Use multiple threads for xBRZ upscaling

Implementation relies on Concurrency Runtime, Grand Central Dispatch aka libdispatch or OpenMP depending on their availability

src/CMakeLists.txt

@@ -265,6 +265,16 @@ else( X64 )
 	set( CMAKE_CXX_FLAGS ${SAFE_CMAKE_CXX_FLAGS} )
 endif( X64 )
 
+CHECK_CXX_SOURCE_COMPILES("#include <ppl.h>
+	int main() { concurrency::parallel_for(0, 1, 1, [](int) { } ); }"
+	HAVE_PARALLEL_FOR)
+
+if( NOT HAVE_PARALLEL_FOR )
+	CHECK_CXX_SOURCE_COMPILES("#include <dispatch/dispatch.h>
+		int main() { dispatch_apply(1, dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0), ^(size_t) { }); }"
+		HAVE_DISPATCH_APPLY)
+endif()
+
 # Set up flags for MSVC
 if (MSVC)
 	set( CMAKE_CXX_FLAGS "/MP ${CMAKE_CXX_FLAGS}" )
@@ -567,6 +577,23 @@ if( HAVE_MMX )
 	endif( ZD_CMAKE_COMPILER_IS_GNUCXX_COMPATIBLE )
 endif( HAVE_MMX )
 
+if( HAVE_PARALLEL_FOR )
+	add_definitions( -DHAVE_PARALLEL_FOR=1 )
+elseif( HAVE_DISPATCH_APPLY )
+	add_definitions( -DHAVE_DISPATCH_APPLY=1 )
+else()
+	option( NO_OPENMP "Disable usage of OpenMP" OFF )
+
+	if( NOT NO_OPENMP )
+		include( FindOpenMP )
+
+		if( OPENMP_FOUND )
+			set( CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}" )
+			set( CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}" )
+		endif( OPENMP_FOUND )
+	endif( NOT NO_OPENMP )
+endif()
+
 add_custom_command( OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/xlat_parser.c ${CMAKE_CURRENT_BINARY_DIR}/xlat_parser.h
 	COMMAND lemon -C${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR}/xlat/xlat_parser.y
 	DEPENDS lemon ${CMAKE_CURRENT_SOURCE_DIR}/xlat/xlat_parser.y )

src/gl/textures/gl_hqresize.cpp

@@ -46,16 +46,44 @@
 #include "gl/xbr/xbrz.h"
 #include "gl/xbr/xbrz_old.h"
 
-#ifdef __APPLE__
+#ifdef HAVE_PARALLEL_FOR
-#	include <AvailabilityMacros.h>
-#	if MAC_OS_X_VERSION_MAX_ALLOWED >= 1060
-#		define GZ_USE_LIBDISPATCH
-#	endif // MAC_OS_X_VERSION_MAX_ALLOWED >= 1060
-#endif // __APPLE__
 
-#ifdef GZ_USE_LIBDISPATCH
+#include <ppl.h>
-#	include <dispatch/dispatch.h>
+
-#endif // GZ_USE_LIBDISPATCH
+template <typename Index, typename Function>
+inline void parallel_for(const Index count, const Index step, const Function& function)
+{
+	concurrency::parallel_for(0, count, step, function);
+}
+
+#elif defined HAVE_DISPATCH_APPLY
+
+#include <dispatch/dispatch.h>
+
+template <typename Index, typename Function>
+inline void parallel_for(const Index count, const Index step, const Function& function)
+{
+	const dispatch_queue_t queue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
+
+	dispatch_apply(count / step + 1, queue, ^(size_t sliceY)
+	{
+		function(sliceY * step);
+	});
+}
+
+#else
+
+template <typename Index, typename Function>
+inline void parallel_for(const Index count, const Index step, const Function& function)
+{
+#pragma omp parallel for
+	for (Index i = 0; i < count; i += step)
+	{
+		function(i);
+	}
+}
+
+#endif // HAVE_PARALLEL_FOR
 
 CUSTOM_CVAR(Int, gl_texture_hqresize, 0, CVAR_ARCHIVE | CVAR_GLOBALCONFIG | CVAR_NOINITCALL)
 {
@@ -87,7 +115,6 @@ CVAR (Flag, gl_texture_hqresize_textures, gl_texture_hqresize_targets, 1);
 CVAR (Flag, gl_texture_hqresize_sprites, gl_texture_hqresize_targets, 2);
 CVAR (Flag, gl_texture_hqresize_fonts, gl_texture_hqresize_targets, 4);
 
-#ifdef GZ_USE_LIBDISPATCH
 CVAR(Bool, gl_texture_hqresize_multithread, true, CVAR_ARCHIVE | CVAR_GLOBALCONFIG);
 
 CUSTOM_CVAR(Int, gl_texture_hqresize_mt_width, 16, CVAR_ARCHIVE | CVAR_GLOBALCONFIG)
@@ -101,7 +128,6 @@ CUSTOM_CVAR(Int, gl_texture_hqresize_mt_height, 4, CVAR_ARCHIVE | CVAR_GLOBALCON
 	if (self < 2)    self = 2;
 	if (self > 1024) self = 1024;
 }
-#endif // GZ_USE_LIBDISPATCH
 
 
 static void scale2x ( uint32_t* inputBuffer, uint32_t* outputBuffer, int inWidth, int inHeight )
@@ -289,7 +315,6 @@ static unsigned char *xbrzHelper( void (*xbrzFunction) ( size_t, const uint32_t*
 
 	unsigned char * newBuffer = new unsigned char[outWidth*outHeight*4];
 	
-#ifdef GZ_USE_LIBDISPATCH
 	const int thresholdWidth  = gl_texture_hqresize_mt_width;
 	const int thresholdHeight = gl_texture_hqresize_mt_height;
 	
@@ -297,16 +322,13 @@ static unsigned char *xbrzHelper( void (*xbrzFunction) ( size_t, const uint32_t*
 		&& inWidth  > thresholdWidth
 		&& inHeight > thresholdHeight)
 	{
-		const dispatch_queue_t queue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
+		parallel_for(inHeight, thresholdHeight, [=](int sliceY)
-		
-		dispatch_apply(inHeight / thresholdHeight + 1, queue, ^(size_t sliceY)
 		{
 			xbrzFunction(N, reinterpret_cast<uint32_t*>(inputBuffer), reinterpret_cast<uint32_t*>(newBuffer),
-				inWidth, inHeight, xbrz::ARGB, xbrz::ScalerCfg(), sliceY * thresholdHeight, (sliceY + 1) * thresholdHeight);
+				inWidth, inHeight, xbrz::ARGB, xbrz::ScalerCfg(), sliceY, sliceY + thresholdHeight);
 		});
 	}
 	else
-#endif // GZ_USE_LIBDISPATCH
 	{
 		xbrzFunction(N, reinterpret_cast<uint32_t*>(inputBuffer), reinterpret_cast<uint32_t*>(newBuffer),
 			inWidth, inHeight, xbrz::ARGB, xbrz::ScalerCfg(), 0, std::numeric_limits<int>::max());