Disclaimer Active development of the RISC-V BOOM Project Template is now located at https://github.com/riscv-boom/boom-template/
This is a starter template for your own RISC-V BOOM project.
BOOM is a superscalar, out-of-order processor that implements RISC-V RV64G ISA. BOOM is a synthesizable core that targets ASIC processes, and is written in the Chisel hardware construction language.
Feature | BOOM |
---|---|
ISA | RISC-V (RV64G) |
Synthesizable | √ |
FPGA | √ |
Parameterized | √ |
Floating Point (IEEE 754-2008) | √ |
Atomic Memory Op Support | √ |
Caches | √ |
Viritual Memory | √ |
Boots Linux | √ |
Privileged Arch v1.11 | √ |
External Debug | √ |
You can find more information about BOOM here: (github.com/ucb-bar/riscv-boom).
Google group: (https://groups.google.com/forum/#!forum/riscv-boom)
The submodules and subdirectories for the project template are organized as follows.
- rocket-chip - contains code for the RocketChip generator, Chisel HCL, and FIRRTL
- rocket-chip/riscv-tools - contains the code for the compiler toolchain and other infrastructure
- boom - contains code for the BOOM core and tile.
- scripts - bash scripts for initializing repo, building RISC-V toolchain
- testchipip - contains the serial adapter, block device, and associated verilog and C++ code
- verisim - directory in which Verilator simulations are compiled and run
- vsim - directory in which Synopsys VCS simulations are compiled and run
- src/main/scala - scala source files for your project extension can go here
After cloning this repo, you will need to initialize all of the submodules
git clone https://github.com/ccelio/boom-template.git
cd boom-template
./scripts/init-submodules.sh
The tools repo contains the cross-compiler toolchain, frontend server, and proxy kernel, which you will need in order to compile code to RISC-V instructions and run them on your design. There are detailed instructions at https://github.com/riscv/riscv-tools. But to get a basic installation that will work with BOOM, just the following steps are necessary.
# You may want to add the following two lines to your shell profile
export RISCV=/path/to/install/dir
export PATH=$RISCV/bin:$PATH
cd boom-template
./scripts/build-tools.sh
To compile a BOOM simulator, run make
in the "verisim" directory.
This will elaborate the BoomConfig from the boom.system project.
cd verisim
make
An executable called simulator-boom-system-BoomConfig will be produced. You can then use this executable to run any compatible RV64G code. For instance, to run one of the riscv-tools assembly tests.
make output/rv64ui-p-simple.out
Or execute the entire riscv-tests
suite:
make run
Or just a smaller regression suite:
make run-regression-tests
If you would like to get a .vpd
waveform, you can instead use:
make output/rv64ui-p-simple.vpd
Or:
make run-debug
If you later create your own project, you can use environment variables to build an alternate configuration. The different variables are
- PROJECT: The package that contains your test harness class
- CFG_PROJECT: The package that contains your config class
- GENERATOR_PROJECT: The package that contains your Generator class
- MODEL: The class name of your test harness
- CONFIG: The class name of your config
You can manually override them like this
make PROJECT=yourproject CONFIG=YourConfig
./simulator-yourproject-YourConfig ...
RISC-V Torture is included as a submodule and includes the ability to test BOOM. You can run a single test like so:
make rgentest R_SIM=../vsim/simv-boom.system-BoomConfig
You can run a nightly test, which runs for a set amount of time or a set number of failures like this:
make rnight R_SIM=../vsim/simv-boom.system-BoomConfig OPTIONS="-C config/default.config -t 5 -m 30"
The boom and rocket-chip repositories are included as git submodules. This allows us to tag specific versions that should work together. You can think of git submodules as "pointers" to specific commits.
To update this repository and then update
cd boom-template
git pull # update this repo and the pointers to the submodules.
git submodule update --recursive # update the contents of the submodules (and their submodules).
git status # check boom/rocket-chip don't have "new commits".
As rocket-chip itself has additional git submodules, you will need to perform a recursive submodule update. This may take a while.
The verisim
directory manages the Verilator build and run process.
The vsim
directory manages the VCS build and run process.
In either directory you can build and then run the riscv-tests
using make && make debug
.
The testharness/build-system is currently set up to provide a vpd
or a fsdb
waveform file.
To get a vpd
dump, go to the verisim
or vsim
directories and
instead of invoking make
, you can invoke make debug
to build a
vpd waveform-output-enabled BOOM simulator. The simulator will now be
suffixed with -debug
. Warning: Verilator takes a very long time
to compile with waveform output enabled.
To run all of the riscv-tests with vpd
waveform output, you can invoke make run-debug
.
Individually, you run a specific test as make output/rv64ui-p-simple.out
to run a regular test
or make output/rv64ui-p-simple.vpd
to generate a vpd
waveform.
To get a fsdb
dump, go to the vsim
directory and invoke make fsdb_debug
to build a fsdb waveform-output-enabled BOOM
simulator. The simulator will now be suffixed with fsdb-debug
. Do
not do this under the verisim
directory. You cannot generate a fsdb
enabled Verilator simulator.
You can generate fsdb
waveforms only when running tests individually with commands of the form
make output/rv64ui-p-simple.fsdb
. To invoke the verdi
verification tool, run
make output/rv64ui-p-simple.verdi
. If the required fsdb file is not available, this make
command will generate it first before starting verdi.
Read the Makefile to find all of the special targets.
You can add to your bash profile:
export MAKEFLAGS="-j `echo \`nproc\`*2/2|bc`"
This will spawn as many threads as you have cores when invoking make
, speed up compilation,
and run the riscv-tests
in parallel.
Warning: if you are performing VCS simulation you will burn through your precious licenses.
Warning: if you are writing a lot of data to the *.out files or *.vpd files you may hose your file system.
Unfortunately, many of the structures in an OOO processor scale worse than linearly. Also, some compilers struggle with large functions that arise when you turn the auto-generated Verilog into straight-line C++ code that is flattened across the whole design. As such, VCS compiles much faster than Verilator (but runs much slower).
Here are some times as measured on my machine using the verilator simulator:
-
make CONFIG=MegaBoomConfig run
takes 57 minutes. -
make CONFIG=BoomConfig run
takes 39 minutes. -
make CONFIG=SmallBoomConfig run
takes 15 minutes.
To improve the speed of your run-debug loop, you can instead invoke a smaller set of tests:
make run-regression-tests
First verify the software is not an issue. Run spike first:
# Verify it works on spike.
spike --isa=rv64imafd my_program
# Then we can run on BOOM.
./emulator-freechips.rocketchip.system-SmallBoomConfig my_program
Also verify the riscv-tools you built is the one pointed to within the boom-template/rocket-chip/riscv-tools repository. Otherwise a version mismatch can easily occur!
I recommend opening up the waveform and starting with the following signals,
located in TestDriver.testHarness.dut.tile.core
:
- debug_tsc_reg (cycle counter)
- debug_irt_reg (retired instruction counter)
- rob_io_commit_valids_* (the commit signals)
- csr_io_pc (roughly the commit pc)
- br_unit_brinfo_valid (was a branch/jalr resolved?)
- br_unit_brinfo_mispredict (was a branch/jalr mispredicted?)
- dis_valids_* (are any instructions being dispatched to the issue units?)
- dis_uops_*_pc (what are the PCs of the dispatched instructions?)
Go to boom/src/main/scala/common/consts.scala
and change COMMIT_LOG_PRINTF
to true
. That will output a log of committed instructions.
You can rebuild the spike
ISA simulator to also print out a commit log to
compare against.
cd rocket-chip/riscv-tools/riscv-isa-sim;
mkdir build
cd build
../configure --prefix=$RISCV --with-fesvr=$RISCV --with-isa=rv64imafd --enable-commitlog
Your new spike will ALWAYS print out a commit log to stderr
. I recommend
you change the prefix
to a different directory (and also build a new riscv-fesvr
to be placed in this same directory), rename your spike to something else
(e.g., lspike
), and add this prefix/bin
to your path. In this manner, you won't
overwrite your regular spike
binary.
Frustrating, right?
By default, the BOOM simulators built within the verisim
and vsim
directories are tethered to the riscv-fesvr (to handle binary loading
and proxying syscalls). The actual binary loading is performed using
the Debug Transport Module (DTM) which implements the RISC-V External
Debug Specification. The riscv-fesvr
magically sends signals to the
DTM to interrupt BOOM and have it execute out of the Debug Program
Buffer. In this manner, BOOM slowly loads the binary into its target
memory.
To proxy syscalls, the DTM occasionally interrupts the BOOM core to
have it read a special tohost
memory location. If the tohost
value
is non-zero, the BOOM core has a message for the riscv-fesvr to handle.
Spike is also tethered to the riscv-fesvr, but it can instantly load the test binary directly (and magically) into its target memory. There is no invocation of the Debug specification to do this.
Eventually, BOOM and spike will converge after they have each loaded the
test binary into their memories. They will again diverge on ocassion to
send proxy syscalls to the host, spinning for indeterminate amounts of
time while waiting for a fromhost
message (by repeatedly reading an
agreed-upon fromhost
address in memory).
There are other ways to load binaries and control the design-under-test.
The (https://github.com/ucb-bar/testchipip) repository provides a Tethered
Serial Interface (TSI) which can directly and coherently write into the target
memory to load programs and query for tohost
communications. This requires
a different bootrom
as well as different top-level I/O connections to
wire-up this TSI. This is less invasive than repeatedly polling through the
DTM, but also more special-purpose built.
Or, one could build a self-hosting BOOM system. But you need to provide the
appropriate drivers, IP blocks, and bootrom
to communicate with the system
in some manner. An example self-hosted rocket-chip system that follows the
project-template
layout is the (https://github.com/sifive/freedom) platform.
Processors are hard.
Boom-template/rocket-chip/boom comes with a quite a project hierarchy that may be hard to keep track of in its entirety. Here's some steps to get started with IntelliJ. This section is a work-in-progress, so please share your own tips and hints on the mailing list.
Step 1: Install a JDK (you have probably already done this, but you need to find it). If you are using OSX, I recommend:
brew tap caskroom/versions
brew update
brew cask install java8
This will give you a jdk that is probably in /Library/Java/JavaVirtualMachines
.
Step 2: Follow the above instructions to check out boom-template
and build a BOOM emulator.
Run it through some tests to make sure everything works. This is important, as this will
build a lib
file full of jar
files of Chisel, FIRRTL, RocketChip, BOOM, and others.
Step 3: Import a project in IntelliJ. Select boom-template/build.sbt
. Use Java 1.8.
Step 4: Right-click on boom-template/lib
in your project hierarchy and Add as Library
.
Step 5: Right-click on boom-template/boom/src
and Mark Directory as -> Sources Root
.
IntelliJ can't find source code that is not under boom-template/src/main/scala
without help.
Step 6: Right-click on boom/src/main/scala/system/Generator
and Run
. It will fail.
Step 7: Open Run->Edit Configurations..
and add the following arguments to your
program arguments
(main class should be boom.system.Generator
). Replace ${TEMPLATE}
with what is appropriate for your system.
${TEMPLATE}/verisim/generated-src boom.system TestHarness boom.system BoomConfig
That string comes from Makefrag-variables
. You can change BoomConfig to match your desired
${CONFIG}
.
Step 8: Run
your Generator
target. It should now generate firrtl code. However,
the rest of the boom-template
system relies on Makefiles to string everything together.
You can either build additional Run
configurations, or connect to the existing Makefiles.
Step 9: Go to Preferences->Plugins
and add Makefile support
.
Step 10: Right-click on on boom-template/verisim/Makefile
and run Makefile.
This should compile a BOOM verilator-based emulator.
Step 11: Edit your Run->Edit Configurations..
to add new Run configurations for
additional Make targets and ${CONFIG}
options. For development,
make $CONFIG=SmallBoomConfig run-regression-tests
is one recommended command.
You can also choose to invoke Make manually in your terminal.
Step 12: Browse the boom source code. Notice that you can highlight a variable and see
its definition (or see its instantiations). You can also browse the View->Tool Windows->Structure
to see what variables are in scope for a particular class. Click on show inherited
to see more,
in particular, what the traits
might be bringing in to scope!
Step 13: Provide feedback and pull requests to this README (or elsewhere). I am hoping that some find this useful and can provide more guidance on how to improve the IntelliJ/ boom-template flow for those that may follow.
The original source of boom-template derives from (https://github.com/ucb-bar/project-template). Please visit it to learn more about adding your own devices or submodules to this template.