Skip to main content

Spice Models and Results Comparisons

Project description

SpiceCmp

A Python library for making comparisons between Spice-class simulators.

Running

SpiceCmp's primary object of comparison is the CompareMe (names, we know), which includes two "PDK plus simulator combinations" (PdkSimCombos), DUTs for comparison, and comparison conditions:

@dataclass
class CompareMe:
    """ Pair of Pdk-Simulator Combos to be Compared. 
    One always uses spectre, and the other always uses xyce. 
    
    Exposes two primary methods, both of which operate over a list of `TestCase`s: 

    * `run` netlists and runs the test-case simulations, as well as it can. 
    * `compare` gathers the results of each `TestCase`, compares them measurement-by-measurement, creates and saves a summary table. 
    """

    # The PDK + Simulator combinations
    spectre: PdkSimCombo
    xyce: PdkSimCombo

    # Comparison name, for labeling and run-directory naming
    name: str

    # DUTs for comparison
    xtors: List[MosModel]
    ro_cells: List[Callable]

    # Other shared data for comparison
    tempers: List[int]
    corners: List[Corner]

Each CompareMe has two primary API methods:

  • run generates netlists and invokes simulation
  • compare collects simulation results, runs post-processesing, and collates results into a summary table

Both take as their sole argument a list of TestCases. Each test-case includes the test conditions (e.g. temperature, process corner) and a shared Test object:

@dataclass
class TestCase:
    """ # Test Case 
    Combination of a `Test`, and the conditions under which it is run. """

    test: Test  # Reference to the parent `Test`
    dut: Any  # Device under test, or generator thereof
    corner: Corner  # PVT Corner
    temper: int  # Temperature

@dataclass
class Test:
    """ # Simulation Comparison Test """

    name: str  # Test Name
    run_func: Callable  # Run-Function
    meas_func: Callable  # Measurement-Manipulation Function

Each Test, in turn, is principally comprised of two functions:

  • A run_func which produces a simulatable DUT and invokes simulation, and
  • A meas_func which post-processes results which come back from said simulation

The post-processing measurement functions (meas_funcs) operate on solely on Spice measurement data. Measurements are typically mappings from string measurement-names to scalar, numeric result values. The required signature for each meas_func is therefore:

def meas_func(inp: Dict[str, float]) -> Dict[str, float]:
    """ Convert a "raw" input measurement-dictionary to post-processed form. """

In Python typing notation, the type of each meas_func is therefore:

meas_func: Callable[[Dict[str, float]], Dict[str, float]]

Run-functions take as their sole argument a TestCaseRun, the combination of a TestCase, PDK, and simulator to run it against, along with other metadata.

@dataclass
class TestCaseRun:
    """ # Test Case Run
    Execution of a `TestCase` with a particular PDK & simulator. """

    testcase: TestCase
    pdk: Pdk
    simulator: Simulator
    parentdir: Path
    errormode: ErrorMode

Their signature takes a TestCaseRun as input and returns nothing:

def run_func(run: TestCaseRun) -> None:
    """ Execute the `TestCaseRun` """

And again in typing module terms, run-functions are of type:

run_func: Callable[["TestCaseRun"], None]

Comparison Results

Results for each TestCase-measurement combination are collated into a MeasComparison including summary information about the test, conditions, and DUT.

@dataclass
class MeasComparison:
    """ Comparison of a Measurement in one of our Tests. 

    Serves as the row-type for the comparison table. 
    Yes, these field-names are non-Pythonic, 
    but they are designed to be nice header-fields in a table. """

    Test: str  # Test/ Test-Bench
    Dut: str  # Device Under Test
    Corner: str  # PVT Corner
    Temper: int  # Temperature
    Measurement: str  # Measurement Name

    Xyce: float  # Xyce Result
    Spectre: float  # Spectre Result
    Diff: float  # Difference (divided by average)

CompareMe.compare also generally saves the combined set of these MeasComparisons to a tabular CSV-format file, with column-names equal to the field-names of MeasComparison. An example such result:

Test,   Dut,        Corner, Temper, Measurement,    Xyce,           Spectre,        Diff
MosIv,  NMOS_STD,   TT,     -25,    idsat,          0.0005419505,   0.000541951,    -9.225930586166412e-07
MosIv,  NMOS_STD,   FF,     -25,    idsat,          0.0006205647,   0.000620565,    -4.834305391191293e-07
# ...

Note the Diff field of each comparison is relative: it reports the difference between the two simulator results, divided by their average value.

Development

To set up a dev install:

pip install -e ".[dev]"

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

spicecmp-6.0.0.tar.gz (13.5 kB view details)

Uploaded Source

File details

Details for the file spicecmp-6.0.0.tar.gz.

File metadata

  • Download URL: spicecmp-6.0.0.tar.gz
  • Upload date:
  • Size: 13.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/5.1.0 CPython/3.11.9

File hashes

Hashes for spicecmp-6.0.0.tar.gz
Algorithm Hash digest
SHA256 4e5eaaa31dd18d8cca3f239249f62023a5c65599c6152617333f42f2e73c1bc5
MD5 f00e513f85e744411debe0350303bcf6
BLAKE2b-256 6fc7084380fc8ed0a550752f93a8f0b8a8b20e6172f33ea556b8e5feec95dc60

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page