Makefile optimization for quickly compiling a library as quickly as possible? - c

Question

Are there any methods I can use to shave seconds off my library's compilation?
My old Makefile took 1.86s to run, up to 5.147 total.
After I changed
%.o: %.c
$(CC) $(FLAGS) -c -o $# $<
Into:
#$(CC) $(FLAGS) -c $(OBJ:.o=.c)
And added # to silence all the output, my new Makefile took 1.40s to run, up to 3.481 s total.
I'm wondering if there's any ways I can optimize my makefile to simply make it run faster or make it more "useful"? Where do I go to learn about micro-adjustments to get a slightly faster compiled program?
https://github.com/Hourai42/libft/blob/master/Makefile
NAME = libft.a
FLAGS = -Wall -Wextra -Werror
CC = gcc
OBJ = ft_strrchr.o ft_putstr.o ft_strcmp.o ft_strlcat.o ft_memcpy.o \
ft_strequm.o ft_memmove.o ft_strsplit.o ft_strncpy.o ft_lstmap.o \
ft_lstadd.o ft_striter.o ft_strstr.o ft_isdigit.o ft_putnbr.o \
ft_memccpy.o ft_strdel.o ft_nmbwrd.o ft_memchr.o ft_bzero.o \
ft_isalnum.o ft_putstr_fd.o ft_lstiter.o ft_lstdelone.o ft_toupper.o
\
ft_strcpy.o ft_lstnew.o ft_strdup.o ft_putnbr_fd.o ft_striteri.o \
ft_strmap.o ft_putendl_fd.o ft_memdel.o ft_strnstr.o ft_putchar.o \
ft_isascii.o ft_strlen.o ft_strsub.o ft_strnew.o ft_memalloc.o
ft_putendl.o \
ft_strncat.o ft_itoa.o ft_strncmp.o ft_memcmp.o ft_strtrim.o
ft_strequ.o
\
ft_putchar_fd.o ft_memset.o ft_isalpha.o ft_strcat.o ft_tolower.o \
ft_strnequ.o ft_strjoin.o ft_atoi.o ft_strclr.o ft_strmapi.o \
ft_isprint.o ft_lstdel.o ft_strchr.o ft_mallocwrd.o ft_countplace.o \
ft_isprime.o ft_hassqrt.o ft_fibonacci.o ft_recursive_power.o \
ft_recursive_factorial.o
all: $(NAME)
$(NAME):
#$(CC) $(FLAGS) -c $(OBJ:.o=.c)
#ar rc $(NAME) $(OBJ)
#ranlib $(NAME)
clean:
#/bin/rm -f $(OBJ)
fclean: clean
#/bin/rm -f $(NAME)
re: fclean all
.PHONY: clean fclean re

Answer 1

Redo only what needs to (huge improvement).
Use parallel make (significant improvement if run on a multicore CPU).
Invoke only one single shell per recipe (moderate to minor improvement).
Example:
NAME = libft.a
FLAGS = -Wall -Wextra -Werror
CC = gcc
SRCS = $(wildcard *.c)
OBJS = $(patsubst %.c,%.o,$(SRCS))
.PHONY: all clean ultraclean
all: $(NAME)
%.o: %.c
#$(CC) $(FLAGS) -c -o $# $<
$(NAME): $(OBJS)
#ar rc $(NAME) $^ && \
ranlib $(NAME)
clean:
#/bin/rm -f $(OBJS)
ultraclean:
#/bin/rm -f $(OBJS) $(NAME)
And then, if you have, let's say, 6 cores and hyper-threading:
$ make -j12
Note: of course, with parallel make, do not parallelize cleaning targets (clean, ultraclean) and building targets (all), as you do with your re target.