Skip to content

Latest commit

 

History

History
460 lines (286 loc) · 20.1 KB

README.adoc

File metadata and controls

460 lines (286 loc) · 20.1 KB

RISC-V Architectural Testing Framework

1. Introduction

1.1. About

This document describes the RISC-V Architectural Testing framework which is used to test if a RISC-V device’s has understood and implemented the specifications correctly

  • It explains the framework around the tests, the running of individual tests, and the suites of tests.

  • It explains how to set up targets to run the tests.

This document is made freely available under a [app_cc_by_4.0].

1.2. Intent of the architectural test suite

The RISC-V Architectural Tests are an evolving set of tests that are created to help ensure that software written for a given RISC-V Profile/Specification will run on all implementations that comply with that profile.

These tests also help ensure that the implementer has both understood and implemented the specification correctly.

The RISC-V Architectural Test suite is a minimal filter. Passing the tests and having the results approved by RISC-V International is a prerequisite to licensing the RISC-V trademarks in connection with the design. Passing the RISC-V Architectural Tests does not mean that the design complies with the RISC-V Architecture. These are only a basic set of tests checking important aspects of the specification without focusing on details.

The RISC-V Architectural Tests are not a substitute for rigorous design verification.

The result that the architecture tests provide to the user is an assurance that the specification has been interpreted correctly and the implementation under test (DUT) can be declared as RISC-V Architecture Test compliant.

1.3. Intended audience

This document is intended for design and verification engineers who wish to check if their RISC-V implementation (simulation models, HDL models, etc.) is compliant to the RISC-V specification.

For those who wish to develop new architectural tests and/or to write or adapt their own test framework are suggested to read the Test Format Spec.

1.4. Contribute

You are encouraged to contribute to this repository (including changes to this document) by submitting pull requests and by commenting on pull requests submitted by other people as described in the README.md file in the top level directory.

While submitting a pull request note that some directories use ChangeLog files to track changes in the code and documentation. Please honor these, keeping them up to date and including the ChangeLog entry in the git commit message.

Make sure to also include a comment with the SPDX license identifier in all source files, for example:

// SPDX-License-Identifier: BSD-3-Clause
Note
Don’t forget to add your own name to the list of contributors in the document.

2. Licensing

In general:

  • code is licensed under the BSD 3-clause license (SPDX license identifier BSD-3-Clause);

  • documentation is licensed under the Creative Commons Attribution 4.0 International license (SPDX license identifier CC-BY-4.0).

The files COPYING.BSD and COPYING.CC in the top level directory contain the complete text of these licenses.

Note
The riscv-ovpsim simulator is licensed under an Imperas license. There is no dependency on this and it is included as a convenience to users.

2.1. AsciiDoc

This is a structured text format used by this document. Simple usage should be fairly self evident.

2.2. Installing tools

To generate the documentation as HTML you need asciidoctor and to generate as PDF you need asciidoctor-pdf.

To spell check you need aspell installed.

2.3. Building the documentation

To build HTML:

make html

To build PDF:

make pdf

To build both:

make

To check the spelling (excludes any listing or code phrases):

make spell

Any custom words for spell checking should be added to custom.wordlist.

2.4. Contributors

This document has been created by the following people (in alphabetical order of surname).

  • Jeremy Bennett

  • Mary Bennett

  • Simon Davidmann

  • Neel Gala

  • Radek Hajek

  • Lee Moore

  • Milan Nostersky

  • Marcela Zachariasova

2.5. Document history

Revision Date Author Modification

1.16 Draft

23 September 2020

Neel Gala

Changed Compliance to Architecture/Architectural. Refined the steps to port a new target. Removed simulator and target specific docs. Added vocabulary from the Test Format Specification

1.15 Draft

14 March 2019

Prashanth Mundkur

Added support and instructions for using the C and OCaml simulators from the Sail RISC-V formal model as targets.

1.14 Draft

21 February 2019

Deborah Soung

Documented how to use SiFive’s RISC-V ISA Formal Specification model as a target.

1.13 Draft

29 January 2019

Deborah Soung

Added documentation on how to use Rocket Chip generated cores as targets.

1.12 Draft

22 November 2018

Simon Davidmann

Updated notes on Test Suites.

1.11 Draft

21 November 2018

Neel Gala

Added new signature format specs .

1.10 Draft

20 June 2018

Simon Davidmann, Lee Moore

Cleaned up description of updated framework and inclusion of riscvOVPsim.

1.9 Draft

12 June 2018

Jeremy Bennett

Update Future work section to take account of Codasip changes. Remove diagrammatic directory structure.

1.8 Draft

12 June 2018

Jeremy Bennett

Add Future work section.

1.7 Draft

12 June 2018

Jeremy Bennett

Add CC license as an appendix.

1.6 Draft

10 June 2018

Jeremy Bennett

Tidy up areas that are flawed in HTML version.

1.5 Draft

8 June 2018

Jeremy Bennett

General tidy up.

1.4 Draft

8 June 2018

Jeremy Bennett

Added license preamble.

1.3 Draft

5 June 2018

Simon Davidmann

Updated to reflect directory structure and trace macros.

1.2 Draft

3 June 2018

Jeremy Bennett

Converted to AsciiDoc, cleaned up and restructured.

1.1 Draft

1 June 2018

Simon Davidmann Lee Moore

Revised format and expand to describe framework, usage of many tests groups, and different Targets

1.0

24 December 2017

Radek Hajek Milan Nostersky Marcela Zachariasova

First version of the document.

3. Repository structure

The top level directory contains :

  • a README.md file providing high-level details about the RISC-V Architecture Testing Framework.

  • top level Makefile for running the tests on a RISC-V implementation.

  • ChangeLog for logging changes with respect to code.

  • complete license files for the Creative Commons and BSD licenses used by the task group.

There are then five top level directories.

doc

All the documentation for the project, written using AsciiDoc.

coverage

This directory contains a sub-directory structure similar to the riscv-test-suite directory. The coverage reports of each test-suite are available in the corresponding directories. The reports are available in the html and yaml formats. The directory also contains multiple coverpoint YAMLs which capture all the coverpoints of interest which are covered by all the tests in the riscv-test-suite directory.

riscv-target

Contains a further subdirectory for each target, within which are placed the model_test.h header for that target and a device directory for all the devices of that target. If the $TARGETDIR environment variable is set to another directory, the scripts will search this directory for targets instead.

riscv-test-env

This contains the verify.sh script which checks if the signatures generated the riscv-target device match the statically hosted reference signatures at the end of simulation. The directory also contains symbolic-links to the arch_test.h and encoding.h files present in the riscv-test-suite/env directory.

riscv-test-suite

This contains a further env subdirectory which contains the arch_test.h and the encoding.h file which includes common assembly macros and routines which is used across the tests in the suite. Further directories are present which contain the actual architectural tests. The names and structure of these directories is based on the link:../test-pool structure guideline highlighted above.

riscv-ovpsim

This contains a README pointing to the Imperas OVP riscvOVPsim simulator for use in architectural testing.

4. Vocabulary

4.1. The architectural test

At the heart of the testing infrastructure is the detailed architectural test available as .S assembly files. The specification and format of the tests is defined in the Test Format Specification.

4.2. The architectural test pool

The architectural tests are grouped into different functional test suites targeting the different subsets of the RISC-V specification. For more details on the test-pool structure, hierarchy and conventions please refer to the Test Format Specification.

For information on the currently supported different test suites, look here: ../riscv-test-suite/README.md

4.3. The test signature

Each test in the architectural test pool generates a test signature, which represents the data written into specific memory locations during the execution of the test. The signature typically will record values (or sanitised values) of the operations carried out in the test. More details on the format and nature of the signatures is available in the Test Format Specification

4.4. The reference signature

In order to claim that a device/implementation has passed the RISC-V Architecture Tests, the test signatures obtained from the execution of the tests on the implementation need to be compared against a set of golden reference signature. These reference signatures are currently generated by the RISC-V SAIL formal model and statically hosted in the repository for each test.

4.5. The test target

The test target can be either a RISC-V Instruction Set Simulator (ISS), a RISC-V emulator, a RISC-V RTL model running on an HDL simulator, a RISC-V FPGA implementation or a physical chip. Each of the target types offers specific features and represents specific interface challenges. It is a role of the target shell to handle different targets while using the same architectural test pool as a test source.

4.6. The target shell

The target shell is the software and hardware environment around the test target that enables it to communicate with the framework, including assembling and linking tests, loading tests into memory, executing tests, and extracting the signature. The input to the target shell is a .S architectural test file, and the output is a test signature.

5. Porting a new target

In this section, a short tutorial on how to add a user target to the RISC-V Architectural Test Framework is provided.

The following steps demonstrate an example in which a target was replaced by the RISCV-ISA-SIM (a.k.a Spike). In a similar way, any RISC-V ISA simulator or any RTL simulation model of the RISC-V processor can be ported as a potential target for testing.

5.1. Setup environment variables

Note
ROOTDIR will always point to the riscv-arch-test repo.
  1. Clone the repository:


    git clone https://github.com/riscv/riscv-arch-test.git
    cd riscv-arch-test

  2. Open the Makefile.include available in the root folder of the repository and edit the following variables based on your target:

    1. TARGETDIR : set TARGETDIR to point to the directory which contains a sub-folder in the same name as the target. For example, the arch-test repo includes the targets: sail-riscv-c and spike under the riscv-target folder, in which case we set the TARGETDIR to riscv-target as shown below. One can set this completely arbitrary paths as suitable by the user.


      export TARGETDIR ?= $(ROOTDIR)/riscv-target

    2. XLEN: set XLEN to max supported XLEN. Allowed values are 32 and 64.


      export XLEN 							?= 64

    3. RISCV_TARGET: set this variable to the name of the target. A folder of the same name must exist in the TARGETDIR directory


      export RISCV_TARGET       ?= spike

    4. RISCV_DEVICE: set the RISCV_DEVICE environment to the extension you want to compile, simulate and verify. Leave this blank if you want to iterate through all the supported extensions of the target. Allowed values are the individual names of the extensions supported by your target like: I, M, C or Zifencei, etc. Multiple extensions are not be provided.


      export RISCV_DEVICE       ?=

    5. RISCV_TARGET_FLAGS: set this to a string which needs to be passed to your target’s Makefile.include files


      export RISCV_TARGET_FLAGS ?=

    6. RISCV_ASSERT: set this if you want to enable assertions on the test-suites. Currently no tests use assertions.


      export RISCV_ASSERT       ?= 0

    7. JOBS: set the number of parallel jobs (along with any other arguments) you would like to execute. Note that the target needs to be coded in such a way to support parallel execution. Some targets use common intermediate files, rather than unique files, which makes them unsuitable for parallel execution, these targets will need to be re-coded.


      JOBS= -j1

  3. Now inside your TARGETDIR/RISCV_TARGET directory you will need to create the following files:

    1. model_test.h: A header file containing the definition of the various target specific assembly macros that are required to compile and simulate the tests. The list and definition of the required target specific macros is available in the Test Format Specification

    2. link.ld: A linker script to compile the tests for your target.

    3. Any other files required by the target (configuration scripts, logs, etc.) can also be placed in this directory.

  4. Inside the TARGETDIR/RISCV_TARGET directory create a new folder named: device. If your device is a 32-bit target then create a directory device/rv32i_m. If your device is a 64-bit target then create a directory device/rv64i_m. If your target is configurable on the XLEN parameter then both the folders need to be created.

  5. Within the rv32i_m/rv64i_m directories sub-folders in the name of the extensions supported by the target need to be created. For eg. A target supporting the ISA RV32IMC_Zifence will have the following directory structure:


    • rv32i_m/I

    • rv32i_m/M

    • rv32i_m/C

    • rv32i_m/privilege

    • rv32i_m/Zifencei


  6. Each of the above extension directories will now need to include a file: Makefile.include which defines the following Makefile variables:

    1. RUN_TARGET:: This variable needs to include commands and steps to execute an ELF on target device. Note here that this variable should include all the necessary steps and arguments to run that specific test-suite. For example, in case of spike for the rv32i_m/C test-suite the corresponding Makefile.include has the --isa=rv32ic argument as opposed to just --isa=rv32i for the base rv32i_m/I test-suite. This variable should also include other steps to extract and sanitize the signature file as well for each test. The only argument available to this variable is the compiled elf file.

    2. COMPILE_TARGET:: This variable should include the commands and steps required to compile an assembly test for the target for each extension mentioned above. Note, currently only the GCC compiler is supported. This compiler takes march and mabi arguments from the corresponding architectural suite framework. COMPILE_TARGET will more or less be the same across test-suites. The only argument available to COMPILE_TARGET is the assembly file of one architectural test.

The following figure depicts the final directory structure of a target device that should get created at the end of the above steps:

TargetStruct
Figure 1. File Structure of the Target directory

5.2. Generating Signature

As previously mentioned the execution of each test on the target must generate a signature file. The name of the signature file should be <test-file-name.signature_output>. The signature file should follow the guidelines mentioned in the Test Format Specification.

In case of spike, we have ensured that the signature region is bounded by the labels: begin_signature and end_signature. This is enforced in the RVMODEL_DATA_BEGIN and RVMODEL_DATA_END macros defined in the riscv-target/spike/model_test.h file. Additionally, if you look closely at the RUN_TARGET variable defined in the Makefile.include files of the spike target, it includes a few bash commands to sanitize the signature produced from spike. This is done to conform the final signature file to the specification defined in the Test Format Specification.

5.3. Compile, simulate and verify the tests

Once you have ported your target to the riscv-arch-test framework by following the above steps, you are now ready to compile, simulate and verify the tests on your target

If you would like to compile, simulate and verify all the extension tests applicable to your target simply run make from the ROOTDIR.

Note
For the above to work the RISCV_DEVICE in ROOTDIR/Makefile.include must be left empty.

If you would like to only compile the tests for a particular extension you can use the following command.

make RISCV_DEVICE=M compile
make RISCV_DEVICE=C compile

The arguments to the RISCV_DEVICE variable must be the extensions supported by the target. NOTE: If RISCV_DEVICE is not defined/empty it will default to the I extension which is necessary for all targets.

To simulate the compiled tests on your target:

make RISCV_DEVICE=M simulate
make RISCV_DEVICE=Zifencei simulate
Note
If RISCV_DEVICE is not defined/empty it will default to the I extension which is necessary for all targets.

To verify if the generated signatures match the corresponding reference signatures.

make RISCV_DEVICE=M verify
Note
If RISCV_DEVICE is not defined/empty it will default to the I extension which is necessary for all targets.

All the above steps create and modify files in the work directory created in ROOTDIR folder. To clean the workdir simple run :

make clean

By default the working directory is set to ROOTDIR/work. This can be overwritten by assigning a new working directory path via the command line. Note, the path must be absolute and not relative :

make WORK=/home/me/my_path/mywork clean compile simulate