Some small fixes with Windows/DX usage (#2797)

* Correct case of windows.h includes

* Use Slang::SharedLibrary to load directx dlls

* s/max/std::max/

* Factor common OS code in calcHasApi

* Add DXIL test for compute/simple

* s/false/FALSE for calls to WinAPI functions

* Factor common OS code in gfxGetAdapters

* 2 missing headers

d3d12sdklayers for ID3DDebug

climits for UINT_MAX

* Define out unused function on Linux

* Only try to load Vulkan and CUDA on Windows or Linux

* simplify D3DUtil::getDxgiModule

* Remove WIN32_LEAN_AND_MEAN &co from source files

Add a global define

* Set WIN32_LEAN_AND_MEAN &friends in headers

Restore previous state also

* regenerate vs projects

21 files changed, 98 insertions, 68 deletions

diff --git a/tools/gfx-unit-test/gfx-test-texture-util.cpp b/tools/gfx-unit-test/gfx-test-texture-util.cpp
index ff1ef39d0..ed651475d 100644
--- a/tools/gfx-unit-test/gfx-test-texture-util.cpp
+++ b/tools/gfx-unit-test/gfx-test-texture-util.cpp
@@ -21,8 +21,7 @@
 
 #if GFX_ENABLE_RENDERDOC_INTEGRATION
 #    include "external/renderdoc_app.h"
-#    define WIN32_LEAN_AND_MEAN
-#    include <Windows.h>
+#    include <windows.h>
 #endif
 
 using namespace Slang;
diff --git a/tools/gfx-unit-test/gfx-test-util.cpp b/tools/gfx-unit-test/gfx-test-util.cpp
index fcebdda9f..298283a4a 100644
--- a/tools/gfx-unit-test/gfx-test-util.cpp
+++ b/tools/gfx-unit-test/gfx-test-util.cpp
@@ -7,8 +7,7 @@
 
 #if GFX_ENABLE_RENDERDOC_INTEGRATION
 #    include "external/renderdoc_app.h"
-#    define WIN32_LEAN_AND_MEAN
-#    include <Windows.h>
+#    include <windows.h>
 #endif
 
 using Slang::ComPtr;
diff --git a/tools/gfx/d3d/d3d-util.cpp b/tools/gfx/d3d/d3d-util.cpp
index 05a948773..eece71c16 100644
--- a/tools/gfx/d3d/d3d-util.cpp
+++ b/tools/gfx/d3d/d3d-util.cpp
@@ -10,6 +10,7 @@
 #include <stdio.h>
 
 #include "core/slang-basic.h"
+#include "core/slang-platform.h"
 
 namespace gfx {
 using namespace Slang;
@@ -498,15 +499,22 @@ bool D3DUtil::isTypeless(DXGI_FORMAT format)
     }
 }
 
-/* static */HMODULE D3DUtil::getDxgiModule()
+/* static */SharedLibrary::Handle D3DUtil::getDxgiModule()
 {
-    static HMODULE s_dxgiModule = LoadLibraryA("dxgi.dll");
-    if (!s_dxgiModule)
-    {
-        fprintf(stderr, "error: failed load 'dxgi.dll'\n");
-        return nullptr;
-    }
-
+#if SLANG_WINDOWS_FAMILY
+    const char* const libPath = "dxgi";
+#else
+    const char* const libPath = "dxvk_dxgi";
+#endif
+    static SharedLibrary::Handle s_dxgiModule = [&](){
+        SharedLibrary::Handle h = nullptr;
+        SharedLibrary::load(libPath, h);
+        if (!h)
+        {
+            fprintf(stderr, "error: failed to load dll '%s'\n", libPath);
+        }
+        return h;
+    }();
     return s_dxgiModule;
 }
 
@@ -522,7 +530,7 @@ bool D3DUtil::isTypeless(DXGI_FORMAT format)
     typedef HRESULT(WINAPI *PFN_DXGI_CREATE_FACTORY_2)(UINT Flags, REFIID riid, _COM_Outptr_ void **ppFactory);
 
     {
-        auto createFactory2 = (PFN_DXGI_CREATE_FACTORY_2)::GetProcAddress(dxgiModule, "CreateDXGIFactory2");
+        auto createFactory2 = (PFN_DXGI_CREATE_FACTORY_2)SharedLibrary::findSymbolAddressByName(dxgiModule, "CreateDXGIFactory2");
         if (createFactory2)
         {
             UINT dxgiFlags = 0;
@@ -541,7 +549,7 @@ bool D3DUtil::isTypeless(DXGI_FORMAT format)
     }
 
     {
-        auto createFactory = (PFN_DXGI_CREATE_FACTORY)::GetProcAddress(dxgiModule, "CreateDXGIFactory");
+        auto createFactory = (PFN_DXGI_CREATE_FACTORY)SharedLibrary::findSymbolAddressByName(dxgiModule, "CreateDXGIFactory");
         if (!createFactory)
         {
             fprintf(stderr, "error: failed load symbol '%s'\n", "CreateDXGIFactory");
diff --git a/tools/gfx/d3d/d3d-util.h b/tools/gfx/d3d/d3d-util.h
index 617ae4af2..1869d7de6 100644
--- a/tools/gfx/d3d/d3d-util.h
+++ b/tools/gfx/d3d/d3d-util.h
@@ -7,13 +7,14 @@
 
 #include "slang-com-ptr.h"
 #include "core/slang-basic.h"
+#include "core/slang-platform.h"
 
 #include "../flag-combiner.h"
 
 #include "slang-gfx.h"
 
-#include <D3Dcommon.h>
-#include <DXGIFormat.h>
+#include <d3dcommon.h>
+#include <dxgiformat.h>
 #include <dxgi.h>
 #include <d3d12.h>
 
@@ -82,7 +83,7 @@ class D3DUtil
     static SlangResult createFactory(DeviceCheckFlags flags, Slang::ComPtr<IDXGIFactory>& outFactory);
 
         /// Get the dxgiModule
-    static HMODULE getDxgiModule();
+    static Slang::SharedLibrary::Handle getDxgiModule();
 
         /// Find adapters
     static SlangResult findAdapters(DeviceCheckFlags flags, const AdapterLUID* adapterLUID, IDXGIFactory* dxgiFactory, Slang::List<Slang::ComPtr<IDXGIAdapter>>& dxgiAdapters);
diff --git a/tools/gfx/d3d11/d3d11-base.h b/tools/gfx/d3d11/d3d11-base.h
index d47682f18..ce38bb174 100644
--- a/tools/gfx/d3d11/d3d11-base.h
+++ b/tools/gfx/d3d11/d3d11-base.h
@@ -13,11 +13,15 @@
 #include "slang-com-ptr.h"
 #include "../flag-combiner.h"
 
-#define WIN32_LEAN_AND_MEAN
-#define NOMINMAX
-#include <Windows.h>
+#pragma push_macro("WIN32_LEAN_AND_MEAN")
+#pragma push_macro("NOMINMAX")
 #undef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN
 #undef NOMINMAX
+#define NOMINMAX
+#include <windows.h>
+#pragma pop_macro("NOMINMAX")
+#pragma pop_macro("WIN32_LEAN_AND_MEAN")
 
 #include <d3d11_2.h>
 #include <d3dcompiler.h>
diff --git a/tools/gfx/d3d11/d3d11-device.cpp b/tools/gfx/d3d11/d3d11-device.cpp
index cc2eda089..e6a348dc1 100644
--- a/tools/gfx/d3d11/d3d11-device.cpp
+++ b/tools/gfx/d3d11/d3d11-device.cpp
@@ -46,15 +46,20 @@ SlangResult DeviceImpl::initialize(const Desc& desc)
     m_desc = desc;
 
     // Rather than statically link against D3D, we load it dynamically.
-    HMODULE d3dModule = LoadLibraryA("d3d11.dll");
-    if (!d3dModule)
+    SharedLibrary::Handle d3dModule;
+#if SLANG_WINDOWS_FAMILY
+    const char* libName = "d3d11";
+#else
+    const char* libName = "dxvk_d3d11";
+#endif
+    if (SLANG_FAILED(SharedLibrary::load(libName, d3dModule)))
     {
-        fprintf(stderr, "error: failed load 'd3d11.dll'\n");
+        fprintf(stderr, "error: failed to load '%s'\n", libName);
         return SLANG_FAIL;
     }
 
     PFN_D3D11_CREATE_DEVICE_AND_SWAP_CHAIN D3D11CreateDeviceAndSwapChain_ =
-        (PFN_D3D11_CREATE_DEVICE_AND_SWAP_CHAIN)GetProcAddress(d3dModule, "D3D11CreateDeviceAndSwapChain");
+        (PFN_D3D11_CREATE_DEVICE_AND_SWAP_CHAIN)SharedLibrary::findSymbolAddressByName(d3dModule, "D3D11CreateDeviceAndSwapChain");
     if (!D3D11CreateDeviceAndSwapChain_)
     {
         fprintf(stderr,
@@ -63,7 +68,7 @@ SlangResult DeviceImpl::initialize(const Desc& desc)
     }
 
     PFN_D3D11_CREATE_DEVICE D3D11CreateDevice_ =
-        (PFN_D3D11_CREATE_DEVICE)GetProcAddress(d3dModule, "D3D11CreateDevice");
+        (PFN_D3D11_CREATE_DEVICE)SharedLibrary::findSymbolAddressByName(d3dModule, "D3D11CreateDevice");
     if (!D3D11CreateDevice_)
     {
         fprintf(stderr,
diff --git a/tools/gfx/d3d11/d3d11-helper-functions.cpp b/tools/gfx/d3d11/d3d11-helper-functions.cpp
index 4e1df9879..6774b7d93 100644
--- a/tools/gfx/d3d11/d3d11-helper-functions.cpp
+++ b/tools/gfx/d3d11/d3d11-helper-functions.cpp
@@ -337,7 +337,7 @@ namespace d3d11
             default: assert(!"Unknown dimension");
             }
 
-            descOut.Texture2DArray.ArraySize = max(textureDesc.size.depth, arraySize);
+            descOut.Texture2DArray.ArraySize = std::max(textureDesc.size.depth, arraySize);
             descOut.Texture2DArray.MostDetailedMip = 0;
             descOut.Texture2DArray.MipLevels = textureDesc.numMipLevels;
             descOut.Texture2DArray.FirstArraySlice = 0;
diff --git a/tools/gfx/d3d12/d3d12-base.h b/tools/gfx/d3d12/d3d12-base.h
index 2446151a2..52e8d4623 100644
--- a/tools/gfx/d3d12/d3d12-base.h
+++ b/tools/gfx/d3d12/d3d12-base.h
@@ -14,12 +14,19 @@
 #include "d3d12-descriptor-heap.h"
 #include "d3d12-resource.h"
 
-#define _CRT_SECURE_NO_WARNINGS
-#define WIN32_LEAN_AND_MEAN
-#define NOMINMAX
-#include <Windows.h>
+#pragma push_macro("WIN32_LEAN_AND_MEAN")
+#pragma push_macro("NOMINMAX")
+#pragma push_macro("_CRT_SECURE_NO_WARNINGS")
 #undef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN
 #undef NOMINMAX
+#define NOMINMAX
+#undef _CRT_SECURE_NO_WARNINGS
+#define _CRT_SECURE_NO_WARNINGS
+#include <windows.h>
+#pragma pop_macro("_CRT_SECURE_NO_WARNINGS")
+#pragma pop_macro("NOMINMAX")
+#pragma pop_macro("WIN32_LEAN_AND_MEAN")
 
 #include <d3d12.h>
 #include <dxgi1_4.h>
diff --git a/tools/gfx/d3d12/d3d12-device.cpp b/tools/gfx/d3d12/d3d12-device.cpp
index 312c81d75..9bf54a01b 100644
--- a/tools/gfx/d3d12/d3d12-device.cpp
+++ b/tools/gfx/d3d12/d3d12-device.cpp
@@ -405,8 +405,13 @@ Result DeviceImpl::initialize(const Desc& desc)
 
     // Rather than statically link against D3D, we load it dynamically.
 
-    HMODULE d3dModule = LoadLibraryA("d3d12.dll");
-    if (!d3dModule)
+    SharedLibrary::Handle d3dModule;
+#if SLANG_WINDOWS_FAMILY
+    const char* libName = "d3d11";
+#else
+    const char* libName = "vkd3d-proton-d3d12";
+#endif
+    if (SLANG_FAILED(SharedLibrary::load(libName, d3dModule)))
     {
         getDebugCallback()->handleMessage(
             DebugMessageType::Error, DebugMessageSource::Layer, "error: failed load 'd3d12.dll'\n");
@@ -1990,7 +1995,7 @@ void DeviceImpl::submitResourceCommandsAndWait(const DeviceImpl::ResourceCommand
     m_resourceCommandTransientHeap->synchronizeAndReset();
 }
 
-void DeviceImpl::processExperimentalFeaturesDesc(void* d3dModule, void* inDesc)
+void DeviceImpl::processExperimentalFeaturesDesc(SharedLibrary::Handle d3dModule, void* inDesc)
 {
     typedef HRESULT(WINAPI* PFN_D3D12_ENABLE_EXPERIMENTAL_FEATURES)(
         UINT      NumFeatures,
@@ -2002,7 +2007,7 @@ void DeviceImpl::processExperimentalFeaturesDesc(void* d3dModule, void* inDesc)
     D3D12ExperimentalFeaturesDesc desc = {};
     memcpy(&desc, inDesc, sizeof(desc));
     auto enableExperimentalFeaturesFunc =
-        (PFN_D3D12_ENABLE_EXPERIMENTAL_FEATURES)loadProc((HMODULE)d3dModule, "D3D12EnableExperimentalFeatures");
+        (PFN_D3D12_ENABLE_EXPERIMENTAL_FEATURES)loadProc(d3dModule, "D3D12EnableExperimentalFeatures");
     if (!enableExperimentalFeaturesFunc)
     {
         getDebugCallback()->handleMessage(
@@ -2173,9 +2178,9 @@ Result DeviceImpl::createCommandQueueImpl(CommandQueueImpl** outQueue)
     return SLANG_OK;
 }
 
-PROC DeviceImpl::loadProc(HMODULE module, char const* name)
+void* DeviceImpl::loadProc(SharedLibrary::Handle module, char const* name)
 {
-    PROC proc = ::GetProcAddress(module, name);
+    void* proc = SharedLibrary::findSymbolAddressByName(module, name);
     if (!proc)
     {
         fprintf(stderr, "error: failed load symbol '%s'\n", name);
diff --git a/tools/gfx/d3d12/d3d12-device.h b/tools/gfx/d3d12/d3d12-device.h
index fb54e5543..e77693ebe 100644
--- a/tools/gfx/d3d12/d3d12-device.h
+++ b/tools/gfx/d3d12/d3d12-device.h
@@ -6,6 +6,7 @@
 #include "d3d12-texture.h"
 
 #include <d3d12.h>
+#include <d3d12sdklayers.h>
 
 namespace gfx
 {
@@ -207,7 +208,7 @@ public:
         const RayTracingPipelineStateDesc& desc, IPipelineState** outState) override;
 
 public:
-    static PROC loadProc(HMODULE module, char const* name);
+    static void* loadProc(SharedLibrary::Handle module, char const* name);
 
     Result createCommandQueueImpl(CommandQueueImpl** outQueue);
 
@@ -248,7 +249,7 @@ public:
     ResourceCommandRecordInfo encodeResourceCommands();
     void submitResourceCommandsAndWait(const ResourceCommandRecordInfo& info);
 private:
-    void processExperimentalFeaturesDesc(void* d3dModule, void* desc);
+    void processExperimentalFeaturesDesc(SharedLibrary::Handle d3dModule, void* desc);
 };
 
 } // namespace d3d12
diff --git a/tools/gfx/d3d12/d3d12-pipeline-state.cpp b/tools/gfx/d3d12/d3d12-pipeline-state.cpp
index 35313f676..754c07e5c 100644
--- a/tools/gfx/d3d12/d3d12-pipeline-state.cpp
+++ b/tools/gfx/d3d12/d3d12-pipeline-state.cpp
@@ -11,6 +11,8 @@
 #include "d3d12-shader-program.h"
 #include "d3d12-vertex-layout.h"
 
+#include <climits>
+
 namespace gfx
 {
 namespace d3d12
diff --git a/tools/gfx/d3d12/d3d12-query.cpp b/tools/gfx/d3d12/d3d12-query.cpp
index 319976bbc..6b2e92980 100644
--- a/tools/gfx/d3d12/d3d12-query.cpp
+++ b/tools/gfx/d3d12/d3d12-query.cpp
@@ -73,7 +73,7 @@ Result QueryPoolImpl::init(const IQueryPool::Desc& desc, DeviceImpl* device)
     m_commandQueue = device->m_resourceCommandQueue->m_d3dQueue;
 
     // Create wait event.
-    m_waitEvent = CreateEventEx(nullptr, false, 0, EVENT_ALL_ACCESS);
+    m_waitEvent = CreateEventEx(nullptr, FALSE, 0, EVENT_ALL_ACCESS);
 
     return SLANG_OK;
 }
diff --git a/tools/gfx/d3d12/d3d12-resource.h b/tools/gfx/d3d12/d3d12-resource.h
index ab9244470..cd56793cc 100644
--- a/tools/gfx/d3d12/d3d12-resource.h
+++ b/tools/gfx/d3d12/d3d12-resource.h
@@ -1,11 +1,15 @@
 // d3d12-resource.h
 #pragma once
 
-#define WIN32_LEAN_AND_MEAN
-#define NOMINMAX
-#include <Windows.h>
+#pragma push_macro("WIN32_LEAN_AND_MEAN")
+#pragma push_macro("NOMINMAX")
 #undef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN
 #undef NOMINMAX
+#define NOMINMAX
+#include <windows.h>
+#pragma pop_macro("NOMINMAX")
+#pragma pop_macro("WIN32_LEAN_AND_MEAN")
 
 #include <dxgi1_4.h>
 #include <d3d12.h>
diff --git a/tools/gfx/d3d12/d3d12-swap-chain.cpp b/tools/gfx/d3d12/d3d12-swap-chain.cpp
index 39a4565e0..57825a323 100644
--- a/tools/gfx/d3d12/d3d12-swap-chain.cpp
+++ b/tools/gfx/d3d12/d3d12-swap-chain.cpp
@@ -26,7 +26,7 @@ Result SwapchainImpl::init(
     {
         m_frameEvents.add(CreateEventEx(
             nullptr,
-            false,
+            FALSE,
             CREATE_EVENT_INITIAL_SET | CREATE_EVENT_MANUAL_RESET,
             EVENT_ALL_ACCESS));
     }
diff --git a/tools/gfx/d3d12/d3d12-transient-heap.cpp b/tools/gfx/d3d12/d3d12-transient-heap.cpp
index 11b03dc12..9139f6d8a 100644
--- a/tools/gfx/d3d12/d3d12-transient-heap.cpp
+++ b/tools/gfx/d3d12/d3d12-transient-heap.cpp
@@ -64,7 +64,7 @@ TransientResourceHeapImpl::QueueWaitInfo& TransientResourceHeapImpl::getQueueWai
     for (auto i = oldCount; i < m_waitInfos.getCount(); i++)
     {
         m_waitInfos[i].waitValue = 0;
-        m_waitInfos[i].fenceEvent = CreateEventEx(nullptr, false, 0, EVENT_ALL_ACCESS);
+        m_waitInfos[i].fenceEvent = CreateEventEx(nullptr, FALSE, 0, EVENT_ALL_ACCESS);
     }
     return m_waitInfos[queueIndex];
 }
diff --git a/tools/gfx/nvapi/nvapi-include.h b/tools/gfx/nvapi/nvapi-include.h
index c213e0bfb..513977048 100644
--- a/tools/gfx/nvapi/nvapi-include.h
+++ b/tools/gfx/nvapi/nvapi-include.h
@@ -7,11 +7,15 @@
 // On windows if we include NVAPI, we must include windows.h first
 
 #   ifdef _WIN32
+#       pragma push_macro("WIN32_LEAN_AND_MEAN")
+#       pragma push_macro("NOMINMAX")
+#       undef WIN32_LEAN_AND_MEAN
 #       define WIN32_LEAN_AND_MEAN
+#       undef NOMINMAX
 #       define NOMINMAX
-#       include <Windows.h>
-#       undef WIN32_LEAN_AND_MEAN
-#   undef NOMINMAX
+#       include <windows.h>
+#       pragma pop_macro("NOMINMAX")
+#       pragma pop_macro("WIN32_LEAN_AND_MEAN")
 #   endif
 
 #   include <nvapi.h>
diff --git a/tools/gfx/open-gl/render-gl.cpp b/tools/gfx/open-gl/render-gl.cpp
index 9167266e9..057828d58 100644
--- a/tools/gfx/open-gl/render-gl.cpp
+++ b/tools/gfx/open-gl/render-gl.cpp
@@ -22,11 +22,8 @@
 // TODO(tfoley): eventually we should be able to run these
 // tests on non-Windows targets to confirm that cross-compilation
 // at least *works* on those platforms...
-#define WIN32_LEAN_AND_MEAN
-#define NOMINMAX
-#include <Windows.h>
-#undef WIN32_LEAN_AND_MEAN
-#undef NOMINMAX
+
+#include <windows.h>
 
 #ifdef _MSC_VER
 #include <stddef.h>
diff --git a/tools/gfx/render.cpp b/tools/gfx/render.cpp
index 048eb8224..2f10fa67c 100644
--- a/tools/gfx/render.cpp
+++ b/tools/gfx/render.cpp
@@ -256,13 +256,9 @@ extern "C"
             break;
         case DeviceType::OpenGl:
             return SLANG_E_NOT_IMPLEMENTED;
-        case DeviceType::Vulkan:
-            SLANG_RETURN_ON_FAIL(getVKAdapters(adapters));
-            break;
-        case DeviceType::CUDA:
-            SLANG_RETURN_ON_FAIL(getCUDAAdapters(adapters));
-            break;
-#elif SLANG_LINUX_FAMILY && !defined(__CYGWIN__)
+#endif
+#if SLANG_WINDOWS_FAMILY || SLANG_LINUX_FAMILY
+        // Assume no Vulkan or CUDA on MacOS or Cygwin
         case DeviceType::Vulkan:
             SLANG_RETURN_ON_FAIL(getVKAdapters(adapters));
             break;
@@ -304,10 +300,6 @@ extern "C"
             {
                 return createVKDevice(desc, outDevice);
             }
-        case DeviceType::CUDA:
-            {
-                return createCUDADevice(desc, outDevice);
-            }
         case DeviceType::Default:
             {
                 IDevice::Desc newDesc = *desc;
@@ -332,11 +324,11 @@ extern "C"
         {
             return createVKDevice(desc, outDevice);
         }
+#endif
         case DeviceType::CUDA:
         {
             return createCUDADevice(desc, outDevice);
         }
-#endif
         case DeviceType::CPU:
             {
                 return createCPUDevice(desc, outDevice);
diff --git a/tools/platform/gui.cpp b/tools/platform/gui.cpp
index 8bfff2d11..a4602e4f2 100644
--- a/tools/platform/gui.cpp
+++ b/tools/platform/gui.cpp
@@ -353,5 +353,9 @@ GUI::~GUI()
 #include "external/imgui/imgui.cpp"
 #include "external/imgui/imgui_draw.cpp"
 #ifdef _WIN32
+// imgui_impl_win32 defines these, so make sure it doesn't error because
+// they're already there
+#undef WIN32_LEAN_AND_MEAN
+#undef NOMINMAX
 #include "external/imgui/examples/imgui_impl_win32.cpp"
 #endif
diff --git a/tools/platform/windows/win-window.cpp b/tools/platform/windows/win-window.cpp
index f0217e04c..d785b0fb7 100644
--- a/tools/platform/windows/win-window.cpp
+++ b/tools/platform/windows/win-window.cpp
@@ -2,8 +2,7 @@
 
 #include "../window.h"
 
-#define WIN32_LEAN_AND_MEAN
-#include <Windows.h>
+#include <windows.h>
 #include <windowsx.h>
 using namespace Slang;
 
diff --git a/tools/render-test/render-test-main.cpp b/tools/render-test/render-test-main.cpp
index d3c284132..f21a5ff07 100644
--- a/tools/render-test/render-test-main.cpp
+++ b/tools/render-test/render-test-main.cpp
@@ -26,8 +26,7 @@
 
 #if ENABLE_RENDERDOC_INTEGRATION
 #    include "external/renderdoc_app.h"
-#    define WIN32_LEAN_AND_MEAN
-#    include <Windows.h>
+#    include <windows.h>
 #endif
 
 namespace renderer_test {