# Copyright (c) 2022-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
#
# NVIDIA CORPORATION, its affiliates and licensors retain all intellectual
# property and proprietary rights in and to this material, related
# documentation and any modifications thereto. Any use, reproduction,
# disclosure or distribution of this material and related documentation
# without an express license agreement from NVIDIA CORPORATION or
# its affiliates is strictly prohibited.
#

# Location of common libraries
LIB_DIR = /usr/lib/aarch64-linux-gnu
# NOTE: This directory PATH will be moved from "tegra" to "nvidia".
TEGRA_LIB_DIR ?= /usr/lib/aarch64-linux-gnu/tegra

# Location of the CUDA Toolkit
CUDA_PATH ?= /usr/local/cuda

# Location of NVSCI header
NVSCI_HEADER_DIR ?= /usr/include/nvsci_headers
NVSCI_LIB_DIR = $(TEGRA_LIB_DIR)

GCC ?= g++
NVCC := $(CUDA_PATH)/bin/nvcc -ccbin $(GCC)

# internal flags
NVCCFLAGS   :=
CCFLAGS     :=
LDFLAGS     :=

# Extra user flags
EXTRA_NVCCFLAGS   ?=
EXTRA_LDFLAGS     ?=
EXTRA_CCFLAGS     ?=

override abi := aarch64
LDFLAGS += --dynamic-linker=/lib/ld-linux-aarch64.so.1

# Debug build flags
dbg = 0
ifeq ($(dbg),1)
      NVCCFLAGS += -g -G
      TARGET := debug
else
      TARGET := release
endif

ALL_CCFLAGS :=
ALL_CCFLAGS += $(NVCCFLAGS)
ALL_CCFLAGS += $(EXTRA_NVCCFLAGS)
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(CCFLAGS))
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(EXTRA_CCFLAGS))

ALL_LDFLAGS :=
ALL_LDFLAGS += $(ALL_CCFLAGS)
ALL_LDFLAGS += $(addprefix -Xlinker ,$(LDFLAGS))
ALL_LDFLAGS += $(addprefix -Xlinker ,$(EXTRA_LDFLAGS))

################################################################################

# Common includes and paths
INCLUDES  := -I./
LIBRARIES := -L$(LIB_DIR)

CUDA_SEARCH_PATH ?= $(CUDA_PATH)/lib64/stubs
CUDA_SEARCH_PATH += $(CUDA_PATH)/lib/stubs
CUDA_SEARCH_PATH += $(CUDA_PATH)/targets/aarch64-linux/lib/stubs

CUDALIB ?= $(shell find -L $(CUDA_SEARCH_PATH) -maxdepth 1 -name libcuda.so 2> /dev/null)
ifeq ("$(CUDALIB)","")
  $(error ERROR - libcuda.so not found, CUDA Driver is not installed or CUDA_PATH is not correctly set.)
else
  CUDALIB := $(shell echo $(CUDALIB) | sed "s/ .*//" | sed "s/\/libcuda.so//" )
  LIBRARIES += -L$(CUDALIB) -lcuda -lrt
endif

# Includes and paths for NVSCI libraries
NVSCIBUFHEADER  := $(shell find -L $(NVSCI_HEADER_DIR) -name nvscibuf.h -print 2>/dev/null)
NVSCISYNCHEADER  := $(shell find -L $(NVSCI_HEADER_DIR) -name nvscisync.h -print 2>/dev/null)
NVSCISTREAMHEADER  := $(shell find -L $(NVSCI_HEADER_DIR) -name nvscistream.h -print 2>/dev/null)
NVSCIEVENTHEADER  := $(shell find -L $(NVSCI_HEADER_DIR) -name nvscievent.h -print 2>/dev/null)
NVSCIIPCHEADER  := $(shell find -L $(NVSCI_HEADER_DIR) -name nvsciipc.h -print 2>/dev/null)

ifeq ("$(NVSCIBUFHEADER)","")
    $(error ERROR - nvscibuf.h not found in $(NVSCI_HEADER_DIR))
endif
ifeq ("$(NVSCISYNCHEADER)","")
    $(error ERROR - nvscisync.h not found in $(NVSCI_HEADER_DIR))
endif
ifeq ("$(NVSCISTREAMHEADER)","")
    $(error ERROR - nvscistream.h not found in $(NVSCI_HEADER_DIR))
endif
ifeq ("$(NVSCIEVENTHEADER)","")
    $(error ERROR - nvscievent.h not found in $(NVSCI_HEADER_DIR))
endif
ifeq ("$(NVSCIIPCHEADER)","")
    $(error ERROR - nvsciipc.h not found in $(NVSCI_HEADER_DIR))
endif

INCLUDES  += -I$(NVSCI_HEADER_DIR)
LIBRARIES += -L$(NVSCI_LIB_DIR) -lnvscibuf -lnvscisync -lnvscievent -lnvsciipc -lnvscistream

ALL_CCFLAGS += --std=c++11 --threads 0

# CUDA code generation flags
# Gencode arguments
SMS ?= 53 61 70 72 75 80 86 87

ifeq ($(GENCODE_FLAGS),)
# Generate SASS code for each SM architecture listed in $(SMS)
$(foreach sm,$(SMS),$(eval GENCODE_FLAGS += -gencode arch=compute_$(sm),code=sm_$(sm)))

# Generate PTX code from the highest SM architecture in $(SMS) to guarantee forward-compatibility
HIGHEST_SM := $(lastword $(sort $(SMS)))
ifneq ($(HIGHEST_SM),)
GENCODE_FLAGS += -gencode arch=compute_$(HIGHEST_SM),code=compute_$(HIGHEST_SM)
endif
endif

################################################################################

# Target rules
OUTPUT := nvscistream_event_sample

all: build

build: $(OUTPUT)

OBJ := main.o
OBJ += block_common.o
OBJ += block_c2c.o
OBJ += block_consumer_uc1.o
OBJ += block_ipc.o
OBJ += block_limiter.o
OBJ += block_multicast.o
OBJ += block_pool.o
OBJ += block_presentsync.o
OBJ += block_producer_uc1.o
OBJ += block_queue.o
OBJ += block_returnsync.o
OBJ += event_loop_service.o
OBJ += event_loop_threads.o
OBJ += util.o


%.o: %.c
	$(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<

$(OUTPUT): $(OBJ)
	$(NVCC) $(ALL_LDFLAGS) $(GENCODE_FLAGS) -o $@ $+ $(LIBRARIES)

run: build
	$(OUTPUT)

testrun: build

clean:
	rm -f $(OBJ) $(OUTPUT)

clobber: clean