clx_render: Slightly optimize `GetSkipSize`

This gives a very slight FPS boost.
1140 to 1143 FPS on my machine as measured by:

```bash
tools/linux_reduced_cpu_variance_run.sh tools/measure_timedemo_performance.py -n 5 --binary build-rel/devilutionx
```

Source/engine/render/clx_render.cpp

@@ -70,25 +70,6 @@ struct RenderSrc {
 	uint_fast16_t width;
 };
 
-struct SkipSize {
-	int_fast16_t wholeLines;
-	int_fast16_t xOffset;
-};
-
-DVL_ALWAYS_INLINE DVL_ATTRIBUTE_HOT SkipSize GetSkipSize(int_fast16_t remainingWidth, int_fast16_t srcWidth)
-{
-	if (remainingWidth < 0) {
-		// If `remainingWidth` is negative, `-remainingWidth` is the overrun.
-		const int_fast16_t overrunLines = -remainingWidth / srcWidth;
-		return {
-			static_cast<int_fast16_t>(1 + overrunLines),
-			static_cast<int_fast16_t>(-remainingWidth - srcWidth * overrunLines)
-		};
-	}
-	// If `remainingWidth` is non-negative, then it is 0, meaning we drew a whole line.
-	return { 1, 0 };
-}
-
 DVL_ALWAYS_INLINE DVL_ATTRIBUTE_HOT const uint8_t *SkipRestOfLineWithOverrun(
     const uint8_t *src, int_fast16_t srcWidth, SkipSize &skipSize)
 {

Source/utils/cl2_to_clx.cpp

@@ -11,24 +11,6 @@
 
 namespace devilution {
 
-namespace {
-
-constexpr size_t FrameHeaderSize = 10;
-
-struct SkipSize {
-	int_fast16_t wholeLines;
-	int_fast16_t xOffset;
-};
-SkipSize GetSkipSize(int_fast16_t overrun, int_fast16_t srcWidth)
-{
-	SkipSize result;
-	result.wholeLines = overrun / srcWidth;
-	result.xOffset = overrun - srcWidth * result.wholeLines;
-	return result;
-}
-
-} // namespace
-
 uint16_t Cl2ToClx(const uint8_t *data, size_t size,
     PointerOrValue<uint16_t> widthOrWidths, std::vector<uint8_t> &clxData)
 {
@@ -74,6 +56,7 @@ uint16_t Cl2ToClx(const uint8_t *data, size_t size,
 			const uint16_t frameWidth = widthOrWidths.HoldsPointer() ? widthOrWidths.AsPointer()[frame - 1] : widthOrWidths.AsValue();
 
 			const size_t frameHeaderPos = clxData.size();
+			constexpr size_t FrameHeaderSize = 10;
 			clxData.resize(clxData.size() + FrameHeaderSize);
 			WriteLE16(&clxData[frameHeaderPos], FrameHeaderSize);
 			WriteLE16(&clxData[frameHeaderPos + 2], frameWidth);
@@ -110,14 +93,9 @@ uint16_t Cl2ToClx(const uint8_t *data, size_t size,
 					}
 				}
 
-				++frameHeight;
+				const auto skipSize = GetSkipSize(remainingWidth, static_cast<int_fast16_t>(frameWidth));
-				if (remainingWidth < 0) {
+				xOffset = skipSize.xOffset;
-					const auto skipSize = GetSkipSize(-remainingWidth, static_cast<int_fast16_t>(frameWidth));
+				frameHeight += skipSize.wholeLines;
-					xOffset = skipSize.xOffset;
-					frameHeight += skipSize.wholeLines;
-				} else {
-					xOffset = 0;
-				}
 			}
 			if (!pixels.empty()) {
 				AppendClxPixelsOrFillRun(pixels.data(), pixels.size(), clxData);

Source/utils/clx_decode.hpp

@@ -2,6 +2,9 @@
 
 #include <cstdint>
 
+#include "appfat.h"
+#include "utils/attributes.h"
+
 namespace devilution {
 
 [[nodiscard]] constexpr bool IsClxOpaque(uint8_t control)
@@ -27,4 +30,29 @@ namespace devilution {
 	return ClxFillEnd - control;
 }
 
+struct SkipSize {
+	int_fast16_t wholeLines;
+	int_fast16_t xOffset;
+};
+
+// Returns the number of lines and the x-offset by which the rendering has overrun
+// the current line (when a CLX command overruns the current line).
+//
+// Requires: remainingWidth <= 0.
+DVL_ALWAYS_INLINE SkipSize GetSkipSize(int_fast16_t remainingWidth, int_fast16_t srcWidth)
+{
+	// If `remainingWidth` is negative, `-remainingWidth` is the overrun.
+	// Otherwise, `remainingWidth` is always 0.
+
+	// Remaining width of 0 (= no overrun) is a common case.
+	// The calculation below would result in the same result.
+	// However, checking for 0 and skipping it entirely turns out to be faster.
+	if (remainingWidth == 0) return { 1, 0 };
+
+	const auto overrun = static_cast<uint_fast16_t>(-remainingWidth);
+	const uint_fast16_t overrunLines = overrun / srcWidth + 1;
+	const uint_fast16_t xOffset = overrun % srcWidth;
+	return { static_cast<int_fast16_t>(overrunLines), static_cast<int_fast16_t>(xOffset) };
+}
+
 } // namespace devilution