Merge recent changes from the witness.

2010-08-31 01:39:08 +00:00 · 2010-08-31 01:39:08 +00:00 · e8dc679874
commit e8dc679874
parent 47e8f23e9f
44 changed files with 2465 additions and 2216 deletions
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@ -1,4 +1,4 @@
-CMAKE_MINIMUM_REQUIRED(VERSION 2.6.0)
+CMAKE_MINIMUM_REQUIRED(VERSION 2.8.0)
 PROJECT(NV)
 ENABLE_TESTING()

--- a/cmake/FindCUDA.cmake
+++ b/cmake/FindCUDA.cmake
--- a/cmake/FindGLEW.cmake
+++ b/cmake/FindGLEW.cmake
@ -13,25 +13,13 @@ IF (WIN32)
 		${GLEW_ROOT_DIR}/include
 		DOC "The directory where GL/glew.h resides")

-	IF (NV_SYSTEM_PROCESSOR STREQUAL "AMD64")
-		FIND_LIBRARY( GLEW_LIBRARY
-			NAMES glew64 glew64s
-			PATHS
-			$ENV{PROGRAMFILES}/GLEW/lib
-			${PROJECT_SOURCE_DIR}/src/nvgl/glew/bin
-	    	${PROJECT_SOURCE_DIR}/src/nvgl/glew/lib
-			DOC "The GLEW library (64-bit)"
-		)
-	ELSE(NV_SYSTEM_PROCESSOR STREQUAL "AMD64")
-		FIND_LIBRARY( GLEW_LIBRARY
-			NAMES glew GLEW glew32 glew32s
-			PATHS
-			$ENV{PROGRAMFILES}/GLEW/lib
-			${PROJECT_SOURCE_DIR}/src/nvgl/glew/bin
-	    	${PROJECT_SOURCE_DIR}/src/nvgl/glew/lib
-			DOC "The GLEW library"
-		)
-	ENDIF(NV_SYSTEM_PROCESSOR STREQUAL "AMD64")
+    FIND_LIBRARY( GLEW_LIBRARY
+        NAMES glew GLEW glew32 glew32s
+        PATHS
+        $ENV{PROGRAMFILES}/GLEW/lib
+        ${PROJECT_SOURCE_DIR}/src/nvgl/glew/bin
+        ${PROJECT_SOURCE_DIR}/src/nvgl/glew/lib
+        DOC "The GLEW library")
 ELSE (WIN32)
 	FIND_PATH( GLEW_INCLUDE_PATH GL/glew.h
 		/usr/include
@ -59,4 +47,3 @@ IF (GLEW_INCLUDE_PATH AND GLEW_LIBRARY)
 	SET(GLEW_LIBRARIES ${GLEW_LIBRARY})
 	SET(GLEW_FOUND "YES")
 ENDIF (GLEW_INCLUDE_PATH AND GLEW_LIBRARY)
-
--- a/cmake/make2cmake.cmake
+++ b/cmake/make2cmake.cmake
@ -1,79 +1,61 @@
-#  James Bigler, NVIDIA Corp (nvidia.com - jbigler)
-#  Abe Stephens, SCI Institute -- http://www.sci.utah.edu/~abe/FindCuda.html
-#
-#  Copyright (c) 2008 - 2009 NVIDIA Corporation.  All rights reserved.
-#
-#  Copyright (c) 2007-2009
-#  Scientific Computing and Imaging Institute, University of Utah
-#
-#  This code is licensed under the MIT License.  See the FindCUDA.cmake script
-#  for the text of the license.
-
-# The MIT License
-#
-# License for the specific language governing rights and limitations under
-# Permission is hereby granted, free of charge, to any person obtaining a
-# copy of this software and associated documentation files (the "Software"),
-# to deal in the Software without restriction, including without limitation
-# the rights to use, copy, modify, merge, publish, distribute, sublicense,
-# and/or sell copies of the Software, and to permit persons to whom the
-# Software is furnished to do so, subject to the following conditions:
-#
-# The above copyright notice and this permission notice shall be included
-# in all copies or substantial portions of the Software.
-#
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
-# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
-# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
-# DEALINGS IN THE SOFTWARE.
-#
-
-#######################################################################
-# This converts a file written in makefile syntax into one that can be included
-# by CMake.
-
-file(READ ${input_file} depend_text)
-
-if (${depend_text} MATCHES ".+")
-
-  # message("FOUND DEPENDS")
-
-  # Remember, four backslashes is escaped to one backslash in the string.
-  string(REGEX REPLACE "\\\\ " " " depend_text ${depend_text})
-
-  # This works for the nvcc -M generated dependency files.
-  string(REGEX REPLACE "^.* : " "" depend_text ${depend_text})
-  string(REGEX REPLACE "[ \\\\]*\n" ";" depend_text ${depend_text})
-
-  set(dependency_list "")
-
-  foreach(file ${depend_text})
-
-    string(REGEX REPLACE "^ +" "" file ${file})
-
-    if(NOT IS_DIRECTORY ${file})
-      # If softlinks start to matter, we should change this to REALPATH.  For now we need
-      # to flatten paths, because nvcc can generate stuff like /bin/../include instead of
-      # just /include.
-      get_filename_component(file_absolute "${file}" ABSOLUTE)
-      list(APPEND dependency_list "${file_absolute}")
-    endif(NOT IS_DIRECTORY ${file})
-
-  endforeach(file)
-
-else()
-  # message("FOUND NO DEPENDS")
-endif()
-
-# Remove the duplicate entries and sort them.
-list(REMOVE_DUPLICATES dependency_list)
-list(SORT dependency_list)
-
-foreach(file ${dependency_list})
-  set(cuda_nvcc_depend "${cuda_nvcc_depend} \"${file}\"\n")
-endforeach()
-
-file(WRITE ${output_file} "# Generated by: make2cmake.cmake\nSET(CUDA_NVCC_DEPEND\n ${cuda_nvcc_depend})\n\n")
+
+#  For more information, please see: http://software.sci.utah.edu
+#
+#  The MIT License
+#
+#  Copyright (c) 2007
+#  Scientific Computing and Imaging Institute, University of Utah
+#
+#  License for the specific language governing rights and limitations under
+#  Permission is hereby granted, free of charge, to any person obtaining a
+#  copy of this software and associated documentation files (the "Software"),
+#  to deal in the Software without restriction, including without limitation
+#  the rights to use, copy, modify, merge, publish, distribute, sublicense,
+#  and/or sell copies of the Software, and to permit persons to whom the
+#  Software is furnished to do so, subject to the following conditions:
+#
+#  The above copyright notice and this permission notice shall be included
+#  in all copies or substantial portions of the Software.
+#
+#  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+#  OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+#  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+#  THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+#  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+#  FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+#  DEALINGS IN THE SOFTWARE.
+
+# Make2cmake CMake Script
+# Abe Stephens and James Bigler
+# (c) 2007 Scientific Computing and Imaging Institute, University of Utah
+# Note that the REGEX expressions may need to be tweaked for different dependency generators.
+
+file(READ ${input_file} depend_text)
+
+if (${depend_text} MATCHES ".+")
+
+  # message("FOUND DEPENDS")
+
+  # Remember, four backslashes is escaped to one backslash in the string.
+  string(REGEX REPLACE "\\\\ " " " depend_text ${depend_text})
+
+  # This works for the nvcc -M generated dependency files.
+  string(REGEX REPLACE "^.* : " "" depend_text ${depend_text})
+  string(REGEX REPLACE "[ \\\\]*\n" ";" depend_text ${depend_text})
+
+  foreach(file ${depend_text})
+
+    string(REGEX REPLACE "^ +" "" file ${file})
+
+    if(NOT IS_DIRECTORY ${file})
+      set(cuda_nvcc_depend "${cuda_nvcc_depend} \"${file}\"\n")
+    endif(NOT IS_DIRECTORY ${file})
+
+  endforeach(file)
+
+else()
+  # message("FOUND NO DEPENDS")
+endif()
+
+
+file(WRITE ${output_file} "# Generated by: make2cmake.cmake\nSET(CUDA_NVCC_DEPEND\n ${cuda_nvcc_depend})\n\n")
--- a/cmake/parse_cubin.cmake
+++ b/cmake/parse_cubin.cmake
@ -1,39 +1,32 @@
-#  James Bigler, NVIDIA Corp (nvidia.com - jbigler)
-#  Abe Stephens, SCI Institute -- http://www.sci.utah.edu/~abe/FindCuda.html
+#  For more information, please see: http://software.sci.utah.edu
 #
-#  Copyright (c) 2008 - 2009 NVIDIA Corporation.  All rights reserved.
+#  The MIT License
 #
-#  Copyright (c) 2007-2009
+#  Copyright (c) 2007
 #  Scientific Computing and Imaging Institute, University of Utah
 #
-#  This code is licensed under the MIT License.  See the FindCUDA.cmake script
-#  for the text of the license.
-
-# The MIT License
-#
-# License for the specific language governing rights and limitations under
-# Permission is hereby granted, free of charge, to any person obtaining a
-# copy of this software and associated documentation files (the "Software"),
-# to deal in the Software without restriction, including without limitation
-# the rights to use, copy, modify, merge, publish, distribute, sublicense,
-# and/or sell copies of the Software, and to permit persons to whom the
-# Software is furnished to do so, subject to the following conditions:
-#
-# The above copyright notice and this permission notice shall be included
-# in all copies or substantial portions of the Software.
-#
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
-# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
-# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
-# DEALINGS IN THE SOFTWARE.
-#
-
-#######################################################################
-# Parses a .cubin file produced by nvcc and reports statistics about the file.
+#  License for the specific language governing rights and limitations under
+#  Permission is hereby granted, free of charge, to any person obtaining a
+#  copy of this software and associated documentation files (the "Software"),
+#  to deal in the Software without restriction, including without limitation
+#  the rights to use, copy, modify, merge, publish, distribute, sublicense,
+#  and/or sell copies of the Software, and to permit persons to whom the
+#  Software is furnished to do so, subject to the following conditions:
+#
+#  The above copyright notice and this permission notice shall be included
+#  in all copies or substantial portions of the Software.
+#
+#  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+#  OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+#  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+#  THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+#  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+#  FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+#  DEALINGS IN THE SOFTWARE.

+# .cubin Parsing CMake Script
+# Abe Stephens
+# (c) 2007 Scientific Computing and Imaging Institute, University of Utah

 file(READ ${input_file} file_text)

@ -75,26 +68,26 @@ if (${file_text} MATCHES ".+")
        if(NOT skip)

          # Registers
-          if (${entry} MATCHES "reg([ ]+)=([ ]+)([^ ]+)")
-            string(REGEX REPLACE ".*([ ]+)=([ ]+)([^ ]+)" "\\3" entry ${entry})
+          if (${entry} MATCHES "reg = ([^ ]+)")
+            string(REGEX REPLACE ".* = ([^ ]+)" "\\1" entry ${entry})
            message("Registers: ${entry}")
-          endif()
+          endif(${entry} MATCHES "reg = ([^ ]+)")

          # Local memory
-          if (${entry} MATCHES "lmem([ ]+)=([ ]+)([^ ]+)")
-            string(REGEX REPLACE ".*([ ]+)=([ ]+)([^ ]+)" "\\3" entry ${entry})
+          if (${entry} MATCHES "lmem = ([^ ]+)")
+            string(REGEX REPLACE ".* = ([^ ]+)" "\\1" entry ${entry})
            message("Local:     ${entry}")
-          endif()
+          endif(${entry} MATCHES "lmem = ([^ ]+)")

          # Shared memory
-          if (${entry} MATCHES "smem([ ]+)=([ ]+)([^ ]+)")
-            string(REGEX REPLACE ".*([ ]+)=([ ]+)([^ ]+)" "\\3" entry ${entry})
+          if (${entry} MATCHES "smem = ([^ ]+)")
+            string(REGEX REPLACE ".* = ([^ ]+)" "\\1" entry ${entry})
            message("Shared:    ${entry}")
-          endif()
+          endif(${entry} MATCHES "smem = ([^ ]+)")

          if (${entry} MATCHES "^}")
            message("")
-          endif()
+          endif(${entry} MATCHES "^}")

        endif(NOT skip)

--- a/cmake/run_nvcc.cmake
+++ b/cmake/run_nvcc.cmake
@ -1,280 +1,227 @@
-#  James Bigler, NVIDIA Corp (nvidia.com - jbigler)
-#
-#  Copyright (c) 2008 - 2009 NVIDIA Corporation.  All rights reserved.
-#
-#  This code is licensed under the MIT License.  See the FindCUDA.cmake script
-#  for the text of the license.
-
-# The MIT License
-#
-# License for the specific language governing rights and limitations under
-# Permission is hereby granted, free of charge, to any person obtaining a
-# copy of this software and associated documentation files (the "Software"),
-# to deal in the Software without restriction, including without limitation
-# the rights to use, copy, modify, merge, publish, distribute, sublicense,
-# and/or sell copies of the Software, and to permit persons to whom the
-# Software is furnished to do so, subject to the following conditions:
-#
-# The above copyright notice and this permission notice shall be included
-# in all copies or substantial portions of the Software.
-#
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
-# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
-# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
-# DEALINGS IN THE SOFTWARE.
-
-
-##########################################################################
-# This file runs the nvcc commands to produce the desired output file along with
-# the dependency file needed by CMake to compute dependencies.  In addition the
-# file checks the output of each command and if the command fails it deletes the
-# output files.
-
-# Input variables
-#
-# verbose:BOOL=<>          OFF: Be as quiet as possible (default)
-#                          ON : Describe each step
-#
-# build_configuration:STRING=<> Typically one of Debug, MinSizeRel, Release, or
-#                               RelWithDebInfo, but it should match one of the
-#                               entries in CUDA_HOST_FLAGS. This is the build
-#                               configuration used when compiling the code.  If
-#                               blank or unspecified Debug is assumed as this is
-#                               what CMake does.
-#
-# generated_file:STRING=<> File to generate.  This argument must be passed in.
-#
-# generated_cubin_file:STRING=<> File to generate.  This argument must be passed
-#                                                   in if build_cubin is true.
-
-if(NOT generated_file)
-  message(FATAL_ERROR "You must specify generated_file on the command line")
-endif()
-
-# Set these up as variables to make reading the generated file easier
-set(CMAKE_COMMAND "@CMAKE_COMMAND@")
-set(source_file "@source_file@")
-set(NVCC_generated_dependency_file "@NVCC_generated_dependency_file@")
-set(cmake_dependency_file "@cmake_dependency_file@")
-set(CUDA_make2cmake "@CUDA_make2cmake@")
-set(CUDA_parse_cubin "@CUDA_parse_cubin@")
-set(build_cubin @build_cubin@)
-# We won't actually use these variables for now, but we need to set this, in
-# order to force this file to be run again if it changes.
-set(generated_file_path "@generated_file_path@")
-set(generated_file_internal "@generated_file@")
-set(generated_cubin_file_internal "@generated_cubin_file@")
-
-set(CUDA_NVCC_EXECUTABLE "@CUDA_NVCC_EXECUTABLE@")
-set(CUDA_NVCC_FLAGS "@CUDA_NVCC_FLAGS@;;@CUDA_WRAP_OPTION_NVCC_FLAGS@")
-@CUDA_NVCC_FLAGS_CONFIG@
-set(nvcc_flags "@nvcc_flags@")
-set(CUDA_NVCC_INCLUDE_ARGS "@CUDA_NVCC_INCLUDE_ARGS@")
-set(format_flag "@format_flag@")
-
-if(build_cubin AND NOT generated_cubin_file)
-  message(FATAL_ERROR "You must specify generated_cubin_file on the command line")
-endif()
-
-# This is the list of host compilation flags.  It C or CXX should already have
-# been chosen by FindCUDA.cmake.
-@CUDA_HOST_FLAGS@
-
-# Take the compiler flags and package them up to be sent to the compiler via -Xcompiler
-set(nvcc_host_compiler_flags "")
-# If we weren't given a build_configuration, use Debug.
-if(NOT build_configuration)
-  set(build_configuration Debug)
-endif()
-string(TOUPPER "${build_configuration}" build_configuration)
-#message("CUDA_NVCC_HOST_COMPILER_FLAGS = ${CUDA_NVCC_HOST_COMPILER_FLAGS}")
-foreach(flag ${CMAKE_HOST_FLAGS} ${CMAKE_HOST_FLAGS_${build_configuration}})
-  # Extra quotes are added around each flag to help nvcc parse out flags with spaces.
-  set(nvcc_host_compiler_flags "${nvcc_host_compiler_flags},\"${flag}\"")
-endforeach()
-if (nvcc_host_compiler_flags)
-  set(nvcc_host_compiler_flags "-Xcompiler" ${nvcc_host_compiler_flags})
-endif()
-#message("nvcc_host_compiler_flags = \"${nvcc_host_compiler_flags}\"")
-# Add the build specific configuration flags
-list(APPEND CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS_${build_configuration}})
-
-if(DEFINED CCBIN)
-  set(CCBIN -ccbin "${CCBIN}")
-endif()
-
-# cuda_execute_process - Executes a command with optional command echo and status message.
-#
-#   status  - Status message to print if verbose is true
-#   command - COMMAND argument from the usual execute_process argument structure
-#   ARGN    - Remaining arguments are the command with arguments
-#
-#   CUDA_result - return value from running the command
-#
-# Make this a macro instead of a function, so that things like RESULT_VARIABLE
-# and other return variables are present after executing the process.
-macro(cuda_execute_process status command)
-  set(_command ${command})
-  if(NOT _command STREQUAL "COMMAND")
-    message(FATAL_ERROR "Malformed call to cuda_execute_process.  Missing COMMAND as second argument. (command = ${command})")
-  endif()
-  if(verbose)
-    execute_process(COMMAND "${CMAKE_COMMAND}" -E echo -- ${status})
-    # Now we need to build up our command string.  We are accounting for quotes
-    # and spaces, anything else is left up to the user to fix if they want to
-    # copy and paste a runnable command line.
-    set(cuda_execute_process_string)
-    foreach(arg ${ARGN})
-      # If there are quotes, excape them, so they come through.
-      string(REPLACE "\"" "\\\"" arg ${arg})
-      # Args with spaces need quotes around them to get them to be parsed as a single argument.
-      if(arg MATCHES " ")
-        list(APPEND cuda_execute_process_string "\"${arg}\"")
-      else()
-        list(APPEND cuda_execute_process_string ${arg})
-      endif()
-    endforeach()
-    # Echo the command
-    execute_process(COMMAND ${CMAKE_COMMAND} -E echo ${cuda_execute_process_string})
-  endif(verbose)
-  # Run the command
-  execute_process(COMMAND ${ARGN} RESULT_VARIABLE CUDA_result )
-endmacro()
-
-# Delete the target file
-cuda_execute_process(
-  "Removing ${generated_file}"
-  COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}"
-  )
-
-# For CUDA 2.3 and below, -G -M doesn't work, so remove the -G flag
-# for dependency generation and hope for the best.
-set(depends_CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}")
-set(CUDA_VERSION @CUDA_VERSION@)
-if(CUDA_VERSION VERSION_LESS "3.0")
-  cmake_policy(PUSH)
-  # CMake policy 0007 NEW states that empty list elements are not
-  # ignored.  I'm just setting it to avoid the warning that's printed.
-  cmake_policy(SET CMP0007 NEW)
-  # Note that this will remove all occurances of -G.
-  list(REMOVE_ITEM depends_CUDA_NVCC_FLAGS "-G")
-  cmake_policy(POP)
-endif()
-
-# nvcc doesn't define __CUDACC__ for some reason when generating dependency files.  This
-# can cause incorrect dependencies when #including files based on this macro which is
-# defined in the generating passes of nvcc invokation.  We will go ahead and manually
-# define this for now until a future version fixes this bug.
-set(CUDACC_DEFINE -D__CUDACC__)
-
-# Generate the dependency file
-cuda_execute_process(
-  "Generating dependency file: ${NVCC_generated_dependency_file}"
-  COMMAND "${CUDA_NVCC_EXECUTABLE}"
-  -M
-  ${CUDACC_DEFINE}
-  "${source_file}"
-  -o "${NVCC_generated_dependency_file}"
-  ${CCBIN}
-  ${nvcc_flags}
-  ${nvcc_host_compiler_flags}
-  ${depends_CUDA_NVCC_FLAGS}
-  -DNVCC
-  ${CUDA_NVCC_INCLUDE_ARGS}
-  )
-
-if(CUDA_result)
-  message(FATAL_ERROR "Error generating ${generated_file}")
-endif()
-
-# Generate the cmake readable dependency file to a temp file.  Don't put the
-# quotes just around the filenames for the input_file and output_file variables.
-# CMake will pass the quotes through and not be able to find the file.
-cuda_execute_process(
-  "Generating temporary cmake readable file: ${cmake_dependency_file}.tmp"
-  COMMAND "${CMAKE_COMMAND}"
-  -D "input_file:FILEPATH=${NVCC_generated_dependency_file}"
-  -D "output_file:FILEPATH=${cmake_dependency_file}.tmp"
-  -P "${CUDA_make2cmake}"
-  )
-
-if(CUDA_result)
-  message(FATAL_ERROR "Error generating ${generated_file}")
-endif()
-
-# Copy the file if it is different
-cuda_execute_process(
-  "Copy if different ${cmake_dependency_file}.tmp to ${cmake_dependency_file}"
-  COMMAND "${CMAKE_COMMAND}" -E copy_if_different "${cmake_dependency_file}.tmp" "${cmake_dependency_file}"
-  )
-
-if(CUDA_result)
-  message(FATAL_ERROR "Error generating ${generated_file}")
-endif()
-
-# Delete the temporary file
-cuda_execute_process(
-  "Removing ${cmake_dependency_file}.tmp and ${NVCC_generated_dependency_file}"
-  COMMAND "${CMAKE_COMMAND}" -E remove "${cmake_dependency_file}.tmp" "${NVCC_generated_dependency_file}"
-  )
-
-if(CUDA_result)
-  message(FATAL_ERROR "Error generating ${generated_file}")
-endif()
-
-# Generate the code
-cuda_execute_process(
-  "Generating ${generated_file}"
-  COMMAND "${CUDA_NVCC_EXECUTABLE}"
-  "${source_file}"
-  ${format_flag} -o "${generated_file}"
-  ${CCBIN}
-  ${nvcc_flags}
-  ${nvcc_host_compiler_flags}
-  ${CUDA_NVCC_FLAGS}
-  -DNVCC
-  ${CUDA_NVCC_INCLUDE_ARGS}
-  )
-
-if(CUDA_result)
-  # Since nvcc can sometimes leave half done files make sure that we delete the output file.
-  cuda_execute_process(
-    "Removing ${generated_file}"
-    COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}"
-    )
-  message(FATAL_ERROR "Error generating file ${generated_file}")
-else()
-  if(verbose)
-    message("Generated ${generated_file} successfully.")
-  endif()
-endif()
-
-# Cubin resource report commands.
-if( build_cubin )
-  # Run with -cubin to produce resource usage report.
-  cuda_execute_process(
-    "Generating ${generated_cubin_file}"
-    COMMAND "${CUDA_NVCC_EXECUTABLE}"
-    "${source_file}"
-    ${CUDA_NVCC_FLAGS}
-    ${nvcc_flags}
-    ${CCBIN}
-    ${nvcc_host_compiler_flags}
-    -DNVCC
-    -cubin
-    -o "${generated_cubin_file}"
-    ${CUDA_NVCC_INCLUDE_ARGS}
-    )
-
-  # Execute the parser script.
-  cuda_execute_process(
-    "Executing the parser script"
-    COMMAND  "${CMAKE_COMMAND}"
-    -D "input_file:STRING=${generated_cubin_file}"
-    -P "${CUDA_parse_cubin}"
-    )
-
-endif( build_cubin )
+# This file runs the nvcc commands to produce the desired output file along with
+# the dependency file needed by CMake to compute dependencies.  In addition the
+# file checks the output of each command and if the command fails it deletes the
+# output files.
+
+# Input variables
+#
+# verbose:BOOL=<>          OFF: Be as quiet as possible (default)
+#                          ON : Describe each step
+#
+# build_configuration:STRING=<> Typically one of Debug, MinSizeRel, Release, or
+#                               RelWithDebInfo, but it should match one of the
+#                               entries in CUDA_HOST_FLAGS. This is the build
+#                               configuration used when compiling the code.  If
+#                               blank or unspecified Debug is assumed as this is
+#                               what CMake does.
+#
+# generated_file:STRING=<> File to generate.  This argument must be passed in.
+#
+# generated_cubin_file:STRING=<> File to generate.  This argument must be passed
+#                                                   in if build_cubin is true.
+
+if(NOT generated_file)
+  message(FATAL_ERROR "You must specify generated_file on the command line")
+endif()
+
+# Set these up as variables to make reading the generated file easier
+set(CMAKE_COMMAND "@CMAKE_COMMAND@")
+set(source_file "@source_file@")
+set(NVCC_generated_dependency_file "@NVCC_generated_dependency_file@")
+set(cmake_dependency_file "@cmake_dependency_file@")
+set(CUDA_make2cmake "@CUDA_make2cmake@")
+set(CUDA_parse_cubin "@CUDA_parse_cubin@")
+set(build_cubin @build_cubin@)
+# We won't actually use these variables for now, but we need to set this, in
+# order to force this file to be run again if it changes.
+set(generated_file_path "@generated_file_path@")
+set(generated_file_internal "@generated_file@")
+set(generated_cubin_file_internal "@generated_cubin_file@")
+
+set(CUDA_NVCC_EXECUTABLE "@CUDA_NVCC_EXECUTABLE@")
+set(CUDA_NVCC_FLAGS "@CUDA_NVCC_FLAGS@;@CUDA_WRAP_OPTION_NVCC_FLAGS@")
+@CUDA_NVCC_FLAGS_CONFIG@
+set(nvcc_flags "@nvcc_flags@")
+set(CUDA_NVCC_INCLUDE_ARGS "@CUDA_NVCC_INCLUDE_ARGS@")
+set(format_flag "@format_flag@")
+
+if(build_cubin AND NOT generated_cubin_file)
+  message(FATAL_ERROR "You must specify generated_cubin_file on the command line")
+endif()
+
+# This is the list of host compilation flags.  It C or CXX should already have
+# been chosen by FindCUDA.cmake.
+@CUDA_HOST_FLAGS@
+
+# Take the compiler flags and package them up to be sent to the compiler via -Xcompiler
+set(nvcc_host_compiler_flags "")
+# If we weren't given a build_configuration, use Debug.
+if(NOT build_configuration)
+  set(build_configuration Debug)
+endif()
+string(TOUPPER "${build_configuration}" build_configuration)
+#message("CUDA_NVCC_HOST_COMPILER_FLAGS = ${CUDA_NVCC_HOST_COMPILER_FLAGS}")
+foreach(flag ${CMAKE_HOST_FLAGS} ${CMAKE_HOST_FLAGS_${build_configuration}})
+  # Extra quotes are added around each flag to help nvcc parse out flags with spaces.
+  set(nvcc_host_compiler_flags "${nvcc_host_compiler_flags},\"${flag}\"")
+endforeach()
+if (nvcc_host_compiler_flags)
+  set(nvcc_host_compiler_flags "-Xcompiler" ${nvcc_host_compiler_flags})
+endif()
+#message("nvcc_host_compiler_flags = \"${nvcc_host_compiler_flags}\"")
+# Add the build specific configuration flags
+list(APPEND CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS_${build_configuration}})
+
+if(DEFINED CCBIN)
+  set(CCBIN -ccbin "${CCBIN}")
+endif()
+
+# cuda_execute_process - Executes a command with optional command echo and status message.
+#
+#   status  - Status message to print if verbose is true
+#   command - COMMAND argument from the usual execute_process argument structure
+#   ARGN    - Remaining arguments are the command with arguments
+#
+#   CUDA_result - return value from running the command
+#
+# Make this a macro instead of a function, so that things like RESULT_VARIABLE
+# and other return variables are present after executing the process.
+macro(cuda_execute_process status command)
+  set(_command ${command})
+  if(NOT _command STREQUAL "COMMAND")
+    message(FATAL_ERROR "Malformed call to cuda_execute_process.  Missing COMMAND as second argument. (command = ${command})")
+  endif()
+  if(verbose)
+    execute_process(COMMAND "${CMAKE_COMMAND}" -E echo -- ${status})
+    # Now we need to build up our command string.  We are accounting for quotes
+    # and spaces, anything else is left up to the user to fix if they want to
+    # copy and paste a runnable command line.
+    set(cuda_execute_process_string)
+    foreach(arg ${ARGN})
+      # If there are quotes, excape them, so they come through.
+      string(REPLACE "\"" "\\\"" arg ${arg})
+      # Args with spaces need quotes around them to get them to be parsed as a single argument.
+      if(arg MATCHES " ")
+        list(APPEND cuda_execute_process_string "\"${arg}\"")
+      else()
+        list(APPEND cuda_execute_process_string ${arg})
+      endif()
+    endforeach()
+    # Echo the command
+    execute_process(COMMAND ${CMAKE_COMMAND} -E echo ${cuda_execute_process_string})
+  endif(verbose)
+  # Run the command
+  execute_process(COMMAND ${ARGN} RESULT_VARIABLE CUDA_result )
+endmacro()
+
+# Delete the target file
+cuda_execute_process(
+  "Removing ${generated_file}"
+  COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}"
+  )
+
+# Generate the dependency file
+cuda_execute_process(
+  "Generating dependency file: ${NVCC_generated_dependency_file}"
+  COMMAND "${CUDA_NVCC_EXECUTABLE}"
+  "${source_file}"
+  ${CUDA_NVCC_FLAGS}
+  ${nvcc_flags}
+  ${CCBIN}
+  ${nvcc_host_compiler_flags}
+  -DNVCC
+  -M
+  -o "${NVCC_generated_dependency_file}"
+  ${CUDA_NVCC_INCLUDE_ARGS}
+  )
+
+if(CUDA_result)
+  message(FATAL_ERROR "Error generating ${generated_file}")
+endif()
+
+# Generate the cmake readable dependency file to a temp file.  Don't put the
+# quotes just around the filenames for the input_file and output_file variables.
+# CMake will pass the quotes through and not be able to find the file.
+cuda_execute_process(
+  "Generating temporary cmake readable file: ${cmake_dependency_file}.tmp"
+  COMMAND "${CMAKE_COMMAND}"
+  -D "input_file:FILEPATH=${NVCC_generated_dependency_file}"
+  -D "output_file:FILEPATH=${cmake_dependency_file}.tmp"
+  -P "${CUDA_make2cmake}"
+  )
+
+if(CUDA_result)
+  message(FATAL_ERROR "Error generating ${generated_file}")
+endif()
+
+# Copy the file if it is different
+cuda_execute_process(
+  "Copy if different ${cmake_dependency_file}.tmp to ${cmake_dependency_file}"
+  COMMAND "${CMAKE_COMMAND}" -E copy_if_different "${cmake_dependency_file}.tmp" "${cmake_dependency_file}"
+  )
+
+if(CUDA_result)
+  message(FATAL_ERROR "Error generating ${generated_file}")
+endif()
+
+# Delete the temporary file
+cuda_execute_process(
+  "Removing ${cmake_dependency_file}.tmp and ${NVCC_generated_dependency_file}"
+  COMMAND "${CMAKE_COMMAND}" -E remove "${cmake_dependency_file}.tmp" "${NVCC_generated_dependency_file}"
+  )
+
+if(CUDA_result)
+  message(FATAL_ERROR "Error generating ${generated_file}")
+endif()
+
+# Generate the code
+cuda_execute_process(
+  "Generating ${generated_file}"
+  COMMAND "${CUDA_NVCC_EXECUTABLE}"
+  "${source_file}"
+  ${CUDA_NVCC_FLAGS}
+  ${nvcc_flags}
+  ${CCBIN}
+  ${nvcc_host_compiler_flags}
+  -DNVCC
+  ${format_flag} -o "${generated_file}"
+  ${CUDA_NVCC_INCLUDE_ARGS}
+  )
+
+if(CUDA_result)
+  # Since nvcc can sometimes leave half done files make sure that we delete the output file.
+  cuda_execute_process(
+    "Removing ${generated_file}"
+    COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}"
+    )
+  message(FATAL_ERROR "Error generating file ${generated_file}")
+else()
+  message("Generated ${generated_file} successfully.")
+endif()
+
+# Cubin resource report commands.
+if( build_cubin )
+  # Run with -cubin to produce resource usage report.
+  cuda_execute_process(
+    "Generating ${generated_cubin_file}"
+    COMMAND "${CUDA_NVCC_EXECUTABLE}"
+    "${source_file}"
+    ${CUDA_NVCC_FLAGS}
+    ${nvcc_flags}
+    ${CCBIN}
+    ${nvcc_host_compiler_flags}
+    -DNVCC
+    -cubin
+    -o "${generated_cubin_file}"
+    ${CUDA_NVCC_INCLUDE_ARGS}
+    )
+
+  # Execute the parser script.
+  cuda_execute_process(
+    "Executing the parser script"
+    COMMAND  "${CMAKE_COMMAND}"
+    -D "input_file:STRING=${generated_cubin_file}"
+    -P "${CUDA_parse_cubin}"
+    )
+
+endif( build_cubin )
--- a/project/vc9/imperativeapi/imperativeapi.vcproj
+++ b/project/vc9/imperativeapi/imperativeapi.vcproj
@ -89,80 +89,6 @@
 				Name="VCPostBuildEventTool"
 			/>
 		</Configuration>
-		<Configuration
-			Name="Release|Win32"
-			OutputDirectory="$(ConfigurationName)\$(PlatformName)"
-			IntermediateDirectory="$(ConfigurationName)\$(PlatformName)"
-			ConfigurationType="1"
-			InheritedPropertySheets="$(SolutionDir)\nvtt.vsprops"
-			CharacterSet="2"
-			WholeProgramOptimization="1"
-			>
-			<Tool
-				Name="VCPreBuildEventTool"
-			/>
-			<Tool
-				Name="VCCustomBuildTool"
-			/>
-			<Tool
-				Name="VCXMLDataGeneratorTool"
-			/>
-			<Tool
-				Name="VCWebServiceProxyGeneratorTool"
-			/>
-			<Tool
-				Name="VCMIDLTool"
-			/>
-			<Tool
-				Name="VCCLCompilerTool"
-				Optimization="2"
-				EnableIntrinsicFunctions="true"
-				RuntimeLibrary="2"
-				EnableFunctionLevelLinking="true"
-				WarningLevel="3"
-				DebugInformationFormat="3"
-			/>
-			<Tool
-				Name="VCManagedResourceCompilerTool"
-			/>
-			<Tool
-				Name="VCResourceCompilerTool"
-			/>
-			<Tool
-				Name="VCPreLinkEventTool"
-			/>
-			<Tool
-				Name="VCLinkerTool"
-				AdditionalDependencies="libpng.lib jpeg.lib tiff.lib FreeImage.lib"
-				OutputFile="$(SolutionDir)\$(ConfigurationName).$(PlatformName)\bin\$(ProjectName).exe"
-				AdditionalLibraryDirectories="$(GnuWinDir)\lib; $(FreeImageDir)"
-				GenerateDebugInformation="true"
-				OptimizeReferences="2"
-				EnableCOMDATFolding="2"
-				TargetMachine="1"
-			/>
-			<Tool
-				Name="VCALinkTool"
-			/>
-			<Tool
-				Name="VCManifestTool"
-			/>
-			<Tool
-				Name="VCXDCMakeTool"
-			/>
-			<Tool
-				Name="VCBscMakeTool"
-			/>
-			<Tool
-				Name="VCFxCopTool"
-			/>
-			<Tool
-				Name="VCAppVerifierTool"
-			/>
-			<Tool
-				Name="VCPostBuildEventTool"
-			/>
-		</Configuration>
 		<Configuration
 			Name="Debug|x64"
 			OutputDirectory="$(ConfigurationName)\$(PlatformName)"
@ -234,6 +160,85 @@
 				Name="VCPostBuildEventTool"
 			/>
 		</Configuration>
+		<Configuration
+			Name="Release|Win32"
+			OutputDirectory="$(ConfigurationName)\$(PlatformName)"
+			IntermediateDirectory="$(ConfigurationName)\$(PlatformName)"
+			ConfigurationType="1"
+			InheritedPropertySheets="$(SolutionDir)\nvtt.vsprops"
+			CharacterSet="2"
+			WholeProgramOptimization="1"
+			>
+			<Tool
+				Name="VCPreBuildEventTool"
+			/>
+			<Tool
+				Name="VCCustomBuildTool"
+			/>
+			<Tool
+				Name="VCXMLDataGeneratorTool"
+			/>
+			<Tool
+				Name="VCWebServiceProxyGeneratorTool"
+			/>
+			<Tool
+				Name="VCMIDLTool"
+			/>
+			<Tool
+				Name="VCCLCompilerTool"
+				Optimization="3"
+				InlineFunctionExpansion="0"
+				EnableIntrinsicFunctions="true"
+				FavorSizeOrSpeed="0"
+				OmitFramePointers="true"
+				StringPooling="true"
+				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
+				WarningLevel="3"
+				DebugInformationFormat="3"
+			/>
+			<Tool
+				Name="VCManagedResourceCompilerTool"
+			/>
+			<Tool
+				Name="VCResourceCompilerTool"
+			/>
+			<Tool
+				Name="VCPreLinkEventTool"
+			/>
+			<Tool
+				Name="VCLinkerTool"
+				AdditionalDependencies="libpng.lib jpeg.lib tiff.lib FreeImage.lib"
+				OutputFile="$(SolutionDir)\$(ConfigurationName).$(PlatformName)\bin\$(ProjectName).exe"
+				AdditionalLibraryDirectories="$(GnuWinDir)\lib; $(FreeImageDir)"
+				GenerateDebugInformation="true"
+				OptimizeReferences="2"
+				EnableCOMDATFolding="2"
+				TargetMachine="1"
+			/>
+			<Tool
+				Name="VCALinkTool"
+			/>
+			<Tool
+				Name="VCManifestTool"
+			/>
+			<Tool
+				Name="VCXDCMakeTool"
+			/>
+			<Tool
+				Name="VCBscMakeTool"
+			/>
+			<Tool
+				Name="VCFxCopTool"
+			/>
+			<Tool
+				Name="VCAppVerifierTool"
+			/>
+			<Tool
+				Name="VCPostBuildEventTool"
+			/>
+		</Configuration>
 		<Configuration
 			Name="Release|x64"
 			OutputDirectory="$(ConfigurationName)\$(PlatformName)"
@ -261,8 +266,11 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
-				Optimization="2"
+				Optimization="3"
 				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
+				StringPooling="true"
 				RuntimeLibrary="2"
 				EnableFunctionLevelLinking="true"
 				WarningLevel="3"
--- a/project/vc9/nvassemble/nvassemble.vcproj
+++ b/project/vc9/nvassemble/nvassemble.vcproj
@ -192,11 +192,16 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
+				InlineFunctionExpansion="0"
+				EnableIntrinsicFunctions="true"
+				FavorSizeOrSpeed="0"
+				OmitFramePointers="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVTT_SHARED;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 			/>
 			<Tool
 				Name="VCManagedResourceCompilerTool"
@ -266,9 +271,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVTT_SHARED;__SSE2__;__SSE__;__MMX__"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 			/>
 			<Tool
 				Name="VCManagedResourceCompilerTool"
--- a/project/vc9/nvcompress/nvcompress.vcproj
+++ b/project/vc9/nvcompress/nvcompress.vcproj
@ -202,16 +202,17 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
-				InlineFunctionExpansion="2"
+				InlineFunctionExpansion="0"
 				EnableIntrinsicFunctions="true"
-				FavorSizeOrSpeed="1"
+				FavorSizeOrSpeed="0"
 				OmitFramePointers="true"
 				EnableFiberSafeOptimizations="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVTT_SHARED;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
@ -287,9 +288,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVDXT_SHARED"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 			/>
--- a/project/vc9/nvcore/nvcore.vcproj
+++ b/project/vc9/nvcore/nvcore.vcproj
@ -173,16 +173,17 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
-				InlineFunctionExpansion="2"
+				InlineFunctionExpansion="0"
 				EnableIntrinsicFunctions="true"
-				FavorSizeOrSpeed="1"
+				FavorSizeOrSpeed="0"
 				OmitFramePointers="true"
 				EnableFiberSafeOptimizations="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_LIB;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
@ -242,9 +243,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_LIB;__SSE2__;__SSE__;__MMX__"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
@ -281,106 +288,98 @@
 	<References>
 	</References>
 	<Files>
-		<Filter
-			Name="Source Files"
-			Filter="cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx"
-			UniqueIdentifier="{4FC737F1-C7A5-4376-A066-2A32D752A2FF}"
+		<File
+			RelativePath="..\..\..\src\nvcore\Algorithms.h"
 			>
-			<File
-				RelativePath="..\..\..\src\nvcore\Debug.cpp"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\FileSystem.cpp"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\Library.cpp"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\Memory.cpp"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\StrLib.cpp"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\TextReader.cpp"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\TextWriter.cpp"
-				>
-			</File>
-		</Filter>
-		<Filter
-			Name="Header Files"
-			Filter="h;hpp;hxx;hm;inl;inc;xsd"
-			UniqueIdentifier="{93995380-89BD-4b04-88EB-625FBE52EBFB}"
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Array.h"
 			>
-			<File
-				RelativePath="..\..\..\src\nvcore\Algorithms.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\Containers.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\Debug.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\DefsVcWin32.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\FileSystem.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\HashMap.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\Library.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\Memory.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\nvcore.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\Ptr.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\RefCounted.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\StrLib.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\TextReader.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\TextWriter.h"
-				>
-			</File>
-			<File
-				RelativePath="..\..\..\src\nvcore\Timer.h"
-				>
-			</File>
-		</Filter>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Containers.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Debug.cpp"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Debug.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\DefsVcWin32.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\FileSystem.cpp"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\FileSystem.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\HashMap.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Library.cpp"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Library.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Memory.cpp"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Memory.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\nvcore.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Ptr.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\RefCounted.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\StrLib.cpp"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\StrLib.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\TextReader.cpp"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\TextReader.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\TextWriter.cpp"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\TextWriter.h"
+			>
+		</File>
+		<File
+			RelativePath="..\..\..\src\nvcore\Timer.h"
+			>
+		</File>
 	</Files>
 	<Globals>
 	</Globals>
--- a/project/vc9/nvddsinfo/nvddsinfo.vcproj
+++ b/project/vc9/nvddsinfo/nvddsinfo.vcproj
@ -192,11 +192,16 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
+				InlineFunctionExpansion="0"
+				EnableIntrinsicFunctions="true"
+				FavorSizeOrSpeed="0"
+				OmitFramePointers="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVTT_SHARED;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 			/>
 			<Tool
 				Name="VCManagedResourceCompilerTool"
@ -265,9 +270,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVTT_SHARED;__SSE2__;__SSE__;__MMX__"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 			/>
 			<Tool
 				Name="VCManagedResourceCompilerTool"
--- a/project/vc9/nvdecompress/nvdecompress.vcproj
+++ b/project/vc9/nvdecompress/nvdecompress.vcproj
@ -202,11 +202,16 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
+				InlineFunctionExpansion="0"
+				EnableIntrinsicFunctions="true"
+				FavorSizeOrSpeed="0"
+				OmitFramePointers="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVTT_SHARED;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
@ -283,9 +288,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVDXT_SHARED"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 			/>
--- a/project/vc9/nvimage/nvimage.vcproj
+++ b/project/vc9/nvimage/nvimage.vcproj
@ -173,14 +173,16 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
-				InlineFunctionExpansion="2"
+				InlineFunctionExpansion="0"
 				EnableIntrinsicFunctions="true"
+				FavorSizeOrSpeed="0"
 				OmitFramePointers="true"
 				AdditionalIncludeDirectories="$(GnuWinDir)/include; $(FreeImageDir)"
 				PreprocessorDefinitions="WIN32;NDEBUG;_LIB;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
@ -240,9 +242,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_LIB;__SSE2__;__SSE__;__MMX__"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
--- a/project/vc9/nvimgdiff/nvimgdiff.vcproj
+++ b/project/vc9/nvimgdiff/nvimgdiff.vcproj
@ -202,11 +202,16 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
+				InlineFunctionExpansion="0"
+				EnableIntrinsicFunctions="true"
+				FavorSizeOrSpeed="0"
+				OmitFramePointers="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVTT_SHARED;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
@ -283,9 +288,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;NVDXT_SHARED"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 			/>
--- a/project/vc9/nvmath/nvmath.vcproj
+++ b/project/vc9/nvmath/nvmath.vcproj
@ -173,16 +173,17 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
-				InlineFunctionExpansion="2"
+				InlineFunctionExpansion="0"
 				EnableIntrinsicFunctions="true"
-				FavorSizeOrSpeed="1"
+				FavorSizeOrSpeed="0"
 				OmitFramePointers="true"
 				EnableFiberSafeOptimizations="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_LIB;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
@ -242,9 +243,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_LIB;__SSE2__;__SSE__;__MMX__"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
--- a/project/vc9/nvtt.sln
+++ b/project/vc9/nvtt.sln
@ -6,7 +6,7 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "nvtt", "nvtt\nvtt.vcproj",
 		{CE017322-01FC-4851-9C8B-64E9A8E26C38} = {CE017322-01FC-4851-9C8B-64E9A8E26C38}
 		{F143D180-D4C4-4037-B3DE-BE89A21C8D1D} = {F143D180-D4C4-4037-B3DE-BE89A21C8D1D}
 		{4046F392-A18B-4C66-9639-3EABFFF5D531} = {4046F392-A18B-4C66-9639-3EABFFF5D531}
-		{C33787E3-5564-4834-9FE3-A9020455A669} = {C33787E3-5564-4834-9FE3-A9020455A669}
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153} = {4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}
 		{50C465FE-B308-42BC-894D-89484482AF06} = {50C465FE-B308-42BC-894D-89484482AF06}
 	EndProjectSection
 EndProject
@ -81,7 +81,7 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "imperativeapi", "imperative
 		{1AEB7681-57D8-48EE-813D-5C41CC38B647} = {1AEB7681-57D8-48EE-813D-5C41CC38B647}
 	EndProjectSection
 EndProject
-Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "bc6h", "bc6h\bc6h.vcproj", "{C33787E3-5564-4834-9FE3-A9020455A669}"
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "bc6h", "bc6h\bc6h.vcproj", "{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}"
 EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
@ -429,30 +429,30 @@ Global
 		{7DCF280E-702B-49F3-84A7-AE7E146384D6}.Release|Win32.Build.0 = Release|Win32
 		{7DCF280E-702B-49F3-84A7-AE7E146384D6}.Release|x64.ActiveCfg = Release|x64
 		{7DCF280E-702B-49F3-84A7-AE7E146384D6}.Release|x64.Build.0 = Release|x64
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug (no cuda)|Mixed Platforms.ActiveCfg = Debug|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug (no cuda)|Mixed Platforms.Build.0 = Debug|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug (no cuda)|Win32.ActiveCfg = Debug|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug (no cuda)|Win32.Build.0 = Debug|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug (no cuda)|x64.ActiveCfg = Debug|x64
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug (no cuda)|x64.Build.0 = Debug|x64
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug|Mixed Platforms.ActiveCfg = Debug|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug|Mixed Platforms.Build.0 = Debug|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug|Win32.ActiveCfg = Debug|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug|Win32.Build.0 = Debug|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug|x64.ActiveCfg = Debug|x64
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Debug|x64.Build.0 = Debug|x64
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release (no cuda)|Mixed Platforms.ActiveCfg = Release|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release (no cuda)|Mixed Platforms.Build.0 = Release|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release (no cuda)|Win32.ActiveCfg = Release|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release (no cuda)|Win32.Build.0 = Release|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release (no cuda)|x64.ActiveCfg = Release|x64
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release (no cuda)|x64.Build.0 = Release|x64
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release|Mixed Platforms.ActiveCfg = Release|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release|Mixed Platforms.Build.0 = Release|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release|Win32.ActiveCfg = Release|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release|Win32.Build.0 = Release|Win32
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release|x64.ActiveCfg = Release|x64
-		{C33787E3-5564-4834-9FE3-A9020455A669}.Release|x64.Build.0 = Release|x64
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug (no cuda)|Mixed Platforms.ActiveCfg = Debug|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug (no cuda)|Mixed Platforms.Build.0 = Debug|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug (no cuda)|Win32.ActiveCfg = Debug|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug (no cuda)|Win32.Build.0 = Debug|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug (no cuda)|x64.ActiveCfg = Debug|x64
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug (no cuda)|x64.Build.0 = Debug|x64
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug|Mixed Platforms.ActiveCfg = Debug|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug|Mixed Platforms.Build.0 = Debug|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug|Win32.ActiveCfg = Debug|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug|Win32.Build.0 = Debug|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug|x64.ActiveCfg = Debug|x64
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Debug|x64.Build.0 = Debug|x64
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release (no cuda)|Mixed Platforms.ActiveCfg = Release|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release (no cuda)|Mixed Platforms.Build.0 = Release|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release (no cuda)|Win32.ActiveCfg = Release|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release (no cuda)|Win32.Build.0 = Release|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release (no cuda)|x64.ActiveCfg = Release|x64
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release (no cuda)|x64.Build.0 = Release|x64
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release|Mixed Platforms.ActiveCfg = Release|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release|Mixed Platforms.Build.0 = Release|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release|Win32.ActiveCfg = Release|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release|Win32.Build.0 = Release|Win32
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release|x64.ActiveCfg = Release|x64
+		{4C16A4CC-8B4E-470C-B9AE-B40AF26C5153}.Release|x64.Build.0 = Release|x64
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE
--- a/project/vc9/nvtt.vsprops
+++ b/project/vc9/nvtt.vsprops
@ -12,16 +12,24 @@
 		Name="SrcDir"
 		Value="$(SolutionDir)..\..\src"
 	/>
+	<UserMacro
+		Name="ExternDir"
+		Value="$(SolutionDir)..\..\extern"
+	/>
 	<UserMacro
 		Name="PoshDir"
-		Value="$(SolutionDir)..\..\extern\poshlib"
+		Value="$(ExternDir)\poshlib"
 	/>
 	<UserMacro
 		Name="GnuWinDir"
-		Value="$(SolutionDir)..\..\extern\gnuwin32"
+		Value="$(ExternDir)\gnuwin32"
 	/>
 	<UserMacro
 		Name="FreeImageDir"
-		Value="$(SolutionDir)..\..\extern\FreeImage"
+		Value="$(ExternDir)\FreeImage"
+	/>
+	<UserMacro
+		Name="GlewDir"
+		Value="$(ExternDir)\glew"
 	/>
 </VisualStudioPropertySheet>
--- a/project/vc9/nvtt/nvtt.vcproj
+++ b/project/vc9/nvtt/nvtt.vcproj
@ -207,16 +207,17 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
-				InlineFunctionExpansion="2"
+				InlineFunctionExpansion="0"
 				EnableIntrinsicFunctions="true"
-				FavorSizeOrSpeed="1"
+				FavorSizeOrSpeed="0"
 				OmitFramePointers="true"
 				EnableFiberSafeOptimizations="true"
 				AdditionalIncludeDirectories="..;..\..\..\src;..\..\..\extern\gnuwin32\include;$(CUDA_INC_PATH)"
 				PreprocessorDefinitions="WIN32;NDEBUG;_WINDOWS;_USRDLL;NVTT_EXPORTS;NVTT_SHARED;HAVE_CUDA;__SSE2__;__SSE__;__MMX__"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="0"
@ -295,11 +296,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
 				EnableIntrinsicFunctions="true"
 				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories="..;..\..\..\src;..\..\..\extern\gnuwin32\include;$(CUDA_INC_PATH)"
 				PreprocessorDefinitions="WIN32;NDEBUG;_WINDOWS;_USRDLL;NVTT_EXPORTS;NVTT_SHARED;HAVE_CUDA;__SSE2__;__SSE__;__MMX__"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="0"
@ -904,10 +909,6 @@
 			Filter="h;hpp;hxx;hm;inl;inc;xsd"
 			UniqueIdentifier="{93995380-89BD-4b04-88EB-625FBE52EBFB}"
 			>
-			<File
-				RelativePath="..\..\..\src\nvtt\cuda\Bitmaps.h"
-				>
-			</File>
 			<File
 				RelativePath="..\..\..\src\nvtt\CompressionOptions.h"
 				>
--- a/project/vc9/squish/squish.vcproj
+++ b/project/vc9/squish/squish.vcproj
@ -173,16 +173,17 @@
 			<Tool
 				Name="VCCLCompilerTool"
 				Optimization="3"
-				InlineFunctionExpansion="2"
+				InlineFunctionExpansion="0"
 				EnableIntrinsicFunctions="true"
-				FavorSizeOrSpeed="1"
+				FavorSizeOrSpeed="0"
 				OmitFramePointers="true"
 				EnableFiberSafeOptimizations="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_LIB;__SSE2__;__SSE__;__MMX__;SQUISH_USE_SSE = 2"
 				StringPooling="true"
 				RuntimeLibrary="2"
-				EnableEnhancedInstructionSet="0"
+				EnableFunctionLevelLinking="true"
+				EnableEnhancedInstructionSet="2"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
@ -242,9 +243,15 @@
 			/>
 			<Tool
 				Name="VCCLCompilerTool"
+				Optimization="3"
+				EnableIntrinsicFunctions="true"
+				OmitFramePointers="true"
+				WholeProgramOptimization="true"
 				AdditionalIncludeDirectories=""
 				PreprocessorDefinitions="WIN32;NDEBUG;_LIB;__SSE2__;__SSE__;__MMX__"
+				StringPooling="true"
 				RuntimeLibrary="2"
+				EnableFunctionLevelLinking="true"
 				UsePrecompiledHeader="0"
 				WarningLevel="3"
 				DebugInformationFormat="3"
--- a/src/nvcore/Array.h
+++ b/src/nvcore/Array.h
@ -235,13 +235,13 @@ namespace nv
        }

        /// Return true if element found.
-        bool find(const T & element, uint * index)
+        bool find(const T & element, uint * index) const
        {
            return find(element, 0, m_size, index);
        }

        /// Return true if element found within the given range.
-        bool find(const T & element, uint first, uint count, uint * index)
+        bool find(const T & element, uint first, uint count, uint * index) const
        {
            for (uint i = first; i < first+count; i++) {
                if (m_buffer[i] == element) {
--- a/src/nvcore/Debug.cpp
+++ b/src/nvcore/Debug.cpp
@ -21,6 +21,7 @@
 #           include <dbghelp.h>
 #       endif
 #   endif
+#   pragma comment(lib,"dbghelp.lib")
 #endif

 #if !NV_OS_WIN32 && defined(HAVE_SIGNAL_H)
@ -76,55 +77,191 @@ namespace

 #if NV_OS_WIN32 && NV_CC_MSVC

-    // TODO write minidump
-
-    static LONG WINAPI nvTopLevelFilter( struct _EXCEPTION_POINTERS * pExceptionInfo)
+    static bool writeMiniDump(EXCEPTION_POINTERS * pExceptionInfo)
    {
-        NV_UNUSED(pExceptionInfo);
-        /* BOOL (WINAPI * Dump) (HANDLE, DWORD, HANDLE, MINIDUMP_TYPE, PMINIDUMP_EXCEPTION_INFORMATION, PMINIDUMP_USER_STREAM_INFORMATION, PMINIDUMP_CALLBACK_INFORMATION );
-
-        AutoString dbghelp_path(512);
-        getcwd(dbghelp_path, 512);
-        dbghelp_path.Append("\\DbgHelp.dll");
-        nvTranslatePath(dbghelp_path);
-
-        PiLibrary DbgHelp_lib(dbghelp_path, true);
-
-        if( !DbgHelp_lib.IsValid() ) {
-        nvDebug("*** 'DbgHelp.dll' not found.\n");
-        return EXCEPTION_CONTINUE_SEARCH;
-        }
-
-        if( !DbgHelp_lib.BindSymbol( (void **)&Dump, "MiniDumpWriteDump" ) ) {
-        nvDebug("*** 'DbgHelp.dll' too old.\n");
-        return EXCEPTION_CONTINUE_SEARCH;
-        }
-
        // create the file
-        HANDLE hFile = ::CreateFile( "nv.dmp", GENERIC_WRITE, FILE_SHARE_WRITE, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL );
-        if( hFile == INVALID_HANDLE_VALUE ) {
-        nvDebug("*** Failed to create dump file.\n");
-        return EXCEPTION_CONTINUE_SEARCH;
+        HANDLE hFile = CreateFile("crash.dmp", GENERIC_WRITE, FILE_SHARE_WRITE, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL);
+        if (hFile == INVALID_HANDLE_VALUE) {
+            nvDebug("*** Failed to create dump file.\n");
+            return false;
        }

-
-        _MINIDUMP_EXCEPTION_INFORMATION ExInfo;
-
+        MINIDUMP_EXCEPTION_INFORMATION ExInfo;
        ExInfo.ThreadId = ::GetCurrentThreadId();
        ExInfo.ExceptionPointers = pExceptionInfo;
        ExInfo.ClientPointers = NULL;

        // write the dump
-        bool ok = Dump( GetCurrentProcess(), GetCurrentProcessId(), hFile, MiniDumpNormal, &ExInfo, NULL, NULL )!=0;
-        ::CloseHandle(hFile);
+        BOOL ok = MiniDumpWriteDump(GetCurrentProcess(), GetCurrentProcessId(), hFile, MiniDumpNormal, &ExInfo, NULL, NULL) != 0;
+        CloseHandle(hFile);

-        if( !ok ) {
-        nvDebug("*** Failed to save dump file.\n");
-        return EXCEPTION_CONTINUE_SEARCH;
+        if (ok == FALSE) {
+            nvDebug("*** Failed to save dump file.\n");
+            return false;
        }

-        nvDebug("--- Dump file saved.\n");
-        */
+        nvDebug("\nDump file saved.\n");
+
+        return true;
+    }
+
+    static bool hasStackTrace() {
+        return true;
+    }
+
+    /*static NV_NOINLINE int backtrace(void * trace[], int maxcount) {
+	
+        // In Windows XP and Windows Server 2003, the sum of the FramesToSkip and FramesToCapture parameters must be less than 63.
+        int xp_maxcount = min(63-1, maxcount);
+
+        int count = RtlCaptureStackBackTrace(1, xp_maxcount, trace, NULL);
+        nvDebugCheck(count <= maxcount);
+
+        return count;
+    }*/
+
+    static NV_NOINLINE int backtraceWithSymbols(CONTEXT * ctx, void * trace[], int maxcount, int skip = 0) {
+		
+        // Init the stack frame for this function
+        STACKFRAME64 stackFrame = { 0 };
+
+    #if NV_CPU_X86_64
+        DWORD dwMachineType = IMAGE_FILE_MACHINE_AMD64;
+        stackFrame.AddrPC.Offset = ctx->Rip;
+        stackFrame.AddrFrame.Offset = ctx->Rbp;
+        stackFrame.AddrStack.Offset = ctx->Rsp;
+    #elif NV_CPU_X86
+        DWORD dwMachineType = IMAGE_FILE_MACHINE_I386;
+        stackFrame.AddrPC.Offset = ctx->Eip;
+        stackFrame.AddrFrame.Offset = ctx->Ebp;
+        stackFrame.AddrStack.Offset = ctx->Esp;
+    #else
+        #error "Platform not supported!"
+    #endif
+        stackFrame.AddrPC.Mode = AddrModeFlat;
+        stackFrame.AddrFrame.Mode = AddrModeFlat;
+        stackFrame.AddrStack.Mode = AddrModeFlat;
+
+        // Walk up the stack
+        const HANDLE hThread = GetCurrentThread();
+        const HANDLE hProcess = GetCurrentProcess();
+        int i;
+        for (i = 0; i < maxcount; i++)
+        {
+            // walking once first makes us skip self
+            if (!StackWalk64(dwMachineType, hProcess, hThread, &stackFrame, ctx, NULL, &SymFunctionTableAccess64, &SymGetModuleBase64, NULL)) {
+                break;
+            }
+
+            /*if (stackFrame.AddrPC.Offset == stackFrame.AddrReturn.Offset || stackFrame.AddrPC.Offset == 0) {
+                break;
+            }*/
+
+            if (i >= skip) {
+                trace[i - skip] = (PVOID)stackFrame.AddrPC.Offset;
+            }
+        }
+
+        return i - skip;
+    }
+
+    static NV_NOINLINE int backtrace(void * trace[], int maxcount) {
+        CONTEXT ctx = { 0 };
+#if NV_CPU_X86 && !NV_CPU_X86_64
+        ctx.ContextFlags = CONTEXT_CONTROL;
+        _asm {
+             call x
+          x: pop eax
+             mov ctx.Eip, eax
+             mov ctx.Ebp, ebp
+             mov ctx.Esp, esp
+        }
+#else
+        RtlCaptureContext(&ctx);
+#endif
+
+        return backtraceWithSymbols(&ctx, trace, maxcount, 1);
+    }
+
+
+    static NV_NOINLINE void printStackTrace(void * trace[], int size, int start=0)
+    {
+	    HANDLE hProcess = GetCurrentProcess();
+    	
+        nvDebug( "\nDumping stacktrace:\n" );
+
+	    // Resolve PC to function names
+	    for (int i = start; i < size; i++)
+	    {
+		    // Check for end of stack walk
+		    DWORD64 ip = (DWORD64)trace[i];
+		    if (ip == NULL)
+			    break;
+
+		    // Get function name
+		    #define MAX_STRING_LEN	(512)
+		    unsigned char byBuffer[sizeof(IMAGEHLP_SYMBOL64) + MAX_STRING_LEN] = { 0 };
+		    IMAGEHLP_SYMBOL64 * pSymbol = (IMAGEHLP_SYMBOL64*)byBuffer;
+		    pSymbol->SizeOfStruct = sizeof(IMAGEHLP_SYMBOL64);
+		    pSymbol->MaxNameLength = MAX_STRING_LEN;
+
+		    DWORD64 dwDisplacement;
+    		
+		    if (SymGetSymFromAddr64(hProcess, ip, &dwDisplacement, pSymbol))
+		    {
+			    pSymbol->Name[MAX_STRING_LEN-1] = 0;
+    			
+			    /*
+			    // Make the symbol readable for humans
+			    UnDecorateSymbolName( pSym->Name, lpszNonUnicodeUnDSymbol, BUFFERSIZE, 
+				    UNDNAME_COMPLETE | 
+				    UNDNAME_NO_THISTYPE |
+				    UNDNAME_NO_SPECIAL_SYMS |
+				    UNDNAME_NO_MEMBER_TYPE |
+				    UNDNAME_NO_MS_KEYWORDS |
+				    UNDNAME_NO_ACCESS_SPECIFIERS );
+			    */
+    			
+			    // pSymbol->Name
+			    const char * pFunc = pSymbol->Name;
+    			
+			    // Get file/line number
+			    IMAGEHLP_LINE64 theLine = { 0 };
+			    theLine.SizeOfStruct = sizeof(theLine);
+
+			    DWORD dwDisplacement;
+			    if (!SymGetLineFromAddr64(hProcess, ip, &dwDisplacement, &theLine))
+			    {
+				    nvDebug("unknown(%08X) : %s\n", (uint32)ip, pFunc);
+			    }
+			    else
+			    {
+				    /*
+				    const char* pFile = strrchr(theLine.FileName, '\\');
+				    if ( pFile == NULL ) pFile = theLine.FileName;
+				    else pFile++;
+				    */
+				    const char * pFile = theLine.FileName;
+    				
+				    int line = theLine.LineNumber;
+    				
+				    nvDebug("%s(%d) : %s\n", pFile, line, pFunc);
+			    }
+		    }
+	    }
+    }
+
+
+    // Write mini dump and print stack trace.
+    static LONG WINAPI topLevelFilter(EXCEPTION_POINTERS * pExceptionInfo)
+    {
+        void * trace[64];
+        
+        int size = backtraceWithSymbols(pExceptionInfo->ContextRecord, trace, 64);
+        printStackTrace(trace, size, 0);
+
+        writeMiniDump(pExceptionInfo);
+
        return EXCEPTION_CONTINUE_SEARCH;
    }

@ -132,7 +269,7 @@ namespace

 #if defined(HAVE_EXECINFO_H) // NV_OS_LINUX

-    static bool nvHasStackTrace() {
+    static bool hasStackTrace() {
 #if NV_OS_DARWIN
        return backtrace != NULL;
 #else
@ -140,7 +277,7 @@ namespace
 #endif
    }

-    static void nvPrintStackTrace(void * trace[], int size, int start=0) {
+    static void printStackTrace(void * trace[], int size, int start=0) {
        char ** string_array = backtrace_symbols(trace, size);

        nvDebug( "\nDumping stacktrace:\n" );
@ -260,7 +397,7 @@ namespace
        }

 #if defined(HAVE_EXECINFO_H)
-        if (nvHasStackTrace()) // in case of weak linking
+        if (hasStackTrace()) // in case of weak linking
        {
            void * trace[64];
            int size = backtrace(trace, 64);
@ -270,7 +407,7 @@ namespace
                trace[1] = pnt;
            }

-            nvPrintStackTrace(trace, size, 1);
+            printStackTrace(trace, size, 1);
        }
 #endif // defined(HAVE_EXECINFO_H)

@ -289,18 +426,14 @@ namespace
        // Code from Daniel Vogel.
        static bool isDebuggerPresent()
        {
-            bool result = false;
-
-            HINSTANCE kern_lib = LoadLibraryExA( "kernel32.dll", NULL, 0 );
-            if( kern_lib ) {
-                FARPROC lIsDebuggerPresent = GetProcAddress( kern_lib, "IsDebuggerPresent" );
-                if( lIsDebuggerPresent && lIsDebuggerPresent() ) {
-                    result = true;
+            HINSTANCE kernel32 = GetModuleHandle("kernel32.dll");
+            if (kernel32) {
+                FARPROC IsDebuggerPresent = GetProcAddress(kernel32, "IsDebuggerPresent");
+                if (IsDebuggerPresent != NULL && IsDebuggerPresent()) {
+                    return true;
                }
-
-                FreeLibrary( kern_lib );
            }
-            return result;
+            return false;
        }

        // Flush the message queue. This is necessary for the message box to show up.
@ -322,11 +455,11 @@ namespace
            StringBuilder error_string;
            if( func != NULL ) {
                error_string.format( "*** Assertion failed: %s\n    On file: %s\n    On function: %s\n    On line: %d\n ", exp, file, func, line );
-                nvDebug( error_string );
+                nvDebug( error_string.str() );
            }
            else {
                error_string.format( "*** Assertion failed: %s\n    On file: %s\n    On line: %d\n ", exp, file, line );
-                nvDebug( error_string );
+                nvDebug( error_string.str() );
            }

            if (isDebuggerPresent()) {
@ -334,7 +467,7 @@ namespace
            }

            flushMessageQueue();
-            int action = MessageBoxA(NULL, error_string, "Assertion failed", MB_ABORTRETRYIGNORE|MB_ICONERROR);
+            int action = MessageBoxA(NULL, error_string.str(), "Assertion failed", MB_ABORTRETRYIGNORE|MB_ICONERROR);
            switch( action ) {
            case IDRETRY:
                ret = NV_ABORT_DEBUG;
@ -352,7 +485,7 @@ namespace
            }*/

            if( ret == NV_ABORT_EXIT ) {
-                // Exit cleanly.
+                 // Exit cleanly.
                throw "Assertion failed";
            }

@ -402,16 +535,16 @@ namespace
 #endif

 #if defined(HAVE_EXECINFO_H)
-            if (nvHasStackTrace())
+            if (hasStackTrace())
            {
                void * trace[64];
                int size = backtrace(trace, 64);
-                nvPrintStackTrace(trace, size, 2);
+                printStackTrace(trace, size, 2);
            }
 #endif

            // Exit cleanly.
-            throw std::runtime_error("Assertion failed");
+            throw "Assertion failed";
        }
    };

@ -453,14 +586,12 @@ void NV_CDECL nvDebugPrint(const char *msg, ...)
 /// Dump debug info.
 void debug::dumpInfo()
 {
-#if !NV_OS_WIN32 && defined(HAVE_SIGNAL_H) && defined(HAVE_EXECINFO_H)
-    if (nvHasStackTrace())
+    if (hasStackTrace())
    {
        void * trace[64];
        int size = backtrace(trace, 64);
-        nvPrintStackTrace(trace, size, 1);
+        printStackTrace(trace, size, 1);
    }
-#endif
 }


@ -497,7 +628,12 @@ void debug::enableSigHandler()

 #if NV_OS_WIN32 && NV_CC_MSVC

-    s_old_exception_filter = ::SetUnhandledExceptionFilter( nvTopLevelFilter );
+    s_old_exception_filter = ::SetUnhandledExceptionFilter( topLevelFilter );
+
+    // SYMOPT_DEFERRED_LOADS make us not take a ton of time unless we actual log traces
+    SymSetOptions(SYMOPT_DEFERRED_LOADS|SYMOPT_FAIL_CRITICAL_ERRORS|SYMOPT_LOAD_LINES|SYMOPT_UNDNAME);
+
+    SymInitialize(GetCurrentProcess(), NULL, TRUE);

 #elif !NV_OS_WIN32 && defined(HAVE_SIGNAL_H)

@ -526,6 +662,8 @@ void debug::disableSigHandler()
    ::SetUnhandledExceptionFilter( s_old_exception_filter );
    s_old_exception_filter = NULL;

+    SymCleanup(GetCurrentProcess());
+
 #elif !NV_OS_WIN32 && defined(HAVE_SIGNAL_H)

    sigaction(SIGSEGV, &s_old_sigsegv, NULL);
--- a/src/nvcore/Memory.cpp
+++ b/src/nvcore/Memory.cpp
@ -3,35 +3,54 @@
 #include "Memory.h"
 #include "Debug.h"

-//#if HAVE_MALLOC_H
-//#include <malloc.h>
-//#endif
-
 #include <stdlib.h>

+#define USE_EFENCE 0
+
+#if USE_EFENCE
+extern "C" void *EF_malloc(size_t size);
+extern "C" void *EF_realloc(void * oldBuffer, size_t newSize);
+extern "C" void EF_free(void * address);
+#endif

 using namespace nv;

 void * nv::mem::malloc(size_t size)
 {
+#if USE_EFENCE
+    return EF_malloc(size);
+#else
    return ::malloc(size);
+#endif
 }

 void * nv::mem::malloc(size_t size, const char * file, int line)
 {
    NV_UNUSED(file);
    NV_UNUSED(line);
+#if USE_EFENCE
+    return EF_malloc(size);
+#else
    return ::malloc(size);
+#endif
 }

 void nv::mem::free(const void * ptr)
 {
+#if USE_EFENCE
+    return EF_free(const_cast<void *>(ptr));
+#else
    ::free(const_cast<void *>(ptr));
+#endif
 }

 void * nv::mem::realloc(void * ptr, size_t size)
 {
    nvDebugCheck(ptr != NULL || size != 0); // undefined realloc behavior.
+#if USE_EFENCE
+    return EF_realloc(ptr, size);
+#else
    return ::realloc(ptr, size);
+#endif
 }

--- a/src/nvcore/StrLib.cpp
+++ b/src/nvcore/StrLib.cpp
@ -211,9 +211,7 @@ StringBuilder::StringBuilder( const char * s, int extra_size_hint/*=0*/ ) : m_si
 /** Delete the string. */
 StringBuilder::~StringBuilder()
 {
-    m_size = 0;
    strFree(m_str);
-    m_str = NULL;
 }


@ -237,7 +235,7 @@ StringBuilder & StringBuilder::formatList( const char * fmt, va_list arg )
 {
    nvDebugCheck(fmt != NULL);

-    if( m_size == 0 ) {
+    if (m_size == 0) {
        m_size = 64;
        m_str = strAlloc( m_size );
    }
@ -287,21 +285,19 @@ StringBuilder & StringBuilder::append( const char * s )

    const uint slen = uint(strlen( s ));

-    if( m_str == NULL ) {
+    if (m_str == NULL) {
        m_size = slen + 1;
        m_str = strAlloc(m_size);
-        strCpy( m_str, m_size, s );
+        memcpy(m_str, s, m_size + 1);
    }
    else {
-
        const uint len = uint(strlen( m_str ));
-
-        if( m_size < len + slen + 1 ) {
+        if (m_size < len + slen + 1) {
            m_size = len + slen + 1;
            m_str = strReAlloc(m_str, m_size);
        }

-        strCat( m_str, m_size, s );
+        memcpy(m_str + len, s, slen + 1);
    }

    return *this;
@ -338,7 +334,7 @@ StringBuilder & StringBuilder::appendFormatList( const char * fmt, va_list arg )
    else {
        StringBuilder tmp_str;
        tmp_str.formatList( fmt, tmp );
-        append( tmp_str );
+        append( tmp_str.str() );
    }

    va_end(tmp);
@ -391,7 +387,7 @@ StringBuilder & StringBuilder::number( uint i, int base )
 StringBuilder & StringBuilder::reserve( uint size_hint )
 {
    nvCheck(size_hint != 0);
-    if( size_hint > m_size ) {
+    if (size_hint > m_size) {
        m_str = strReAlloc(m_str, size_hint);
        m_size = size_hint;
    }
@ -403,9 +399,9 @@ StringBuilder & StringBuilder::reserve( uint size_hint )
 StringBuilder & StringBuilder::copy( const char * s, int extra_size/*=0*/ )
 {
    nvCheck( s != NULL );
-    uint str_size = uint(strlen( s )) + 1;
+    const uint str_size = uint(strlen( s )) + 1;
    reserve(str_size + extra_size);
-    strCpy( m_str, str_size, s );
+    memcpy(m_str, s, str_size);
    return *this;
 }

@ -413,11 +409,9 @@ StringBuilder & StringBuilder::copy( const char * s, int extra_size/*=0*/ )
 /** Copy an StringBuilder. */
 StringBuilder & StringBuilder::copy( const StringBuilder & s )
 {
-    if( s.m_str == NULL ) {
+    if (s.m_str == NULL) {
        nvCheck( s.m_size == 0 );
-        m_size = 0;
-        strFree( m_str );
-        m_str = NULL;
+        reset();
    }
    else {
        reserve( s.m_size );
@ -428,8 +422,8 @@ StringBuilder & StringBuilder::copy( const StringBuilder & s )

 bool StringBuilder::endsWith(const char * str) const
 {
-    size_t l = strlen(str);
-    size_t ml = strlen(m_str);
+    uint l = uint(strlen(str));
+    uint ml = uint(strlen(m_str));
    if (ml < l) return false;
    return strncmp(m_str + ml - l, str, l) == 0;
 }
@ -596,7 +590,7 @@ void String::setString(const StringBuilder & str)
        data =	NULL;
    }
    else {
-        allocString(str);
+        allocString(str.str());
        addRef();
    }
 }	
--- a/src/nvcore/StrLib.h
+++ b/src/nvcore/StrLib.h
@ -89,8 +89,8 @@ namespace nv
        bool isNull() const { return m_size == 0; }

        // const char * accessors
-        operator const char * () const { return m_str; }
-        operator char * () { return m_str; }
+        //operator const char * () const { return m_str; }
+        //operator char * () { return m_str; }
        const char * str() const { return m_str; }
        char * str() { return m_str; }

--- a/src/nvimage/DirectDrawSurface.cpp
+++ b/src/nvimage/DirectDrawSurface.cpp
@ -42,6 +42,8 @@ using namespace nv;
    (uint(uint8(ch2)) << 16) | (uint(uint8(ch3)) << 24 ))
 #endif

+const uint nv::FOURCC_NVTT = MAKEFOURCC('N', 'V', 'T', 'T');
+
 namespace
 {
    static const uint FOURCC_DDS = MAKEFOURCC('D', 'D', 'S', ' ');
@ -58,6 +60,8 @@ namespace

    static const uint FOURCC_DX10 = MAKEFOURCC('D', 'X', '1', '0');

+    static const uint FOURCC_UVER = MAKEFOURCC('U', 'V', 'E', 'R');
+
    // 32 bit RGB formats.
    static const uint D3DFMT_R8G8B8 = 20;
    static const uint D3DFMT_A8R8G8B8 = 21;
@ -480,63 +484,62 @@ namespace nv

 } // nv namespace

-/* Not used!
 namespace
 {
-struct FormatDescriptor
-{
-uint format;
-uint bitcount;
-uint rmask;
-uint gmask;
-uint bmask;
-uint amask;
-};
+    struct FormatDescriptor
+    {
+        uint format;
+        uint bitcount;
+        uint rmask;
+        uint gmask;
+        uint bmask;
+        uint amask;
+    };

-static const FormatDescriptor s_d3dFormats[] =
-{
-{ D3DFMT_R8G8B8,		24, 0xFF0000,   0xFF00,	    0xFF,       0 },
-{ D3DFMT_A8R8G8B8,		32, 0xFF0000,   0xFF00,     0xFF,       0xFF000000 },  // DXGI_FORMAT_B8G8R8A8_UNORM
-{ D3DFMT_X8R8G8B8,		32, 0xFF0000,   0xFF00,     0xFF,       0 },           // DXGI_FORMAT_B8G8R8X8_UNORM
-{ D3DFMT_R5G6B5,		16,	0xF800,     0x7E0,      0x1F,       0 },           // DXGI_FORMAT_B5G6R5_UNORM
-{ D3DFMT_X1R5G5B5,		16, 0x7C00,     0x3E0,      0x1F,       0 },
-{ D3DFMT_A1R5G5B5,		16, 0x7C00,     0x3E0,      0x1F,       0x8000 },      // DXGI_FORMAT_B5G5R5A1_UNORM
-{ D3DFMT_A4R4G4B4,		16, 0xF00,      0xF0,       0xF,        0xF000 },
-{ D3DFMT_R3G3B2,		8,  0xE0,       0x1C,       0x3,	    0 },
-{ D3DFMT_A8,			8,  0,          0,          0,		    8 },           // DXGI_FORMAT_A8_UNORM
-{ D3DFMT_A8R3G3B2,		16, 0xE0,       0x1C,       0x3,        0xFF00 },
-{ D3DFMT_X4R4G4B4,		16, 0xF00,      0xF0,       0xF,        0 },
-{ D3DFMT_A2B10G10R10,	32, 0x3FF,      0xFFC00,    0x3FF00000, 0xC0000000 },  // DXGI_FORMAT_R10G10B10A2
-{ D3DFMT_A8B8G8R8,		32, 0xFF,       0xFF00,     0xFF0000,   0xFF000000 },  // DXGI_FORMAT_R8G8B8A8_UNORM
-{ D3DFMT_X8B8G8R8,		32, 0xFF,       0xFF00,     0xFF0000,   0 },
-{ D3DFMT_G16R16,		32, 0xFFFF,     0xFFFF0000, 0,          0 },           // DXGI_FORMAT_R16G16_UNORM
-{ D3DFMT_A2R10G10B10,	32, 0x3FF00000, 0xFFC00,    0x3FF,      0xC0000000 },
+    static const FormatDescriptor s_d3dFormats[] =
+    {
+        { D3DFMT_R8G8B8,		24, 0xFF0000,   0xFF00,	    0xFF,       0 },
+        { D3DFMT_A8R8G8B8,		32, 0xFF0000,   0xFF00,     0xFF,       0xFF000000 },  // DXGI_FORMAT_B8G8R8A8_UNORM
+        { D3DFMT_X8R8G8B8,		32, 0xFF0000,   0xFF00,     0xFF,       0 },           // DXGI_FORMAT_B8G8R8X8_UNORM
+        { D3DFMT_R5G6B5,		16,	0xF800,     0x7E0,      0x1F,       0 },           // DXGI_FORMAT_B5G6R5_UNORM
+        { D3DFMT_X1R5G5B5,		16, 0x7C00,     0x3E0,      0x1F,       0 },
+        { D3DFMT_A1R5G5B5,		16, 0x7C00,     0x3E0,      0x1F,       0x8000 },      // DXGI_FORMAT_B5G5R5A1_UNORM
+        { D3DFMT_A4R4G4B4,		16, 0xF00,      0xF0,       0xF,        0xF000 },
+        { D3DFMT_R3G3B2,		8,  0xE0,       0x1C,       0x3,	    0 },
+        { D3DFMT_A8,			8,  0,          0,          0,		    8 },           // DXGI_FORMAT_A8_UNORM
+        { D3DFMT_A8R3G3B2,		16, 0xE0,       0x1C,       0x3,        0xFF00 },
+        { D3DFMT_X4R4G4B4,		16, 0xF00,      0xF0,       0xF,        0 },
+        { D3DFMT_A2B10G10R10,	32, 0x3FF,      0xFFC00,    0x3FF00000, 0xC0000000 },  // DXGI_FORMAT_R10G10B10A2
+        { D3DFMT_A8B8G8R8,		32, 0xFF,       0xFF00,     0xFF0000,   0xFF000000 },  // DXGI_FORMAT_R8G8B8A8_UNORM
+        { D3DFMT_X8B8G8R8,		32, 0xFF,       0xFF00,     0xFF0000,   0 },
+        { D3DFMT_G16R16,		32, 0xFFFF,     0xFFFF0000, 0,          0 },           // DXGI_FORMAT_R16G16_UNORM
+        { D3DFMT_A2R10G10B10,	32, 0x3FF00000, 0xFFC00,    0x3FF,      0xC0000000 },
+        { D3DFMT_A2B10G10R10,	32, 0x3FF,      0xFFC00,    0x3FF00000, 0xC0000000 },

-{ D3DFMT_L8,			8,  8,          0,          0,          0 },           // DXGI_FORMAT_R8_UNORM 
-{ D3DFMT_L16,			16, 16,         0,          0,          0 },           // DXGI_FORMAT_R16_UNORM
-};
+        { D3DFMT_L8,			8,  8,          0,          0,          0 },           // DXGI_FORMAT_R8_UNORM 
+        { D3DFMT_L16,			16, 16,         0,          0,          0 },           // DXGI_FORMAT_R16_UNORM
+    };

-static const uint s_d3dFormatCount = sizeof(s_d3dFormats) / sizeof(s_d3dFormats[0]);
+    static const uint s_d3dFormatCount = sizeof(s_d3dFormats) / sizeof(s_d3dFormats[0]);

-static uint findD3D9Format(uint bitcount, uint rmask, uint gmask, uint bmask, uint amask)
-{
-for (int i = 0; i < s_d3dFormatCount; i++)
-{
-if (s_d3dFormats[i].bitcount == bitcount &&
-s_d3dFormats[i].rmask == rmask &&
-s_d3dFormats[i].gmask == gmask &&
-s_d3dFormats[i].bmask == bmask &&
-s_d3dFormats[i].amask == amask)
-{
-return s_d3dFormats[i].format;
-}
-}
+    static uint findD3D9Format(uint bitcount, uint rmask, uint gmask, uint bmask, uint amask)
+    {
+        for (int i = 0; i < s_d3dFormatCount; i++)
+        {
+            if (s_d3dFormats[i].bitcount == bitcount &&
+                s_d3dFormats[i].rmask == rmask &&
+                s_d3dFormats[i].gmask == gmask &&
+                s_d3dFormats[i].bmask == bmask &&
+                s_d3dFormats[i].amask == amask)
+            {
+                return s_d3dFormats[i].format;
+            }
+        }

-return 0;
-}
+        return 0;
+    }

-} // nv namespace
-*/
+} // namespace

 DDSHeader::DDSHeader()
 {
@ -551,7 +554,7 @@ DDSHeader::DDSHeader()
    memset(this->reserved, 0, sizeof(this->reserved));

    // Store version information on the reserved header attributes.
-    this->reserved[9] = MAKEFOURCC('N', 'V', 'T', 'T');
+    this->reserved[9] = FOURCC_NVTT;
    this->reserved[10] = (2 << 16) | (1 << 8) | (0); // major.minor.revision

    this->pf.size = 32;
@ -598,7 +601,7 @@ void DDSHeader::setMipmapCount(uint count)
    if (count == 0 || count == 1)
    {
        this->flags &= ~DDSD_MIPMAPCOUNT;
-        this->mipmapcount = 0;
+        this->mipmapcount = 1;

        if (this->caps.caps2 == 0) {
            this->caps.caps1 = DDSCAPS_TEXTURE;
@ -725,15 +728,13 @@ void DDSHeader::setPixelFormat(uint bitcount, uint rmask, uint gmask, uint bmask
        }
    }

-    nvCheck(bitcount > 0 && bitcount <= 32);
-
-    // Align to 8.
-    if (bitcount <= 8) bitcount = 8;
-    else if (bitcount <= 16) bitcount = 16;
-    else if (bitcount <= 24) bitcount = 24;
-    else bitcount = 32;
-
+    // D3DX functions do not like this:
    this->pf.fourcc = 0; //findD3D9Format(bitcount, rmask, gmask, bmask, amask);
+    /*if (this->pf.fourcc) {
+        this->pf.flags |= DDPF_FOURCC;
+    }*/
+
+    nvCheck(bitcount > 0 && bitcount <= 32);
    this->pf.bitcount = bitcount;
    this->pf.rmask = rmask;
    this->pf.gmask = gmask;
@ -760,6 +761,12 @@ void DDSHeader::setHasAlphaFlag(bool b)
    else this->pf.flags &= ~DDPF_ALPHAPIXELS;
 }

+void DDSHeader::setUserVersion(int version)
+{
+    this->reserved[7] = FOURCC_UVER;
+    this->reserved[8] = version;
+}
+
 void DDSHeader::swapBytes()
 {
    this->fourcc = POSH_LittleU32(this->fourcc);
@ -798,10 +805,48 @@ void DDSHeader::swapBytes()

 bool DDSHeader::hasDX10Header() const
 {
-    return this->pf.fourcc == FOURCC_DX10;  // @@ This is according to AMD
-    //return this->pf.flags == 0;             // @@ This is according to MS
+    return this->pf.fourcc == FOURCC_DX10;
 }

+uint DDSHeader::signature() const
+{
+    return this->reserved[9];
+}
+
+uint DDSHeader::toolVersion() const
+{
+    return this->reserved[10];
+}
+
+uint DDSHeader::userVersion() const
+{
+    if (this->reserved[7] == FOURCC_UVER) {
+        return this->reserved[8];
+    }
+    return 0;
+}
+
+bool DDSHeader::isNormalMap() const
+{
+    return (pf.flags & DDPF_NORMAL) != 0;
+}
+
+bool DDSHeader::hasAlpha() const
+{
+    return (pf.flags & DDPF_ALPHAPIXELS) != 0;
+}
+
+uint DDSHeader::d3d9Format() const
+{
+    if (pf.flags & DDPF_FOURCC) {
+        return pf.fourcc;
+    }
+    else {
+        return findD3D9Format(pf.bitcount, pf.rmask, pf.gmask, pf.bmask, pf.amask);
+    }
+}
+
+

 DirectDrawSurface::DirectDrawSurface() : stream(NULL)
 {
@ -955,7 +1000,7 @@ bool DirectDrawSurface::hasAlpha() const
            }
            else
            {
-                // @@ Here we could check the ALPHA_PIXELS flag, but nobody sets it.
+                // @@ Here we could check the ALPHA_PIXELS flag, but nobody sets it. (except us?)
                return true;
            }
        }
@ -1047,6 +1092,11 @@ void DirectDrawSurface::setHasAlphaFlag(bool b)
    header.setHasAlphaFlag(b);
 }

+void DirectDrawSurface::setUserVersion(int version)
+{
+    nvDebugCheck(isValid());
+    header.setUserVersion(version);
+}

 void DirectDrawSurface::mipmap(Image * img, uint face, uint mipmap)
 {
@ -1093,6 +1143,28 @@ void DirectDrawSurface::mipmap(Image * img, uint face, uint mipmap)
    }
 }

+void * DirectDrawSurface::readData(uint * sizePtr)
+{
+    uint header_size = 128; // sizeof(DDSHeader);
+
+    if (header.hasDX10Header())
+    {
+        header_size += 20; // sizeof(DDSHeader10);
+    }
+
+    stream->seek(header_size);
+
+    int size = stream->size() - header_size;
+    *sizePtr = size;
+
+    void * data = new unsigned char [size];
+    
+    size = stream->serialize(data, size);
+    nvDebugCheck(size == *sizePtr);
+
+    return data;
+}
+
 void DirectDrawSurface::readLinearImage(Image * img)
 {
    nvDebugCheck(stream != NULL);
@ -1411,26 +1483,33 @@ void DirectDrawSurface::printInfo() const
    if (header.pf.flags & DDPF_ALPHAPREMULT) printf("\t\tDDPF_ALPHAPREMULT\n");
    if (header.pf.flags & DDPF_NORMAL) printf("\t\tDDPF_NORMAL\n");

-    printf("\tFourCC: '%c%c%c%c'\n",
-        ((header.pf.fourcc >> 0) & 0xFF),
-        ((header.pf.fourcc >> 8) & 0xFF),
-        ((header.pf.fourcc >> 16) & 0xFF),
-        ((header.pf.fourcc >> 24) & 0xFF));
-    if ((header.pf.fourcc & DDPF_FOURCC) && (header.pf.bitcount != 0))
+    if (header.pf.fourcc != 0) { 
+        // Display fourcc code even when DDPF_FOURCC flag not set.
+        printf("\tFourCC: '%c%c%c%c' (0x%.8X)\n",
+            ((header.pf.fourcc >> 0) & 0xFF),
+            ((header.pf.fourcc >> 8) & 0xFF),
+            ((header.pf.fourcc >> 16) & 0xFF),
+            ((header.pf.fourcc >> 24) & 0xFF), 
+            header.pf.fourcc);
+    }
+
+    if ((header.pf.flags & DDPF_FOURCC) && (header.pf.bitcount != 0))
    {
-        printf("\tSwizzle: '%c%c%c%c'\n", 
+        printf("\tSwizzle: '%c%c%c%c' (0x%.8X)\n", 
            (header.pf.bitcount >> 0) & 0xFF,
            (header.pf.bitcount >> 8) & 0xFF,
            (header.pf.bitcount >> 16) & 0xFF,
-            (header.pf.bitcount >> 24) & 0xFF);
+            (header.pf.bitcount >> 24) & 0xFF,
+            header.pf.bitcount);
    }
    else
    {
        printf("\tBit count: %d\n", header.pf.bitcount);
    }
-    printf("\tRed mask: 0x%.8X\n", header.pf.rmask);
+
+    printf("\tRed mask:   0x%.8X\n", header.pf.rmask);
    printf("\tGreen mask: 0x%.8X\n", header.pf.gmask);
-    printf("\tBlue mask: 0x%.8X\n", header.pf.bmask);
+    printf("\tBlue mask:  0x%.8X\n", header.pf.bmask);
    printf("\tAlpha mask: 0x%.8X\n", header.pf.amask);

    printf("Caps:\n");
@ -1467,7 +1546,7 @@ void DirectDrawSurface::printInfo() const
        printf("\tArray size: %u\n", header.header10.arraySize);
    }

-    if (header.reserved[9] == MAKEFOURCC('N', 'V', 'T', 'T'))
+    if (header.reserved[9] == FOURCC_NVTT)
    {
        int major = (header.reserved[10] >> 16) & 0xFF;
        int minor = (header.reserved[10] >> 8) & 0xFF;
@ -1476,5 +1555,10 @@ void DirectDrawSurface::printInfo() const
        printf("Version:\n");
        printf("\tNVIDIA Texture Tools %d.%d.%d\n", major, minor, revision);
    }
+
+    if (header.reserved[7] == FOURCC_UVER)
+    {
+        printf("User Version: %d\n", header.reserved[8]);
+    }
 }

--- a/src/nvimage/DirectDrawSurface.h
+++ b/src/nvimage/DirectDrawSurface.h
@ -33,6 +33,8 @@ namespace nv
    class Stream;
    struct ColorBlock;

+    extern const uint FOURCC_NVTT;
+
    struct NVIMAGE_CLASS DDSPixelFormat
    {
        uint size;
@ -100,10 +102,17 @@ namespace nv
        void setDX10Format(uint format);
        void setNormalFlag(bool b);
        void setHasAlphaFlag(bool b);
+        void setUserVersion(int version);

        void swapBytes();

        bool hasDX10Header() const;
+        uint signature() const;
+        uint toolVersion() const;
+        uint userVersion() const;
+        bool isNormalMap() const;
+        bool hasAlpha() const;
+        uint d3d9Format() const;
    };

    NVIMAGE_API Stream & operator<< (Stream & s, DDSHeader & header);
@ -137,9 +146,11 @@ namespace nv

        void setNormalFlag(bool b);
        void setHasAlphaFlag(bool b);
+        void setUserVersion(int version);

        void mipmap(Image * img, uint f, uint m);
        //	void mipmap(FloatImage * img, uint f, uint m);
+        void * readData(uint * sizePtr);

        void printInfo() const;

--- a/src/nvimage/FloatImage.cpp
+++ b/src/nvimage/FloatImage.cpp
@ -199,12 +199,12 @@ void FloatImage::normalize(uint base_component)

 void FloatImage::packNormals(uint base_component)
 {
-    scaleBias(base_component, 3, 0.5f, 1.0f);
+    scaleBias(base_component, 3, 0.5f, 0.5f);
 }

 void FloatImage::expandNormals(uint base_component)
 {
-    scaleBias(base_component, 3, 2, -0.5);
+    scaleBias(base_component, 3, 2, -1.0);
 }

 void FloatImage::scaleBias(uint base_component, uint num, float scale, float bias)
@ -215,7 +215,7 @@ void FloatImage::scaleBias(uint base_component, uint num, float scale, float bia
        float * ptr = this->channel(base_component + c);

        for(uint i = 0; i < size; i++) {
-            ptr[i] = scale * (ptr[i] + bias);
+            ptr[i] = scale * ptr[i] + bias;
        }
    }
 }
--- a/src/nvimage/ImageIO.cpp
+++ b/src/nvimage/ImageIO.cpp
@ -5,8 +5,10 @@
 #include "FloatImage.h"
 #include "TgaFile.h"
 #include "PsdFile.h"
+#include "DirectDrawSurface.h"

 #include "nvmath/Color.h"
+#include "nvmath/Half.h"

 #include "nvcore/Ptr.h"
 #include "nvcore/Utils.h"
@ -93,6 +95,8 @@ namespace nv

 	#endif // defined(HAVE_FREEIMAGE)

+        static FloatImage * loadFloatDDS(Stream & s);
+
 	} // ImageIO namespace
 } // nv namespace

@ -207,12 +211,18 @@ FloatImage * nv::ImageIO::loadFloat(const char * fileName, Stream & s)
    
    const uint spos = s.tell(); // Save stream position.

+	if (strCaseCmp(extension, ".dds") == 0) {
+        floatImage = loadFloatDDS(s);
+    }
+
    // Try to load as a floating point image.
 #if defined(HAVE_FREEIMAGE)
-	FREE_IMAGE_FORMAT fif = FreeImage_GetFIFFromFilename(fileName);
-	if (fif != FIF_UNKNOWN && FreeImage_FIFSupportsReading(fif)) {
-		floatImage = loadFloatFreeImage(fif, s);
-	}
+    if (floatImage == NULL) {
+	    FREE_IMAGE_FORMAT fif = FreeImage_GetFIFFromFilename(fileName);
+	    if (fif != FIF_UNKNOWN && FreeImage_FIFSupportsReading(fif)) {
+		    floatImage = loadFloatFreeImage(fif, s);
+	    }
+    }
 #else // defined(HAVE_FREEIMAGE)
 #pragma message(NV_FILE_LINE "TODO: Load TIFF and EXR files from stream.")
 #if defined(HAVE_TIFF)
@ -1697,3 +1707,44 @@ bool nv::ImageIO::saveFloatEXR(const char * fileName, const FloatImage * fimage,
 #endif // defined(HAVE_OPENEXR)

 #endif // defined(HAVE_FREEIMAGE)
+
+FloatImage * nv::ImageIO::loadFloatDDS(Stream & s)
+{
+	nvCheck(s.isLoading());
+	nvCheck(!s.isError());
+
+    DDSHeader header;
+    s << header;
+
+    static const uint D3DFMT_A16B16G16R16F = 113;
+
+    // @@ We only support RGBA16F for now.
+    if (header.pf.fourcc == D3DFMT_A16B16G16R16F) {
+        const int size = header.width * header.height;
+        uint16 * const data = new uint16[size * 4];
+
+        s.serialize(data, size * 4 * sizeof(uint16));
+
+        FloatImage * img = new FloatImage;
+        img->allocate(4, header.width, header.height);
+
+        uint32 * r = (uint32 *)img->channel(0);
+        uint32 * g = (uint32 *)img->channel(1);
+        uint32 * b = (uint32 *)img->channel(2);
+        uint32 * a = (uint32 *)img->channel(3);
+
+        uint16 * ptr = data;
+        for (int i = 0; i < size; i++) {
+            *r++ = half_to_float( *ptr++ );
+            *g++ = half_to_float( *ptr++ );
+            *b++ = half_to_float( *ptr++ );
+            *a++ = half_to_float( *ptr++ );
+        }
+
+        delete [] data;
+
+        return img;
+    }
+ 
+    return NULL;
+}
--- a/src/nvtt/CompressionOptions.cpp
+++ b/src/nvtt/CompressionOptions.cpp
@ -59,7 +59,9 @@ void CompressionOptions::reset()
 	m.gsize = 8;
 	m.bsize = 8;
 	m.asize = 8;
-	m.pixelType = PixelType_UnsignedNorm;
+	
+    m.pixelType = PixelType_UnsignedNorm;
+    m.pitchAlignment = 1;

 	m.enableColorDithering = false;
 	m.enableAlphaDithering = false;
@ -98,10 +100,10 @@ void CompressionOptions::setColorWeights(float red, float green, float blue, flo


 /// Set color mask to describe the RGB/RGBA format.
-void CompressionOptions::setPixelFormat(uint bitcount, uint rmask, uint gmask, uint bmask, uint amask)
+void CompressionOptions::setPixelFormat(uint bitCount, uint rmask, uint gmask, uint bmask, uint amask)
 {
 	// Validate arguments.
-	nvCheck(bitcount == 8 || bitcount == 16 || bitcount == 24 || bitcount == 32);
+    nvCheck(bitCount <= 32);
 	nvCheck((rmask & gmask) == 0);
 	nvCheck((rmask & bmask) == 0);
 	nvCheck((rmask & amask) == 0);
@ -109,16 +111,16 @@ void CompressionOptions::setPixelFormat(uint bitcount, uint rmask, uint gmask, u
 	nvCheck((gmask & amask) == 0);
 	nvCheck((bmask & amask) == 0);

-	if (bitcount != 32)
+	if (bitCount != 32)
 	{
-		uint maxMask = (1 << bitcount);
+		uint maxMask = (1 << bitCount);
 		nvCheck(maxMask > rmask);
 		nvCheck(maxMask > gmask);
 		nvCheck(maxMask > bmask);
 		nvCheck(maxMask > amask);
 	}

-	m.bitcount = bitcount;
+	m.bitcount = bitCount;
 	m.rmask = rmask;
 	m.gmask = gmask;
 	m.bmask = bmask;
@ -153,6 +155,14 @@ void CompressionOptions::setPixelType(PixelType pixelType)
 }


+/// Set pitch alignment in bytes.
+void CompressionOptions::setPitchAlignment(int pitchAlignment)
+{
+    nvDebugCheck(pitchAlignment > 0 && isPowerOfTwo(pitchAlignment));
+	m.pitchAlignment = pitchAlignment;
+}
+
+
 /// Use external compressor.
 void CompressionOptions::setExternalCompressor(const char * name)
 {
--- a/src/nvtt/CompressionOptions.h
+++ b/src/nvtt/CompressionOptions.h
@ -51,7 +51,8 @@ namespace nvtt
 		uint8 asize;
 		
 		PixelType pixelType;
-		
+        uint pitchAlignment;
+
 		nv::String externalCompressor;

 		// Quantization.
--- a/src/nvtt/CompressorRGB.cpp
+++ b/src/nvtt/CompressorRGB.cpp
@ -40,12 +40,9 @@ using namespace nvtt;
 namespace 
 {

-	inline uint computePitch(uint w, uint bitsize)
+	inline uint computePitch(uint w, uint bitsize, uint alignment)
 	{
-		uint p = w * ((bitsize + 7) / 8);
-
-		// Align to 32 bits.
-		return ((p + 3) / 4) * 4;
+		return ((w * bitsize +  8 * alignment - 1) / (8 * alignment)) * alignment;
 	}

 	inline void convert_to_a8r8g8b8(const void * src, void * dst, uint w)
@ -65,6 +62,67 @@ namespace
        return half_from_float(c.u);
    }

+    struct BitStream
+    {
+        BitStream(uint8 * ptr) : ptr(ptr), buffer(0), bits(0) {
+        }
+
+        void putBits(uint p, int bitCount)
+        {
+            nvDebugCheck(bits < 8);
+            nvDebugCheck(bitCount <= 32);
+
+            uint64 buffer = (this->buffer << bitCount) | p;
+            uint bits = this->bits + bitCount;
+
+            while (bits >= 8)
+            {
+                *ptr++ = (buffer & 0xFF);
+                
+                buffer >>= 8;
+                bits -= 8;
+            }
+
+            this->buffer = (uint8)buffer;
+            this->bits = bits;
+        }
+
+        void putFloat(float f)
+        {
+            nvDebugCheck(bits == 0);
+            *((float *)ptr) = f;
+            ptr += 4;
+        }
+
+        void putHalf(float f)
+        {
+            nvDebugCheck(bits == 0);
+            *((uint16 *)ptr) = to_half(f);
+            ptr += 2;
+        }
+
+        void flush()
+        {
+            nvDebugCheck(bits < 8);
+            if (bits) {
+                *ptr++ = buffer;
+                buffer = 0;
+                bits = 0;
+            }
+        }
+
+        void align(int alignment)
+        {
+            nvDebugCheck(alignment >= 1);
+            flush();
+            putBits(0, ((size_t)ptr % alignment) * 8);
+        }
+
+        uint8 * ptr;
+        uint8 buffer;
+        uint8 bits;
+    };
+
 } // namespace


@ -96,7 +154,7 @@ void PixelFormatConverter::compress(nvtt::InputFormat inputFormat, nvtt::AlphaMo
 	    if (compressionOptions.bitcount != 0)
 	    {
 		    bitCount = compressionOptions.bitcount;
-		    nvCheck(bitCount == 8 || bitCount == 16 || bitCount == 24 || bitCount == 32);
+		    nvCheck(bitCount <= 32);

 		    rmask = compressionOptions.rmask;
 		    gmask = compressionOptions.gmask;
@ -130,20 +188,16 @@ void PixelFormatConverter::compress(nvtt::InputFormat inputFormat, nvtt::AlphaMo
 	    }
    }

-	uint byteCount = (bitCount + 7) / 8;
-    uint pitch = computePitch(w, bitCount);
-
-    uint srcPitch = w;
-    uint srcPlane = w * h;
-
+    const uint pitch = computePitch(w, bitCount, compressionOptions.pitchAlignment);
+    const uint wh = w * h;

    // Allocate output scanline.
-	uint8 * dst = (uint8 *)mem::malloc(pitch + 4);
+	uint8 * const dst = (uint8 *)mem::malloc(pitch);

 	for (uint y = 0; y < h; y++)
 	{
-        const uint * src = (const uint *)data + y * srcPitch;
-        const float * fsrc = (const float *)data + y * srcPitch;
+        const uint * src = (const uint *)data + y * w;
+        const float * fsrc = (const float *)data + y * w;

 	    if (inputFormat == nvtt::InputFormat_BGRA_8UB && compressionOptions.pixelType == nvtt::PixelType_UnsignedNorm && bitCount == 32 && rmask == 0xFF0000 && gmask == 0xFF00 && bmask == 0xFF && amask == 0xFF000000)
 	    {
@ -151,7 +205,7 @@ void PixelFormatConverter::compress(nvtt::InputFormat inputFormat, nvtt::AlphaMo
        }
        else
        {
-            uint8 * ptr = dst;
+            BitStream stream(dst);

 		    for (uint x = 0; x < w; x++)
 		    {
@ -171,29 +225,25 @@ void PixelFormatConverter::compress(nvtt::InputFormat inputFormat, nvtt::AlphaMo
 			        //g = ((float *)src)[4 * x + 1];
 			        //b = ((float *)src)[4 * x + 2];
 			        //a = ((float *)src)[4 * x + 3];
-			        r = fsrc[x + 0 * srcPlane];
-			        g = fsrc[x + 1 * srcPlane];
-			        b = fsrc[x + 2 * srcPlane];
-			        a = fsrc[x + 3 * srcPlane];
+			        r = fsrc[x + 0 * wh];
+			        g = fsrc[x + 1 * wh];
+			        b = fsrc[x + 2 * wh];
+			        a = fsrc[x + 3 * wh];
                }

                if (compressionOptions.pixelType == nvtt::PixelType_Float)
                {
-			        if (rsize == 32) *((float *)ptr) = r;
-			        else if (rsize == 16) *((uint16 *)ptr) = to_half(r);
-			        ptr += rsize / 8;
+			        if (rsize == 32) stream.putFloat(r);
+			        else if (rsize == 16) stream.putHalf(r);

-			        if (gsize == 32) *((float *)ptr) = g;
-			        else if (gsize == 16) *((uint16 *)ptr) = to_half(g);
-			        ptr += gsize / 8;
+			        if (gsize == 32) stream.putFloat(g);
+			        else if (gsize == 16) stream.putHalf(g);

-			        if (bsize == 32) *((float *)ptr) = b;
-			        else if (bsize == 16) *((uint16 *)ptr) = to_half(b);
-			        ptr += bsize / 8;
+			        if (bsize == 32) stream.putFloat(b);
+			        else if (bsize == 16) stream.putHalf(b);

-			        if (asize == 32) *((float *)ptr) = a;
-			        else if (asize == 16) *((uint16 *)ptr) = to_half(a);
-			        ptr += asize / 8;
+			        if (asize == 32) stream.putFloat(a);
+			        else if (asize == 16) stream.putHalf(a);
                }
                else
                {
@ -212,25 +262,27 @@ void PixelFormatConverter::compress(nvtt::InputFormat inputFormat, nvtt::AlphaMo
 				    p |= PixelFormat::convert(c.b, 8, bsize) << bshift;
 				    p |= PixelFormat::convert(c.a, 8, asize) << ashift;
    				
+                    stream.putBits(p, bitCount);
+
 				    // Output one byte at a time.
-				    for (uint i = 0; i < byteCount; i++)
+				    /*for (uint i = 0; i < byteCount; i++)
 				    {
 					    *(dst + x * byteCount + i) = (p >> (i * 8)) & 0xFF;
-				    }
+				    }*/
                }
            }

 		    // Zero padding.
-		    for (uint x = w * byteCount; x < pitch; x++)
+            stream.align(compressionOptions.pitchAlignment);
+            nvDebugCheck(stream.ptr == dst + pitch);
+
+		    /*for (uint x = w * byteCount; x < pitch; x++)
 		    {
 			    *(dst + x) = 0;
-		    }
+		    }*/
        }

-		if (outputOptions.outputHandler != NULL)
-		{
-			outputOptions.outputHandler->writeData(dst, pitch);
-		}
+		outputOptions.writeData(dst, pitch);
    }

 	mem::free(dst);
--- a/src/nvtt/Context.cpp
+++ b/src/nvtt/Context.cpp
@ -89,18 +89,15 @@ namespace
        return 0;
    }

-    inline uint computePitch(uint w, uint bitsize)
-    {
-        uint p = w * ((bitsize + 7) / 8);
+	inline uint computePitch(uint w, uint bitsize, uint alignment)
+	{
+        return ((w * bitsize +  8 * alignment - 1) / (8 * alignment)) * alignment;
+	}

-        // Align to 32 bits.
-        return ((p + 3) / 4) * 4;
-    }
-
-    static int computeImageSize(uint w, uint h, uint d, uint bitCount, Format format)
+    static int computeImageSize(uint w, uint h, uint d, uint bitCount, uint alignment, Format format)
    {
        if (format == Format_RGBA) {
-            return d * h * computePitch(w, bitCount);
+            return d * h * computePitch(w, bitCount, alignment);
        }
        else {
            // @@ Handle 3D textures. DXT and VTC have different behaviors.
@ -324,7 +321,7 @@ int Compressor::estimateSize(int w, int h, int d, const CompressionOptions & com

    uint bitCount = co.getBitCount();

-    return computeImageSize(w, h, d, bitCount, format);
+    return computeImageSize(w, h, d, bitCount, co.pitchAlignment, format);
 }


@ -413,7 +410,20 @@ bool Compressor::Private::outputHeader(const InputOptions::Private & inputOption
    if (outputOptions.container == Container_DDS || outputOptions.container == Container_DDS10)
    {
        DDSHeader header;
-	
+
+        header.setUserVersion(outputOptions.version);
+
+        if (inputOptions.textureType == TextureType_2D) {
+            header.setTexture2D();
+        }
+        else if (inputOptions.textureType == TextureType_Cube) {
+            header.setTextureCube();
+        }
+        /*else if (inputOptions.textureType == TextureType_3D) {
+            header.setTexture3D();
+            header.setDepth(inputOptions.targetDepth);
+        }*/
+
        header.setWidth(inputOptions.targetWidth);
        header.setHeight(inputOptions.targetHeight);

@ -499,7 +509,7 @@ bool Compressor::Private::outputHeader(const InputOptions::Private & inputOption
            if (compressionOptions.format == Format_RGBA)
            {
                // Get output bit count.
-                header.setPitch(computePitch(inputOptions.targetWidth, compressionOptions.getBitCount()));
+                header.setPitch(computePitch(inputOptions.targetWidth, compressionOptions.getBitCount(), compressionOptions.pitchAlignment));

                if (compressionOptions.pixelType == PixelType_Float)
                {
@ -564,7 +574,7 @@ bool Compressor::Private::outputHeader(const InputOptions::Private & inputOption
            }
            else
            {
-                header.setLinearSize(computeImageSize(inputOptions.targetWidth, inputOptions.targetHeight, inputOptions.targetDepth, compressionOptions.bitcount, compressionOptions.format));
+                header.setLinearSize(computeImageSize(inputOptions.targetWidth, inputOptions.targetHeight, inputOptions.targetDepth, compressionOptions.bitcount, compressionOptions.pitchAlignment, compressionOptions.format));

                if (compressionOptions.format == Format_DXT1 || compressionOptions.format == Format_DXT1a || compressionOptions.format == Format_DXT1n) {
                    header.setFourCC('D', 'X', 'T', '1');
@ -618,17 +628,6 @@ bool Compressor::Private::outputHeader(const InputOptions::Private & inputOption
            return false;
        }

-        if (inputOptions.textureType == TextureType_2D) {
-            header.setTexture2D();
-        }
-        else if (inputOptions.textureType == TextureType_Cube) {
-            header.setTextureCube();
-        }
-        /*else if (inputOptions.textureType == TextureType_3D) {
-            header.setTexture3D();
-            header.setDepth(inputOptions.targetDepth);
-        }*/
-
        // Swap bytes if necessary.
        header.swapBytes();

@ -669,6 +668,19 @@ bool Compressor::Private::outputHeader(const TexImage & tex, int mipmapCount, co
    {
        DDSHeader header;

+        header.setUserVersion(outputOptions.version);
+
+        if (tex.textureType() == TextureType_2D) {
+            header.setTexture2D();
+        }
+        else if (tex.textureType() == TextureType_Cube) {
+            header.setTextureCube();
+        }
+        /*else if (tex.textureType() == TextureType_3D) {
+            header.setTexture3D();
+            header.setDepth(tex.depth());
+        }*/
+
        header.setWidth(tex.width());
        header.setHeight(tex.height());
        header.setMipmapCount(mipmapCount);
@ -750,7 +762,7 @@ bool Compressor::Private::outputHeader(const TexImage & tex, int mipmapCount, co
            if (compressionOptions.format == Format_RGBA)
            {
                // Get output bit count.
-                header.setPitch(computePitch(tex.width(), compressionOptions.getBitCount()));
+                header.setPitch(computePitch(tex.width(), compressionOptions.getBitCount(), compressionOptions.pitchAlignment));

                if (compressionOptions.pixelType == PixelType_Float)
                {
@ -815,7 +827,7 @@ bool Compressor::Private::outputHeader(const TexImage & tex, int mipmapCount, co
            }
            else
            {
-                header.setLinearSize(computeImageSize(tex.width(), tex.height(), tex.depth(), compressionOptions.bitcount, compressionOptions.format));
+                header.setLinearSize(computeImageSize(tex.width(), tex.height(), tex.depth(), compressionOptions.bitcount, compressionOptions.pitchAlignment, compressionOptions.format));

                if (compressionOptions.format == Format_DXT1 || compressionOptions.format == Format_DXT1a || compressionOptions.format == Format_DXT1n) {
                    header.setFourCC('D', 'X', 'T', '1');
@ -869,17 +881,6 @@ bool Compressor::Private::outputHeader(const TexImage & tex, int mipmapCount, co
            return false;
        }

-        if (tex.textureType() == TextureType_2D) {
-            header.setTexture2D();
-        }
-        else if (tex.textureType() == TextureType_Cube) {
-            header.setTextureCube();
-        }
-        /*else if (tex.textureType() == TextureType_3D) {
-            header.setTexture3D();
-            header.setDepth(tex.depth());
-        }*/
-
        // Swap bytes if necessary.
        header.swapBytes();

@ -890,7 +891,7 @@ bool Compressor::Private::outputHeader(const TexImage & tex, int mipmapCount, co
            headerSize = 128 + 20;
        }

-        bool writeSucceed = outputOptions.outputHandler->writeData(&header, headerSize);
+        bool writeSucceed = outputOptions.writeData(&header, headerSize);
        if (!writeSucceed)
        {
            outputOptions.error(Error_FileWrite);
@ -916,7 +917,7 @@ bool Compressor::Private::compressMipmaps(uint f, const InputOptions::Private &

    for (uint m = 0; m < mipmapCount; m++)
    {
-        int size = computeImageSize(w, h, d, compressionOptions.getBitCount(), compressionOptions.format);
+        int size = computeImageSize(w, h, d, compressionOptions.getBitCount(), compressionOptions.pitchAlignment, compressionOptions.format);
        outputOptions.beginImage(size, w, h, d, f, m);

        if (!initMipmap(mipmap, inputOptions, w, h, d, f, m))
@ -1553,12 +1554,12 @@ int Compressor::Private::estimateSize(const InputOptions::Private & inputOptions
 {
    const Format format = compressionOptions.format;

-    uint bitCount = compressionOptions.bitcount;
-    if (format == Format_RGBA && bitCount == 0) bitCount = compressionOptions.rsize + compressionOptions.gsize + compressionOptions.bsize + compressionOptions.asize;
+    const uint bitCount = compressionOptions.getBitCount();
+    const uint pitchAlignment = compressionOptions.pitchAlignment;

    inputOptions.computeTargetExtents();

-    uint mipmapCount = inputOptions.realMipmapCount();
+    const uint mipmapCount = inputOptions.realMipmapCount();

    int size = 0;

@ -1570,7 +1571,7 @@ int Compressor::Private::estimateSize(const InputOptions::Private & inputOptions

        for (uint m = 0; m < mipmapCount; m++)
        {
-            size += computeImageSize(w, h, d, bitCount, format);
+            size += computeImageSize(w, h, d, bitCount, pitchAlignment, format);

            // Compute extents of next mipmap:
            w = max(1U, w / 2);
--- a/src/nvtt/OutputOptions.cpp
+++ b/src/nvtt/OutputOptions.cpp
@ -49,13 +49,20 @@ void OutputOptions::reset()

 	m.outputHeader = true;
 	m.container = Container_DDS;
+    m.version = 0;
 }


 /// Set output file name.
 void OutputOptions::setFileName(const char * fileName)
 {
-	m.fileName = fileName; // @@ Do we need to record filename?
+	if (!m.fileName.isNull())
+	{
+        // To close the file and avoid leak.
+		delete m.outputHandler;
+	}
+
+	m.fileName = fileName;
 	m.outputHandler = NULL;

 	DefaultOutputHandler * oh = new DefaultOutputHandler(fileName);
@ -94,6 +101,11 @@ void OutputOptions::setContainer(Container container)
 	m.container = container;
 }

+/// Set user version.
+void OutputOptions::setUserVersion(int version)
+{
+	m.version = version;
+}

 bool OutputOptions::Private::hasValidOutputHandler() const
 {
--- a/src/nvtt/OutputOptions.h
+++ b/src/nvtt/OutputOptions.h
@ -35,9 +35,7 @@ namespace nvtt
 	{
 		DefaultOutputHandler(const char * fileName) : stream(fileName) {}
 		
-		virtual ~DefaultOutputHandler()
-		{
-		}
+		virtual ~DefaultOutputHandler() {}
 		
 		virtual void beginImage(int size, int width, int height, int depth, int face, int miplevel)
 		{
@ -66,6 +64,7 @@ namespace nvtt

 		bool outputHeader;
 		Container container;
+        int version;
 		
 		bool hasValidOutputHandler() const;

--- a/src/nvtt/TexImage.cpp
+++ b/src/nvtt/TexImage.cpp
@ -271,23 +271,19 @@ float TexImage::alphaTestCoverage(float alphaRef/*= 0.5*/) const

 bool TexImage::load(const char * fileName)
 {
-#pragma message(NV_FILE_LINE "TODO: Add support for DDS textures in TexImage::load().")
+    AutoPtr<FloatImage> img(ImageIO::loadFloat(fileName));
+    if (img == NULL) {
+	    return false;
+    }

-	AutoPtr<FloatImage> img(ImageIO::loadFloat(fileName));
+    detach();

-	if (img == NULL)
-	{
-		return false;
-	}
+    img->resizeChannelCount(4);

-	detach();
+    m->imageArray.resize(1);
+    m->imageArray[0] = img.release();

-	img->resizeChannelCount(4);
-
-	m->imageArray.resize(1);
-	m->imageArray[0] = img.release();
-
-	return true;
+    return true;
 }

 bool TexImage::save(const char * fileName) const
@ -560,25 +556,25 @@ void TexImage::resize(int w, int h, ResizeFilter filter)
 			if (filter == ResizeFilter_Box)
 			{
 				BoxFilter filter;
-				m->imageArray[i]->resize(filter, w, h, wrapMode, 3);
+				img = img->resize(filter, w, h, wrapMode, 3);
 			}
 			else if (filter == ResizeFilter_Triangle)
 			{
 				TriangleFilter filter;
-				m->imageArray[i]->resize(filter, w, h, wrapMode, 3);
+				img = img->resize(filter, w, h, wrapMode, 3);
 			}
 			else if (filter == ResizeFilter_Kaiser)
 			{
 				//KaiserFilter filter(inputOptions.kaiserWidth);
 				//filter.setParameters(inputOptions.kaiserAlpha, inputOptions.kaiserStretch);
 				KaiserFilter filter(3);
-				m->imageArray[i]->resize(filter, w, h, wrapMode, 3);
+				img = img->resize(filter, w, h, wrapMode, 3);
 			}
 			else //if (filter == ResizeFilter_Mitchell)
 			{
 				nvDebugCheck(filter == ResizeFilter_Mitchell);
 				MitchellFilter filter;
-				m->imageArray[i]->resize(filter, w, h, wrapMode, 3);
+				img = img->resize(filter, w, h, wrapMode, 3);
 			}
 		}
 		else
@ -586,27 +582,30 @@ void TexImage::resize(int w, int h, ResizeFilter filter)
 			if (filter == ResizeFilter_Box)
 			{
 				BoxFilter filter;
-				m->imageArray[i]->resize(filter, w, h, wrapMode);
+				img = img->resize(filter, w, h, wrapMode);
 			}
 			else if (filter == ResizeFilter_Triangle)
 			{
 				TriangleFilter filter;
-				m->imageArray[i]->resize(filter, w, h, wrapMode);
+				img = img->resize(filter, w, h, wrapMode);
 			}
 			else if (filter == ResizeFilter_Kaiser)
 			{
 				//KaiserFilter filter(inputOptions.kaiserWidth);
 				//filter.setParameters(inputOptions.kaiserAlpha, inputOptions.kaiserStretch);
 				KaiserFilter filter(3);
-				m->imageArray[i]->resize(filter, w, h, wrapMode);
+				img = img->resize(filter, w, h, wrapMode);
 			}
 			else //if (filter == ResizeFilter_Mitchell)
 			{
 				nvDebugCheck(filter == ResizeFilter_Mitchell);
 				MitchellFilter filter;
-				m->imageArray[i]->resize(filter, w, h, wrapMode);
+				img = img->resize(filter, w, h, wrapMode);
 			}
 		}
+
+        delete m->imageArray[i];
+        m->imageArray[i] = img;
 	}
 }

@ -813,6 +812,18 @@ void TexImage::scaleBias(int channel, float scale, float bias)
 	}
 }

+void TexImage::clamp(int channel, float low, float high)
+{
+	detach();
+
+	foreach (i, m->imageArray)
+	{
+		if (m->imageArray[i] == NULL) continue;
+
+		m->imageArray[i]->clamp(channel, 1, low, high);
+	}
+}
+
 void TexImage::packNormal()
 {
 	scaleBias(0, 0.5f, 0.5f);
--- a/src/nvtt/nvtt.h
+++ b/src/nvtt/nvtt.h
@ -48,7 +48,7 @@
 #	define NVTT_API
 #endif

-#define NVTT_VERSION 020100
+#define NVTT_VERSION 20100

 #define NVTT_FORBID_COPY(Class) \
 	private: \
@ -144,6 +144,8 @@ namespace nvtt
 		
 		NVTT_API void setPixelType(PixelType pixelType);

+        NVTT_API void setPitchAlignment(int pitchAlignment);
+
 		NVTT_API void setQuantization(bool colorDithering, bool alphaDithering, bool binaryAlpha, int alphaThreshold = 127);
 	};

@ -345,6 +347,7 @@ namespace nvtt

 		NVTT_API void setOutputHeader(bool outputHeader);
 		NVTT_API void setContainer(Container container);
+        NVTT_API void setUserVersion(int version);
 	};


@ -432,6 +435,7 @@ namespace nvtt
 		NVTT_API void transform(const float w0[4], const float w1[4], const float w2[4], const float w3[4], const float offset[4]);
 		NVTT_API void swizzle(int r, int g, int b, int a);
 		NVTT_API void scaleBias(int channel, float scale, float bias);
+        NVTT_API void clamp(int channel, float low = 0.0f, float high = 1.0f);
 		NVTT_API void packNormal();
 		NVTT_API void expandNormal();
 		NVTT_API void blend(float r, float g, float b, float a, float t);
--- a/src/nvtt/nvtt_wrapper.h
+++ b/src/nvtt/nvtt_wrapper.h
@ -47,7 +47,7 @@
 #	define NVTT_API
 #endif

-#define NVTT_VERSION 020100
+#define NVTT_VERSION 20100

 #ifdef __cplusplus
 typedef struct nvtt::InputOptions NvttInputOptions;
--- a/src/nvtt/tests/imperativeapi.cpp
+++ b/src/nvtt/tests/imperativeapi.cpp
@ -55,7 +55,7 @@ int main(int argc, char *argv[])
    outputFileName.stripExtension();
    outputFileName.append(".dds");

-    outputOptions.setFileName(outputFileName);
+    outputOptions.setFileName(outputFileName.str());

    // Output compressed image.
    context.outputHeader(image, image.countMipmaps(), compressionOptions, outputOptions);
--- a/src/nvtt/tests/testsuite.cpp
+++ b/src/nvtt/tests/testsuite.cpp
@ -423,7 +423,7 @@ int main(int argc, char *argv[])

 	Path csvFileName;
 	csvFileName.format("%s/result.csv", outPath);
-	StdOutputStream csvStream(csvFileName);
+	StdOutputStream csvStream(csvFileName.str());
 	TextWriter csvWriter(&csvStream);

 	float totalTime = 0;
@ -465,7 +465,7 @@ int main(int argc, char *argv[])
 		outputFileName.format("%s/%s", outPath, fileNames[i]);
 		outputFileName.stripExtension();
 		outputFileName.append(".png");
-		if (!ImageIO::save(outputFileName, img_out.ptr()))
+		if (!ImageIO::save(outputFileName.str(), img_out.ptr()))
 		{
 			printf("Error saving file '%s'.\n", outputFileName.str());
 		}
--- a/src/nvtt/tools/assemble.cpp
+++ b/src/nvtt/tools/assemble.cpp
@ -114,7 +114,7 @@ int main(int argc, char *argv[])

 	for (uint i = 0; i < imageCount; i++)
 	{
-		if (!images[i].load(files[i]))
+		if (!images[i].load(files[i].str()))
 		{
 			printf("*** error loading file\n");
 			return 1;
@ -138,7 +138,7 @@ int main(int argc, char *argv[])
 	}
 	
 	
-	nv::StdOutputStream stream(output);
+	nv::StdOutputStream stream(output.str());
 	if (stream.isError()) {
 		printf("Error opening '%s' for writting\n", output.str());
 		return 1;
--- a/src/nvtt/tools/compress.cpp
+++ b/src/nvtt/tools/compress.cpp
@ -295,11 +295,12 @@ int main(int argc, char *argv[])
    }

    const uint version = nvtt::version();
-    const uint major = version / 100;
-    const uint minor = version % 100;
+    const uint major = version / 100 / 100;
+    const uint minor = (version / 100) % 100;
+    const uint rev = version % 100;


-    printf("NVIDIA Texture Tools %u.%u - Copyright NVIDIA Corporation 2007\n\n", major, minor);
+    printf("NVIDIA Texture Tools %u.%u.%u - Copyright NVIDIA Corporation 2007\n\n", major, minor, rev);

    if (input.isNull())
    {
@ -351,7 +352,7 @@ int main(int argc, char *argv[])
    if (nv::strCaseCmp(input.extension(), ".dds") == 0)
    {
        // Load surface.
-        nv::DirectDrawSurface dds(input);
+        nv::DirectDrawSurface dds(input.str());
        if (!dds.isValid())
        {
            fprintf(stderr, "The file '%s' is not a valid DDS file.\n", input.str());
@ -400,7 +401,7 @@ int main(int argc, char *argv[])

        if (loadAsFloat)
        {
-            nv::AutoPtr<nv::FloatImage> image(nv::ImageIO::loadFloat(input));
+            nv::AutoPtr<nv::FloatImage> image(nv::ImageIO::loadFloat(input.str()));

            if (image == NULL)
            {
@ -420,7 +421,7 @@ int main(int argc, char *argv[])
        {
            // Regular image.
            nv::Image image;
-            if (!image.load(input))
+            if (!image.load(input.str()))
            {
                fprintf(stderr, "The file '%s' is not a supported image type.\n", input.str());
                return 1;
@ -449,6 +450,9 @@ int main(int argc, char *argv[])
        inputOptions.setAlphaMode(nvtt::AlphaMode_None);
    }

+    inputOptions.setRoundMode(nvtt::RoundMode_ToNearestPowerOfTwo);
+
+
    if (normal)
    {
        setNormalMap(inputOptions);
@ -524,7 +528,7 @@ int main(int argc, char *argv[])


    MyErrorHandler errorHandler;
-    MyOutputHandler outputHandler(output);
+    MyOutputHandler outputHandler(output.str());
    if (outputHandler.stream->isError())
    {
        fprintf(stderr, "Error opening '%s' for writting\n", output.str());
--- a/src/nvtt/tools/decompress.cpp
+++ b/src/nvtt/tools/decompress.cpp
@ -112,7 +112,7 @@ int main(int argc, char *argv[])
 	}

 	// Load surface.
-	nv::DirectDrawSurface dds(input);
+	nv::DirectDrawSurface dds(input.str());
 	if (!dds.isValid())
 	{
 		fprintf(stderr, "The file '%s' is not a valid DDS file.\n", input.str());
@ -179,7 +179,7 @@ int main(int argc, char *argv[])
 				return 1;
 			}
 			
-			nv::ImageIO::save(name, stream, &mipmap);
+			nv::ImageIO::save(name.str(), stream, &mipmap);
 		}
 	}

--- a/src/nvtt/tools/imgdiff.cpp
+++ b/src/nvtt/tools/imgdiff.cpp
@ -204,8 +204,8 @@ int main(int argc, char *argv[])
 	}

 	nv::Image image0, image1;
-	if (!loadImage(image0, input0)) return 0;
-	if (!loadImage(image1, input1)) return 0;
+	if (!loadImage(image0, input0.str())) return 0;
+	if (!loadImage(image1, input1.str())) return 0;

 	const uint w0 = image0.width();
 	const uint h0 = image0.height();
--- a/src/nvtt/tools/resize.cpp
+++ b/src/nvtt/tools/resize.cpp
@ -164,7 +164,7 @@ int main(int argc, char *argv[])
 	}

 	nv::Image image;
-	if (!loadImage(image, input)) return 0;
+	if (!loadImage(image, input.str())) return 0;

 	nv::FloatImage fimage(&image);
 	fimage.toLinear(0, 3, gamma);