Moved compressed blitting into its own class for readability & easier…

… maintenance.

softlight/CMakeLists.txt

@@ -123,6 +123,7 @@ set(SL_LIB_HEADERS
     include/softlight/SL_AnimationProperty.hpp
     include/softlight/SL_Atlas.hpp
     include/softlight/SL_BlitProcesor.hpp
+    include/softlight/SL_BlitCompressedProcesor.hpp
     include/softlight/SL_BoundingBox.hpp
     include/softlight/SL_Camera.hpp
     include/softlight/SL_ClearProcesor.hpp
@@ -213,6 +214,7 @@ set(SL_LIB_SOURCES
     src/SL_AnimationPlayer.cpp
     src/SL_Atlas.cpp
     src/SL_BlitProcessor.cpp
+    src/SL_BlitCompressedProcessor.cpp
     src/SL_BoundingBox.cpp
     src/SL_Camera.cpp
     src/SL_ClearProcessor.cpp

softlight/include/softlight/SL_BlitCompressedProcesor.hpp

@@ -0,0 +1,86 @@
+
+#ifndef SL_BLIT_COMPRESSED_PROCESSOR_HPP
+#define SL_BLIT_COMPRESSED_PROCESSOR_HPP
+
+#include <cstdint>
+
+
+
+/*-----------------------------------------------------------------------------
+ * Forward Declarations
+-----------------------------------------------------------------------------*/
+class SL_Texture;
+
+
+
+/**----------------------------------------------------------------------------
+ * @brief The Blit Processor helps to perform texture blitting to the native
+ * window backbuffer on another thread.
+ *
+ * Much of the blitting routines are templated to support conversion between
+ * possible texture types and the backbuffer (which is an 8-bit RGBA buffer).
+ *
+ * Texture blitting uses nearest-neighbor filtering to increase or decrease the
+ * resolution and fit the backbuffer. Fixed-point calculation is used to avoid
+ * precision errors and increase ALU throughput. Benchmarks on x86 and ARM has
+ * shown that floating-point logic performs worse in this area.
+-----------------------------------------------------------------------------*/
+struct SL_BlitCompressedProcessor
+{
+    enum : uint_fast32_t
+    {
+        NUM_FIXED_BITS = 16u
+    };
+
+    // 32 bits
+    uint16_t mThreadId;
+    uint16_t mNumThreads;
+
+    // 64-bits
+    uint16_t srcX0;
+    uint16_t srcY0;
+    uint16_t srcX1;
+    uint16_t srcY1;
+
+    // 64-bits
+    uint16_t dstX0;
+    uint16_t dstY0;
+    uint16_t dstX1;
+    uint16_t dstY1;
+
+    // 64-128 bits
+    const SL_Texture* mTexture;
+    SL_Texture* mBackBuffer;
+
+    // 224-288 bits total, 28-36 bytes
+
+    // Blit a single R channel
+    template<typename inColor_type>
+    void blit_src_r() noexcept;
+
+    // Blit a texture with only RG color channels
+    template<typename inColor_type>
+    void blit_src_rg() noexcept;
+
+    // Blit an RGB texture
+    template<typename inColor_type>
+    void blit_src_rgb() noexcept;
+
+    // Blit all 4 color components
+    template<typename inColor_type>
+    void blit_src_rgba() noexcept;
+
+    // Blit compressed color components
+    template<typename inColor_type>
+    void blit_src_compressed() noexcept;
+
+    // Blit all 4 color components
+    template<class BlitOp>
+    void blit_nearest() noexcept;
+
+    void execute() noexcept;
+};
+
+
+
+#endif /* SL_BLIT_COMPRESSED_PROCESSOR_HPP */