NVAPI improvements (#1512)

* First pass at incorporating nvapi into test harness.

* D3d12 Atomic Float Add via NVAPI working

* Dx12 atomic float appears to work.

* Atomic float add on Dx12.

* Added atomic64 feature addition to vk.
Fix correct output for atomic-float-byte-address.slang

* Disable atomic float failing tests.

* Upgraded VK headers.

* Detect atomic float availability on VK.

* Try to get test working for in64 atomic.

* Made HLSL prelude controlled via the render-test requirements.

* Added -enable-nvapi to premake.

* Fix D3D12Renderer when NVAPI is not available.

* Small improvements to VKRenderer.

* Improve atomic documentation in target-compatibility.md.

* Fixed NVAPI working on D3D12.

* Test for specific NVAPI features.

* Remove requiredFeatures from Renderer::Desc as was ignored. Tried to document more around nvapiExtnSlot.

* Readded requiredFeatures to Renderer::Desc

* Improve comments in the tests.

5 files changed, 83 insertions, 39 deletions

diff --git a/tools/gfx/d3d11/render-d3d11.cpp b/tools/gfx/d3d11/render-d3d11.cpp
index 8eafd24b4..30c982ab3 100644
--- a/tools/gfx/d3d11/render-d3d11.cpp
+++ b/tools/gfx/d3d11/render-d3d11.cpp
@@ -514,6 +514,19 @@ D3D11Renderer::ScopeNVAPI::~ScopeNVAPI()
 
 // !!!!!!!!!!!!!!!!!!!!!!!!!!!! Renderer interface !!!!!!!!!!!!!!!!!!!!!!!!!!
 
+static bool _isSupportedNVAPIOp(IUnknown* dev, uint32_t op)
+{
+#ifdef GFX_NVAPI
+    {
+        bool isSupported;
+        NvAPI_Status status = NvAPI_D3D11_IsNvShaderExtnOpCodeSupported(dev, NvU32(op), &isSupported);
+        return status == NVAPI_OK && isSupported;
+    }
+#else
+    return false;
+#endif
+}
+
 SlangResult D3D11Renderer::initialize(const Desc& desc, void* inWindowHandle)
 {
     auto windowHandle = (HWND)inWindowHandle;
@@ -647,29 +660,32 @@ SlangResult D3D11Renderer::initialize(const Desc& desc, void* inWindowHandle)
     }
 
     // NVAPI
+    if (desc.nvapiExtnSlot >= 0)
     {
-        const char* features[] = { "nvapi", "atomic-float", "atomic-int64" };
-        bool needsNvapi = false;
-        for (Index i = 0; i < SLANG_COUNT_OF(features); ++i)
+        if (SLANG_FAILED(NVAPIUtil::initialize()))
         {
-            if (desc.requiredFeatures.indexOf(features[i]) >= 0)
-            {
-                needsNvapi = true;
-                break;
-            }
+            return SLANG_E_NOT_AVAILABLE;
         }
 
-        if (needsNvapi && SLANG_SUCCEEDED(NVAPIUtil::initialize()))
+#ifdef GFX_NVAPI
+        if (NvAPI_D3D11_SetNvShaderExtnSlot(m_device, NvU32(desc.nvapiExtnSlot)) != NVAPI_OK)
         {
-            // TODO(JS): We should test for specific features here.
-            for (Index i = 0; i < SLANG_COUNT_OF(features); ++i)
-            {
-                m_features.add(features[i]);
-            }
-            m_nvapi = true;
+            return SLANG_E_NOT_AVAILABLE;
         }
-    }
 
+        if (_isSupportedNVAPIOp(m_device, NV_EXTN_OP_UINT64_ATOMIC ))
+        {
+            m_features.add("atomic-int64");
+        }
+        if (_isSupportedNVAPIOp(m_device, NV_EXTN_OP_FP32_ATOMIC))
+        {
+            m_features.add("atomic-float");
+        }
+
+        m_nvapi = true;
+#endif
+    }
+    
     // TODO: Add support for debugging to help detect leaks:
     //
     //      ComPtr<ID3D11Debug> gDebug;
diff --git a/tools/gfx/d3d12/render-d3d12.cpp b/tools/gfx/d3d12/render-d3d12.cpp
index ca44aa04d..0f23d8dd4 100644
--- a/tools/gfx/d3d12/render-d3d12.cpp
+++ b/tools/gfx/d3d12/render-d3d12.cpp
@@ -1523,6 +1523,19 @@ Result D3D12Renderer::_createDevice(DeviceCheckFlags deviceCheckFlags, const Uno
     return SLANG_OK;
 }
 
+static bool _isSupportedNVAPIOp(ID3D12Device* dev, uint32_t op)
+{
+#ifdef GFX_NVAPI
+    {
+        bool isSupported;
+        NvAPI_Status status = NvAPI_D3D12_IsNvShaderExtnOpCodeSupported(dev, NvU32(op), &isSupported);
+        return status == NVAPI_OK && isSupported;
+    }
+#else
+    return false;
+#endif
+}
+
 Result D3D12Renderer::initialize(const Desc& desc, void* inWindowHandle)
 {
     m_hwnd = (HWND)inWindowHandle;
@@ -1602,32 +1615,40 @@ Result D3D12Renderer::initialize(const Desc& desc, void* inWindowHandle)
         return SLANG_FAIL;
     }
 
+    // Set the device
+    m_device = m_deviceInfo.m_device;
+
     // NVAPI
+    if (desc.nvapiExtnSlot >= 0)
     {
-        const char* features[] = { "nvapi", "atomic-float", "atomic-int64" };
-        bool needsNvapi = false;
-        for (Index i = 0; i < SLANG_COUNT_OF(features); ++i)
+        if (SLANG_FAILED(NVAPIUtil::initialize()))
         {
-            if (desc.requiredFeatures.indexOf(features[i]) >= 0)
-            {
-                needsNvapi = true;
-                break;
-            }
+            return SLANG_E_NOT_AVAILABLE;
         }
 
-        if (needsNvapi && SLANG_SUCCEEDED(NVAPIUtil::initialize()))
+#ifdef GFX_NVAPI
+        // From DOCS: Applications are expected to bind null UAV to this slot.
+        // NOTE! We don't currently do this, but doesn't seem to be a problem.
+
+        const NvAPI_Status status = NvAPI_D3D12_SetNvShaderExtnSlotSpace(m_device, NvU32(desc.nvapiExtnSlot), NvU32(0));
+        
+        if (status != NVAPI_OK)
         {
-            // TODO(JS): We should test for specific features here.
-            for (Index i = 0; i < SLANG_COUNT_OF(features); ++i)
-            {
-                m_features.add(features[i]);
-            }
-            m_nvapi = true;
+            return SLANG_E_NOT_AVAILABLE;
         }
-    }
 
-    // Set the device
-    m_device = m_deviceInfo.m_device;
+        if (_isSupportedNVAPIOp(m_device, NV_EXTN_OP_UINT64_ATOMIC))
+        {
+            m_features.add("atomic-int64");
+        }
+        if (_isSupportedNVAPIOp(m_device, NV_EXTN_OP_FP32_ATOMIC))
+        {
+            m_features.add("atomic-float");
+        }
+
+        m_nvapi = true;
+#endif
+    }
 
     // Find what features are supported
     {
diff --git a/tools/gfx/nvapi/nvapi-include.h b/tools/gfx/nvapi/nvapi-include.h
index e3674af95..c213e0bfb 100644
--- a/tools/gfx/nvapi/nvapi-include.h
+++ b/tools/gfx/nvapi/nvapi-include.h
@@ -15,5 +15,7 @@
 #   endif
 
 #   include <nvapi.h>
+#   include <nvShaderExtnEnums.h>
+
 #endif
 
diff --git a/tools/gfx/open-gl/render-gl.cpp b/tools/gfx/open-gl/render-gl.cpp
index 5f16f07bc..caf8794c0 100644
--- a/tools/gfx/open-gl/render-gl.cpp
+++ b/tools/gfx/open-gl/render-gl.cpp
@@ -717,9 +717,12 @@ SlangResult GLRenderer::initialize(const Desc& desc, void* inWindowHandle)
         }
     }
 
-    if (m_desc.requiredFeatures.indexOf("nvapi") >= 0 && SLANG_SUCCEEDED(NVAPIUtil::initialize()))
+    if (m_desc.nvapiExtnSlot >= 0)
     {
-        m_features.add("nvapi");
+        if (SLANG_FAILED(NVAPIUtil::initialize()))
+        {
+            return SLANG_E_NOT_AVAILABLE;
+        }
     }
 
     auto extensions = glGetString(GL_EXTENSIONS);
diff --git a/tools/gfx/render.h b/tools/gfx/render.h
index 12ef1a9e9..55beb5774 100644
--- a/tools/gfx/render.h
+++ b/tools/gfx/render.h
@@ -800,12 +800,14 @@ public:
 
     struct Desc
     {
-        int width;                                      ///< Width in pixels
-        int height;                                     ///< height in pixels
+        int width = 0;                                  ///< Width in pixels
+        int height = 0;                                 ///< height in pixels
         Slang::String adapter;                          ///< Name to identify the adapter to use
-        Slang::List<Slang::String> requiredFeatures;    ///< The features enabled on this renderer
+        Slang::List<Slang::String> requiredFeatures;    ///< List of required feature names. 
+        int nvapiExtnSlot = -1;                         ///< The slot (typically UAV) used to identify NVAPI intrinsics. If >=0 NVAPI is required.
     };
 
+        // Will return with SLANG_E_NOT_AVAILABLE if NVAPI can't be initialized and nvapiExtnSlot >= 0
     virtual SlangResult initialize(const Desc& desc, void* inWindowHandle) = 0;
 
     bool hasFeature(const Slang::UnownedStringSlice& feature) { return getFeatures().indexOf(Slang::String(feature)) != Slang::Index(-1); }