Minimal functional interface to AC²E for Hy.
Install with pip:
$ pip install git+https://github.com/augustunderground/hace.gitOr by cloning:
$ git clone https://github.com/augustunderground/hace.git
$ pushd hace
$ pip install .Make sure AC²E and all dependencies are installed.
One way is to create a symlink from the resource directory of the AC²E
repository, which contains all the backends as git submodules, to ~/.ace.
hace will automatically look there if pdk_path and ckt_path are not
specified.
$ ln -s /path/to/ace/resource $HOME/.aceIt should look something like this:
$HOME/.ace
├── sky130-1V8
│ ├── op1
│ │ ├── input.scs
│ │ └── properties.json
│ ├── ...
│ ├── pdk
│ │ ├── cells
│ │ │ ├── nfet_01v8
│ │ │ │ ├── sky130_fd_pr__nfet_01v8__mismatch.corner.scs
│ │ │ │ ├── sky130_fd_pr__nfet_01v8__tt.corner.scs
│ │ │ │ └── sky130_fd_pr__nfet_01v8__tt.pm3.scs
│ │ │ └── pfet_01v8
│ │ │ ├── sky130_fd_pr__pfet_01v8__mismatch.corner.scs
│ │ │ ├── sky130_fd_pr__pfet_01v8__tt.corner.scs
│ │ │ └── sky130_fd_pr__pfet_01v8__tt.pm3.scs
│ │ ├── models
│ │ │ ├── all.scs
│ │ │ ├── corners
│ │ │ │ └── tt
│ │ │ │ └── nofet.scs
│ │ │ ├── parameters
│ │ │ │ └── lod.scs
│ │ │ └── sky130.scs
│ │ ├── README.md
│ │ └── tests
│ │ ├── nfet_01v8_tt.scs
│ │ └── pfet_01v8_tt.scs
└── ...
Alternatively you can set environment variables, telling hace where to find
the pdk and testbenches.
$ export ACE_BACKEND=/path/to/ace/resource
$ export ACE_PDK=/path/to/ace/resource/<tech>/pdkWhere <tech> has to be a valid backend such as xh035-3V3 for example.
Otherwise paths have to be given explicitly to the make_env function via the
kwargs pdk_path and ckt_path.
Please refer to the AC²E documentation for environment IDs and available backends.
import hace as ac
amp = ac.make_env('op2', 'xh035-3V3')
siz = ac.random_sizing(amp)
res = ac.evaluate_circuit(amp, params = siz)Create an ace environment object:
amp = make_env( ace_id: str # ACE Environment ID
, ace_backend: str # ACE Backend ID
, pdk_path: Optional[List[str]] = [] # Path to ace backend
, ckt_path: Optional[str] = None # Path to testbench
, sim_path: Optional[str] = None # Path to store results
) => amplifier # Returns ace env objWhere ace_id can be any supported AC²E environment id such as op1 .. 9,
nand4 or st1. And ace_backend should be a supported/installed backend,
such as sky130-1V8.
res = evaluate_circuit( amp # Amplifier object
, params: Optional[Dict[str, float]] # Sizing parameters
, blocklist: Optional[List[str]] # List of blocked simulations
) => Dict[str, float] # Returns performanceThe blocklist argument may specify a list of simulations that will not be
performed. To get a list of all available simulation analyses one may call
sim = simulation_analyses(amp) => List[str] # List of available analysesTo get the simulation result of the previous run without simulating again:
res = current_performance(amp) => Dict[str, float]The following functions can be used for accessing and indexing the Dictionaries:
sizing_identifiers(amp) => List[str] # Keys in optional params dict
performance_identifiers(amp) => List[str] # Keys in current performance dictGet random sizing parameters:
rng = random_sizing(amp) => Dict[str, float] # Random sizing parametersGet sensible initial sizing parameters:
siz = initial_sizing(amp) => Dict[str, float] # "Good" sizing parametersSave the current state to a file (.json, .yaml and .csv are currently
supported):
dump_state(amp, file_name="file.ext") => Dict[str, float] # Dumps current stateLoad a state (created with dump_state):
load_state(amp, file_name: str) => Dict[str, float] # Loads the given state The concurrent API acts much the same as the default one. Create pooled amplifier object:
amps = ac.make_env_pool( ace_ids: List[str],
, ace_backends: List[str]
, pdk_paths: Optional[List[List[str]]] = [[]]
, ckt_paths: Optional[List[str]] = []
, sim_paths: Optional[List[str]] = ["/tmp", ..]
) => amplifier_poolThere is a short hand for creating a pool of the same environment.
envs = ac.make_same_env_pool( num_envs: int
, ace_id: str,
, ace_backends: str
, pdk_paths: Optional[List[str]] = []
, ckt_paths: Optional[str] = []
, sim_paths: Optional[str] = "/tmp"
) => amplifier_poolThe AcePoolEnvironment is a
namedtuple
with an envs and pool field. Where the former is a dict mapping an ID to an
environment.
Parameters of environments in a pool can be set with a dict mapping an ID to a
dict that is then passed to set_parameters.
envs = ac.set_parameters_pool( pool_env
, pool_params dict[int, dict[str, float]]
) => pool_envAn environment pool can be evaluated in a similar way.
ress = ac.evaluate_circuit_pool( pool_env
, pool_params Optional[dict[int, dict[str, float]]] = {}
, npar: int = len(os.shed_getaffinity(0)) // 2
) => simulation_resultsSo far the blocklist is not supported in pooled environments.
See ACE Documentation.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent