Merge branch 'master' into cache_vk_devices

Author: gtong-nv

.github/actions/common-setup/action.yml

@@ -25,16 +25,13 @@ runs:
 
     - name: Set up MSVC dev tools on Windows
       uses: ilammy/msvc-dev-cmd@v1
-      with:
-        sdk: "10.0.19041.0"
 
     - name: Install dependencies (Linux only)
+      if: inputs.os == 'linux'
       shell: bash
       run: |
-        if [[ "${{ inputs.os }}" == "linux" ]]; then
-            sudo apt-get update
-            sudo apt-get install -y libx11-dev 
-        fi
+        sudo apt-get update
+        sudo apt-get install -y libx11-dev
 
     - name: Setup Node.js (Linux only)
       if: inputs.os == 'linux'

.github/workflows/ci.yml

@@ -228,15 +228,21 @@ jobs:
           echo "Linux Release x64: ${{ needs.test-linux-release-gcc-x86_64.result }}"
           echo "Linux Debug x64: ${{ needs.test-linux-debug-gcc-x86_64.result }}"
 
-          # Check if all required jobs succeeded
-          if [[ "${{ needs.test-windows-release-cl-x86_64-gpu.result }}" != "success" ]] || \
-             [[ "${{ needs.test-windows-debug-cl-x86_64-gpu.result }}" != "success" ]] || \
-             [[ "${{ needs.test-macos-release-clang-aarch64.result }}" != "success" ]] || \
-             [[ "${{ needs.test-macos-debug-clang-aarch64.result }}" != "success" ]] || \
-             [[ "${{ needs.test-linux-release-gcc-x86_64.result }}" != "success" ]] || \
-             [[ "${{ needs.test-linux-debug-gcc-x86_64.result }}" != "success" ]]; then
+          # Check if any required jobs failed (allow success or skipped)
+          if [[ "${{ needs.test-windows-release-cl-x86_64-gpu.result }}" == "failure" ]] || \
+             [[ "${{ needs.test-windows-debug-cl-x86_64-gpu.result }}" == "failure" ]] || \
+             [[ "${{ needs.test-macos-release-clang-aarch64.result }}" == "failure" ]] || \
+             [[ "${{ needs.test-macos-debug-clang-aarch64.result }}" == "failure" ]] || \
+             [[ "${{ needs.test-linux-release-gcc-x86_64.result }}" == "failure" ]] || \
+             [[ "${{ needs.test-linux-debug-gcc-x86_64.result }}" == "failure" ]] || \
+             [[ "${{ needs.test-windows-release-cl-x86_64-gpu.result }}" == "cancelled" ]] || \
+             [[ "${{ needs.test-windows-debug-cl-x86_64-gpu.result }}" == "cancelled" ]] || \
+             [[ "${{ needs.test-macos-release-clang-aarch64.result }}" == "cancelled" ]] || \
+             [[ "${{ needs.test-macos-debug-clang-aarch64.result }}" == "cancelled" ]] || \
+             [[ "${{ needs.test-linux-release-gcc-x86_64.result }}" == "cancelled" ]] || \
+             [[ "${{ needs.test-linux-debug-gcc-x86_64.result }}" == "cancelled" ]]; then
             echo "❌ One or more CI jobs failed or were cancelled"
             exit 1
           fi
 
-          echo "✅ All CI jobs passed successfully!"
+          echo "✅ All CI jobs completed successfully (passed or skipped)!"

.github/workflows/release.yml

@@ -38,7 +38,11 @@ jobs:
               compiler: gcc,
               build-slang-llvm: false,
             }
-          - { os: windows, runs-on: windows-latest, compiler: cl }
+          - {
+              os: windows,
+              runs-on: ["Windows", "self-hosted", "GCP-T4"],
+              compiler: cl,
+            }
           - { os: macos, runs-on: macos-latest, compiler: clang }
 
           - { build-slang-llvm: false }
@@ -58,12 +62,7 @@ jobs:
         with:
           submodules: "recursive"
           fetch-depth: "0"
-      - name: Install dependencies
-        run: |
-          if [[ "${{ matrix.os }}" == "linux" ]]; then
-              sudo apt-get update
-              sudo apt-get install -y libx11-dev
-          fi
+
       - name: Setup
         uses: ./.github/actions/common-setup
         with:
@@ -79,25 +78,6 @@ jobs:
           mkdir build-platform-generators
           cmake --install build --config Release --component generators --prefix build-platform-generators
 
-      - name: Change dev tools to host arch (windows)
-        uses: ilammy/msvc-dev-cmd@v1
-        with:
-          arch: ${{matrix.platform == 'aarch64' && 'amd64_arm64' || 'amd64'}}
-          sdk: "10.0.19041.0"
-
-      - name: Change dev tools to host arch (linux and macos)
-        run: |
-          if [[ "${{matrix.os}}" == linux* && "${{matrix.platform}}" == "aarch64" && "$(uname -m)" != "aarch64" ]]; then
-            export CC=aarch64-linux-gnu-gcc
-            export CXX=aarch64-linux-gnu-g++
-          fi
-          CMAKE_OSX_ARCHITECTURES="${{matrix.platform}}"
-          CMAKE_OSX_ARCHITECTURES=${CMAKE_OSX_ARCHITECTURES//aarch64/arm64}
-
-          echo "CC=$CC" >> "$GITHUB_ENV"
-          echo "CXX=$CXX" >> "$GITHUB_ENV"
-          echo "CMAKE_OSX_ARCHITECTURES=$CMAKE_OSX_ARCHITECTURES" >> "$GITHUB_ENV"
-
       - name: Build Slang
         run: |
           if [[ "${{ matrix.platform }}" == "wasm" ]]; then

CMakeLists.txt

@@ -185,10 +185,16 @@ advanced_option(
     "Build using system glslang library"
     OFF
 )
+if(WIN32)
+    set(SLANG_ENABLE_SPIRV_TOOLS_MIMALLOC_DEFAULT ON)
+else()
+    set(SLANG_ENABLE_SPIRV_TOOLS_MIMALLOC_DEFAULT OFF)
+endif()
+
 advanced_option(
     SLANG_ENABLE_SPIRV_TOOLS_MIMALLOC
     "Enable mimalloc for SPIRV-Tools to improve compilation performance"
-    OFF
+    ${SLANG_ENABLE_SPIRV_TOOLS_MIMALLOC_DEFAULT}
 )
 
 option(

external/CMakeLists.txt

@@ -25,15 +25,6 @@ if(NOT SLANG_ENABLE_EXTERNAL_COMPILER_WARNINGS)
     endif()
 endif()
 
-# Option to enable mimalloc for SPIRV-Tools to improve compilation performance
-if(WIN32)
-    option(
-        SLANG_ENABLE_SPIRV_TOOLS_MIMALLOC
-        "Enable mimalloc for SPIRV-Tools to improve compilation performance"
-        ON
-    )
-endif()
-
 # Unordered dense
 if(NOT ${SLANG_USE_SYSTEM_UNORDERED_DENSE})
     if(NOT SLANG_OVERRIDE_UNORDERED_DENSE_PATH)
@@ -114,6 +105,8 @@ target_include_directories(
 )
 
 # SPIRV-Headers
+# Avoid using ${system} intentionally in `add_subdirectory`, because if we do, it will end up using
+# spirv.h installed in the system directory on MacOS.
 if(NOT TARGET SPIRV-Headers)
     if(SLANG_USE_SYSTEM_SPIRV_HEADERS)
         if(SLANG_OVERRIDE_SPIRV_HEADERS_PATH)
@@ -127,10 +120,9 @@ if(NOT TARGET SPIRV-Headers)
             ${SLANG_OVERRIDE_SPIRV_HEADERS_PATH}/spirv-headers
             spirv-headers
             EXCLUDE_FROM_ALL
-            ${system}
         )
     else()
-        add_subdirectory(spirv-headers EXCLUDE_FROM_ALL ${system})
+        add_subdirectory(spirv-headers EXCLUDE_FROM_ALL)
     endif()
 endif()
 
@@ -214,13 +206,12 @@ if(SLANG_ENABLE_SLANG_GLSLANG)
 
         # Tools
         if(NOT SLANG_OVERRIDE_SPIRV_TOOLS_PATH)
-            add_subdirectory(spirv-tools EXCLUDE_FROM_ALL ${system})
+            add_subdirectory(spirv-tools EXCLUDE_FROM_ALL)
         else()
             add_subdirectory(
                 ${SLANG_OVERRIDE_SPIRV_TOOLS_PATH}/spirv-tools
                 spirv-tools
                 EXCLUDE_FROM_ALL
-                ${system}
             )
         endif()
     endif()

external/unordered_dense

@@ -20,6 +20,7 @@
 #endif
 
 #if SLANG_LINUX_FAMILY
+#include <limits.h>
 #include <unistd.h>
 #endif
 
@@ -128,19 +129,24 @@ SLANG_COMPILE_TIME_ASSERT(E_OUTOFMEMORY == SLANG_E_OUT_OF_MEMORY);
     }
 
     // We try to search the DLL in two different attempts.
-    // First attempt tries on the directories explicitly specified with AddDllDirectory(),
-    // If it failed to find one, we will search over all PATH.
-    // Windows API made two approaches mutually exclusive and we need to try two times.
+    // First attempt - LoadLibraryExW()
+    // If it failed to find one, we will use LoadLibraryW() to search over all PATH.
+    // Search order: 1) The directory that contains the DLL (LOAD_LIBRARY_SEARCH_DLL_LOAD_DIR).
+    //                  This directory is searched only for dependencies of the DLL being loaded.
+    //               2) Application directory
+    //               3) User directories (AddDllDirectory/SetDllDirectory)
+    //               4) System32
+    //               5) PATH environment variable (by the 2nd attempt with LoadLibraryW())
     // https://learn.microsoft.com/en-us/windows/win32/api/libloaderapi/nf-libloaderapi-loadlibraryexw
-    String platformFileNameStr(platformFileName);
-    HMODULE h =
-        LoadLibraryExW(platformFileNameStr.toWString(), nullptr, LOAD_LIBRARY_SEARCH_USER_DIRS);
-    // If LoadLibraryExW failed, try again with LoadLibraryW.
     // https://docs.microsoft.com/en-us/windows/desktop/api/libloaderapi/nf-libloaderapi-loadlibraryw
-    if (!h)
-        h = LoadLibraryW(platformFileNameStr.toWString());
+    String platformFileNameStr(platformFileName);
+    OSString wideFileName = platformFileNameStr.toWString();
+    HMODULE handle = LoadLibraryExW(wideFileName, nullptr, LOAD_LIBRARY_SEARCH_DEFAULT_DIRS);
+
+    if (!handle)
+        handle = LoadLibraryW(wideFileName);
     // If still not found, return an error.
-    if (!h)
+    if (!handle)
     {
         const DWORD lastError = GetLastError();
         switch (lastError)
@@ -163,7 +169,7 @@ SLANG_COMPILE_TIME_ASSERT(E_OUTOFMEMORY == SLANG_E_OUT_OF_MEMORY);
         // Turn to Result, if not one of the well known errors
         return HRESULT_FROM_WIN32(lastError);
     }
-    handleOut = (Handle)h;
+    handleOut = (Handle)handle;
     return SLANG_OK;
 }
 

source/core/slang-platform.cpp

@@ -24,6 +24,11 @@ if(EMSCRIPTEN)
     # To generate binding code
     target_link_options(
         slang-wasm
-        PUBLIC "--bind" "--emit-tsd" "interface.d.ts" "-s" "EXPORT_ES6=1"
+        PUBLIC
+        "--bind"
+        "--emit-tsd" "interface.d.ts"
+        "-sEXPORT_ES6=1"
+        "-sMODULARIZE=1"
+        "-sEXPORTED_RUNTIME_METHODS=['FS']"
     )
 endif()

source/slang-wasm/CMakeLists.txt

@@ -1278,18 +1278,16 @@ struct __none_t
 {
 };
 
-// @hidden: this type is a BaseType since we want it to work with
-// `registerBuiltinDecl`
-__builtin_type($((int)BaseType::AddressSpace))
+/// @category misc_types
+__magic_enum(AddressSpace)
 enum AddressSpace : uint64_t
 {
     Device = $((uint64_t)AddressSpace::UserPointer),
     GroupShared = $((uint64_t)AddressSpace::GroupShared),
 };
 
-// @hidden: this type is a BaseType since we want it to work with
-// `registerBuiltinDecl`
-__builtin_type($((int)BaseType::MemoryScope))
+/// @category misc_types
+__magic_enum(MemoryScope)
 enum MemoryScope : int32_t
 {
     CrossDevice = $((int32_t)MemoryScope::CrossDevice),
@@ -1300,9 +1298,8 @@ enum MemoryScope : int32_t
     QueueFamily = $((int32_t)MemoryScope::QueueFamily),
 }
 
-// @hidden: this type is a BaseType since we want it to work with
-// `registerBuiltinDecl`
-__builtin_type($((int)BaseType::AccessQualifier))
+/// @category misc_types
+__magic_enum(AccessQualifier)
 enum Access : uint64_t
 {
     ReadWrite = $((uint64_t)AccessQualifier::ReadWrite),

source/slang/core.meta.slang

@@ -1914,7 +1914,7 @@ public vector<T.Element,4> texture(Sampler1D<T> sampler, float p)
 __generic<T:ITexelElement>
 [ForceInline]
 [require(cpp_glsl_hlsl_spirv, texture_sm_4_0_fragment)]
-public vector<T.Element,4> texture(Sampler1D<T> sampler, float p, constexpr float bias)
+public vector<T.Element,4> texture(Sampler1D<T> sampler, float p, float bias)
 {
     return __vectorReshape2<T.Element, 4>(sampler.SampleBias(p, bias));
 }
@@ -1950,7 +1950,7 @@ public vector<T.Element,4> texture(_Texture<
         0, // isShadow
         1, // isCombined
         format
-    > sampler, vector<float,Shape.dimensions+isArray> p, constexpr float bias)
+    > sampler, vector<float,Shape.dimensions+isArray> p, float bias)
 {
     return __vectorReshape2<T.Element,4>(sampler.SampleBias(p, bias));
 }
@@ -3891,7 +3891,7 @@ public vector<T.Element,4> textureGatherOffset(_Texture<
         0, // isShadow
         1, // isCombined
         format
-    > sampler, vector<float,2+isArray> p, constexpr vector<int,2> offset, int comp = 0)
+    > sampler, vector<float,2+isArray> p, vector<int,2> offset, int comp = 0)
 {
     switch (comp)
     {
@@ -3915,11 +3915,12 @@ public vec4 textureGatherOffset(_Texture<
         1, // isShadow
         1, // isCombined
         format
-    > sampler, vector<float,2+isArray> p, float refZ, constexpr vector<int,2> offset)
+    > sampler, vector<float,2+isArray> p, float refZ, vector<int,2> offset)
 {
     return sampler.GatherCmp(p, refZ, offset);
 }
 
+
 // -------------------
 // textureGatherOffsets
 // -------------------