use for example trianglestrip instead of patches 
with 3
controlpoints automatically ignores the hull/domain shaders.

Maybe this is the reason, why MS want not correct the bug so fast.

Original comment by remi2...@googlemail.com on 17 Mar 2010 at 9:26

GoogleCodeExporter commented 9 years ago

Original comment by josh.petrie on 18 Mar 2010 at 5:25