Merge branch 'develop'

Conflicts:
	src/CMakeLists.txt

Update the revision number to 2.4 to resolve the conflict

README.md

@@ -2,14 +2,12 @@ clFFT
 =====
 [![Build Status](https://travis-ci.org/clMathLibraries/clFFT.png)](https://travis-ci.org/clMathLibraries/clFFT)
 
-clMath is a software library containing FFT and BLAS functions written
+clFFT is a software library containing FFT functions written
 in OpenCL. In addition to GPU devices, the libraries also support
-running on CPU devices to facilitate debugging and multicore
+running on CPU devices to facilitate debugging and heterogeneous
 programming.
 
-clMath 2.1 is the latest version and is available as source only.
-clMath's predecessor APPML 1.10 has pre-built binaries available for
-download on both Linux and Windows platforms.
+Pre-built binaries are available [here][binary_release].
 
 ## Introduction to clFFT
 
@@ -210,3 +208,4 @@ int main( void )
   [clmath-developers@googlegroups.com]: https://github.com/clMathLibraries/clFFT/wiki/Build
   [Contributing]: CONTRIBUTING.md
   [Apache License, Version 2.0]: http://www.apache.org/licenses/LICENSE-2.0
+  [binary_release]: https://github.com/clMathLibraries/clFFT/releases

src/CMakeLists.txt

@@ -1,12 +1,12 @@
 # ########################################################################
 # Copyright 2013 Advanced Micro Devices, Inc.
-# 
+#
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at
-# 
+#
 # http://www.apache.org/licenses/LICENSE-2.0
-# 
+#
 # Unless required by applicable law or agreed to in writing, software
 # distributed under the License is distributed on an "AS IS" BASIS,
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
@@ -40,7 +40,7 @@ if( NOT DEFINED CLFFT_VERSION_MAJOR )
 endif( )
 
 if( NOT DEFINED CLFFT_VERSION_MINOR )
-    set( CLFFT_VERSION_MINOR 2 )
+    set( CLFFT_VERSION_MINOR 4 )
 endif( )
 
 if( NOT DEFINED CLFFT_VERSION_PATCH )
@@ -72,7 +72,7 @@ option( BUILD_TEST "Build the library testing suite (dependency on google test,
 option( BUILD_LOADLIBRARIES "Build the optional dynamic load libraries that the FFT runtime will search for" ON )
 option( BUILD_SHARED_LIBRARY "Build shared libraries." ON)
 
-# If BOOST_ROOT is defined as an environment value, use that value and cache it so it's visible in the cmake-gui.  
+# If BOOST_ROOT is defined as an environment value, use that value and cache it so it's visible in the cmake-gui.
 # Otherwise, create a sensible default that the user can change
 if( DEFINED ENV{BOOST_ROOT} )
 	set( BOOST_ROOT $ENV{BOOST_ROOT} CACHE PATH "Environment variable defining the root of the Boost installation" )
@@ -105,8 +105,8 @@ else()
 	endif()
 endif()
 
-# These variables are meant to contain string which should be appended to the installation paths 
-# of library and executable binaries, respectively.  They are meant to be user configurable/overridable.  
+# These variables are meant to contain string which should be appended to the installation paths
+# of library and executable binaries, respectively.  They are meant to be user configurable/overridable.
 set( SUFFIX_LIB_DEFAULT "" )
 set( SUFFIX_BIN_DEFAULT "" )
 
@@ -155,7 +155,7 @@ find_package( OpenCL )
 # This will define FFTW_FOUND
 find_package( FFTW )
 
-if( (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} VERSION_LESS 2.8) ) 
+if( (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} VERSION_LESS 2.8) )
 	message( STATUS "Cmake version 2.8 or greater needed to use GTest" )
 else()
 	# This will define GTEST_FOUND
@@ -195,7 +195,7 @@ if( MSVC )
 	# CMake sets huge stack frames for windows, for whatever reason.  We go with compiler default.
 	string( REGEX REPLACE "/STACK:[0-9]+" "" CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}" )
 	string( REGEX REPLACE "/STACK:[0-9]+" "" CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS}" )
-	string( REGEX REPLACE "/STACK:[0-9]+" "" CMAKE_MODULE_LINKER_FLAGS "${CMAKE_MODULE_LINKER_FLAGS}" ) 
+	string( REGEX REPLACE "/STACK:[0-9]+" "" CMAKE_MODULE_LINKER_FLAGS "${CMAKE_MODULE_LINKER_FLAGS}" )
 
 elseif( CMAKE_COMPILER_IS_GNUCXX )
 	message( STATUS "Detected GNU fortran compiler." )
@@ -208,7 +208,12 @@ elseif( CMAKE_COMPILER_IS_GNUCXX )
 
 	set( CMAKE_CXX_FLAGS "-pthread ${CMAKE_CXX_FLAGS}" )
 	set( CMAKE_C_FLAGS "-pthread ${CMAKE_C_FLAGS}" )
-
+
+	# For linux debug builds, define the same preprocessing symbols as win to keep it simple
+	if( CMAKE_BUILD_TYPE MATCHES "Debug" )
+		add_definitions( "/D_DEBUG" )
+	endif( )
+
 	if( BUILD64 )
 		set( CMAKE_CXX_FLAGS "-m64 ${CMAKE_CXX_FLAGS}" )
 		set( CMAKE_C_FLAGS "-m64 ${CMAKE_C_FLAGS}" )
@@ -237,12 +242,12 @@ message( STATUS "CMAKE_EXE_LINKER link flags: " ${CMAKE_EXE_LINKER_FLAGS} )
 
 # configure a header file to pass the CMake version settings to the source, and package the header files in the output archive
 configure_file( "${PROJECT_SOURCE_DIR}/include/clFFT.version.h.in" "${PROJECT_BINARY_DIR}/include/clFFT.version.h" )
-install( FILES 
-			"${PROJECT_BINARY_DIR}/include/clFFT.version.h" 
+install( FILES
+			"${PROJECT_BINARY_DIR}/include/clFFT.version.h"
 			"include/clFFT.h"
 			"include/clAmdFft.h"
-			"include/clAmdFft.version.h" 
-		DESTINATION 
+			"include/clAmdFft.version.h"
+		DESTINATION
 			"./include" )
 
 
@@ -278,7 +283,7 @@ else( )
 	message( "GoogleTest unit tests will NOT be built" )
 endif( )
 
-# The following code is setting variables to control the behavior of CPack to generate our 
+# The following code is setting variables to control the behavior of CPack to generate our
 if( WIN32 )
 	set( CPACK_SOURCE_GENERATOR "ZIP" )
 	set( CPACK_GENERATOR "ZIP" )

src/FindOpenCL.cmake

@@ -56,6 +56,8 @@ find_path(OPENCL_INCLUDE_DIRS
     PATHS
         /usr/include
         /usr/local/include
+	/usr/local/cuda/include
+	/opt/cuda/include
     DOC "OpenCL header file path"
 )
 mark_as_advanced( OPENCL_INCLUDE_DIRS )
@@ -70,6 +72,8 @@ if( LIB64 )
             ${OPENCL_ROOT}/lib
             $ENV{AMDAPPSDKROOT}/lib
             $ENV{CUDA_PATH}/lib
+	    /usr/local/cuda/lib
+	    /opt/cuda/lib
         DOC "OpenCL dynamic library path"
         PATH_SUFFIXES x86_64 x64
         PATHS
@@ -82,6 +86,8 @@ else( )
             ${OPENCL_ROOT}/lib
             $ENV{AMDAPPSDKROOT}/lib
             $ENV{CUDA_PATH}/lib
+	    /usr/local/cuda/lib
+	    /opt/cuda/lib
         DOC "OpenCL dynamic library path"
         PATH_SUFFIXES x86 Win32
         PATHS

src/library/private.h

@@ -270,6 +270,8 @@ inline tstring clfftErrorStatusAsString( const cl_int& status )
 
 //	This is used to either wrap an OpenCL function call, or to explicitly check a variable for an OpenCL error condition.
 //	If an error occurs, we issue a return statement to exit the calling function.
+#if defined( _DEBUG )
+
 #define OPENCL_V( fn, msg ) \
 { \
 	clfftStatus vclStatus = static_cast< clfftStatus >( fn ); \
@@ -288,6 +290,23 @@ inline tstring clfftErrorStatusAsString( const cl_int& status )
 	} \
 }
 
+#else
+
+#define OPENCL_V( fn, msg ) \
+{ \
+	clfftStatus vclStatus = static_cast< clfftStatus >( fn ); \
+	switch( vclStatus ) \
+	{ \
+		case	CL_SUCCESS:		/**< No error */ \
+			break; \
+		default: \
+		{ \
+			return	vclStatus; \
+		} \
+	} \
+}
+#endif
+
 static inline bool IsPo2 (size_t u) {
 	return (u != 0) &&  (0 == (u & (u-1)));
 }

src/tests/cl_transform.h

@@ -151,7 +151,6 @@ class clfft {
 	//	OpenCL resources that need to be carefully managed
 	std::unique_ptr< _cl_context, clContext_deleter > context;
 	std::unique_ptr< _cl_command_queue, clCommQueue_deleter > queue;
-	std::unique_ptr< _cl_event, clEvent_deleter > an_event;
 	std::vector< std::unique_ptr< _cl_mem, clMem_deleter > > cl_mem_input;
 	std::vector< std::unique_ptr< _cl_mem, clMem_deleter > > cl_mem_output;
 	std::vector< cl_device_id >	device_id;
@@ -630,7 +629,6 @@ class clfft {
 
 		// In order to call clfftEnqueueTransform, we need to pass naked pointers
 		cl_command_queue tempQueue = queue.get( );
-		cl_event tempEvent = an_event.get( );
 		size_t buffer_size = 0;
 
 		EXPECT_EQ( CLFFT_SUCCESS, clfftBakePlan(*plan_handle, 1, &tempQueue, NULL, NULL ));
@@ -683,7 +681,7 @@ class clfft {
 
 		// In order to call clfftEnqueueTransform, we need to pass naked pointers
 		cl_command_queue tempQueue = queue.get( );
-		cl_event tempEvent = an_event.get( );
+		std::unique_ptr< _cl_event, clEvent_deleter > tempEvent;
 		std::unique_ptr< _cl_mem, clMem_deleter > intermediate_buffer;
 
 		throw_if_total_memory_footprint_is_too_large_for_device();
@@ -726,6 +724,7 @@ class clfft {
 			for( cl_uint i = 0; i < cl_mem_output.size( ); ++i )
 				tempOutput[ i ] = cl_mem_output[ i ].get( );
 
+			cl_event tevent = NULL;
 			if( buffer_size )
 			{
 				status = clfftEnqueueTransform(*plan_handle,
@@ -734,7 +733,7 @@ class clfft {
 								  &tempQueue,
 								  0,
 								  NULL,
-								  &tempEvent,
+								  &tevent,
 								  &tempInput[ 0 ],
 								  &tempOutput[ 0 ],
 								  intermediate_buffer.get() );
@@ -747,24 +746,26 @@ class clfft {
 								  &tempQueue,
 								  0,
 								  NULL,
-								  &tempEvent,
+								  &tevent,
 								  &tempInput[ 0 ],
 								  &tempOutput[ 0 ],
 								  NULL );
 			}
             clFinish(tempQueue);
+			tempEvent.reset(tevent); tevent = NULL;
 
 			if( status != CLFFT_SUCCESS )
 			{
 				throw std::runtime_error(prettyPrintclFFTStatus(status).c_str());
 			}
 
 			// wait for the kernel call to finish execution
-            cl_int wait_status = clWaitForEvents(1, &tempEvent);
+			const cl_event revent = tempEvent.get();
+            cl_int wait_status = clWaitForEvents(1, &revent);
 			if( wait_status == CL_EXEC_STATUS_ERROR_FOR_EVENTS_IN_WAIT_LIST )
 			{
 				cl_int error_code;
-				clGetEventInfo( tempEvent, CL_EVENT_COMMAND_EXECUTION_STATUS, sizeof(cl_int), &error_code, NULL );
+				clGetEventInfo( revent, CL_EVENT_COMMAND_EXECUTION_STATUS, sizeof(cl_int), &error_code, NULL );
 				throw std::runtime_error(prettyPrintclFFTStatus(error_code).c_str());
 			}
             else if( wait_status != CL_SUCCESS )