URL
https://opencores.org/ocsvn/openrisc_me/openrisc_me/trunk
Subversion Repositories openrisc_me
[/] [openrisc/] [trunk/] [orpsocv2/] [sim/] [bin/] [Makefile] - Rev 131
Go to most recent revision | Compare with Previous | Blame | View Log
######################################################################
#### ####
#### ORPSoCv2 Testbenches Makefile ####
#### ####
#### Description ####
#### ORPSoCv2 Testbenches Makefile, containing rules for ####
#### configuring and running different tests on the current ####
#### ORPSoC(v2) design. ####
#### ####
#### To do: ####
#### * Test if each software test file gets made properly ####
#### before it's run in whatever model we're using ####
#### * Expand software test-suite (uClibc, ecos tests, LTP?) ####
#### ####
#### Author(s): ####
#### - jb, jb@orsoc.se ####
#### ####
#### ####
######################################################################
#### ####
#### Copyright (C) 2009 Authors and OPENCORES.ORG ####
#### ####
#### This source file may be used and distributed without ####
#### restriction provided that this copyright statement is not ####
#### removed from the file and that any derivative work contains ####
#### the original copyright notice and the associated disclaimer. ####
#### ####
#### This source file is free software; you can redistribute it ####
#### and/or modify it under the terms of the GNU Lesser General ####
#### Public License as published by the Free Software Foundation; ####
#### either version 2.1 of the License, or (at your option) any ####
#### later version. ####
#### ####
#### This source is distributed in the hope that it will be ####
#### useful, but WITHOUT ANY WARRANTY; without even the implied ####
#### warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR ####
#### PURPOSE. See the GNU Lesser General Public License for more ####
#### details. ####
#### ####
#### You should have received a copy of the GNU Lesser General ####
#### Public License along with this source; if not, download it ####
#### from http://www.opencores.org/lgpl.shtml ####
#### ####
######################################################################
# Usage:
#
# make rtl-tests
#
# Run the software tests in the RTL model of the ORPSoC being
# simulated with an event-driven simulator like Icarus. It's also
# possible to use Modelsim's vsim and Cadence's Verilog simulators.
#
# make vlt-tests
#
# Run all the software tests in the RTL model which has been
# converted into a cycle-accurate SystemC model with Verilator.
#
# make sim-tests
#
# Run all the software tests in the architectural simulator
#
#
# Debugging modes:
#
# make rtl-debug
#
# Enable a GDB stub integrated into the simulation via VPI. This will
# start a simulation, then the GDB server, and allow the user to connect
# using the OpenRISC GDB port. It should provide the same functionality
# as GDB to a physical target, although a little slower.
# It is provided here as an example of how to compile and run an OpenRISC
# model at RTL level with support for debugging from GDB.
# UART output from printf() is enabled by default. The model loads with
# the dhrystone test running as default, but can be changed by defining
# VPI_TEST_SW at the command line. Logging of the processor's execution
# is also disabled by default to speed up simulation.
#
# Simulation results:
#
# The results and output of the event-driven simulations are in the
# results path, in parallel to the simulation run and bin paths.
# Specific tests:
#
# To run an individual test, specify it in the variable TESTS when
# calling make, eg:
#
# make rtl-tests TESTS="mmu-nocache mul-idcd-O2"
# UART printf:
#
# It is possible to enable printf to the console via the UART when
# running the event-driven simulators. To do this define UART_PRINTF=1
# when calling make. The SystemC cycle-acccurate model uses this by
# default.
# Also note when switching between runs with and without UART printf
# enabled, run a clean-sw so the library files are recompiled when
# the tests are run - this is not done automatically.
# VCDs:
#
# VCD (value change dumps, usable in a waveform viewer, such as gtkwave
# to inspect the internals of the system graphically) files can be
# generated by defining a variable VCD, eg.
#
# make rtl-tests VCD=1
#
# and a dump file will be created in the simulation results directory,
# and named according to the test run which generated it. This is
# possible for both event-driven and cycle-accurate simulations.
# However the cycle-accurate
# NO_SIM_LOGGING:
#
# It is possible to speed up the event-driven simulation slightly by
# disabling log output of the processor's state to files by defining
# NO_SIM_LOGGING, eg:
#
# make rtl-tests TESTS=except-icdc NO_SIM_LOGGING=1
#
# Cleaning:
# A simple "make clean" cleans everything - software and all temporary
# simulation files and directories. To clean just the software run:
#
# make clean-sw
#
# and to clean just the temporary simulation files (including VCDs,
# results logs - everything under, and including, sim/results/, run
#
# make clean-sim
#
# Note:
#
# The way each of the test loops is written is probably a bit overly complex
# but this is to save maintaining, and calling, multiple files.
#
# Model configuration:
#
# Currently, the ORPSoCv2, by default, contains an internal SRAM (configurable
# size - check the defparam in rtl/verilog/orpsoc_top.v), standard OR1200 (check
# the config in rtl/verilog/or1200_defines.v) and UART.
# Switches can be passed to enable certain parts of the design if testing with
# these is desired.
#
# SDRAM and controller
#
# To enable the use of SDRAM, define USE_SDRAM when calling the sim -this
# only has an effect in the event-driven simulators as the external SDRAM model
# is not availble in SystemC format. eg:
#
# make rtl-tests USE_SDRAM=1
#
# This not only enables SDRAM but also enables the booting from external SPI
# interfaced flash memory. This causes significant increase in the time taken
# for simulation as the program to test is first loaded out of SPI flash memory
# and into SDRAM before it is executed. Although this more closely mimics the
# behaviour of the hardware, for simulation purposes it is purely time-consuming
# however it may be useful to track down any problems with this boot-loading
# process. Therefore, becuase it enables SDRAM memory, it also enables the flash
# memory model and SPI controller inside ORPSoC.
#
# Ethernet
#
# Ethernet is disabled by default. This is due to the fact that it is not
# supported in the verilator/systemC model. Also, there is currently no software
# which tests it in any meaningful way.
#
#
# Event-driven simulation compilation
#
# The way the event-driven simulations are compiled is simply using the
# configuration script file in this directory, currently called icarus.scr -
# however it is first processesed to replace the variables, beginning with $'s,
# with the appropriate paths. Instead of naming each file to be compiled, the
# paths to be searched for each module are instead defined ( -y paths), and
# only the toplevel testbench and library source files are explicitly named.
# This simplifies the script, and also requires that the name of each verilog
# source file is the same as the module it contains (a good convention
# regardless.) In addition to the script/command file, defines are passed to
# the compiler via the command line in the EVENT_SIM_FLAGS variable.
# Additionally, a source file, test_define.v, is created with some defines
# that cannot be passed to the compiled reliably (there are differences between
# the way, for instance, icarus and ncverilog parse strings +define+'d on the
# command line). This file is then included at the appropriate places.
# It is probably not ideal that the entire design be re-compiled for each test,
# but currently the design is small enough so that this doesn't cause a
# significant overhead, unlike the cycle-accurate model compile time.
#
#
# SystemC cycle-accurate model compilation
#
# A new addition to ORPSoC v2 is the cycle-accurate model. The primary enabler
# behind this is verilator, which processes the RTL source and generates a c++
# description of the system. This c++ description is then compiled, with a
# SystemC wrapper. Finally a top-level SystemC testbench instantiates the
# model, and other useful modules - in this case a reset generation, UART
# decoder, and monitor module are included at the top level. These additional
# modules and models are written in SystemC. Finally, everything is linked with
# the cycle-accurate ORPSoC model to create the simulation executable. This
# executable is the cycle-representation of the system.
#
# Run the resulting executable with the -h switch for usage.
#
# The compilation is all done with the GNU c++ compiler, g++.
#
# The compilation process is a little more complicated than the event-driven
# simulator. It proceeds basically by generating the makefiles for compiling
# the design with verilator, running these makes which produces a library
# containing the cycle-accurate ORPSoC design, compiling the additional
# top-level, and testbench, systemC models into a library, and then linking it
# all together into the simulation executable.
#
# The major advantage of the cycle-accurate model is that it is quicker, in
# terms of simulated cycles/second, when compared with event-driven simulators.
# It is, of course, less accurate in that it cannot model propegation delays.
# However this is usually not an issue for simulating a design which is known
# to synthesize and run OK. It is very useful for running complex software,
# such as the linux kernel and real-time OS applications, which generally
# result in long simulation times.
#
# Currently the cycle-accurate model being used doesn't contain much more than
# the processor and a UART, however it's exepected in future this will be
# expanded on and more complex software test suites will be implemented to put
# the system through its paces.
#
#
#
# Name of the directory we're currently in
CUR_DIR=$(shell pwd)
# The root path of the whole project
PROJECT_ROOT ?=$(CUR_DIR)/../..
# Tests is only defined if it wasn't already defined when make was called
# This is the default list of every test that is currently possible
TESTS ?= basic-nocache cbasic-nocache-O2 dhry-nocache-O2 except-nocache mmu-nocache mul-nocache-O2 syscall-nocache tick-nocache uart-nocache basic-icdc cbasic-icdc-O2 dhry-icdc-O2 except-icdc mmu-icdc mul-icdc-O2 syscall-icdc tick-icdc uart-icdc
# Paths to other important parts of this test suite
SIM_DIR ?=$(PROJECT_ROOT)/sim
SIM_RUN_DIR=$(SIM_DIR)/run
SIM_BIN_DIR=$(SIM_DIR)/bin
SIM_RESULTS_DIR=$(SIM_DIR)/results
SIM_VLT_DIR=$(SIM_DIR)/vlt
BENCH_DIR=$(PROJECT_ROOT)/bench
BACKEND_DIR ?=$(PROJECT_ROOT)/backend
BENCH_VERILOG_DIR=$(BENCH_DIR)/verilog
BENCH_TOP_VERILOG_DIR ?= $(BENCH_DIR)/verilog
BENCH_SYSC_DIR=$(BENCH_DIR)/sysc
BENCH_SYSC_SRC_DIR=$(BENCH_SYSC_DIR)/src
BENCH_SYSC_INCLUDE_DIR=$(BENCH_SYSC_DIR)/include
RTL_VERILOG_DIR=$(PROJECT_ROOT)/rtl/verilog
SW_DIR=$(PROJECT_ROOT)/sw
ICARUS=iverilog
ICARUS_VVP=vvp
VSIM_COMP=vlog
VSIM=vsim
NCVERILOG=ncverilog
SILOS=silos
ICARUS_COMMAND_FILE=icarus.scr
VLT_COMMAND_FILE=verilator.scr
SIM_SUCCESS_MESSAGE=deaddead
MGC_COMMAND_FILE=modelsim.scr
ARCH_SIM_EXE=or32-elf-sim
ARCH_SIM_CFG_FILE=or1ksim-orpsocv2.cfg
# Set V=1 when calling make to enable verbose output
# mainly for debugging purposes.
ifeq ($(V), 1)
Q=
else
Q=@
endif
# If USE_SDRAM is defined we'll add it to the simulator's defines on the
# command line becuase it's used by many different modules and it's easier
# to do it this way than make them all include a file.
ifdef USE_SDRAM
EVENT_SIM_FLAGS +=USE_SDRAM=$(USE_SDRAM)
endif
# Enable ethernet if defined on the command line
ifdef USE_ETHERNET
EVENT_SIM_FLAGS +=USE_ETHERNET=$(USE_ETHERNET) USE_ETHERNET_IO=$(USE_ETHERNET)
# Extra tests we do if ethernet is enabled
TESTS += eth-basic eth-int
endif
DASH_D_EVENT_SIM_FLAGS=$(shell for flag in $(EVENT_SIM_FLAGS); do echo "-D "$$flag; done)
PLUS_DEFINE_EVENT_SIM_FLAGS=$(shell for flag in $(EVENT_SIM_FLAGS); do echo "+define+"$$flag; done)
#Default simulator is Icarus Verilog
# Set SIMULATOR=vsim to use Modelsim
# Set SIMULATOR=ncverilog to use Cadence's NC-Verilog
SIMULATOR ?= $(ICARUS)
# Set the command file to use, simulator dependent
ifeq ($(SIMULATOR), $(ICARUS))
# Icarus Verilog Simulator
SIM_COMMANDFILE=$(ICARUS_COMMAND_FILE)
endif
ifeq ($(SIMULATOR), $(VSIM))
# Modelsim has own command file (it's a little more stupid than Icarus & NC)
SIM_COMMANDFILE=$(MGC_COMMAND_FILE)
endif
ifeq ($(SIMULATOR), $(NCVERILOG))
# NCVerilog uses same command file as Icarus
SIM_COMMANDFILE=$(ICARUS_COMMAND_FILE)
endif
ifeq ($(SIMULATOR), $(SILOS))
# SILOS uses same command file as Icarus (this should be default)
SIM_COMMANDFILE=$(ICARUS_COMMAND_FILE)
endif
GENERATED_COMMANDFILE=$(SIM_COMMANDFILE).generated
# When Modelsim is selected as simulator, we compile
# the ORPSoC system into one library called orpsoc and
# then simply re-compile the testbench and or1200_monitor
# whenever we run the simulation, so just that part is
# recompiled for every test, instead of the whole thing.
MGC_ORPSOC_LIB=orpsoc
MGC_ORPSOC_LIB_DIR=$(SIM_RUN_DIR)/$(MGC_ORPSOC_LIB)
# If VCD dump is desired, tell Modelsim not to optimise
# away everything.
ifeq ($(VCD), 1)
VOPT_ARGS=-voptargs="+acc=rnp"
endif
# RTL testbench toplevel name
RTL_TESTBENCH_TOP ?= orpsoc_testbench
# Simulation compile and run commands, depending on your
# simulator.
# Icarus Verilog
ifeq ($(SIMULATOR), $(ICARUS))
# Icarus Verilog Simulator compile and run commands
SIM_COMMANDCOMPILE=rm -f $(SIM_RUN_DIR)/a.out; $(ICARUS) -s$(RTL_TESTBENCH_TOP) -c $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE) $(DASH_D_EVENT_SIM_FLAGS)
# Icarus Verilog run command
SIM_COMMANDRUN=$(ICARUS_VVP) -l $(SIM_RESULTS_DIR)/$$TEST-vvp-out.log a.out
endif
# Modelsim
ifeq ($(SIMULATOR), $(VSIM))
# Line to compile the orpsoc design into a modelsim library.
SIM_COMMANDCOMPILE=if [ ! -e work ]; then vlib work; vlib $(MGC_ORPSOC_LIB); vlog -work $(MGC_ORPSOC_LIB) -f $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE) $(PLUS_DEFINE_EVENT_SIM_FLAGS); fi
# Final modelsim compile, done each time, pulling in or1200
# monitor and the new test_defines.v file:
VSIM_COMPILE_TB=vlog +incdir+. +incdir+$(BENCH_VERILOG_DIR) -y $(BENCH_VERILOG_DIR) +libext+.v +incdir+$(BENCH_TOP_VERILOG_DIR) +incdir+$(RTL_VERILOG_DIR) +define+TEST_DEFINE_FILE $(PLUS_DEFINE_EVENT_SIM_FLAGS) $(BENCH_TOP_VERILOG_DIR)/$(RTL_TESTBENCH_TOP).v
# Simulation run command:
SIM_COMMANDRUN=$(VSIM_COMPILE_TB); $(VSIM) -c -quiet +nowarnTFMPC -L $(MGC_ORPSOC_LIB) $(VOPT_ARGS) -do "run -all; exit" $(RTL_TESTBENCH_TOP)
endif
# NCVerilog
ifeq ($(SIMULATOR), $(NCVERILOG))
SIM_COMMANDCOMPILE=echo
SIM_COMMANDRUN=$(NCVERILOG) -f $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE) -Q -l $(SIM_RESULTS_DIR)/$$TEST-$(NCVERILOG)-out.log $(EVENT_SIM_FLAGS)
endif
# Silos
ifeq ($(SIMULATOR), $(SILOS))
SIM_COMMANDCOMPILE=echo
SIM_COMMANDRUN=$(SILOS) -b -w +width_mistmatches -f $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE) -l $(SIM_RESULTS_DIR)/$$TEST-$(SILOS)-out.log $(EVENT_SIM_FLAGS)
endif
# Names of memory files used in simulation
SIM_FLASH_MEM_FILE="flash.in"
FLASH_MEM_FILE_SUFFIX="-twobyte-sizefirst.hex"
SIM_SRAM_MEM_FILE="sram.vmem"
TESTS_PASSED=0
TESTS_PERFORMED=0;
################################################################################
# Event-driven simulator build rules
################################################################################
$(RTL_VERILOG_DIR)/components/wb_sdram_ctrl/wb_sdram_ctrl_fsm.v:
@cd $(RTL_VERILOG_DIR)/components/wb_sdram_ctrl && perl fizzim.pl -encoding onehot -terse < wb_sdram_ctrl_fsm.fzm > wb_sdram_ctrl_fsm.v
.PHONY: prepare-rtl
prepare-rtl: $(RTL_VERILOG_DIR)/components/wb_sdram_ctrl/wb_sdram_ctrl_fsm.v
$(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE): $(SIM_BIN_DIR)/$(SIM_COMMANDFILE)
$(Q)sed < $(SIM_BIN_DIR)/$(SIM_COMMANDFILE) > $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE) \
-e s!\$$BENCH_DIR!$(BENCH_VERILOG_DIR)! \
-e s!\$$RTL_DIR!$(RTL_VERILOG_DIR)! \
-e s!\$$BACKEND_DIR!$(BACKEND_DIR)! \
-e \\!^//.*\$$!d -e \\!^\$$!d ; \
echo "+define+TEST_DEFINE_FILE=\"test_define.v\"" >> $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE); \
if [ ! -z $$VCD ]; \
then echo "+define+VCD" >> $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE); \
if [ $(SIMULATOR) = $(NCVERILOG) ]; \
then echo "+access+r" >> $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE); \
fi; \
fi; \
if [ ! -z $$UART_PRINTF ]; \
then echo "+define+UART_PRINTF" >> $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE); \
fi; \
if [ $(SIMULATOR) = $(NCVERILOG) ]; \
then echo "+nocopyright" >> $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE); \
echo "+nowarn+MACRDF" >> $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE); \
fi
ifdef UART_PRINTF
TEST_SW_MAKE_OPTS="UART_PRINTF=1"
endif
.PHONY: prepare-sw
prepare-sw:
$(Q)$(MAKE) -C $(SW_DIR)/support all $(TEST_SW_MAKE_OPTS)
$(Q)$(MAKE) -C $(SW_DIR)/utils all
# A rule with UART_PRINTF hard defined ... used by verilator make sw
prepare-sw-uart-printf:
$(Q)$(MAKE) -C $(SW_DIR)/support all UART_PRINTF=1 $(TEST_SW_MAKE_OPTS)
$(Q)$(MAKE) -C $(SW_DIR)/utils all
prepare-dirs:
$(Q)if [ ! -d $(SIM_RESULTS_DIR) ]; then mkdir -p $(SIM_RESULTS_DIR); fi
#
# Rough guide to how event driven simulation test loop works:
#
# 1. Compile software support programs.
# 2. Generate RTL compilation script file
# 3. For each test listed in $(TESTS), loop and
# a) Compile software
# b) Create appropriate image to be loaded into sim
# c) Create a verilog file to be included by top level
# d) Compile the RTL design
# e) Run the RTL design in the chosen simulator
# f) Check the output (files in ../results)
#
# Default setup is:
# * Event-driven simulation with Icarus Verilog
# * Internal SRAM memory, preloaded with application
# * Ethernet disabled
# * VCD generation disabled
# * printf() via UART disabled
# * Logging enabled
#
# Options:
# SIMULATOR=vsim
# Use Mentor Graphics Modelsim simulator
# SIMULATOR=ncverilog
# Use Cadence's NC-Verilog
# USE_SDRAM=1
# Enable use of SDRAM - changes boot sequence and takes
# a lot longer due to application being loaded out of
# external FLASH memory and into SDRAM before execution
# from the SDRAM.
# VCD=1
# Enable VCD generation. These files are output to
# ../results
# USE_ETHERNET=1
# Turns on ethernet core inclusion. There are currently
# some tests, but not included by default. Check the sw
# directory
# UART_PRINTF=1
# Make the software use the UART core to print out
# printf() calls.
# NO_SIM_LOGGING=1
# Turn off generation of logging files in the ../results
# directory.
#
rtl-tests: $(SIM_RUN_DIR)/$(GENERATED_COMMANDFILE) prepare-sw prepare-rtl prepare-dirs
@echo
@echo "Beginning loop that will complete the following tests: $(TESTS)"
@echo
$(Q)for TEST in $(TESTS); do \
echo "################################################################################"; \
echo; \
echo "\t#### Current test: $$TEST ####"; echo; \
echo "\t#### Compiling software ####"; echo; \
CURRENT_TEST_SW_DIR=$(SW_DIR)/`echo $$TEST | cut -d "-" -f 1`; \
$(MAKE) -C $$CURRENT_TEST_SW_DIR $$TEST $(TEST_SW_MAKE_OPTS); \
rm -f $(SIM_RUN_DIR)/$(SIM_FLASH_MEM_FILE); \
rm -f $(SIM_RUN_DIR)/$(SIM_SRAM_MEM_FILE); \
ln -s $$CURRENT_TEST_SW_DIR/$$TEST$(FLASH_MEM_FILE_SUFFIX) $(SIM_RUN_DIR)/$(SIM_FLASH_MEM_FILE); \
ln -s $$CURRENT_TEST_SW_DIR/$$TEST.vmem $(SIM_RUN_DIR)/$(SIM_SRAM_MEM_FILE); \
echo "\`define TEST_NAME_STRING \"$$TEST\"" > $(SIM_RUN_DIR)/test_define.v; \
echo "\`define TEST_RESULTS_DIR \"$(SIM_RESULTS_DIR)/\" " >> $(SIM_RUN_DIR)/test_define.v; \
if [ ! -z $$VCD ]; \
then echo "\`define VCD" >> $(SIM_RUN_DIR)/test_define.v; \
fi; \
if [ ! -z $$UART_PRINTF ]; \
then echo "\`define UART_PRINTF" >> $(SIM_RUN_DIR)/test_define.v; \
fi; \
if echo $$TEST | grep -q -i ^eth; then \
echo "\`define ENABLE_ETH_STIM" >> $(SIM_RUN_DIR)/test_define.v; \
echo "\`define ETH_PHY_VERBOSE" >> $(SIM_RUN_DIR)/test_define.v; \
fi; \
if [ -z $$NO_SIM_LOGGING ]; then \
echo "\`define OR1200_DISPLAY_ARCH_STATE" >> $(SIM_RUN_DIR)/test_define.v; \
fi; \
echo ; \
echo "\t#### Compiling RTL ####"; \
$(SIM_COMMANDCOMPILE); \
echo; \
echo "\t#### Beginning simulation ####"; \
time -p $(SIM_COMMANDRUN) ; \
if [ "$$SIMULATOR" != "$$SILOS" ]; then if [ $$? -gt 0 ]; then exit $$?; fi; fi; \
TEST_RESULT=`cat $(SIM_RESULTS_DIR)/$$TEST-general.log | grep report | grep $(SIM_SUCCESS_MESSAGE) -c`; \
echo; echo "\t####"; \
if [ $$TEST_RESULT -gt 0 ]; then \
echo "\t#### Test $$TEST PASSED ####";TESTS_PASSED=`expr $$TESTS_PASSED + 1`;\
else echo "\t#### Test $$TEST FAILED ####";\
fi; \
echo "\t####"; echo; \
TESTS_PERFORMED=`expr $$TESTS_PERFORMED + 1`;\
done; \
echo "Test results: "$$TESTS_PASSED" out of "$$TESTS_PERFORMED" tests passed"; echo
################################################################################
# RTL simulation in Icarus with GDB stub via VPI for debugging
################################################################################
# This compiles a version of the system which starts up the dhrystone nocache
# test, and launches the simulator with a VPI module that provides a GDB stub
# allowing the OpenRISC compatible GDB to connect and debug the system.
# The launched test can be changed by defining VPI_TEST_SW on the make line
VPI_DIR=$(BENCH_VERILOG_DIR)/vpi
VPI_C_DIR=$(VPI_DIR)/c
VPI_VERILOG_DIR=$(VPI_DIR)/verilog
VPI_LIB_NAME=jp_vpi
ICARUS_VPI_OPTS=-M$(VPI_C_DIR) -m$(VPI_LIB_NAME)
VPI_TEST_SW ?= dhry-nocache-O2
prepare-vpi:
## Build the VPI library
$(MAKE) -C $(VPI_C_DIR) $(VPI_LIB_NAME)
clean-vpi:
$(MAKE) -C $(VPI_C_DIR) clean
rtl-debug: prepare-sw-uart-printf prepare-rtl prepare-vpi prepare-dirs
## Prepare the software for the test
@echo "\t#### Compiling software ####"; echo; \
CURRENT_TEST_SW_DIR=$(SW_DIR)/`echo $(VPI_TEST_SW) | cut -d "-" -f 1`; \
$(MAKE) -C $$CURRENT_TEST_SW_DIR $(VPI_TEST_SW) $(TEST_SW_MAKE_OPTS); \
rm -f $(SIM_RUN_DIR)/$(SIM_FLASH_MEM_FILE); \
rm -f $(SIM_RUN_DIR)/$(SIM_SRAM_MEM_FILE); \
ln -s $$CURRENT_TEST_SW_DIR/$(VPI_TEST_SW)$(FLASH_MEM_FILE_SUFFIX) $(SIM_RUN_DIR)/$(SIM_FLASH_MEM_FILE); \
ln -s $$CURRENT_TEST_SW_DIR/$(VPI_TEST_SW).vmem $(SIM_RUN_DIR)/$(SIM_SRAM_MEM_FILE)
## Generate the icarus script we'll compile with
$(Q)sed < $(SIM_BIN_DIR)/$(ICARUS_COMMAND_FILE) > $(SIM_RUN_DIR)/$(ICARUS_COMMAND_FILE).generated \
-e s!\$$BENCH_DIR!$(BENCH_VERILOG_DIR)! \
-e s!\$$RTL_DIR!$(RTL_VERILOG_DIR)! \
-e s!\$$BACKEND_DIR!$(BACKEND_DIR)! \
-e \\!^//.*\$$!d -e \\!^\$$!d
## Add a couple of extra defines to the icarus compile script
$(Q)echo "+define+TEST_DEFINE_FILE=\"test_define.v\"" >> $(SIM_RUN_DIR)/$(ICARUS_COMMAND_FILE).generated
## The define that enables the VPI debug module
$(Q)echo "+define+VPI_DEBUG_ENABLE" >> $(SIM_RUN_DIR)/$(ICARUS_COMMAND_FILE).generated
$(Q)if [ ! -z $$VCD ];then echo "+define+VCD" >> $(SIM_RUN_DIR)/$(ICARUS_COMMAND_FILE).generated;fi
## Unless NO_UART_PRINTF=1 we use printf via the UART
$(Q)if [ -z $$NO_UART_PRINTF ];then echo "+define+UART_PRINTF" >> $(SIM_RUN_DIR)/$(ICARUS_COMMAND_FILE).generated; fi
$(Q)echo "\`define TEST_NAME_STRING \"$(VPI_TEST_SW)-vpi\"" > $(SIM_RUN_DIR)/test_define.v
$(Q)echo "\`define TEST_RESULTS_DIR \"$(SIM_RESULTS_DIR)/\" " >> $(SIM_RUN_DIR)/test_define.v
$(Q)if [ -z $$NO_SIM_LOGGING ]; then echo "\`define OR1200_DISPLAY_ARCH_STATE" >> $(SIM_RUN_DIR)/test_define.v; fi
@echo
@echo "\t#### Compiling RTL ####"
$(Q)rm -f $(SIM_RUN_DIR)/a.out
$(Q)$(ICARUS) -s$(RTL_TESTBENCH_TOP) -c $(SIM_RUN_DIR)/$(ICARUS_COMMAND_FILE).generated $(EVENT_SIM_FLAGS)
@echo
@echo "\t#### Beginning simulation with VPI debug module enabled ####"; echo
$(Q)$(ICARUS_VVP) $(ICARUS_VPI_OPTS) -l $(SIM_RESULTS_DIR)/$(VPI_TEST_SW)-vvp-out.log a.out
################################################################################
# Verilator model build rules
################################################################################
SYSC_LIB_ARCH_DIR=$(shell ls $$SYSTEMC | grep "lib-")
# List of System C models - use this list to link the sources into the Verilator
# build directory
SYSC_MODELS=OrpsocAccess MemoryLoad
ifdef VLT_DEBUG
VLT_DEBUG_COMPILE_FLAGS = -g
# Enabling the following generates a TON of debugging
# when running verilator. Not so helpful.
#VLT_DEBUG_OPTIONS = --debug --dump-tree
VLT_SYSC_DEBUG_DEFINE = VLT_DEBUG=1
endif
# If set on the command line we build the cycle accurate model which will generate verilator-specific profiling information. This is useful for checking the efficiency of the model - not really useful for checking code or the function of the model.
ifdef VLT_ORPSOC_PROFILING
VLT_CPPFLAGS +=-pg
VLT_DEBUG_OPTIONS +=-profile-cfuncs
else
VLT_CPPFLAGS +=-fprofile-use -Wcoverage-mismatch
#VLT_CPPFLAGS=-Wall
endif
# Set VLT_IN_GDB=1 when making if going to run the cycle accurate model executable in GDB to check suspect behavior. This also removes optimisation.
ifdef VLT_IN_GDB
VLT_CPPFLAGS +=-g -O0
else
# The default optimisation flag applied to all of the cycle accurate model files
VLT_CPPFLAGS +=-O3
endif
ifdef VLT_DO_PROFILING
VLT_CPPFLAGS +=-ftest-coverage -fprofile-arcs -fprofile-generate
endif
# VCD Enabled by default when building, enable it at runtime
#ifdef VCD
VLT_FLAGS +=-trace
TRACE_FLAGS=-DVM_TRACE=1 -I${SYSTEMPERL}/src
#endif
# Only need the trace target if we are tracing
#ifneq (,$(findstring -trace, $(VLT_FLAGS)))
VLT_TRACEOBJ = verilated_vcd_c
#endif
# This is the list of extra models we'll issue make commands for
# Included is the SystemPerl trace model
SYSC_MODELS_BUILD=$(SYSC_MODELS) $(VLT_TRACEOBJ)
prepare-vlt: prepare-rtl vlt-model-links $(SIM_VLT_DIR)/Vorpsoc_top
@echo;echo "\tCycle-accurate model compiled successfully"
@echo;echo "\tRun the executable with the -h option for usage instructions:";echo
$(SIM_VLT_DIR)/Vorpsoc_top -h
@echo;echo
$(SIM_VLT_DIR)/Vorpsoc_top: $(SIM_VLT_DIR)/libVorpsoc_top.a $(SIM_VLT_DIR)/OrpsocMain.o
# Final linking of the simulation executable. Order of libraries here is important!
@echo; echo "\tGenerating simulation executable"; echo
cd $(SIM_VLT_DIR) && g++ $(VLT_DEBUG_COMPILE_FLAGS) $(VLT_CPPFLAGS) -I$(BENCH_SYSC_INCLUDE_DIR) -I$(SIM_VLT_DIR) -I$(VERILATOR_ROOT)/include -I$(SYSTEMC)/include -o Vorpsoc_top -L. -L$(BENCH_SYSC_SRC_DIR) -L$(SYSTEMC)/$(SYSC_LIB_ARCH_DIR) OrpsocMain.o -lVorpsoc_top -lmodules -lsystemc
# Now compile the top level systemC "testbench" module from the systemC source path
$(SIM_VLT_DIR)/OrpsocMain.o: $(BENCH_SYSC_SRC_DIR)/OrpsocMain.cpp
@echo; echo "\tCompiling top level SystemC testbench"; echo
cd $(SIM_VLT_DIR) && g++ $(VLT_DEBUG_COMPILE_FLAGS) $(VLT_CPPFLAGS) $(TRACE_FLAGS) -I$(BENCH_SYSC_INCLUDE_DIR) -I$(SIM_VLT_DIR) -I$(VERILATOR_ROOT)/include -I$(SYSTEMC)/include -c $(BENCH_SYSC_SRC_DIR)/OrpsocMain.cpp
$(SIM_VLT_DIR)/libVorpsoc_top.a: $(SIM_VLT_DIR)/Vorpsoc_top__ALL.a vlt-modules-compile $(SIM_VLT_DIR)/verilated.o
# Now archive all of the libraries from verilator witht he other modules we might have
@echo; echo "\tArchiving libraries into libVorpsoc_top.a"; echo
$(Q)cd $(SIM_VLT_DIR) && \
cp Vorpsoc_top__ALL.a libVorpsoc_top.a && \
ar rcs libVorpsoc_top.a verilated.o; \
for SYSCMODEL in $(SYSC_MODELS_BUILD); do \
ar rcs libVorpsoc_top.a $$SYSCMODEL.o; \
done
$(SIM_VLT_DIR)/verilated.o:
@echo; echo "\tCompiling verilated.o"; echo
$(Q)cd $(SIM_VLT_DIR) && \
export CXXFLAGS=$(VLT_DEBUG_COMPILE_FLAGS); \
export USER_CPPFLAGS="$(VLT_CPPFLAGS)"; \
export USER_LDDFLAGS="$(VLT_CPPFLAGS)"; \
$(MAKE) -f Vorpsoc_top.mk verilated.o
.PHONY: vlt-modules-compile
vlt-modules-compile:
# Compile the module files
@echo; echo "\tCompiling SystemC models"
$(Q)cd $(SIM_VLT_DIR) && \
for SYSCMODEL in $(SYSC_MODELS_BUILD); do \
echo;echo "\t$$SYSCMODEL"; echo; \
export CXXFLAGS=$(VLT_DEBUG_COMPILE_FLAGS); \
export USER_CPPFLAGS="$(VLT_CPPFLAGS) -I$(BENCH_SYSC_INCLUDE_DIR)"; \
export USER_LDDFLAGS="$(VLT_CPPFLAGS)"; \
$(MAKE) -f Vorpsoc_top.mk $$SYSCMODEL.o; \
done
$(SIM_VLT_DIR)/Vorpsoc_top__ALL.a: $(SIM_VLT_DIR)/Vorpsoc_top.mk
@echo; echo "\tCompiling main design"; echo
$(Q)cd $(SIM_VLT_DIR) && \
export USER_CPPFLAGS="$(VLT_CPPFLAGS)"; \
export USER_LDDFLAGS="$(VLT_CPPFLAGS)"; \
$(MAKE) -f Vorpsoc_top.mk Vorpsoc_top__ALL.a
$(SIM_VLT_DIR)/Vorpsoc_top.mk: $(SIM_VLT_DIR)/$(VLT_COMMAND_FILE).generated $(SIM_VLT_DIR)/libmodules.a
# Now call verilator to generate the .mk files
@echo; echo "\tGenerating makefiles with Verilator"; echo
cd $(SIM_VLT_DIR) && \
verilator -language 1364-2001 -Wno-lint --top-module orpsoc_top $(VLT_DEBUG_OPTIONS) -Mdir . -sc $(VLT_FLAGS) -I$(BENCH_SYSC_INCLUDE_DIR) -I$(BENCH_SYSC_SRC_DIR) -f $(VLT_COMMAND_FILE).generated
# SystemC modules library
$(SIM_VLT_DIR)/libmodules.a:
@echo; echo "\tCompiling SystemC modules"; echo
$(Q)export VLT_CPPFLAGS="$(VLT_CPPFLAGS)"; \
$(MAKE) -C $(BENCH_SYSC_SRC_DIR) -f $(BENCH_SYSC_SRC_DIR)/Modules.make $(VLT_SYSC_DEBUG_DEFINE)
ALL_VLOG=$(shell find $(RTL_VERILOG_DIR) -name "*.v")
# Verilator command script
# Generate the compile script to give Verilator - make it sensitive to the RTL
$(SIM_VLT_DIR)/$(VLT_COMMAND_FILE).generated: $(ALL_VLOG)
@echo; echo "\tGenerating verilator compile script"; echo
$(Q)sed < $(SIM_BIN_DIR)/$(VLT_COMMAND_FILE) > $(SIM_VLT_DIR)/$(VLT_COMMAND_FILE).generated \
-e s!\$$BENCH_DIR!$(BENCH_VERILOG_DIR)! \
-e s!\$$RTL_DIR!$(RTL_VERILOG_DIR)! \
-e s!\$$BACKEND_DIR!$(BACKEND_DIR)! \
-e \\!^//.*\$$!d -e \\!^\$$!d;
.PHONY: vlt-model-links
vlt-model-links:
# Link all the required system C model files into the verilator work dir
@echo; echo "\tLinking SystemC model source to verilator build path"; echo
@if [ ! -d $(SIM_VLT_DIR) ]; then mkdir $(SIM_VLT_DIR); fi
$(Q)cd $(SIM_VLT_DIR) && \
for SYSCMODEL in $(SYSC_MODELS); do \
if [ ! -e $$SYSCMODEL.cpp ]; then \
ln -s $(BENCH_SYSC_SRC_DIR)/$$SYSCMODEL.cpp .; \
ln -s $(BENCH_SYSC_INCLUDE_DIR)/$$SYSCMODEL.h .; \
fi; \
done
################################################################################
# Verilator test loop
################################################################################
# Verilator defaults to internal memories
vlt-tests: prepare-sw prepare-rtl prepare-dirs prepare-vlt
@echo
@echo "Beginning loop that will complete the following tests: $(TESTS)"
@echo
$(Q)for TEST in $(TESTS); do \
echo "################################################################################"; \
echo; \
echo "\t#### Current test: $$TEST ####"; echo; \
echo "\t#### Compiling software ####"; echo; \
CURRENT_TEST_SW_DIR=$(SW_DIR)/`echo $$TEST | cut -d "-" -f 1`; \
$(MAKE) -C $$CURRENT_TEST_SW_DIR $$TEST $(TEST_SW_MAKE_OPTS) UART_PRINTF=1; \
rm -f $(SIM_RUN_DIR)/$(SIM_SRAM_MEM_FILE); \
ln -s $$CURRENT_TEST_SW_DIR/$$TEST.vmem $(SIM_RUN_DIR)/$(SIM_SRAM_MEM_FILE); \
echo "\t#### Beginning simulation ####"; \
time -p $(SIM_VLT_DIR)/Vorpsoc_top $$TEST; \
if [ $$? -gt 0 ]; then exit $$?; fi; \
TEST_RESULT=1; \
echo; echo "\t####"; \
if [ $$TEST_RESULT -gt 0 ]; then \
echo "\t#### Test $$TEST PASSED ####";TESTS_PASSED=`expr $$TESTS_PASSED + 1`;\
else echo "\t#### Test $$TEST FAILED ####";\
fi; \
echo "\t####"; echo; \
TESTS_PERFORMED=`expr $$TESTS_PERFORMED + 1`;\
done; \
echo "Test results: "$$TESTS_PASSED" out of "$$TESTS_PERFORMED" tests passed"; echo
###############################################################################
# Verilator profiled module make
###############################################################################
# To run this, first run a "make prepare-vlt VLT_DO_PROFILING=1" then do a
# "make clean" and then a "make prepare-vlt_profiled"
# This new make target copies athe results of the profiling back to the right
# paths before we create everything again
###############################################################################
.PHONY: prepare-vlt-profiled
prepare-vlt-profiled: $(SIM_VLT_DIR)/OrpsocMain.gcda clean vlt-restore-profileoutput prepare-rtl vlt-model-links $(SIM_VLT_DIR)/Vorpsoc_top
$(SIM_VLT_DIR)/OrpsocMain.gcda: $(SIM_VLT_DIR)/Vorpsoc_top-for-profiling prepare-sw-uart-printf
$(MAKE) -C $(SW_DIR)/dhry dhry-nocache-O2 NUM_RUNS=200
$(SIM_VLT_DIR)/Vorpsoc_top -f $(SW_DIR)/dhry/dhry-nocache-O2.or32 -v -l sim.log --crash-monitor
.PHONY: $(SIM_VLT_DIR)/Vorpsoc_top-for-profiling
$(SIM_VLT_DIR)/Vorpsoc_top-for-profiling:
$(MAKE) prepare-vlt VLT_DO_PROFILING=1
.PHONY: vlt-restore-profileoutput
vlt-restore-profileoutput:
@echo;echo "\tRestoring profiling outputs"; echo
$(Q)mkdir -p ../vlt
$(Q)cp /tmp/*.gc* $(SIM_VLT_DIR)
$(Q)cp /tmp/*.gc* $(BENCH_SYSC_SRC_DIR)
################################################################################
# Architectural simulator test loop
################################################################################
# Verilator defaults to internal memories
sim-tests: prepare-sw
@if [ ! -d $(SIM_RESULTS_DIR) ]; then mkdir -p $(SIM_RESULTS_DIR); fi
@echo
@echo "Beginning loop that will complete the following tests: $(TESTS)"
@echo
$(Q)for TEST in $(TESTS); do \
echo "################################################################################"; \
echo; \
echo "\t#### Current test: $$TEST ####"; echo; \
echo "\t#### Compiling software ####"; echo; \
CURRENT_TEST_SW_DIR=$(SW_DIR)/`echo $$TEST | cut -d "-" -f 1`; \
$(MAKE) -C $$CURRENT_TEST_SW_DIR $$TEST $(TEST_SW_MAKE_OPTS) UART_PRINTF=1; \
rm -f $(SIM_RUN_DIR)/$(SIM_SRAM_MEM_FILE); \
ln -s $$CURRENT_TEST_SW_DIR/$$TEST.or32 $(SIM_RUN_DIR)/.; \
echo;echo "\t#### Launching architectural simulator ####"; \
time -p $(ARCH_SIM_EXE) --nosrv -f $(SIM_BIN_DIR)/$(ARCH_SIM_CFG_FILE) $$TEST.or32 > $(SIM_RESULTS_DIR)/$$TEST-or1ksim.log 2>&1; \
if [ $$? -gt 0 ]; then exit $$?; fi; \
if [ `tail -n 10 $(SIM_RESULTS_DIR)/$$TEST-or1ksim.log | grep -c $(SIM_SUCCESS_MESSAGE)` -gt 0 ]; then \
TEST_RESULT=1; \
fi; \
echo; echo "\t####"; \
if [ $$TEST_RESULT -gt 0 ]; then \
echo "\t#### Test $$TEST PASSED ####";TESTS_PASSED=`expr $$TESTS_PASSED + 1`;\
else echo "\t#### Test $$TEST FAILED ####";\
fi; \
echo "\t####"; echo; \
TESTS_PERFORMED=`expr $$TESTS_PERFORMED + 1`;\
unlink $(SIM_RUN_DIR)/$$TEST.or32; \
done; \
echo "Test results: "$$TESTS_PASSED" out of "$$TESTS_PERFORMED" tests passed"; echo
################################################################################
# Cleaning rules
################################################################################
dist-clean: clean
$(MAKE) -C $(SW_DIR)/utils clean
clean: clean-sw clean-sim clean-sysc clean-rtl clean-vpi
clean-sw:
@for SWDIR in `ls $(SW_DIR) | grep -v utils`; do \
echo $$SWDIR; \
$(MAKE) -C $(SW_DIR)/$$SWDIR clean; \
done
clean-sim:
#backup any profiling output files
@if [ -f $(SIM_VLT_DIR)/OrpsocMain.gcda ]; then echo;echo "\tBacking up verilator profiling output to /tmp"; echo; \
cp $(SIM_VLT_DIR)/*.gc* /tmp; \
cp $(BENCH_SYSC_SRC_DIR)/*.gc* /tmp; fi
rm -rf $(SIM_RESULTS_DIR) $(SIM_RUN_DIR)/*.* $(SIM_VLT_DIR) $(MGC_ORPSOC_LIB_DIR) $(SIM_RUN_DIR)/work $(SIM_RUN_DIR)/transcript
clean-sysc:
# Clean away dependency files generated by verilator
$(MAKE) -C $(BENCH_SYSC_SRC_DIR) -f $(BENCH_SYSC_SRC_DIR)/Modules.make clean
clean-rtl:
# Clean away temporary verilog source files
rm -f $(RTL_VERILOG_DIR)/components/wb_sdram_ctrl/wb_sdram_ctrl_fsm.v
Go to most recent revision | Compare with Previous | Blame | View Log