move image scaling functions to their own unit, as -O2

5 files changed, 113 insertions, 102 deletions

diff --git a/source/core/CMakeLists.txt b/source/core/CMakeLists.txt
index fa1d246..6deb473 100644
--- a/source/core/CMakeLists.txt
+++ b/source/core/CMakeLists.txt
@@ -41,6 +41,7 @@ SET (star_core_HEADERS
     StarIdMap.hpp
     StarImage.hpp
     StarImageProcessing.hpp
+    StarImageScaling.hpp
     StarInputEvent.hpp
     StarInterpolation.hpp
     StarRefPtr.hpp
@@ -152,6 +153,7 @@ SET (star_core_SOURCES
     StarIODevice.cpp
     StarImage.cpp
     StarImageProcessing.cpp
+    StarImageScaling.cpp
     StarInputEvent.cpp
     StarJson.cpp
     StarJsonBuilder.cpp
@@ -229,6 +231,10 @@ IF(STAR_PRECOMPILED_HEADERS)
   TARGET_PRECOMPILE_HEADERS (star_core PUBLIC StarPch.hpp)
 ENDIF()
 
+IF(STAR_COMPILER STREQUAL "gnu")
+  SET_SOURCE_FILES_PROPERTIES(StarImageScaling.cpp PROPERTIES COMPILE_FLAGS "-O2")
+ENDIF()
+
 IF(STAR_USE_JEMALLOC AND JEMALLOC_IS_PREFIXED)
   SET_SOURCE_FILES_PROPERTIES(StarMemory.cpp PROPERTIES
           COMPILE_DEFINITIONS STAR_JEMALLOC_IS_PREFIXED
diff --git a/source/core/StarImageProcessing.cpp b/source/core/StarImageProcessing.cpp
index 96edc0b..1cfd847 100644
--- a/source/core/StarImageProcessing.cpp
+++ b/source/core/StarImageProcessing.cpp
@@ -1,4 +1,5 @@
 #include "StarImageProcessing.hpp"
+#include "StarImageScaling.hpp"
 #include "StarMatrix3.hpp"
 #include "StarInterpolation.hpp"
 #include "StarLexicalCast.hpp"
@@ -10,104 +11,6 @@
 
 namespace Star {
 
-Image scaleNearest(Image const& srcImage, Vec2F const& scale) {
-  Vec2U srcSize = srcImage.size();
-  Vec2U destSize = Vec2U::round(vmult(Vec2F(srcSize), scale));
-  destSize[0] = max(destSize[0], 1u);
-  destSize[1] = max(destSize[1], 1u);
-
-  Image destImage(destSize, srcImage.pixelFormat());
-
-  for (unsigned y = 0; y < destSize[1]; ++y) {
-    for (unsigned x = 0; x < destSize[0]; ++x)
-      destImage.set({x, y}, srcImage.clamp(Vec2I::round(vdiv(Vec2F(x, y), scale))));
-  }
-  return destImage;
-}
-
-#pragma GCC push_options
-#pragma GCC optimize("-fno-unsafe-math-optimizations", "-ffloat-store")
-Image scaleBilinear(Image const& srcImage, Vec2F const& scale) {
-  Vec2U srcSize = srcImage.size();
-  Vec2U destSize = Vec2U::round(vmult(Vec2F(srcSize), scale));
-  destSize[0] = max(destSize[0], 1u);
-  destSize[1] = max(destSize[1], 1u);
-
-  Image destImage(destSize, srcImage.pixelFormat());
-
-  auto lerpVec = [](float const& offset, Vec4F f0, Vec4F f1) {
-    return f0 * (1 - offset) + f1 * (offset);
-  };
-
-  for (unsigned y = 0; y < destSize[1]; ++y) {
-    for (unsigned x = 0; x < destSize[0]; ++x) {
-      auto pos = vdiv(Vec2F(x, y), scale);
-      auto ipart = Vec2I::floor(pos);
-      auto fpart = pos - Vec2F(ipart);
-
-      auto result = lerpVec(fpart[1], lerpVec(fpart[0], Vec4F(srcImage.clamp(ipart[0], ipart[1])), Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1]))), lerpVec(fpart[0],
-            Vec4F(srcImage.clamp(ipart[0], ipart[1] + 1)), Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1] + 1))));
-
-      destImage.set({x, y}, Vec4B(result));
-    }
-  }
-
-  return destImage;
-}
-
-Image scaleBicubic(Image const& srcImage, Vec2F const& scale) {
-  Vec2U srcSize = srcImage.size();
-  Vec2U destSize = Vec2U::round(vmult(Vec2F(srcSize), scale));
-  destSize[0] = max(destSize[0], 1u);
-  destSize[1] = max(destSize[1], 1u);
-
-  Image destImage(destSize, srcImage.pixelFormat());
-
-  for (unsigned y = 0; y < destSize[1]; ++y) {
-    for (unsigned x = 0; x < destSize[0]; ++x) {
-      auto pos = vdiv(Vec2F(x, y), scale);
-      auto ipart = Vec2I::floor(pos);
-      auto fpart = pos - Vec2F(ipart);
-
-      Vec4F a = cubic4(fpart[0],
-          Vec4F(srcImage.clamp(ipart[0], ipart[1])),
-          Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1])),
-          Vec4F(srcImage.clamp(ipart[0] + 2, ipart[1])),
-          Vec4F(srcImage.clamp(ipart[0] + 3, ipart[1])));
-
-      Vec4F b = cubic4(fpart[0],
-          Vec4F(srcImage.clamp(ipart[0], ipart[1] + 1)),
-          Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1] + 1)),
-          Vec4F(srcImage.clamp(ipart[0] + 2, ipart[1] + 1)),
-          Vec4F(srcImage.clamp(ipart[0] + 3, ipart[1] + 1)));
-
-      Vec4F c = cubic4(fpart[0],
-          Vec4F(srcImage.clamp(ipart[0], ipart[1] + 2)),
-          Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1] + 2)),
-          Vec4F(srcImage.clamp(ipart[0] + 2, ipart[1] + 2)),
-          Vec4F(srcImage.clamp(ipart[0] + 3, ipart[1] + 2)));
-
-      Vec4F d = cubic4(fpart[0],
-          Vec4F(srcImage.clamp(ipart[0], ipart[1] + 3)),
-          Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1] + 3)),
-          Vec4F(srcImage.clamp(ipart[0] + 2, ipart[1] + 3)),
-          Vec4F(srcImage.clamp(ipart[0] + 3, ipart[1] + 3)));
-
-      auto result = cubic4(fpart[1], a, b, c, d);
-
-      destImage.set({x, y}, Vec4B(
-          clamp(result[0], 0.0f, 255.0f),
-          clamp(result[1], 0.0f, 255.0f),
-          clamp(result[2], 0.0f, 255.0f),
-          clamp(result[3], 0.0f, 255.0f)
-        ));
-    }
-  }
-
-  return destImage;
-}
-#pragma GCC pop_options
-
 StringList colorDirectivesFromConfig(JsonArray const& directives) {
   List<String> result;
 
diff --git a/source/core/StarImageProcessing.hpp b/source/core/StarImageProcessing.hpp
index 8316d18..dfc9420 100644
--- a/source/core/StarImageProcessing.hpp
+++ b/source/core/StarImageProcessing.hpp
@@ -10,10 +10,6 @@ STAR_CLASS(Image);
 
 STAR_EXCEPTION(ImageOperationException, StarException);
 
-Image scaleNearest(Image const& srcImage, Vec2F const& scale);
-Image scaleBilinear(Image const& srcImage, Vec2F const& scale);
-Image scaleBicubic(Image const& srcImage, Vec2F const& scale);
-
 StringList colorDirectivesFromConfig(JsonArray const& directives);
 String paletteSwapDirectivesFromConfig(Json const& swaps);
 
diff --git a/source/core/StarImageScaling.cpp b/source/core/StarImageScaling.cpp
new file mode 100644
index 0000000..96837aa
--- /dev/null
+++ b/source/core/StarImageScaling.cpp
@@ -0,0 +1,98 @@
+#include "StarImage.hpp"
+#include "StarImageScaling.hpp"
+#include "StarInterpolation.hpp"
+
+namespace Star {
+
+Image scaleNearest(Image const& srcImage, Vec2F const& scale) {
+  Vec2U srcSize = srcImage.size();
+  Vec2U destSize = Vec2U::round(vmult(Vec2F(srcSize), scale));
+  destSize[0] = max(destSize[0], 1u);
+  destSize[1] = max(destSize[1], 1u);
+
+  Image destImage(destSize, srcImage.pixelFormat());
+
+  for (unsigned y = 0; y < destSize[1]; ++y) {
+    for (unsigned x = 0; x < destSize[0]; ++x)
+      destImage.set({x, y}, srcImage.clamp(Vec2I::round(vdiv(Vec2F(x, y), scale))));
+  }
+  return destImage;
+}
+
+Image scaleBilinear(Image const& srcImage, Vec2F const& scale) {
+  Vec2U srcSize = srcImage.size();
+  Vec2U destSize = Vec2U::round(vmult(Vec2F(srcSize), scale));
+  destSize[0] = max(destSize[0], 1u);
+  destSize[1] = max(destSize[1], 1u);
+
+  Image destImage(destSize, srcImage.pixelFormat());
+
+  for (unsigned y = 0; y < destSize[1]; ++y) {
+    for (unsigned x = 0; x < destSize[0]; ++x) {
+      auto pos = vdiv(Vec2F(x, y), scale);
+      auto ipart = Vec2I::floor(pos);
+      auto fpart = pos - Vec2F(ipart);
+
+      auto result = lerp(fpart[1], lerp(fpart[0], Vec4F(srcImage.clamp(ipart[0], ipart[1])), Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1]))), lerp(fpart[0],
+            Vec4F(srcImage.clamp(ipart[0], ipart[1] + 1)), Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1] + 1))));
+
+      destImage.set({x, y}, Vec4B(result));
+    }
+  }
+
+  return destImage;
+}
+
+Image scaleBicubic(Image const& srcImage, Vec2F const& scale) {
+  Vec2U srcSize = srcImage.size();
+  Vec2U destSize = Vec2U::round(vmult(Vec2F(srcSize), scale));
+  destSize[0] = max(destSize[0], 1u);
+  destSize[1] = max(destSize[1], 1u);
+
+  Image destImage(destSize, srcImage.pixelFormat());
+
+  for (unsigned y = 0; y < destSize[1]; ++y) {
+    for (unsigned x = 0; x < destSize[0]; ++x) {
+      auto pos = vdiv(Vec2F(x, y), scale);
+      auto ipart = Vec2I::floor(pos);
+      auto fpart = pos - Vec2F(ipart);
+
+      Vec4F a = cubic4(fpart[0],
+          Vec4F(srcImage.clamp(ipart[0], ipart[1])),
+          Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1])),
+          Vec4F(srcImage.clamp(ipart[0] + 2, ipart[1])),
+          Vec4F(srcImage.clamp(ipart[0] + 3, ipart[1])));
+
+      Vec4F b = cubic4(fpart[0],
+          Vec4F(srcImage.clamp(ipart[0], ipart[1] + 1)),
+          Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1] + 1)),
+          Vec4F(srcImage.clamp(ipart[0] + 2, ipart[1] + 1)),
+          Vec4F(srcImage.clamp(ipart[0] + 3, ipart[1] + 1)));
+
+      Vec4F c = cubic4(fpart[0],
+          Vec4F(srcImage.clamp(ipart[0], ipart[1] + 2)),
+          Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1] + 2)),
+          Vec4F(srcImage.clamp(ipart[0] + 2, ipart[1] + 2)),
+          Vec4F(srcImage.clamp(ipart[0] + 3, ipart[1] + 2)));
+
+      Vec4F d = cubic4(fpart[0],
+          Vec4F(srcImage.clamp(ipart[0], ipart[1] + 3)),
+          Vec4F(srcImage.clamp(ipart[0] + 1, ipart[1] + 3)),
+          Vec4F(srcImage.clamp(ipart[0] + 2, ipart[1] + 3)),
+          Vec4F(srcImage.clamp(ipart[0] + 3, ipart[1] + 3)));
+
+      auto result = cubic4(fpart[1], a, b, c, d);
+
+      destImage.set({x, y}, Vec4B(
+          clamp(result[0], 0.0f, 255.0f),
+          clamp(result[1], 0.0f, 255.0f),
+          clamp(result[2], 0.0f, 255.0f),
+          clamp(result[3], 0.0f, 255.0f)
+        ));
+    }
+  }
+
+  return destImage;
+}
+
+}
\ No newline at end of file
diff --git a/source/core/StarImageScaling.hpp b/source/core/StarImageScaling.hpp
new file mode 100644
index 0000000..6a7271b
--- /dev/null
+++ b/source/core/StarImageScaling.hpp
@@ -0,0 +1,8 @@
+namespace Star {
+
+STAR_CLASS(Image);
+Image scaleNearest(Image const& srcImage, Vec2F const& scale);
+Image scaleBilinear(Image const& srcImage, Vec2F const& scale);
+Image scaleBicubic(Image const& srcImage, Vec2F const& scale);
+
+}
\ No newline at end of file