LLVMSRCDIR= $(HOME)/llvm/src
LLVMGCCFLAGS += -I$(LLVMSRCDIR)/include/VectorC -I$(LLVMSRCDIR)/include/SIMD #-Wa,-unroll-threshold=200
SSECFLAGS += -I$(LLVMSRCDIR)/include/SIMD -msse2 -O2
ALTIVECCFLAGS += -I$(LLVMSRCDIR)/include/SIMD -faltivec -O2


# Specify how many runs you want here (for timing). You can also say
# NRUNS=n on the command line.
#
ifndef $(NRUNS)
NRUNS= 1
endif

all:
	# Please specify altivec or sse!

altivec: altivec.handwritten altivec.vectorc
	./altivec.handwritten
	./altivec.vectorc

sse: sse.handwritten sse.vectorc
	./sse.handwritten
	./sse.vectorc

#
# General stuff
#
main.bc : main.c
	llvm-gcc $(LLVMGCCFLAGS) -DNRUNS=$(NRUNS) -c -o $@ $<

$(NAME).vectorc.bc : $(NAME).vectorc.c
	llvm-gcc $(LLVMGCCFLAGS) -c $< -o $@	

$(NAME).raised.bc : $(NAME).vectorc.bc
	opt -raisevectors < $< > $@

%.ll : %.bc
	llvm-dis < $< > $@

clean:
	rm -f altivec.handwritten sse.handwritten altivec.vectorc sse.vectorc *.o *.s *.bc *.exe *.ll *.cbe.c

#
# AltiVec-specific stuff
#
../_malloc.bc : ../_malloc.c
	llvm-gcc $(LLVMGCCFLAGS) -DMEMALIGN=0 -c -o $@ $<

altivec.handwritten: $(NAME).altivec.handwritten.c ../_malloc.c main.c
	gcc $(ALTIVECCFLAGS) -DMEMALIGN=0 -DNRUNS=$(NRUNS) -o $@ $+

$(NAME).altivec.bc : $(NAME).raised.bc
	opt < $< -altivec > $@

altivec.vectorc.bc: $(NAME).altivec.bc ../_malloc.bc main.bc
	llvm-gcc $(LLVMGCCFLAGS) $+ -o altivec.vectorc

altivec.vectorc.cbe.c: altivec.vectorc.bc
	llc -march=altivec-c < $< | sed 's/_2E_/_/g' > $@	

altivec.vectorc: altivec.vectorc.cbe.c
	gcc $(ALTIVECCFLAGS) -o $@ $<

#
# SSE-specific stuff
#
../_malloc.memalign.bc : ../_malloc.c
	llvm-gcc $(LLVMGCCFLAGS) -DMEMALIGN=1 -c -o $@ $<

sse.handwritten: $(NAME).sse.handwritten.c ../_malloc.c main.c
	gcc $(SSECFLAGS) -DMEMALIGN=1 -DNRUNS=$(NRUNS) -o $@ $+

$(NAME).sse.bc : $(NAME).raised.bc
	opt < $< -sse > $@

sse.vectorc.bc: $(NAME).sse.bc ../_malloc.memalign.bc main.bc
	llvm-gcc $(LLVMGCCFLAGS) $+ -o sse.vectorc

sse.vectorc.cbe.c: sse.vectorc.bc
	llc -march=sse-c < $< | sed 's/_2E_/_/g' > $@	

sse.vectorc: sse.vectorc.cbe.c
	gcc $(SSECFLAGS) -o $@ $<