Trying to create Unordered Access View on a 2D texture

Question

I am trying to use the GPU to do some math on 2D arrays, which I supply to the compute shader as simple 2d textures (in R32_FLOAT) without mipmaps, tied to Shader Resource Views. They work, they are passed to the shader, and are addressible.

But I got stuck trying to create an Unordered Access View on a similar 2d texture. The CreateUnorderedAccessView(...) returns E_INVALIDARG, but I don't know how to expand that into more detail. Same problem happens with CreateTexture2D(...) whenever I use D3D_BIND_UNORDERED_ACCESS flag.

My computer supports D3D11_0 and cs_5_0. Here's my code:

HRESULT Create_Simple_Texture_2D(ID3D11Device* device, ID3D11Texture2D** texture, UINT width, UINT height, void* initdata, UINT initsize)
{
    *texture = nullptr; 
    
    UINT elementsize = sizeof(float);
    
    DXGI_FORMAT format = DXGI_FORMAT_R32_FLOAT;

    D3D11_TEXTURE2D_DESC desc = {};
    {

// this works
//        desc.BindFlags = D3D11_BIND_SHADER_RESOURCE;    // input to GPU
//        desc.Usage = D3D11_USAGE_DEFAULT;               // GPU read&write
//        desc.CPUAccessFlags = 0;                        // no CPU access

// this works too
//        desc.BindFlags = D3D11_BIND_SHADER_RESOURCE;  // input to GPU
//        desc.Usage = D3D11_USAGE_DYNAMIC;             // CPU write -> GPU read
//        desc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; // CPU write

// and this
//        desc.BindFlags = 0;                             // 
//        desc.Usage = D3D11_USAGE_STAGING;               //
//        desc.CPUAccessFlags = D3D11_CPU_ACCESS_READ;    // CPU read

// BUT NOT THIS:
        desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_UNORDERED_ACCESS;
        desc.Usage = D3D11_USAGE_DEFAULT;               // GPU read&write
        desc.CPUAccessFlags = 0;                        // no CPU access, except at creation


        desc.Width = width;
        desc.Height = height;
        desc.Format = format;                           // see above
        desc.ArraySize = 1;                             // single texture
        desc.MipLevels = 1;                             // no other mipmaps
        desc.MiscFlags = 0;
        {
            desc.SampleDesc.Count = 1;
            desc.SampleDesc.Quality = D3D11_STANDARD_MULTISAMPLE_PATTERN;
        }
    }

    const UINT expected = desc.Width * desc.Height * desc.ArraySize * elementsize;

    if (initdata && (initsize != expected)) wcprintf("\n\tWARNING: declared array size mismatch in Create_Simple_Texture_2D(...)\n\n");

    D3D11_SUBRESOURCE_DATA sub = {};
    {
        sub.pSysMem = (void*)(new char[expected]);
        sub.SysMemPitch = desc.Width * elementsize;
        sub.SysMemSlicePitch = 0;
        const UINT copysize = min(expected, initsize);
        if (sub.pSysMem && initdata && copysize) memcpy((void*)sub.pSysMem, initdata, copysize);
    }

    HRESULT result = device->CreateTexture2D(&desc, &sub, texture);
    
    if (sub.pSysMem) delete[] sub.pSysMem;

    return result;
}

Answer 1

TL;DR: Enable the Direct3D debug device. See also this blog post.

The problem here is your use of desc.SampleDesc.Quality = D3D11_STANDARD_MULTISAMPLE_PATTERN;.

The Direct3D Debug Layer when enabled tells you this:

D3D11 ERROR: ID3D11Device::CreateTexture2D: Multisampling is not supported with the D3D11_BIND_UNORDERED_ACCESS BindFlag.
SampleDesc.Count must be 1 and SampleDesc.Quality must be 0.
[ STATE_CREATION ERROR #99: CREATETEXTURE2D_INVALIDBINDFLAGS]

You normally use this to mean "no MSAA"

desc.SampleDesc.Count = 1;
desc.SampleDesc.Quality = 0;

Since you already started with a zero-filled struct with D3D11_TEXTURE2D_DESC desc = {};, just remove your set of desc.SampleDesc.Quality entirely and you'll be good to go.

D3D11_TEXTURE2D_DESC desc = {};
desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_UNORDERED_ACCESS;
desc.Usage = D3D11_USAGE_DEFAULT;
desc.Width = width;
desc.Height = height;
desc.Format = DXGI_FORMAT_R32_FLOAT;
desc.ArraySize = desc.MipLevels = desc.SampleDesc.Count = 1;