Hi,
I've got a question about the function D3DX11CreateShaderResourceViewFromMemory()
I tried to combine it with a live camera image, and wasn't able to upload the images coming from the camera to the ShaderResource.
What I tried:
ID3DX11_IMAGE_LOAD_INFO pLoadInfo;
pLoadInfo.Width = 640; pLoadInfo.Height = 480; pLoadInfo.Depth = 1; pLoadInfo.FirstMipLevel = 0;
pLoadInfo.Usage = D3D11_USAGE_DYNAMIC;
pLoadInfo.BindFlags = D3D11_BIND_SHADER_RESOURCE;
pLoadInfo.CpuAcessFlags = D3D11_CPU_ACCESS_WRITE;
pLoadInfo.MiscFlags = 0;
pLoadInfo.Format = DXGI_FORMAT_R8_UINT; // Monochrome, 8bit per pixel bitmap data
pLoadInfo.Filter = D3DX11_FILTER_NONE;
pLoadInfo.MipFilter = D3DX11_FILTER_NONE;
pLoadInfo.pSrcInfo = 0;
D3DX11CreateShaderResourceViewFromMemory(pd3dDevice, (LPCVOID) image.GetData(), size, &pLoadInfo, NULL, &g_Target, &result);
When I tried this, I would result = E_FAIL;
So my suspect would be the pSrcInfo, but I have no idea what it should, or if it is something totally different. I was searching for samples showing the ResourceViewFromMemory, but this seems to be a hardly used function.
Thanks for any feedback!
Edit: fixed the pixelformat, wrote R8G8 by accident
[DX11] CreateShaderResourceViewFromMemory
D3DX11CreateShaderResourceViewFrom* functions are intended for loading an image file and creating a texture from it, not for taking a raw block of data and filling a texture from it. In the case of D3DX11CreateShaderResourceViewFromMemory, it expects an array containing raw data from an image file such as .PNG, or .BMP.
For what you're attempting to do, you shouldn't need this helper function at all. Simply use ID3D11Device::CreateTexture2D and fill the D3D11_TEXTURE2D_DESC with your desired settings. If you have initial data you want to fill the texture with, pass the pointer to that data by filling out the D3D11_SUBRESOURCE_DATA structure and passing it as the pInitialData parameter. Then when you want to change the contents of that texture at runtime, you call ID3D11DeviceContext::Map.
To use the texture as a shader resource, you just create a shader resource view. If you just need to access the whole texture and don't need to do anything fancy, you can just pass NULL as the pDesc parameter.
For what you're attempting to do, you shouldn't need this helper function at all. Simply use ID3D11Device::CreateTexture2D and fill the D3D11_TEXTURE2D_DESC with your desired settings. If you have initial data you want to fill the texture with, pass the pointer to that data by filling out the D3D11_SUBRESOURCE_DATA structure and passing it as the pInitialData parameter. Then when you want to change the contents of that texture at runtime, you call ID3D11DeviceContext::Map.
To use the texture as a shader resource, you just create a shader resource view. If you just need to access the whole texture and don't need to do anything fancy, you can just pass NULL as the pDesc parameter.
Author
Arghs... I knew there must be something really wrong... Thanks for the input, I'll try that!
Author
Okay, I've got one more question.
The data I want to display is an unsigned 8bit monochrome image. Now I set the data, but it tells me .Sample() doesn't work for DXGI_FORMAT_R8_UINT. I tested pd3dDevice->CheckFormatSuppor(), and R8_UINT should be supported by Texture2D.
So what I tried instead of sample is getting the integer value from the texture via load()
Something like int r = g_txDiffuse.load( int3 ( asint(In.TextureUV.x), asint(In.TextureUV.y), 1));
This is wrong, I get an error when trying to load that into the pixel shader, but how am I supposed to get the int value from there?
The other issue I am having seems to be the fact that pd3dImmediateContext->Map() is not blocking, and when Draw() is called, the mapping hasn't finished yet. Can I make this call block until the upload is done somehow?
Also, I am unable to debug this in the first place, when running PIX the code fails at
D3DX11CompileFromFile( str, NULL, NULL, "RenderScenePS", "ps_4_0_level_9_3", dwShaderFlags, 0, NULL, &pPixelShaderBuffer, NULL, NULL )
which is expected since the pixel shader is still wrong, but how can I debug the shader aside from that?
Edit: okay, noticed that the asint() is failing, so I rewrote it to
int3 loc;
loc.x = (int)(In.TextureUV.x * 640);
loc.y = (int)(In.TextureUV.y * 480);
loc.z = 1;
unsigned int r = g_txDiffuse.load(loc);
with
Texture2D <unsigned int> g_txDiffuse : register( t0 );
but it's still failing to load.
The data I want to display is an unsigned 8bit monochrome image. Now I set the data, but it tells me .Sample() doesn't work for DXGI_FORMAT_R8_UINT. I tested pd3dDevice->CheckFormatSuppor(), and R8_UINT should be supported by Texture2D.
So what I tried instead of sample is getting the integer value from the texture via load()
Something like int r = g_txDiffuse.load( int3 ( asint(In.TextureUV.x), asint(In.TextureUV.y), 1));
This is wrong, I get an error when trying to load that into the pixel shader, but how am I supposed to get the int value from there?
The other issue I am having seems to be the fact that pd3dImmediateContext->Map() is not blocking, and when Draw() is called, the mapping hasn't finished yet. Can I make this call block until the upload is done somehow?
Also, I am unable to debug this in the first place, when running PIX the code fails at
D3DX11CompileFromFile( str, NULL, NULL, "RenderScenePS", "ps_4_0_level_9_3", dwShaderFlags, 0, NULL, &pPixelShaderBuffer, NULL, NULL )
which is expected since the pixel shader is still wrong, but how can I debug the shader aside from that?
Edit: okay, noticed that the asint() is failing, so I rewrote it to
int3 loc;
loc.x = (int)(In.TextureUV.x * 640);
loc.y = (int)(In.TextureUV.y * 480);
loc.z = 1;
unsigned int r = g_txDiffuse.load(loc);
with
Texture2D <unsigned int> g_txDiffuse : register( t0 );
but it's still failing to load.
Yeah you don't want asint. It takes the raw data and interprets it as integer, similiar if you were to this in c++:
You just want to cast to int, like you did in your revised code.
If compiliation fails to a shader compilation error, then you'll get the errors back in the buffer set for the ppErrorMsgs parameter. Pass it an ID3D10Blob, and then call GetBufferPointer on the blob and vast the void pointer to char* to get a string containing the shader compilation errors. This is the function I use:
You can obviously take out the exception stuff if you want, and if you're not using wstrings you can skip that MultiByteToWideChar stuff. If you're looking to debug shaders in PIX you'll definitely want to add in those debug flags like I did.
float fval = 1.0f;
int ival = *((int*)(&fval));
You just want to cast to int, like you did in your revised code.
If compiliation fails to a shader compilation error, then you'll get the errors back in the buffer set for the ppErrorMsgs parameter. Pass it an ID3D10Blob, and then call GetBufferPointer on the blob and vast the void pointer to char* to get a string containing the shader compilation errors. This is the function I use:
ID3D10Blob* CompileShader(LPCWSTR path,
LPCSTR functionName,
LPCSTR profile,
CONST D3D10_SHADER_MACRO* defines,
ID3D10Include* includes)
{
// Loop until we succeed, or an exception is thrown
while (true)
{
UINT flags = 0;
#ifdef _DEBUG
flags |= D3D10_SHADER_DEBUG|D3D10_SHADER_SKIP_OPTIMIZATION|D3D10_SHADER_WARNINGS_ARE_ERRORS;
#endif
ID3D10Blob* compiledShader;
ID3D10BlobPtr errorMessages;
HRESULT hr = D3DX11CompileFromFileW(path, defines, includes, functionName, profile,
flags, 0, NULL, &compiledShader, &errorMessages, NULL);
if (FAILED(hr))
{
if (errorMessages)
{
WCHAR message[1024];
message[0] = NULL;
char* blobdata = reinterpret_cast<char*>(errorMessages->GetBufferPointer());
MultiByteToWideChar(CP_ACP, 0, blobdata, static_cast<int>(errorMessages->GetBufferSize()), message, 1024);
std::wstring fullMessage = L"Error compiling shader file \"";
fullMessage += path;
fullMessage += L"\" - ";
fullMessage += message;
#ifdef _DEBUG
// Pop up a message box allowing user to retry compilation
int retVal = MessageBoxW(NULL, fullMessage.c_str(), L"Shader Compilation Error", MB_RETRYCANCEL);
if(retVal != IDRETRY)
throw DXException(hr, fullMessage.c_str());
#else
throw DXException(hr, fullMessage.c_str());
#endif
}
else
{
_ASSERT(false);
throw DXException(hr);
}
}
else
return compiledShader;
}
}
You can obviously take out the exception stuff if you want, and if you're not using wstrings you can skip that MultiByteToWideChar stuff. If you're looking to debug shaders in PIX you'll definitely want to add in those debug flags like I did.
Author
Awesome, thanks a lot. This helps a lot, turns out, it didn't load, because I used
Texture.load, instead of Texture.Load, so it failed on capitalization!
Texture.load, instead of Texture.Load, so it failed on capitalization!
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
