#############################################################################
#
# GNU/Linux x86 Makefile
# 
# Edit this file to enable/disable optimizations
#
# $Id: Makefile.linuxx86,v 1.5 2002-12-14 09:28:56 suxen_drol Exp $
# 
#############################################################################

include Makefile.inc

# You can set/add options to CC and CFLAGS without editing this Makefile
# You have just to export an environment CC/CFLAGS variable before running
# make.
#
# Examples :
# $ export CC='gcc-3.1'
# $ export CFLAGS='-maccumulate-outgoing-args'
# $ make
# ...

# You can choose a specific compiler here removing the sharp symbol
#   - Default is 'cc' which is a symlink to your default distro gcc


# CC = gcc
# CC = gcc-3.1
# CC = icc

# Assembler
AS = nasm
AFLAGS = -f elf

# Constants which should not be modified
CFLAGS += -DARCH_X86
CFLAGS += -DLINUX

# Enable some experimental/debugging code
#   (Use only if you know what you're doing !)
    # PSNR during encoding
# CFLAGS += -D_DEBUG_PSNR

LIBDIR = /usr/local/lib

ifeq ($(CC),icc)

    # -tpp7 might be faster (P4 optmization)...
    CFLAGS += -O3 -ip -tpp6 -xM
    CFLAGS += -I/opt/intel/compiler50/ia32/include
    CFLAGS += -I/opt/intel/compiler50/ia32/substitute_headers/

else

  # Warning level options
      # Shows all warnings
  CFLAGS += -Wall
      # To detect non ANSI C code (xvid is not ANSI C (yet?))
  # CFLAGS += -pedantic


  # General Optimization options
      # on gcc 2.95 using -Os (which implies -O2) might be faster
  CFLAGS += -O2
      # Unroll loops code, this might speed up loops
  CFLAGS += -funroll-loops
      # Generates FPU math code
  CFLAGS += -ffast-math
      # ??? (man gcc)
  CFLAGS += -fstrict-aliasing
      # Omit the x86 push ebp; mov ebp, esp function
      # header and its corresponding function tailer
  CFLAGS += -fomit-frame-pointer
      # Generates Position Independant Code depending on your nasm version,
      # this could generate shared libraries that segfault because of a nasm
      # bug
  CFLAGS += -fPIC

  # Debugging and Profiling options
     # Writes debugging symbols in object files
  # CFLAGS += -g
     # Includes profiling code in generated code, don't forget to disable
     # -fomit-frame-pointer as these 2 options are exclusive.
  # CFLAGS += -pg
     # Precise profiling + test coverage (-pg must be activated)
     # Build the library twice and execute a program using the lib twice too
       # First pass : Will create sourcefile.da files to keep branch statistics
  # CFLAGS += -fprofile-arcs 
       # Second pass : Will create .bb and .bbg files suitable for coverage and
       #               branch probabilities during compilation
  # CFLAGS += -fprofile-arcs -fbranch-probabilities -ftest-coverage


  # Architecture options : Valid for every gcc version
  # Uncomment the one which you want to compile for

  # CFLAGS += -march=i386 -mcpu=i386
  # CFLAGS += -march=i486 -mcpu=i486
  # CFLAGS += -march=i586 -mcpu=i586
  # CFLAGS += -march=i686 -mcpu=i686
  # CFLAGS += -march=pentium -mcpu=pentium
  # CFLAGS += -march=pentiumpro -mcpu=pentiumpro

  # Architecture options : Valid for every gcc >= 3.1
  # Uncomment the one which you want to compile for

  # CFLAGS += -march=pentium-mmx -mcpu=pentium-mmx
  # CFLAGS += -march=pentium3 -mcpu=pentium3
  # CFLAGS += -march=pentium4 -mcpu=pentium4
  # CFLAGS += -march=athlon -mcpu=athlon
  # CFLAGS += -march=k6 -mcpu=k6
  # CFLAGS += -march=k6-2 -mcpu=k6-2
  # CFLAGS += -march=k6-3 -mcpu=k6-3
  # CFLAGS += -march=athlon-tbird -mcpu=athlon-tbird
  # CFLAGS += -march=athlon-xp -mcpu=athlon-xp
  # CFLAGS += -march=athlon-mp -mcpu=athlon-mp

  # x86 specific optimization for every gcc
      # Aligns the stack to 2^4=16bytes boundaries in each function
      # Local variables should be aligned and make the code faster
      # some gcc versions (3.0.4) have problems with this option
      # we can't really trust gcc output :-( You're warned.
  CFLAGS += -mpreferred-stack-boundary=4

  # Architecture specific optimization available in gcc >= 3.1
      # generates floating point sse code (for >= pentium 3)
  # CFLAGS += -mfpmath=sse
      # generates both floating point sse and 387 code, this double the amount
      # of fp registers and could make the code faster
      # (Still experimental in gcc 3.1.0)
  # CFLAGS += -mfpmath=sse,387

endif

#############################################################################
#
# Makefile rules
#
#############################################################################

%.o: %.asm
	$(AS) $(AFLAGS) -I$(<D) $< -o $@

RM = rm -rf


all: $(LIB) $(LIBSO)


$(LIB): $(SRC:.c=.o) $(SRC_INTEL:.asm=.o)
	ar rcs $@ $^


$(LIBSO): $(SRC:.c=.o) $(SRC_INTEL:.asm=.o)
	$(CC) $(CFLAGS) $^ -shared -lc -lm -o $@


clean:	
	$(RM) `find $(SRCDIR) -name "*.o"`
	$(RM) $(LIB)
	$(RM) $(LIBSO)

install: $(LIBSO) $(LIB)
	cp $(LIBSO) $(LIBDIR)                   # you have to bee root for this
	cp $(LIB) $(LIBDIR)
	/sbin/ldconfig 

# if you don't want to overwrite previous compile
install-test: $(LIBSO)
	cp $(LIBSO) $(LIBDIR)/libtestcore.so
	/sbin/ldconfig