Merge pull request #1091 from stenzek/android-armv7

CI armv7 builds

.github/workflows/rolling-release.yml

@@ -148,7 +148,7 @@ jobs:
       shell: bash
       run: |
         sudo apt-get update
-        sudo apt-get -y install gcc-aarch64-linux-gnu g++-aarch64-linux-gnu
+        sudo apt-get -y install gcc-aarch64-linux-gnu g++-aarch64-linux-gnu gcc-arm-linux-gnueabihf g++-arm-linux-gnueabihf
 
     - name: Compile and zip Linux x64 libretro core
       shell: bash
@@ -165,6 +165,21 @@ jobs:
         name: "linux-libretro"
         path: "build-libretro-linux-x64/duckstation_libretro_x64.so.zip"
 
+    - name: Compile and zip Linux armv7 libretro core
+      shell: bash
+      run: |
+        mkdir build-libretro-linux-armv7
+        cd build-libretro-linux-armv7
+        cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_LIBRETRO_CORE=ON -DCMAKE_TOOLCHAIN_FILE=../CMakeModules/armv7-cross-toolchain.cmake ..
+        cmake --build . --parallel 2
+        zip -j duckstation_libretro_linux_armv7.so.zip duckstation_libretro.so
+
+    - name: Upload Linux AArch64 libretro core
+      uses: actions/upload-artifact@v1
+      with:
+        name: "linux-libretro"
+        path: "build-libretro-linux-armv7/duckstation_libretro_linux_armv7.so.zip"
+      
     - name: Compile and zip Linux AArch64 libretro core
       shell: bash
       run: |
@@ -173,13 +188,28 @@ jobs:
         cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_LIBRETRO_CORE=ON -DCMAKE_TOOLCHAIN_FILE=../CMakeModules/aarch64-cross-toolchain.cmake ..
         cmake --build . --parallel 2
         zip -j duckstation_libretro_linux_aarch64.so.zip duckstation_libretro.so
-      
+
     - name: Upload Linux AArch64 libretro core
       uses: actions/upload-artifact@v1
       with:
         name: "linux-libretro"
         path: "build-libretro-linux-aarch64/duckstation_libretro_linux_aarch64.so.zip"
 
+    - name: Compile and zip Android armv7 libretro core
+      shell: bash
+      run: |
+        mkdir build-libretro-android-armv7
+        cd build-libretro-android-armv7
+        cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_LIBRETRO_CORE=ON -DANDROID_ABI=armeabi-v7a -DANDROID_ARM_NEON=ON -DCMAKE_TOOLCHAIN_FILE=${ANDROID_SDK_ROOT}/ndk-bundle/build/cmake/android.toolchain.cmake ..
+        cmake --build . --parallel 2
+        zip -j duckstation_libretro_android_armv7.so.zip duckstation_libretro_android.so
+      
+    - name: Upload Android armv7 libretro core
+      uses: actions/upload-artifact@v1
+      with:
+        name: "linux-libretro"
+        path: "build-libretro-android-armv7/duckstation_libretro_android_armv7.so.zip"
+
     - name: Compile and zip Android AArch64 libretro core
       shell: bash
       run: |
@@ -224,14 +254,14 @@ jobs:
       shell: bash
       run: |
         cd android
-        mv app/build/outputs/apk/release/app-release-unsigned-signed.apk ../duckstation-android-aarch64.apk
+        mv app/build/outputs/apk/release/app-release-unsigned-signed.apk ../duckstation-android.apk
          
     - name: Upload APK
       if: github.ref == 'refs/heads/master'
       uses: actions/upload-artifact@v1
       with:
         name: "android"
-        path: "duckstation-android-aarch64.apk"
+        path: "duckstation-android.apk"
 
   macos-build:
     runs-on: macos-10.15
@@ -348,9 +378,11 @@ jobs:
             linux-x64-appimage-qt/duckstation-qt-x64.AppImage
             linux-x64-appimage-qt-zsync/duckstation-qt-x64.AppImage.zsync
             linux-libretro/duckstation_libretro_x64.so.zip
+            linux-libretro/duckstation_libretro_linux_armv7.so.zip
             linux-libretro/duckstation_libretro_linux_aarch64.so.zip
+            linux-libretro/duckstation_libretro_android_armv7.so.zip
             linux-libretro/duckstation_libretro_android_aarch64.so.zip
-            android/duckstation-android-aarch64.apk
+            android/duckstation-android.apk
             macos-x64/duckstation-mac-release.zip
             macos-x64/duckstation_libretro_mac.dylib.zip
 

CMakeLists.txt

@@ -196,6 +196,15 @@ elseif(${CMAKE_SYSTEM_PROCESSOR} STREQUAL "aarch64")
   set(CPU_ARCH "aarch64")
 elseif(${CMAKE_SYSTEM_PROCESSOR} STREQUAL "arm" OR ${CMAKE_SYSTEM_PROCESSOR} STREQUAL "armv7-a" OR ${CMAKE_SYSTEM_PROCESSOR} STREQUAL "armv7l")
   set(CPU_ARCH "aarch32")
+  if(ANDROID)
+    # Force ARM mode, since apparently ANDROID_ARM_MODE isn't working..
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -marm")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -marm")
+  else()
+    # Enable NEON.
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -marm -march=armv7-a+simd -mfpu=neon-vfpv3")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -marm -march=armv7-a+simd -mfpu=neon-vfpv3")
+  endif()
 else()
   message(FATAL_ERROR "Unknown system processor: " ${CMAKE_SYSTEM_PROCESSOR})
 endif()

CMakeModules/armv7-cross-toolchain.cmake

@@ -0,0 +1,14 @@
+# Source: https://github.com/stenzek/duckstation/issues/626#issuecomment-660718306
+
+# Target system
+SET(CMAKE_SYSTEM_NAME Linux)
+SET(CMAKE_SYSTEM_PROCESSOR armv7l)
+SET(CMAKE_SYSTEM_VERSION 1)
+set(CMAKE_CROSSCOMPILING TRUE)
+
+# Cross compiler
+SET(CMAKE_C_COMPILER   arm-linux-gnueabihf-gcc)
+SET(CMAKE_CXX_COMPILER arm-linux-gnueabihf-g++)
+set(CMAKE_LIBRARY_ARCHITECTURE arm-linux-gnueabihf)
+
+set(THREADS_PTHREAD_ARG "0" CACHE STRING "Result from TRY_RUN" FORCE)

android/app/build.gradle

@@ -30,8 +30,8 @@ android {
     defaultConfig {
         externalNativeBuild {
             cmake {
-                arguments "-DCMAKE_BUILD_TYPE=RelWithDebInfo"
-                abiFilters "arm64-v8a"
+                arguments "-DCMAKE_BUILD_TYPE=Release -DANDROID_ARM_NEON=ON"
+                abiFilters "arm64-v8a", "armeabi-v7a"
             }
         }
     }

src/common/vulkan/util.cpp

@@ -83,6 +83,8 @@ u32 GetTexelSize(VkFormat format)
 
     case VK_FORMAT_R5G5B5A1_UNORM_PACK16:
     case VK_FORMAT_A1R5G5B5_UNORM_PACK16:
+    case VK_FORMAT_R5G6B5_UNORM_PACK16:
+    case VK_FORMAT_B5G6R5_UNORM_PACK16:
       return 2;
 
     case VK_FORMAT_BC1_RGBA_UNORM_BLOCK:

src/core/gpu_sw.cpp

@@ -11,6 +11,8 @@ Log_SetChannel(GPU_SW);
 
 #if defined(CPU_X64)
 #include <emmintrin.h>
+#elif defined(CPU_AARCH32)
+#include <arm_neon.h>
 #elif defined(CPU_AARCH64)
 #ifdef _MSC_VER
 #include <arm64_neon.h>
@@ -155,7 +157,7 @@ ALWAYS_INLINE void CopyOutRow16<HostDisplayPixelFormat::RGBA5551, u16>(const u16
     _mm_storeu_si128(reinterpret_cast<__m128i*>(dst_ptr), value);
     dst_ptr += 8;
   }
-#elif defined(CPU_AARCH64)
+#elif defined(CPU_AARCH32) || defined(CPU_AARCH64)
   const u32 aligned_width = Common::AlignDownPow2(width, 8);
   for (; col < aligned_width; col += 8)
   {
@@ -195,7 +197,7 @@ ALWAYS_INLINE void CopyOutRow16<HostDisplayPixelFormat::RGB565, u16>(const u16*
     _mm_storeu_si128(reinterpret_cast<__m128i*>(dst_ptr), value);
     dst_ptr += 8;
   }
-#elif defined(CPU_AARCH64)
+#elif defined(CPU_AARCH32) || defined(CPU_AARCH64)
   const u32 aligned_width = Common::AlignDownPow2(width, 8);
   const uint16x8_t single_mask = vdupq_n_u16(0x1F);
   for (; col < aligned_width; col += 8)

src/core/gpu_sw_backend.cpp

@@ -102,8 +102,6 @@ void ALWAYS_INLINE_RELEASE GPU_SW_Backend::ShadePixel(const GPUBackendDrawComman
                    (cmd->draw_mode.GetTexturePageBaseY() + ZeroExtend32(texcoord_y)) % VRAM_HEIGHT);
         const u16 palette_index = (palette_value >> ((texcoord_x % 4) * 4)) & 0x0Fu;
 
-        const u32 px = (cmd->palette.GetXBase() + ZeroExtend32(palette_index)) % VRAM_WIDTH;
-        const u32 py = cmd->palette.GetYBase();
         texture_color.bits =
           GetPixel((cmd->palette.GetXBase() + ZeroExtend32(palette_index)) % VRAM_WIDTH, cmd->palette.GetYBase());
       }

src/frontend-common/vulkan_host_display.cpp

@@ -248,7 +248,7 @@ bool VulkanHostDisplay::DownloadTexture(const void* texture_handle, u32 x, u32 y
 }
 
 static constexpr std::array<VkFormat, static_cast<u32>(HostDisplayPixelFormat::Count)> s_display_pixel_format_mapping =
-  {{VK_FORMAT_UNDEFINED, VK_FORMAT_R8G8B8A8_UNORM, VK_FORMAT_B8G8R8A8_UNORM, VK_FORMAT_B5G6R5_UNORM_PACK16,
+  {{VK_FORMAT_UNDEFINED, VK_FORMAT_R8G8B8A8_UNORM, VK_FORMAT_B8G8R8A8_UNORM, VK_FORMAT_R5G6B5_UNORM_PACK16,
     VK_FORMAT_A1R5G5B5_UNORM_PACK16}};
 
 bool VulkanHostDisplay::SupportsDisplayPixelFormat(HostDisplayPixelFormat format) const