slight cleanup of the rendering code and some (minimal) speed

improvement for hardware scaled as well as "high res" mode in 2x2 no-AA
ivanovic · Dec 13, 2011 · 5883cd9 · 5883cd9
1 parent b026090
commit 5883cd9
Show file tree

Hide file tree

Showing 2 changed files with 21 additions and 22 deletions.
diff --git a/unix/pandora_scaling/simple_noAA_scaler.cpp b/unix/pandora_scaling/simple_noAA_scaler.cpp
@@ -3,43 +3,41 @@
 
 void render_x_single_xy(uint16* destination_pointer_address, uint16 screen_pitch_half,
                      uint16* gfx_screen, uint16 source_panewidth, int Width, int Height) {
-	uint16 height_desired = Height/* << 1*/; // height_desired = Height * 2
+	uint16 height_desired = Height;
 	uint16 bytes_per_line = Width << 1; // bytes_per_line = Width * 2
+
+	// sp is the pointer to a pixel in the SDL screen
+	register uint16 *sp16 = gfx_screen;
+	// dp is the pointer to a pixel in the "real" screen
+	register uint16 *dp16 = destination_pointer_address;
+
 	for (register uint16 i = 0; i < height_desired; ++i) {
-		// dp is the pointer to a pixel in the "real" screen
-		register uint16 *dp16 = destination_pointer_address + ( i * screen_pitch_half );
-
-		// sp is the pointer to a pixel in the SDL screen
-		register uint16 *sp16 = gfx_screen;
-		// ( i / 2 ) * source_panewidth; "i/2" for doubling lines, aka "deinterlacing"
-		sp16 += ( ( i /*>> 1*/ ) * source_panewidth );
-
 		//since we apply no scaling in X dimension here, we can simply use a single memcpy instead of a for loop
 		memcpy ( dp16, sp16, bytes_per_line );
-		//for (register uint16 j = 0; j < Width ; ++j, ++sp16) {
-		//	*dp16++ = *sp16;
-		//}
+
+		// update the references to display and sdl screen for the next run
+		dp16 += screen_pitch_half;
+		sp16 += source_panewidth;
 	} // for each height unit
 }
 
 void render_x_single(uint16* destination_pointer_address, uint16 screen_pitch_half,
                      uint16* gfx_screen, uint16 source_panewidth, int Width, int Height) {
 	uint16 height_desired = Height << 1; // height_desired = Height * 2
 	uint16 bytes_per_line = Width << 1; // bytes_per_line = Width * 2
+
+	// dp is the pointer to a pixel in the "real" screen
+	register uint16 *dp16 = destination_pointer_address;
+
 	for (register uint16 i = 0; i < height_desired; ++i) {
-		// dp is the pointer to a pixel in the "real" screen
-		register uint16 *dp16 = destination_pointer_address + ( i * screen_pitch_half );
-
 		// sp is the pointer to a pixel in the SDL screen
-		register uint16 *sp16 = gfx_screen;
-		// ( i / 2 ) * source_panewidth; "i/2" for doubling lines, aka "deinterlacing"
-		sp16 += ( ( i >> 1 ) * source_panewidth );
+		register uint16 *sp16 = gfx_screen + ( ( i >> 1 ) * source_panewidth );
 
 		//since we apply no scaling in X dimension here, we can simply use a single memcpy instead of a for loop
 		memcpy ( dp16, sp16, bytes_per_line );
-		//for (register uint16 j = 0; j < Width ; ++j, ++sp16) {
-		//	*dp16++ = *sp16;
-		//}
+
+		// update the references to display for the next run
+		dp16 += screen_pitch_half;
 	} // for each height unit
 }
 

diff --git a/unix/unix.cpp b/unix/unix.cpp
@@ -1070,7 +1070,8 @@ bool8_32 S9xDeinitUpdate ( int Width, int Height ) {
 					g_fullscreen ? SDL_SWSURFACE|SDL_FULLSCREEN : SDL_SWSURFACE);
 		}
 
-		render_x_single_xy((uint16*)(screen -> pixels) /*destination_pointer_address*/, (screen -> pitch) >> 1 /*screen_pitch_half*/,
+		render_x_single_xy((uint16*)(screen -> pixels) /*destination_pointer_address*/,
+						(screen -> pitch) >> 1 /*screen_pitch_half*/,
 						(uint16*)(GFX.Screen), source_panewidth, Width, Height);
 	}
 	else