Replace broken compressor with the latest version from trunk.

Delete images more efficiently?

ChangeLog

@@ -1,3 +1,10 @@
+NVIDIA Texture Tools version 2.0.7
+ * Output correct exit codes. Fixes issue 92.
+ * Fix thread-safety errors. Fixes issue 90.
+ * Add SIMD power method. Fixes issue 94.
+ * Interact better with applications that already use CUDA.
+ * Faster CPU compression.
+
 NVIDIA Texture Tools version 2.0.6
  * Fix dll version checking.
  * Detect CUDA 2.1 and future CUDA versions correctly.

VERSION

@@ -1 +1 @@
-2.0.6
+2.0.7

src/CMakeLists.txt

@@ -105,7 +105,8 @@ ENDIF(OPENEXR_FOUND)
 FIND_PACKAGE(Qt4)
 
 # Threads
-FIND_PACKAGE(Threads)
+FIND_PACKAGE(Threads REQUIRED)
+MESSAGE(STATUS "Use thread library: ${CMAKE_THREAD_LIBS_INIT}")
 
 # configuration file
 INCLUDE(CheckIncludeFiles)

src/nvcore/Debug.cpp

@@ -38,7 +38,7 @@
 #	include <unistd.h>	// getpid
 #	include <sys/types.h>
 #	include <sys/sysctl.h>	// sysctl
-#	include <ucontext.h>
+#	include <sys/ucontext.h>
 #	undef HAVE_EXECINFO_H
 #	if defined(HAVE_EXECINFO_H) // only after OSX 10.5
 #		include <execinfo.h> // backtrace

src/nvcore/Debug.h

@@ -115,6 +115,7 @@ namespace nv
 	{
 		NVCORE_API void dumpInfo();
 	
+		// These functions are not thread safe.
 		NVCORE_API void setMessageHandler( MessageHandler * messageHandler );
 		NVCORE_API void resetMessageHandler();
 	

src/nvcore/StrLib.cpp

@@ -545,8 +545,6 @@ const char * Path::extension(const char * str)
 }
 
 
-// static
-String String::s_null(String::null);
 
 /// Clone this string
 String String::clone() const
@@ -557,13 +555,13 @@ String String::clone() const
 
 void String::setString(const char * str)
 {
-	if( str == NULL ) {
-		data = s_null.data;
+	if (str == NULL) {
+		data = NULL;
 	}
 	else {
 		allocString( str );
+		addRef();
 	}
-	addRef();
 }
 
 void String::setString(const char * str, int length)
@@ -576,11 +574,11 @@ void String::setString(const char * str, int length)
 
 void String::setString(const StringBuilder & str)
 {
-	if( str.str() == NULL ) {
-		data = s_null.data;
+	if (str.str() == NULL) {
+		data =	NULL;
 	}
 	else {
 		allocString(str);
+		addRef();
 	}
-	addRef();
 }	

src/nvcore/StrLib.h

@@ -151,15 +151,14 @@ namespace nv
 		/// Constructs a null string. @sa isNull()
 		String()
 		{
-			data = s_null.data;
-			addRef();
+			data = NULL;
 		}
 
 		/// Constructs a shared copy of str.
 		String(const String & str)
 		{
 			data = str.data;
-			addRef();
+			if (data != NULL) addRef();
 		}
 
 		/// Constructs a shared string from a standard string.
@@ -183,7 +182,6 @@ namespace nv
 		/// Dtor.
 		~String()
 		{
-			nvDebugCheck(data != NULL);
 			release();
 		}
 
@@ -220,43 +218,49 @@ namespace nv
 		/// Equal operator.
 		bool operator==( const String & str ) const
 		{
-			nvDebugCheck(data != NULL);
-			nvDebugCheck(str.data != NULL);
 			if( str.data == data ) {
 				return true;
 			}
+			if ((data == NULL) != (str.data == NULL)) {
+				return false;
+			}
 			return strcmp(data, str.data) == 0;
 		}
 
 		/// Equal operator.
 		bool operator==( const char * str ) const
 		{
-			nvDebugCheck(data != NULL);
 			nvCheck(str != NULL);	// Use isNull!
+			if (data == NULL) {
+				return false;
+			}
 			return strcmp(data, str) == 0;
 		}
 
 		/// Not equal operator.
 		bool operator!=( const String & str ) const
 		{
-			nvDebugCheck(data != NULL);
-			nvDebugCheck(str.data != NULL);
 			if( str.data == data ) {
 				return false;
 			}
+			if ((data == NULL) != (str.data == NULL)) {
+				return true;
+			}
 			return strcmp(data, str.data) != 0;
 		}
 	
 		/// Not equal operator.
 		bool operator!=( const char * str ) const
 		{
-			nvDebugCheck(data != NULL);
 			nvCheck(str != NULL);	// Use isNull!
+			if (data == NULL) {
+				return false;
+			}
 			return strcmp(data, str) != 0;
 		}
 	
 		/// Returns true if this string is the null string.
-		bool isNull() const { nvDebugCheck(data != NULL); return data == s_null.data; }
+		bool isNull() const { return data == NULL; }
 	
 		/// Return the exact length.
 		uint length() const { nvDebugCheck(data != NULL); return uint(strlen(data)); }
@@ -265,44 +269,45 @@ namespace nv
 		uint hash() const { nvDebugCheck(data != NULL); return strHash(data); }
 	
 		/// const char * cast operator.
-		operator const char * () const { nvDebugCheck(data != NULL); return data; }
+		operator const char * () const { return data; }
 	
 		/// Get string pointer.
-		const char * str() const { nvDebugCheck(data != NULL); return data; }
+		const char * str() const { return data; }
 	
 
 	private:
 
-		enum null_t { null };
-		
-		// Private constructor for null string.
-		String(null_t) {
-			setString("");
-		}
-
 		// Add reference count.
-		void addRef() {
-			nvDebugCheck(data != NULL);
-			setRefCount(getRefCount() + 1);
-		}
-		
-		// Decrease reference count.
-		void release() {
-			nvDebugCheck(data != NULL);
-
-			const uint16 count = getRefCount();
-			setRefCount(count - 1);
-			if( count - 1 == 0 ) {
-				mem::free(data - 2);
-				data = NULL;
+		void addRef()
+		{
+			if (data != NULL)
+			{
+				setRefCount(getRefCount() + 1);
 			}
 		}
 		
-		uint16 getRefCount() const {
+		// Decrease reference count.
+		void release()
+		{
+			if (data != NULL)
+			{
+				const uint16 count = getRefCount();
+				setRefCount(count - 1);
+				if (count - 1 == 0) {
+					mem::free(data - 2);
+					data = NULL;
+				}
+			}
+		}
+		
+		uint16 getRefCount() const
+		{
+			nvDebugCheck(data != NULL);
 			return *reinterpret_cast<const uint16 *>(data - 2);
 		}
 		
 		void setRefCount(uint16 count) {
+			nvDebugCheck(data != NULL);
 			nvCheck(count < 0xFFFF);
 			*reinterpret_cast<uint16 *>(const_cast<char *>(data - 2)) = uint16(count);
 		}
@@ -341,8 +346,6 @@ namespace nv
 	
 	private:
 
-		NVCORE_API static String s_null;
-
 		const char * data;
 		
 	};

src/nvimage/DirectDrawSurface.cpp

@@ -532,7 +532,7 @@ DDSHeader::DDSHeader()
 
 	// Store version information on the reserved header attributes.
 	this->reserved[9] = MAKEFOURCC('N', 'V', 'T', 'T');
-	this->reserved[10] = (2 << 16) | (0 << 8) | (6);	// major.minor.revision
+	this->reserved[10] = (2 << 16) | (0 << 8) | (7);	// major.minor.revision
 
 	this->pf.size = 32;
 	this->pf.flags = 0;

src/nvimage/Image.cpp

@@ -78,7 +78,7 @@ void Image::unwrap()
 
 void Image::free()
 {
-	::free(m_data);
+	nv::mem::free(m_data);
 	m_data = NULL;
 }
 

src/nvmath/Matrix.h

@@ -332,7 +332,7 @@ inline Matrix transpose(Matrix::Arg m)
 	Matrix r;
 	for (int i = 0; i < 4; i++)
 	{
-		for (int j = 0; j < 4; i++)
+		for (int j = 0; j < 4; j++)
 		{
 			r(i, j) = m(j, i);
 		}

src/nvtt/CompressDXT.cpp

@@ -205,9 +205,9 @@ void nv::SlowCompressor::compressDXT1(const CompressionOptions::Private & compre
 	ColorBlock rgba;
 	BlockDXT1 block;
 
-	//squish::WeightedClusterFit fit;
+	squish::WeightedClusterFit fit;
 	//squish::ClusterFit fit;
-	squish::FastClusterFit fit;
+	//squish::FastClusterFit fit;
 	fit.SetMetric(compressionOptions.colorWeight.x(), compressionOptions.colorWeight.y(), compressionOptions.colorWeight.z());
 
 	for (uint y = 0; y < h; y += 4) {
@@ -221,7 +221,7 @@ void nv::SlowCompressor::compressDXT1(const CompressionOptions::Private & compre
 			}
 			else
 			{
-				squish::ColourSet colours((uint8 *)rgba.colors(), 0);
+				squish::ColourSet colours((uint8 *)rgba.colors(), 0, true);
 				fit.SetColourSet(&colours, squish::kDxt1);
 				fit.Compress(&block);
 			}

src/nvtt/Compressor.cpp

@@ -153,7 +153,7 @@ namespace nvtt
 				if (inputOptions.isNormalMap)
 				{
 					// Expand normals to [-1, 1] range.
-				//	floatImage->expandNormals(0);
+					//	floatImage->expandNormals(0);
 				}
 				else if (inputOptions.inputGamma != 1.0f)
 				{
@@ -207,28 +207,16 @@ Compressor::Compressor() : m(*new Compressor::Private())
 {
 	// CUDA initialization.
 	m.cudaSupported = cuda::isHardwarePresent();
-	m.cudaEnabled = m.cudaSupported;
+	m.cudaEnabled = false;
+	m.cudaDevice = -1;
 
-	if (m.cudaEnabled)
-	{
-		// Select fastest CUDA device.
-		int device = cuda::getFastestDevice();
-		cuda::setDevice(device);
-		
-		m.cuda = new CudaCompressor();
-
-		if (!m.cuda->isValid())
-		{
-			m.cudaEnabled = false;
-			m.cuda = NULL;
-		}
-	}
+	enableCudaAcceleration(m.cudaSupported);
 }
 
 Compressor::~Compressor()
 {
+	enableCudaAcceleration(false);
 	delete &m;
-	cuda::exit();
 }
 
 
@@ -237,21 +225,33 @@ void Compressor::enableCudaAcceleration(bool enable)
 {
 	if (m.cudaSupported)
 	{
-		m.cudaEnabled = enable;
-	}
-
-	if (m.cudaEnabled && m.cuda == NULL)
-	{
-		// Select fastest CUDA device.
-		int device = cuda::getFastestDevice();
-		cuda::setDevice(device);
-		
-		m.cuda = new CudaCompressor();
-		
-		if (!m.cuda->isValid())
+		if (m.cudaEnabled && !enable)
 		{
 			m.cudaEnabled = false;
 			m.cuda = NULL;
+
+			if (m.cudaDevice != -1)
+			{
+				// Exit device.
+				cuda::exitDevice();
+			}
+		}
+		else if (!m.cudaEnabled && enable)
+		{
+			// Init the CUDA device. This may return -1 if CUDA was already initialized by the app.
+			m.cudaEnabled = cuda::initDevice(&m.cudaDevice);
+
+			if (m.cudaEnabled)
+			{
+				// Create compressor if initialization succeeds.
+				m.cuda = new CudaCompressor();
+
+				// But cleanup if failed.
+				if (!m.cuda->isValid())
+				{
+					enableCudaAcceleration(false);
+				}
+			}
 		}
 	}
 }
@@ -341,8 +341,8 @@ bool Compressor::Private::outputHeader(const InputOptions::Private & inputOption
 		header.setTextureCube();
 	}		
 	/*else if (inputOptions.textureType == TextureType_3D) {
-		header.setTexture3D();
-		header.setDepth(inputOptions.targetDepth);
+	header.setTexture3D();
+	header.setDepth(inputOptions.targetDepth);
 	}*/
 
 	if (compressionOptions.format == Format_RGBA)
@@ -715,29 +715,29 @@ bool Compressor::Private::compressMipmap(const Mipmap & mipmap, const InputOptio
 #endif
 
 #if defined(HAVE_ATITC)
-		if (compressionOptions.externalCompressor == "ati")
-		{
-			atiCompressDXT1(image, outputOptions);
-		}
-		else
-#endif
-		if (compressionOptions.quality == Quality_Fastest)
-		{
-			fast.compressDXT1(outputOptions);
-		}
-		else
-		{
-			if (useCuda)
+			if (compressionOptions.externalCompressor == "ati")
 			{
-				nvDebugCheck(cudaSupported);
-				cuda->setImage(image, inputOptions.alphaMode);
-				cuda->compressDXT1(compressionOptions, outputOptions);
+				atiCompressDXT1(image, outputOptions);
 			}
 			else
-			{
-				slow.compressDXT1(compressionOptions, outputOptions);
-			}
-		}
+#endif
+				if (compressionOptions.quality == Quality_Fastest)
+				{
+					fast.compressDXT1(outputOptions);
+				}
+				else
+				{
+					if (useCuda)
+					{
+						nvDebugCheck(cudaSupported);
+						cuda->setImage(image, inputOptions.alphaMode);
+						cuda->compressDXT1(compressionOptions, outputOptions);
+					}
+					else
+					{
+						slow.compressDXT1(compressionOptions, outputOptions);
+					}
+				}
 	}
 	else if (compressionOptions.format == Format_DXT1a)
 	{

src/nvtt/Compressor.h

@@ -63,10 +63,12 @@ namespace nvtt
 		bool compressMipmap(const Mipmap & mipmap, const InputOptions::Private & inputOptions, const CompressionOptions::Private & compressionOptions, const OutputOptions::Private & outputOptions) const;
 
 
+
 	public:
 
 		bool cudaSupported;
 		bool cudaEnabled;
+		int cudaDevice;
 
 		nv::AutoPtr<nv::CudaCompressor> cuda;
 

src/nvtt/InputOptions.cpp

@@ -94,7 +94,7 @@ void InputOptions::reset()
 	m.textureType = TextureType_2D;
 	m.inputFormat = InputFormat_BGRA_8UB;
 
-	m.alphaMode = AlphaMode_Transparency;
+	m.alphaMode = AlphaMode_None;
 
 	m.inputGamma = 2.2f;
 	m.outputGamma = 2.2f;

src/nvtt/cuda/CudaUtils.cpp

@@ -41,11 +41,11 @@ using namespace cuda;
 
 static bool isWindowsVista()
 {
-	OSVERSIONINFO osvi;
-	osvi.dwOSVersionInfoSize = sizeof(OSVERSIONINFO);
+OSVERSIONINFO osvi;
+osvi.dwOSVersionInfoSize = sizeof(OSVERSIONINFO);
 
-	::GetVersionEx(&osvi);
-	return osvi.dwMajorVersion >= 6;
+::GetVersionEx(&osvi);
+return osvi.dwMajorVersion >= 6;
 }
 
 
@@ -53,20 +53,20 @@ typedef BOOL (WINAPI *LPFN_ISWOW64PROCESS) (HANDLE, PBOOL);
 
 static bool isWow32()
 {
-	LPFN_ISWOW64PROCESS fnIsWow64Process = (LPFN_ISWOW64PROCESS)GetProcAddress(GetModuleHandle("kernel32"), "IsWow64Process");
+LPFN_ISWOW64PROCESS fnIsWow64Process = (LPFN_ISWOW64PROCESS)GetProcAddress(GetModuleHandle("kernel32"), "IsWow64Process");
 
-	BOOL bIsWow64 = FALSE;
+BOOL bIsWow64 = FALSE;
 
-	if (NULL != fnIsWow64Process)
-	{
-		if (!fnIsWow64Process(GetCurrentProcess(), &bIsWow64))
-		{
-			// Assume 32 bits.
-			return true;
-		}
-	}
+if (NULL != fnIsWow64Process)
+{
+if (!fnIsWow64Process(GetCurrentProcess(), &bIsWow64))
+{
+// Assume 32 bits.
+return true;
+}
+}
 
-	return !bIsWow64;
+return !bIsWow64;
 }
 
 #endif
@@ -158,6 +158,8 @@ bool nv::cuda::isHardwarePresent()
 
 	// @@ Make sure that warp size == 32
 
+	// @@ Make sure available GPU is faster than the CPU.
+
 	return count > 0;
 #else
 	return false;
@@ -180,30 +182,59 @@ int nv::cuda::deviceCount()
 	return 0;
 }
 
+
+// Make sure device meets requirements:
+// - Not an emulation device.
+// - Not an integrated device?
+// - Faster than CPU.
+bool nv::cuda::isValidDevice(int i)
+{
+#if defined HAVE_CUDA
+
+	cudaDeviceProp device_properties;
+	cudaGetDeviceProperties(&device_properties, i);
+	int gflops = device_properties.multiProcessorCount * device_properties.clockRate;
+
+	if (device_properties.major == -1 || device_properties.minor == -1) {
+		// Emulation device.
+		return false;
+	}
+
+#if CUDART_VERSION >= 2030 // 2.3
+	/*if (device_properties.integrated)
+	{
+		// Integrated devices.
+		return false;
+	}*/
+#endif
+
+	return true;
+#else
+	return false;
+#endif
+}
+
 int nv::cuda::getFastestDevice()
 {
-	int max_gflops_device = 0;
+	int max_gflops_device = -1;
 #if defined HAVE_CUDA
 	int max_gflops = 0;
 
 	const int device_count = deviceCount();
-	int current_device = 0;
-	while (current_device < device_count)
+	for (int i = 0; i < device_count; i++)
 	{
-		cudaDeviceProp device_properties;
-		cudaGetDeviceProperties(&device_properties, current_device);
-		int gflops = device_properties.multiProcessorCount * device_properties.clockRate;
-
-		if (device_properties.major != -1 && device_properties.minor != -1)
+		if (isValidDevice(i))
 		{
-			if( gflops > max_gflops )
+			cudaDeviceProp device_properties;
+			cudaGetDeviceProperties(&device_properties, i);
+			int gflops = device_properties.multiProcessorCount * device_properties.clockRate;
+
+			if (gflops > max_gflops)
 			{
 				max_gflops = gflops;
-				max_gflops_device = current_device;
+				max_gflops_device = i;
 			}
 		}
-		
-		current_device++;
 	}
 #endif
 	return max_gflops_device;
@@ -211,23 +242,53 @@ int nv::cuda::getFastestDevice()
 
 
 /// Activate the given devices.
-bool nv::cuda::setDevice(int i)
+bool nv::cuda::initDevice(int * device_ptr)
 {
-	nvCheck(i < deviceCount());
+	nvDebugCheck(device_ptr != NULL);
 #if defined HAVE_CUDA
-	cudaError_t result = cudaSetDevice(i);
 
-	if (result != cudaSuccess) {
-		nvDebug("*** CUDA Error: %s\n", cudaGetErrorString(result));
+#if CUDART_VERSION >= 2030 // 2.3
+
+	// Set device flags to yield in order to play nice with other threads and to find out if CUDA was already active.
+	cudaError_t resul = cudaSetDeviceFlags(cudaDeviceScheduleYield);
+
+#endif
+
+	int device = getFastestDevice();
+
+	if (device == -1)
+	{
+		// No device is fast enough.
+		*device_ptr = -1;
+		return false;
 	}
 
-	return result == cudaSuccess;
+	// Select CUDA device.
+	cudaError_t result = cudaSetDevice(device);
+
+	if (result == cudaErrorSetOnActiveProcess)
+	{
+		int device;
+		result = cudaGetDevice(&device);
+
+		*device_ptr = -1;  // No device to cleanup.
+		return isValidDevice(device); // Return true if device is valid.
+	}
+	else if (result != cudaSuccess)
+	{
+		nvDebug("*** CUDA Error: %s\n", cudaGetErrorString(result));
+		*device_ptr = -1;
+		return false;
+	}
+
+	*device_ptr = device;
+	return true;
 #else
 	return false;
 #endif
 }
 
-void nv::cuda::exit()
+void nv::cuda::exitDevice()
 {
 #if defined HAVE_CUDA
 	cudaError_t result = cudaThreadExit();

src/nvtt/cuda/CudaUtils.h

@@ -32,8 +32,10 @@ namespace nv
 		bool isHardwarePresent();
 		int deviceCount();
 		int getFastestDevice();
-		bool setDevice(int i);
-		void exit();
+		bool isValidDevice(int i);
+
+		bool initDevice(int * device_ptr);
+		void exitDevice();
 	};
 	
 } // nv namespace

src/nvtt/nvtt.h

@@ -73,7 +73,7 @@ namespace nvtt
 		Format_DXT1a,   // DXT1 with binary alpha.
 		Format_DXT3,
 		Format_DXT5,
-		Format_DXT5n,   // Compressed HILO: R=0, G=x, B=0, A=y
+		Format_DXT5n,   // Compressed HILO: R=1, G=y, B=0, A=x
 		
 		// DX10 formats.
 		Format_BC1 = Format_DXT1,
@@ -194,7 +194,7 @@ namespace nvtt
 		// Describe the format of the input.
 		NVTT_API void setFormat(InputFormat format);
 		
-		// Set the way the input alpha channel is interpreted. @@ Not implemented!
+		// Set the way the input alpha channel is interpreted.
 		NVTT_API void setAlphaMode(AlphaMode alphaMode);
 		
 		// Set gamma settings.

src/nvtt/squish/maths.cpp

@@ -24,6 +24,7 @@
    -------------------------------------------------------------------------- */
    
 #include "maths.h"
+#include "simd.h"
 #include <cfloat>
 
 namespace squish {
@@ -60,12 +61,39 @@ Sym3x3 ComputeWeightedCovariance( int n, Vec3 const* points, float const* weight
 }
 
 
+#define POWER_ITERATION_COUNT   8
+
+#if SQUISH_USE_SIMD
+
 Vec3 ComputePrincipleComponent( Sym3x3 const& matrix )
 {
-	const int NUM = 8;
+	Vec4 const row0( matrix[0], matrix[1], matrix[2], 0.0f );
+	Vec4 const row1( matrix[1], matrix[3], matrix[4], 0.0f );
+	Vec4 const row2( matrix[2], matrix[4], matrix[5], 0.0f );
+	Vec4 v = VEC4_CONST( 1.0f );
+	for( int i = 0; i < POWER_ITERATION_COUNT; ++i )
+	{
+		// matrix multiply
+		Vec4 w = row0*v.SplatX();
+		w = MultiplyAdd(row1, v.SplatY(), w);
+		w = MultiplyAdd(row2, v.SplatZ(), w);
 
+		// get max component from xyz in all channels
+		Vec4 a = Max(w.SplatX(), Max(w.SplatY(), w.SplatZ()));
+
+		// divide through and advance
+		v = w*Reciprocal(a);
+	}
+	return v.GetVec3();
+}
+
+#else
+
+Vec3 ComputePrincipleComponent( Sym3x3 const& matrix )
+{
 	Vec3 v(1, 1, 1);
-	for(int i = 0; i < NUM; i++) {
+	for (int i = 0; i < POWER_ITERATION_COUNT; i++)
+	{
 		float x = v.X() * matrix[0] + v.Y() * matrix[1] + v.Z() * matrix[2];
 		float y = v.X() * matrix[1] + v.Y() * matrix[3] + v.Z() * matrix[4];
 		float z = v.X() * matrix[2] + v.Y() * matrix[4] + v.Z() * matrix[5];
@@ -82,5 +110,6 @@ Vec3 ComputePrincipleComponent( Sym3x3 const& matrix )
 	return v;
 }
 
+#endif
 
 } // namespace squish

src/nvtt/tools/compress.cpp

@@ -87,7 +87,10 @@ struct MyErrorHandler : public nvtt::ErrorHandler
 {
 	virtual void error(nvtt::Error e)
 	{
+#if _DEBUG
 		nvDebugBreak();
+#endif
+		printf("Error: '%s'\n", nvtt::errorString(e));
 	}
 };
 
@@ -254,7 +257,12 @@ int main(int argc, char *argv[])
 		}
 	}
 
-	printf("NVIDIA Texture Tools - Copyright NVIDIA Corporation 2007\n\n");
+	const uint version = nvtt::version();
+	const uint major = version / 100;
+	const uint minor = version % 100;
+	
+
+	printf("NVIDIA Texture Tools %u.%u - Copyright NVIDIA Corporation 2007\n\n", major, minor);
 
 	if (input.isNull())
 	{
@@ -281,7 +289,7 @@ int main(int argc, char *argv[])
 		printf("  -bc4     \tBC4 format (ATI1)\n");
 		printf("  -bc5     \tBC5 format (3Dc/ATI2)\n\n");
 		
-		return 1;
+		return EXIT_FAILURE;
 	}
 
 	// @@ Make sure input file exists.
@@ -296,13 +304,13 @@ int main(int argc, char *argv[])
 		if (!dds.isValid())
 		{
 			fprintf(stderr, "The file '%s' is not a valid DDS file.\n", input.str());
-			return 1;
+			return EXIT_FAILURE;
 		}
 		
 		if (!dds.isSupported() || dds.isTexture3D())
 		{
 			fprintf(stderr, "The file '%s' is not a supported DDS file.\n", input.str());
-			return 1;
+			return EXIT_FAILURE;
 		}
 		
 		uint faceCount;
@@ -339,7 +347,7 @@ int main(int argc, char *argv[])
 		if (!image.load(input))
 		{
 			fprintf(stderr, "The file '%s' is not a supported image type.\n", input.str());
-			return 1;
+			return EXIT_FAILURE;
 		}
 		
 		inputOptions.setTextureLayout(nvtt::TextureType_2D, image.width(), image.height());
@@ -402,7 +410,7 @@ int main(int argc, char *argv[])
 	if (outputHandler.stream->isError())
 	{
 		fprintf(stderr, "Error opening '%s' for writting\n", output.str());
-		return 1;
+		return EXIT_FAILURE;
 	}
 
 	nvtt::Compressor compressor;
@@ -430,27 +438,16 @@ int main(int argc, char *argv[])
 //	fflush(stdout);
 //	getchar();
 
-/*	LARGE_INTEGER temp;
-	QueryPerformanceFrequency((LARGE_INTEGER*) &temp);
-	double freq = ((double) temp.QuadPart) / 1000.0;
-
-    LARGE_INTEGER start_time;
-    QueryPerformanceCounter((LARGE_INTEGER*) &start_time);
-*/
 	clock_t start = clock();
 	
-	compressor.process(inputOptions, compressionOptions, outputOptions);
-/*
-	LARGE_INTEGER end_time;
-	QueryPerformanceCounter((LARGE_INTEGER*) &end_time);
-
-	float diff_time = (float) (((double) end_time.QuadPart - (double) start_time.QuadPart) / freq);
-	printf("\rtime taken: %.3f seconds\n", diff_time/1000);
-*/
+	if (!compressor.process(inputOptions, compressionOptions, outputOptions))
+	{
+		return EXIT_FAILURE;
+	}
 
 	clock_t end = clock();
 	printf("\rtime taken: %.3f seconds\n", float(end-start) / CLOCKS_PER_SEC);
 	
-	return 0;
+	return EXIT_SUCCESS;
 }