gzdoom-gles/src/CMakeLists.txt

768 lines
20 KiB
Text
Raw Normal View History

cmake_minimum_required( VERSION 2.4 )
if( COMMAND cmake_policy )
cmake_policy( SET CMP0003 NEW )
endif( COMMAND cmake_policy )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
include( CheckFunctionExists )
include( FindPkgConfig )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
option( NO_ASM "Disable assembly code" )
if( CMAKE_COMPILER_IS_GNUCXX )
option( NO_STRIP "Do not strip Release or MinSizeRel builds" )
endif( CMAKE_COMPILER_IS_GNUCXX )
if( CMAKE_SIZEOF_VOID_P MATCHES "8" )
set( X64 64 )
endif( CMAKE_SIZEOF_VOID_P MATCHES "8" )
if( WIN32 )
if( X64 )
set( WIN_TYPE Win64 )
set( XBITS x64 )
else( X64 )
set( WIN_TYPE Win32 )
set( XBITS x86 )
endif( X64 )
add_definitions( -D_WIN32 )
set( FMOD_SEARCH_PATHS
"C:/Program Files/FMOD SoundSystem/FMOD Programmers API ${WIN_TYPE}/api"
"C:/Program Files (x86)/FMOD SoundSystem/FMOD Programmers API ${WIN_TYPE}/api"
"E:/Program Files (x86)/FMOD SoundSystem/FMOD Programmers API ${WIN_TYPE}/api" )
set( FMOD_INC_PATH_SUFFIXES PATH_SUFFIXES inc )
set( FMOD_LIB_PATH_SUFFIXES PATH_SUFFIXES lib )
set( NASM_NAMES nasmw nasm )
find_path( D3D_INCLUDE_DIR d3d9.h
PATHS ENV DXSDK_DIR
PATH_SUFFIXES Include )
if( NOT D3D_INCLUDE_DIR )
message( SEND_ERROR "Could not find DirectX 9 header files" )
else( NOT D3D_INCLUDE_DIR )
include_directories( ${D3D_INCLUDE_DIR} )
endif( NOT D3D_INCLUDE_DIR )
find_library( DX_ddraw_LIBRARY ddraw
PATHS ENV DXSDK_DIR
PATH_SUFFIXES Lib Lib/${XBITS} )
find_library( DX_dxguid_LIBRARY dxguid
PATHS ENV DXSDK_DIR
PATH_SUFFIXES Lib Lib/${XBITS} )
find_library( DX_dinput8_LIBRARY dinput8
PATHS ENV DXSDK_DIR
PATH_SUFFIXES Lib Lib/${XBITS} )
set( DX_LIBS_FOUND YES )
if( NOT DX_ddraw_LIBRARY )
set( DX_LIBS_FOUND NO )
endif( NOT DX_ddraw_LIBRARY )
if( NOT DX_dxguid_LIBRARY )
set( DX_LIBS_FOUND NO )
endif( NOT DX_dxguid_LIBRARY )
if( NOT DX_dinput8_LIBRARY )
set( DX_LIBS_FOUND NO )
endif( NOT DX_dinput8_LIBRARY )
if( NOT DX_LIBS_FOUND )
message( FATAL_ERROR "Could not find DirectX 9 libraries" )
endif( NOT DX_LIBS_FOUND )
set( ZDOOM_LIBS
wsock32
winmm
"${DX_ddraw_LIBRARY}"
"${DX_dxguid_LIBRARY}"
"${DX_dinput8_LIBRARY}"
ole32
user32
gdi32
comctl32
comdlg32
ws2_32
setupapi )
else( WIN32 )
option( NO_GTK "Disable GTK+ dialogs (Not applicable to Windows)" )
option( VALGRIND "Add special Valgrind sequences to self-modifying code" )
set( FMOD_SEARCH_PATHS
/usr/local/include
/usr/include
/opt/local/include
/opt/include )
set( FMOD_INC_PATH_SUFFIXES PATH_SUFFIXES fmodex )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
set( NASM_NAMES nasm )
# Non-Windows version also needs SDL
find_package( SDL )
if( NOT SDL_FOUND )
message( SEND_ERROR "SDL is required for building." )
endif( NOT SDL_FOUND )
set( ZDOOM_LIBS "${SDL_LIBRARY}" )
include_directories( "${SDL_INCLUDE_DIR}" )
# Use GTK+ for the IWAD picker, if available.
if( NOT NO_GTK )
pkg_check_modules( GTK2 gtk+-2.0 )
if( GTK2_FOUND )
set( ZDOOM_LIBS ${ZDOOM_LIBS} ${GTK2_LIBRARIES} )
include_directories( ${GTK2_INCLUDE_DIRS} )
else( GTK2_FOUND )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
set( NO_GTK ON )
endif( GTK2_FOUND )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
endif( NOT NO_GTK )
if( NO_GTK )
add_definitions( -DNO_GTK=1 )
endif( NO_GTK )
find_path( FPU_CONTROL_DIR fpu_control.h )
if( FPU_CONTROL_DIR )
include_directories( ${FPU_CONTROL_DIR} )
add_definitions( -DHAVE_FPU_CONTROL )
endif( FPU_CONTROL_DIR )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
endif( WIN32 )
# Decide on the name of the FMOD library we want to use.
if( NOT FMOD_LIB_NAME AND MSVC )
set( FMOD_LIB_NAME fmodex${X64}_vc )
endif( NOT FMOD_LIB_NAME AND MSVC )
if( NOT FMOD_LIB_NAME AND BORLAND )
set( FMOD_LIB_NAME fmodex${X64}_bc )
endif( NOT FMOD_LIB_NAME AND BORLAND )
if( NOT FMOD_LIB_NAME )
set( FMOD_LIB_NAME fmodex${X64} )
endif( NOT FMOD_LIB_NAME )
# Search for FMOD include files
find_path( FMOD_INCLUDE_DIR fmod.h
PATHS ${FMOD_SEARCH_PATHS}
${FMOD_INC_PATH_SUFFIXES} )
if( FMOD_INCLUDE_DIR )
message( STATUS "FMOD include files found at ${FMOD_INCLUDE_DIR}" )
else( FMOD_INCLUDE_DIR )
message( SEND_ERROR "Could not find FMOD include files" )
endif( FMOD_INCLUDE_DIR )
# Search for FMOD library
find_library( FMOD_LIBRARY ${FMOD_LIB_NAME}
PATHS ${FMOD_SEARCH_PATHS}
${FMOD_LIB_PATH_SUFFIXES} )
if( FMOD_LIBRARY )
message( STATUS "FMOD library found at ${FMOD_LIBRARY}" )
else( FMOD_LIBRARY )
message( SEND_ERROR "Could not find FMOD library" )
endif( FMOD_LIBRARY )
# Search for NASM
if( NOT NO_ASM )
if( UNIX AND X64 )
find_program( GAS_PATH as )
if( GAS_PATH )
set( ASSEMBLER ${GAS_PATH} )
else( GAS_PATH )
message( STATUS "Could not find as. Disabling assembly code." )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
set( NO_ASM ON )
endif( GAS_PATH )
else( UNIX AND X64 )
find_program( NASM_PATH NAMES ${NASM_NAMES} )
find_program( YASM_PATH yasm )
if( YASM_PATH )
set( ASSEMBLER ${YASM_PATH} )
else( YASM_PATH )
if( X64 )
message( STATUS "Could not find YASM. Disabling assembly code." )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
set( NO_ASM ON )
else( X64 )
if( NOT NASM_PATH )
message( STATUS "Could not find YASM or NASM. Disabling assembly code." )
set( NO_ASM ON )
else( NOT NASM_PATH )
set( ASSEMBLER ${NASM_PATH} )
endif( NOT NASM_PATH )
endif( X64 )
endif( YASM_PATH )
endif( UNIX AND X64 )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
# I think the only reason there was a version requirement was because the
# executable name for Windows changed from 0.x to 2.0, right? This is
# how to do it in case I need to do something similar later.
# execute_process( COMMAND ${NASM_PATH} -v
# OUTPUT_VARIABLE NASM_VER_STRING )
# string( REGEX REPLACE ".*version ([0-9]+[.][0-9]+).*" "\\1" NASM_VER "${NASM_VER_STRING}" )
# if( NOT NASM_VER LESS 2 )
# message( SEND_ERROR "NASM version should be 2 or later. (Installed version is ${NASM_VER}.)" )
# endif( NOT NASM_VER LESS 2 )
endif( NOT NO_ASM )
if( NOT NO_ASM )
# Valgrind support is meaningless without assembly code.
if( VALGRIND )
add_definitions( -DVALGRIND_AWARE=1 )
# If you're Valgrinding, you probably want to keep symbols around.
set( NO_STRIP ON )
endif( VALGRIND )
# Tell CMake how to assemble our files
if( UNIX )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
set( ASM_OUTPUT_EXTENSION .o )
if( X64 )
set( ASM_FLAGS )
set( ASM_SOURCE_EXTENSION .s )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
else( X64 )
set( ASM_FLAGS -f elf -DM_TARGET_LINUX -i${CMAKE_CURRENT_SOURCE_DIR}/ )
set( ASM_SOURCE_EXTENSION .asm )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
endif( X64 )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
else( UNIX )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
set( ASM_OUTPUT_EXTENSION .obj )
set( ASM_SOURCE_EXTENSION .asm )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
if( X64 )
set( ASM_FLAGS -f win64 -DWIN32 -DWIN64 )
else( X64 )
set( ASM_FLAGS -f win32 -DWIN32 -i${CMAKE_CURRENT_SOURCE_DIR}/ )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
endif( X64 )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
endif( UNIX )
if( WIN32 )
set( FIXRTEXT fixrtext )
endif( WIN32 )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
message( STATUS "Selected assembler: ${ASSEMBLER}" )
MACRO( ADD_ASM_FILE indir infile )
set( ASM_OUTPUT_${infile} "${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/zdoom.dir/${indir}/${infile}${ASM_OUTPUT_EXTENSION}" )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
if( WIN32 )
set( FIXRTEXT_${infile} COMMAND ${FIXRTEXT} "${ASM_OUTPUT_${infile}}" )
endif( WIN32 )
add_custom_command( OUTPUT ${ASM_OUTPUT_${infile}}
COMMAND ${CMAKE_COMMAND} -E make_directory ${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/zdoom.dir/${indir}
COMMAND ${ASSEMBLER} ${ASM_FLAGS} -o"${ASM_OUTPUT_${infile}}" "${CMAKE_CURRENT_SOURCE_DIR}/${indir}/${infile}${ASM_SOURCE_EXTENSION}"
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
${FIXRTEXT_${infile}}
DEPENDS ${indir}/${infile}.asm ${FIXRTEXT} )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
set( ASM_SOURCES ${ASM_SOURCES} "${ASM_OUTPUT_${infile}}" )
ENDMACRO( ADD_ASM_FILE )
endif( NOT NO_ASM )
# Set up flags for GCC
if( CMAKE_COMPILER_IS_GNUCXX )
set( REL_CXX_FLAGS "-fno-rtti -fomit-frame-pointer" )
set( CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} ${REL_CXX_FLAGS}" )
set( CMAKE_CXX_FLAGS_MINSIZEREL "${CMAKE_CXX_FLAGS_MINSIZEREL} ${REL_CXX_FLAGS}" )
set( CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELWITHDEBINFO} ${REL_CXX_FLAGS}" )
set( CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wno-unused -fno-strict-aliasing" )
if( NOT NO_STRIP )
set (CMAKE_EXE_LINKER_FLAGS_RELEASE "${CMAKE_EXE_LINKER_FLAGS_RELEASE} -s" )
set (CMAKE_EXE_LINKER_FLAGS_MINSIZEREL "${CMAKE_EXE_LINKER_FLAGS_MINSIZEREL} -s" )
endif( NOT NO_STRIP )
endif( CMAKE_COMPILER_IS_GNUCXX )
# Check for functions that may or may not exist.
CHECK_FUNCTION_EXISTS( filelength FILELENGTH_EXISTS )
if( FILELENGTH_EXISTS )
add_definitions( -DHAVE_FILELENGTH=1 )
endif( FILELENGTH_EXISTS )
CHECK_FUNCTION_EXISTS( strupr STRUPR_EXISTS )
if( NOT STRUPR_EXISTS )
add_definitions( -DNEED_STRUPR=1 )
endif( NOT STRUPR_EXISTS )
CHECK_FUNCTION_EXISTS( stricmp STRICMP_EXISTS )
if( NOT STRICMP_EXISTS )
add_definitions( -Dstricmp=strcasecmp )
endif( NOT STRICMP_EXISTS )
CHECK_FUNCTION_EXISTS( strnicmp STRNICMP_EXISTS )
if( NOT STRNICMP_EXISTS )
add_definitions( -Dstrnicmp=strncasecmp )
endif( NOT STRNICMP_EXISTS )
if( NOT MSVC )
add_definitions( -D__forceinline=inline )
endif( NOT MSVC )
if( UNIX )
CHECK_LIBRARY_EXISTS( rt clock_gettime "" CLOCK_GETTIME_IN_RT )
if( NOT CLOCK_GETTIME_IN_RT )
CHECK_FUNCTION_EXISTS( clock_gettime CLOCK_GETTIME_EXISTS )
if( NOT CLOCK_GETTIME_EXISTS )
message( STATUS "Could not find clock_gettime. Timing statistics will not be available." )
add_definitions( -DNO_CLOCK_GETTIME )
endif( NOT CLOCK_GETTIME_EXISTS )
else( NOT CLOCK_GETTIME_IN_RT )
set( ZDOOM_LIBS ${ZDOOM_LIBS} rt )
endif( NOT CLOCK_GETTIME_IN_RT )
endif( UNIX )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
# Update svnrevision.h
add_custom_target( revision_check ALL
COMMAND ${CMAKE_BINARY_DIR}/tools/updaterevision/updaterevision . src/svnrevision.h
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
DEPENDS updaterevision )
# Libraries ZDoom needs
set( ZDOOM_LIBS ${ZDOOM_LIBS} "${ZLIB_LIBRARIES}" "${JPEG_LIBRARIES}" "${FMOD_LIBRARY}" )
include_directories( "${ZLIB_INCLUDE_DIR}" "${JPEG_INCLUDE_DIR}" "${FMOD_INCLUDE_DIR}" )
# Start defining source files for ZDoom
if( WIN32 )
set( SYSTEM_SOURCES_DIR win32 )
set( SYSTEM_SOURCES
win32/eaxedit.cpp
win32/fb_d3d9.cpp
win32/fb_d3d9_wipe.cpp
win32/fb_ddraw.cpp
win32/hardware.cpp
win32/helperthread.cpp
win32/i_cd.cpp
win32/i_crash.cpp
win32/i_input.cpp
win32/i_main.cpp
win32/i_movie.cpp
win32/i_system.cpp
win32/st_start.cpp
win32/win32video.cpp )
if( CMAKE_COMPILER_IS_GNUCXX )
# CMake is not set up to compile and link rc files with GCC. :(
add_custom_command( OUTPUT zdoom-rc.o
COMMAND windres -o zdoom-rc.o -i ${CMAKE_CURRENT_SOURCE_DIR}/win32/zdoom.rc
DEPENDS win32/zdoom.rc )
set( SYSTEM_SOURCES ${SYSTEM_SOURCES} zdoom-rc.o )
else( CMAKE_COMPILER_IS_GNUCXX )
set( SYSTEM_SOURCES ${SYSTEM_SOURCES} win32/zdoom.rc )
endif( CMAKE_COMPILER_IS_GNUCXX )
else( WIN32 )
set( SYSTEM_SOURCES_DIR sdl )
set( SYSTEM_SOURCES
sdl/crashcatcher.c
sdl/hardware.cpp
sdl/i_cd.cpp
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
sdl/i_input.cpp
sdl/i_main.cpp
sdl/i_movie.cpp
sdl/i_system.cpp
sdl/sdlvideo.cpp
sdl/st_start.cpp )
endif( WIN32 )
if( NO_ASM )
add_definitions( -DNOASM )
else( NO_ASM )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
if( X64 )
ADD_ASM_FILE( asm_x86_64 tmap3 )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
else( X64 )
ADD_ASM_FILE( asm_ia32 a )
ADD_ASM_FILE( asm_ia32 misc )
ADD_ASM_FILE( asm_ia32 tmap )
ADD_ASM_FILE( asm_ia32 tmap2 )
ADD_ASM_FILE( asm_ia32 tmap3 )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
endif( X64 )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
if( WIN32 )
if( NOT X64 )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
ADD_ASM_FILE( win32/wrappers.asm )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
endif( NOT X64 )
endif( WIN32 )
endif( NO_ASM )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
add_custom_command( OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/xlat_parser.c ${CMAKE_CURRENT_BINARY_DIR}/xlat_parser.h
COMMAND ${CMAKE_COMMAND} -E copy_if_different ${CMAKE_CURRENT_SOURCE_DIR}/xlat/xlat_parser.y .
COMMAND ${CMAKE_BINARY_DIR}/tools/lemon/lemon xlat_parser.y
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
DEPENDS lemon ${CMAKE_CURRENT_SOURCE_DIR}/xlat/xlat_parser.y )
add_custom_command( OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/sc_man_scanner.h
COMMAND ${CMAKE_BINARY_DIR}/tools/re2c/re2c --no-generation-date -s -o ${CMAKE_CURRENT_BINARY_DIR}/sc_man_scanner.h ${CMAKE_CURRENT_SOURCE_DIR}/sc_man_scanner.re
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
DEPENDS re2c ${CMAKE_CURRENT_SOURCE_DIR}/sc_man_scanner.re )
include_directories( ${CMAKE_CURRENT_BINARY_DIR} )
add_executable( zdoom WIN32
autostart.cpp
${ASM_SOURCES}
${SYSTEM_SOURCES}
am_map.cpp
b_bot.cpp
b_func.cpp
b_game.cpp
b_move.cpp
b_think.cpp
bbannouncer.cpp
c_bind.cpp
c_cmds.cpp
c_console.cpp
c_cvars.cpp
c_dispatch.cpp
c_expr.cpp
cmdlib.cpp
colormatcher.cpp
configfile.cpp
ct_chat.cpp
d_dehacked.cpp
d_main.cpp
d_net.cpp
d_netinfo.cpp
d_protocol.cpp
decallib.cpp
dobject.cpp
dobjgc.cpp
dobjtype.cpp
doomdef.cpp
doomstat.cpp
dsectoreffect.cpp
dthinker.cpp
f_finale.cpp
f_wipe.cpp
farchive.cpp
files.cpp
g_game.cpp
g_hub.cpp
g_level.cpp
gameconfigfile.cpp
gi.cpp
hu_scores.cpp
i_net.cpp
info.cpp
lumpconfigfile.cpp
m_alloc.cpp
m_argv.cpp
m_bbox.cpp
m_cheat.cpp
m_menu.cpp
m_misc.cpp
m_options.cpp
m_png.cpp
m_random.cpp
mus2midi.cpp
name.cpp
nodebuild.cpp
nodebuild_classify_nosse2.cpp
nodebuild_classify_sse2.cpp
nodebuild_events.cpp
nodebuild_extract.cpp
nodebuild_gl.cpp
nodebuild_utility.cpp
p_3dmidtex.cpp
p_acs.cpp
p_buildmap.cpp
p_ceiling.cpp
p_conversation.cpp
p_doors.cpp
p_effect.cpp
p_enemy.cpp
p_enemy_a_lookex.cpp
p_floor.cpp
p_interaction.cpp
p_lights.cpp
p_linkedsectors.cpp
p_lnspec.cpp
p_map.cpp
p_maputl.cpp
p_mobj.cpp
p_pillar.cpp
p_plats.cpp
p_pspr.cpp
p_saveg.cpp
p_sectors.cpp
p_setup.cpp
p_sight.cpp
p_slopes.cpp
p_spec.cpp
p_switch.cpp
p_teleport.cpp
p_terrain.cpp
p_things.cpp
p_tick.cpp
p_trace.cpp
p_udmf.cpp
p_user.cpp
p_writemap.cpp
p_xlat.cpp
parsecontext.cpp
po_man.cpp
r_anim.cpp
r_bsp.cpp
r_data.cpp
r_draw.cpp
r_drawt.cpp
r_interpolate.cpp
r_main.cpp
r_plane.cpp
r_polymost.cpp
r_segs.cpp
r_sky.cpp
r_things.cpp
r_translate.cpp
s_advsound.cpp
s_environment.cpp
s_playlist.cpp
s_sndseq.cpp
s_sound.cpp
sc_man.cpp
st_stuff.cpp
stats.cpp
stringtable.cpp
tables.cpp
teaminfo.cpp
tempfiles.cpp
v_collection.cpp
v_draw.cpp
v_font.cpp
v_palette.cpp
v_pfx.cpp
v_text.cpp
v_video.cpp
w_wad.cpp
wi_stuff.cpp
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
x86.cpp
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
zstrformat.cpp
zstring.cpp
g_doom/a_arachnotron.cpp
g_doom/a_archvile.cpp
g_doom/a_bossbrain.cpp
g_doom/a_bruiser.cpp
g_doom/a_cacodemon.cpp
g_doom/a_cyberdemon.cpp
g_doom/a_demon.cpp
g_doom/a_doomimp.cpp
g_doom/a_doommisc.cpp
g_doom/a_doomweaps.cpp
g_doom/a_fatso.cpp
g_doom/a_keen.cpp
g_doom/a_lostsoul.cpp
g_doom/a_painelemental.cpp
g_doom/a_possessed.cpp
g_doom/a_revenant.cpp
g_doom/a_scriptedmarine.cpp
g_doom/a_spidermaster.cpp
g_doom/doom_sbar.cpp
g_heretic/a_chicken.cpp
g_heretic/a_dsparil.cpp
g_heretic/a_hereticartifacts.cpp
g_heretic/a_hereticimp.cpp
g_heretic/a_hereticmisc.cpp
g_heretic/a_hereticweaps.cpp
g_heretic/a_ironlich.cpp
g_heretic/a_knight.cpp
g_heretic/a_wizard.cpp
g_heretic/heretic_sbar.cpp
g_hexen/a_bats.cpp
g_hexen/a_bishop.cpp
g_hexen/a_blastradius.cpp
g_hexen/a_boostarmor.cpp
g_hexen/a_centaur.cpp
g_hexen/a_clericflame.cpp
g_hexen/a_clericholy.cpp
g_hexen/a_clericmace.cpp
g_hexen/a_clericstaff.cpp
g_hexen/a_dragon.cpp
g_hexen/a_fighteraxe.cpp
g_hexen/a_fighterhammer.cpp
g_hexen/a_fighterplayer.cpp
g_hexen/a_fighterquietus.cpp
g_hexen/a_firedemon.cpp
g_hexen/a_flechette.cpp
g_hexen/a_fog.cpp
g_hexen/a_healingradius.cpp
g_hexen/a_heresiarch.cpp
g_hexen/a_hexenspecialdecs.cpp
g_hexen/a_iceguy.cpp
g_hexen/a_korax.cpp
g_hexen/a_magecone.cpp
g_hexen/a_magelightning.cpp
g_hexen/a_magestaff.cpp
g_hexen/a_magewand.cpp
g_hexen/a_pig.cpp
g_hexen/a_serpent.cpp
g_hexen/a_spike.cpp
g_hexen/a_summon.cpp
g_hexen/a_teleportother.cpp
g_hexen/a_wraith.cpp
g_hexen/hexen_sbar.cpp
g_raven/a_artitele.cpp
g_raven/a_minotaur.cpp
g_strife/a_acolyte.cpp
g_strife/a_alienspectres.cpp
g_strife/a_coin.cpp
g_strife/a_crusader.cpp
g_strife/a_entityboss.cpp
g_strife/a_inquisitor.cpp
g_strife/a_loremaster.cpp
g_strife/a_macil.cpp
g_strife/a_oracle.cpp
g_strife/a_programmer.cpp
g_strife/a_reaver.cpp
g_strife/a_rebels.cpp
g_strife/a_sentinel.cpp
g_strife/a_spectral.cpp
g_strife/a_stalker.cpp
g_strife/a_strifeitems.cpp
g_strife/a_strifestuff.cpp
g_strife/a_strifeweapons.cpp
g_strife/a_templar.cpp
g_strife/a_thingstoblowup.cpp
g_strife/strife_sbar.cpp
g_shared/a_action.cpp
g_shared/a_armor.cpp
g_shared/a_artifacts.cpp
g_shared/a_bridge.cpp
g_shared/a_camera.cpp
g_shared/a_debris.cpp
g_shared/a_decals.cpp
g_shared/a_flashfader.cpp
g_shared/a_fountain.cpp
g_shared/a_hatetarget.cpp
g_shared/a_keys.cpp
g_shared/a_lightning.cpp
g_shared/a_mapmarker.cpp
g_shared/a_morph.cpp
g_shared/a_movingcamera.cpp
g_shared/a_pickups.cpp
g_shared/a_puzzleitems.cpp
g_shared/a_quake.cpp
g_shared/a_randomspawner.cpp
g_shared/a_secrettrigger.cpp
g_shared/a_sectoraction.cpp
g_shared/a_setcolor.cpp
g_shared/a_skies.cpp
g_shared/a_soundenvironment.cpp
g_shared/a_soundsequence.cpp
g_shared/a_spark.cpp
g_shared/a_specialspot.cpp
g_shared/a_waterzone.cpp
g_shared/a_weaponpiece.cpp
g_shared/a_weapons.cpp
g_shared/hudmessages.cpp
g_shared/sbarinfo_display.cpp
g_shared/sbarinfo_parser.cpp
g_shared/sbar_mugshot.cpp
g_shared/shared_hud.cpp
g_shared/shared_sbar.cpp
oplsynth/fmopl.cpp
oplsynth/mlopl.cpp
oplsynth/mlopl_io.cpp
oplsynth/music_opldumper_mididevice.cpp
oplsynth/music_opl_mididevice.cpp
oplsynth/opl_mus_player.cpp
sound/fmodsound.cpp
sound/i_music.cpp
sound/i_sound.cpp
sound/music_cd.cpp
sound/music_dumb.cpp
sound/music_midistream.cpp
sound/music_midi_base.cpp
sound/music_midi_midiout.cpp
sound/music_midi_timidity.cpp
sound/music_mus_midiout.cpp
sound/music_mus_opl.cpp
sound/music_spc.cpp
sound/music_stream.cpp
sound/music_timidity_mididevice.cpp
sound/music_win_mididevice.cpp
textures/automaptexture.cpp
textures/bitmap.cpp
textures/buildtexture.cpp
textures/canvastexture.cpp
textures/ddstexture.cpp
textures/flattexture.cpp
textures/imgztexture.cpp
textures/jpegtexture.cpp
textures/multipatchtexture.cpp
textures/patchtexture.cpp
textures/pcxtexture.cpp
textures/pngtexture.cpp
textures/rawpagetexture.cpp
textures/texture.cpp
textures/texturemanager.cpp
textures/tgatexture.cpp
textures/warptexture.cpp
thingdef/olddecorations.cpp
thingdef/thingdef.cpp
thingdef/thingdef_codeptr.cpp
thingdef/thingdef_exp.cpp
thingdef/thingdef_main.cpp
thingdef/thingdef_properties.cpp
thingdef/thingdef_states.cpp
timidity/common.cpp
timidity/instrum.cpp
timidity/instrum_dls.cpp
timidity/instrum_font.cpp
timidity/instrum_sf2.cpp
timidity/mix.cpp
timidity/playmidi.cpp
timidity/resample.cpp
timidity/timidity.cpp
xlat/parse_xlat.cpp
autozend.cpp )
set_source_files_properties( xlat/parse_xlat.cpp PROPERTIES OBJECT_DEPENDS "${CMAKE_CURRENT_BINARY_DIR}/xlat_parser.c" )
set_source_files_properties( sc_man.cpp PROPERTIES OBJECT_DEPENDS "${CMAKE_CURRENT_BINARY_DIR}/sc_man_scanner.h" )
target_link_libraries( zdoom ${ZDOOM_LIBS} snes_spc gdtoa dumb )
include_directories( .
g_doom
g_heretic
g_hexen
g_raven
g_strife
g_shared
oplsynth
sound
textures
thingdef
timidity
xlat
../snes_spc/snes_spc
../gdtoa
../dumb/include
${CMAKE_BINARY_DIR}/gdtoa
${SYSTEM_SOURCES_DIR} )
add_dependencies( zdoom revision_check )
# RUNTIME_OUTPUT_DIRECTORY does not exist in CMake 2.4.
# Linux distributions are slow to adopt 2.6. :(
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
set_target_properties( zdoom PROPERTIES RUNTIME_OUTPUT_DIRECTORY ${ZDOOM_OUTPUT_DIR} )
set_target_properties( zdoom PROPERTIES OUTPUT_NAME ${ZDOOM_EXE_NAME} )
if( NOT WIN32 )
FILE( WRITE ${CMAKE_CURRENT_BINARY_DIR}/link-make "if [ ! -e ${ZDOOM_OUTPUT_DIR}/${ZDOOM_EXE_NAME} ]; then ln -sf ${CMAKE_CURRENT_BINARY_DIR}/${ZDOOM_EXE_NAME} ${ZDOOM_OUTPUT_DIR}/${ZDOOM_EXE_NAME}; fi" )
add_custom_command( TARGET zdoom POST_BUILD
COMMAND chmod +x ${CMAKE_CURRENT_BINARY_DIR}/link-make
COMMAND /bin/sh -c ${CMAKE_CURRENT_BINARY_DIR}/link-make )
endif( NOT WIN32 )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
if( CMAKE_COMPILER_IS_GNUCXX )
# GCC misoptimizes this file
set_source_files_properties( oplsynth/fmopl.cpp PROPERTIES COMPILE_FLAGS "-fno-tree-dominator-opts -fno-tree-fre" )
# Compile this one file with SSE2 support.
set_source_files_properties( nodebuild_classify_sse2.cpp PROPERTIES COMPILE_FLAGS "-msse2 -mfpmath=sse" )
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of registers AMD64 provides, this routine still needs to be written as self- modifying code for maximum performance. The additional registers do allow for further optimization over the x86 version by allowing all four pixels to be in flight at the same time. The end result is that AMD64 ASM is about 2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM. (For further comparison, AMD64 C and x86 C are practically the same for this function.) Should I port any more assembly to AMD64, mvlineasm4 is the most likely candidate, but it's not used enough at this point to bother. Also, this may or may not work with Linux at the moment, since it doesn't have the eh_handler metadata. Win64 is easier, since I just need to structure the function prologue and epilogue properly and use some assembler directives/macros to automatically generate the metadata. And that brings up another point: You need YASM to assemble the AMD64 code, because NASM doesn't support the Win64 metadata directives. - Added an SSE version of DoBlending. This is strictly C intrinsics. VC++ still throws around unneccessary register moves. GCC seems to be pretty close to optimal, requiring only about 2 cycles/color. They're both faster than my hand-written MMX routine, so I don't need to feel bad about not hand-optimizing this for x64 builds. - Removed an extra instruction from DoBlending_MMX, transposed two instructions, and unrolled it once, shaving off about 80 cycles from the time required to blend 256 palette entries. Why? Because I tried writing a C version of the routine using compiler intrinsics and was appalled by all the extra movq's VC++ added to the code. GCC was better, but still generated extra instructions. I only wanted a C version because I can't use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit of a pain. (It's a pain because Linux and Windows have different calling conventions, and you need to maintain extra metadata for functions.) So, the assembly version stays and the C version stays out. - Removed all the pixel doubling r_detail modes, since the one platform they were intended to assist (486) actually sees very little benefit from them. - Rewrote CheckMMX in C and renamed it to CheckCPU. - Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache only for AMD processors, so we must not use it on other architectures, or we end up overwriting the L1 cache line size with 0 or some other number we don't actually understand. SVN r1134 (trunk)
2008-08-09 03:13:43 +00:00
# Need to enable intrinsics for this file.
set_source_files_properties( x86.cpp PROPERTIES COMPILE_FLAGS "-msse2 -mmmx" )
About a week's worth of changes here. As a heads-up, I wouldn't be surprised if this doesn't build in Linux right now. The CMakeLists.txt were checked with MinGW and NMake, but how they fair under Linux is an unknown to me at this time. - Converted most sprintf (and all wsprintf) calls to either mysnprintf or FStrings, depending on the situation. - Changed the strings in the wbstartstruct to be FStrings. - Changed myvsnprintf() to output nothing if count is greater than INT_MAX. This is so that I can use a series of mysnprintf() calls and advance the pointer for each one. Once the pointer goes beyond the end of the buffer, the count will go negative, but since it's an unsigned type it will be seen as excessively huge instead. This should not be a problem, as there's no reason for ZDoom to be using text buffers larger than 2 GB anywhere. - Ripped out the disabled bit from FGameConfigFile::MigrateOldConfig(). - Changed CalcMapName() to return an FString instead of a pointer to a static buffer. - Changed startmap in d_main.cpp into an FString. - Changed CheckWarpTransMap() to take an FString& as the first argument. - Changed d_mapname in g_level.cpp into an FString. - Changed DoSubstitution() in ct_chat.cpp to place the substitutions in an FString. - Fixed: The MAPINFO parser wrote into the string buffer to construct a map name when given a Hexen map number. This was fine with the old scanner code, but only a happy coincidence prevents it from crashing with the new code - Added the 'B' conversion specifier to StringFormat::VWorker() for printing binary numbers. - Added CMake support for building with MinGW, MSYS, and NMake. Linux support is probably broken until I get around to booting into Linux again. Niceties provided over the existing Makefiles they're replacing: * All command-line builds can use the same build system, rather than having a separate one for MinGW and another for Linux. * Microsoft's NMake tool is supported as a target. * Progress meters. * Parallel makes work from a fresh checkout without needing to be primed first with a single-threaded make. * Porting to other architectures should be simplified, whenever that day comes. - Replaced the makewad tool with zipdir. This handles the dependency tracking itself instead of generating an external makefile to do it, since I couldn't figure out how to generate a makefile with an external tool and include it with a CMake-generated makefile. Where makewad used a master list of files to generate the package file, zipdir just zips the entire contents of one or more directories. - Added the gdtoa package from netlib's fp library so that ZDoom's printf-style formatting can be entirely independant of the CRT. SVN r1082 (trunk)
2008-07-23 04:57:26 +00:00
endif( CMAKE_COMPILER_IS_GNUCXX )
if( MSVC )
# Compile this one file with SSE2 support.
set_source_files_properties( nodebuild_classify_sse2.cpp PROPERTIES COMPILE_FLAGS "/arch:SSE2" )
endif( MSVC )