D3D12阴影渲染报错：采样数/质量与渲染目标格式不兼容

错误日志 #

D3D12 ERROR: ID3D12Device::CreateGraphicsPipelineState: The specified sample count or quality is not supported with the render target format in slot 1 [ STATE_CREATION ERROR #685: CREATEGRAPHICSPIPELINESTATE_INVALID_SAMPLE_DESC]
D3D12 ERROR: ID3D12Device::CreateGraphicsPipelineState: The specified sample count or quality is not supported with the render target format in slot 2 [ STATE_CREATION ERROR #685: CREATEGRAPHICSPIPELINESTATE_INVALID_SAMPLE_DESC]
D3D12 ERROR: ID3D12Device::CreateGraphicsPipelineState: The specified sample count or quality is not supported with the render target format in slot 3 [ STATE_CREATION ERROR #685: CREATEGRAPHICSPIPELINESTATE_INVALID_SAMPLE_DESC]
D3D12 ERROR: ID3D12Device::CreateGraphicsPipelineState: The specified sample count or quality is not supported with the render target format in slot 4 [ STATE_CREATION ERROR #685: CREATEGRAPHICSPIPELINESTATE_INVALID_SAMPLE_DESC]

问题根源 #

1. 渲染目标数量与格式数组不匹配：
   在Stage::BuildObjects中创建Shader时，统一传入了nRenderTargets=5，但：

   • m_pDepthRenderShader对应的RtvFormats数组仅包含1个格式值，循环赋值时i=1到4会访问数组越界，导致这些位置的格式为内存垃圾值，无效格式无法支持采样数1的配置。
   • m_pShadowShader和m_pShadowMapToViewport传入pdxgiRtvFormats=NULL，会默认将5个渲染目标都设为DXGI_FORMAT_R8G8B8A8_UNORM，但这类Shader实际不需要这么多渲染目标，且格式与后续绑定的资源不匹配。

2. 采样数与格式兼容性：
   越界产生的无效渲染目标格式，本身不支持SampleDesc.Count=1的采样配置，触发D3D12格式兼容性检查错误。

解决方案 #

1. 修正渲染目标数量参数 #

根据每个Shader的实际需求传递正确的nRenderTargets值：

• m_pDepthRenderShader仅需要1个渲染目标，将参数从5改为1；
• m_pShadowShader仅用于深度渲染（无颜色渲染目标输出），将nRenderTargets设为0；
• m_pShadowMapToViewport输出到单个视口，将nRenderTargets设为1。

2. 修改后的代码示例 #

m_ppShaders[0] = pBoxShader;

m_pDepthRenderShader = new CDepthRenderShader(pBoxShader, m_pLights);
DXGI_FORMAT RtvFormats[1] = { DXGI_FORMAT_R32_FLOAT };
// 修正nRenderTargets为1
m_pDepthRenderShader->CreateShader(pd3dDevice, m_pd3dGraphicsRootSignature, D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE, 1, RtvFormats, DXGI_FORMAT_D32_FLOAT);
m_pDepthRenderShader->BuildObjects(pd3dDevice, pd3dCommandList, NULL);

m_pShadowShader = new CShadowMapShader(pBoxShader);
// 修正nRenderTargets为0（无颜色渲染目标）
m_pShadowShader->CreateShader(pd3dDevice, m_pd3dGraphicsRootSignature, D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE, 0, NULL, DXGI_FORMAT_D24_UNORM_S8_UINT);
m_pShadowShader->BuildObjects(pd3dDevice, pd3dCommandList, m_pDepthRenderShader->GetDepthTexture());

m_pShadowMapToViewport = new CTextureToViewportShader();
// 修正nRenderTargets为1（单个视口输出）
m_pShadowMapToViewport->CreateShader(pd3dDevice, m_pd3dGraphicsRootSignature, D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE, 1, NULL, DXGI_FORMAT_D24_UNORM_S8_UINT);
m_pShadowMapToViewport->BuildObjects(pd3dDevice, pd3dCommandList, m_pDepthRenderShader->GetDepthTexture());

3. 额外检查 #

确保CreateShader函数中，当pdxgiRtvFormats不为NULL时，数组长度与nRenderTargets严格匹配，彻底避免越界访问问题。

内容的提问来源于stack exchange，提问作者yoon-0012