Struct Cli 

pub(crate) struct Cli {

    pub(crate) input: Option<PathBuf>,
    pub(crate) target: String,
    pub(crate) output: Option<PathBuf>,
    pub(crate) opt_level: OptLevel,
    pub(crate) emit_ir: bool,
    pub(crate) emit_memory_plan: bool,
    pub(crate) emit_schedule: bool,
    pub(crate) dry_run: bool,
    pub(crate) precision: PrecisionPolicy,
    pub(crate) dynamic_batch: bool,
    pub(crate) symbolic_dim: Option<u32>,
    pub(crate) weights: Option<PathBuf>,
    pub(crate) calibration_data: Option<PathBuf>,
    pub(crate) calibration_method: CalibrationMethod,
    pub(crate) estimate_latency: bool,
    pub(crate) verbose: bool,
    pub(crate) list_targets: bool,
}

NxPU — WGSL to NPU transpiler

Fields

input: Option<PathBuf>

Input WGSL file

target: String

Target backend (default: ir-dump)

output: Option<PathBuf>

Output path (default: stdout)

opt_level: OptLevel

Optimization level: 0, 1, or 2

emit_ir: bool

Dump IR to stderr before backend compilation

emit_memory_plan: bool

Print memory plan to stderr (peak memory, buffer count, reuse ratio)

emit_schedule: bool

Dump operation schedule to stderr (dataflow analysis + scheduling)

dry_run: bool

Validate and optimize without producing output

precision: PrecisionPolicy

Precision policy: keep, f16, bf16, int8, or auto (default: auto)

dynamic_batch: bool

Mark the first dimension of all input tensors as dynamic (variable batch size)

symbolic_dim: Option<u32>

Concrete extent for tensor dimensions the kernel leaves symbolic

A WGSL kernel over array<f32> carries no length, so most dimensions have no size at compile time. ONNX names them and lets the runtime resolve them; TFLite requires a concrete extent and will not load a model without one. Without this flag the smallest valid extent (1) is used, which loads but measures nothing.

weights: Option<PathBuf>

Directory of tensor contents, one <tensor-name>.bin per tensor

A WGSL kernel takes its weights through var<storage, read>, so the source says nothing about what they are — the host binds them per dispatch. An NNAPI driver will not take a convolution on those terms. Measured on a MediaTek MT6899: mtk-neuron_shim accelerates a convolution whose filter is a compile-time constant and refuses the identical convolution, at the same shapes, whose filter is a graph input. Naming a tensor here emits it as a constant instead.

Raw little-endian, in the tensor’s own element type. A file whose size does not match the tensor is an error rather than a truncation.

Not needed for the GPU: TFLite’s own delegate takes a convolution with a runtime filter, and on that phone it accelerates more of what this compiler emits than any NNAPI driver does.

calibration_data: Option<PathBuf>

Directory containing calibration data (.bin files with f32 values)

calibration_method: CalibrationMethod

Calibration method: minmax, percentile, kl-divergence (default: minmax)

estimate_latency: bool

Print per-op cost estimation and roofline latency analysis to stderr

verbose: bool

Verbose output (print calibration statistics, etc.)

list_targets: bool

List all available target backends and exit

Trait Implementations

impl Args for Cli

fn group_id() -> Option<Id>

Report the [ArgGroup::id][crate::ArgGroup::id] for this set of arguments

fn augment_args<'b>(__clap_app: Command) -> Command

Append to [Command] so it can instantiate Self via [FromArgMatches::from_arg_matches_mut]

fn augment_args_for_update<'b>(__clap_app: Command) -> Command

Append to [Command] so it can instantiate self via [FromArgMatches::update_from_arg_matches_mut]

impl CommandFactory for Cli

fn command<'b>() -> Command

Build a [Command] that can instantiate Self.

fn command_for_update<'b>() -> Command

Build a [Command] that can update self.

impl FromArgMatches for Cli

fn from_arg_matches(__clap_arg_matches: &ArgMatches) -> Result<Self, Error>

Instantiate Self from [ArgMatches], parsing the arguments as needed.

fn from_arg_matches_mut( __clap_arg_matches: &mut ArgMatches, ) -> Result<Self, Error>

Instantiate Self from [ArgMatches], parsing the arguments as needed.

fn update_from_arg_matches( &mut self, __clap_arg_matches: &ArgMatches, ) -> Result<(), Error>

Assign values from ArgMatches to self.

fn update_from_arg_matches_mut( &mut self, __clap_arg_matches: &mut ArgMatches, ) -> Result<(), Error>

Assign values from ArgMatches to self.

impl Parser for Cli

fn parse() -> Self

Parse from std::env::args_os(), [exit][Error::exit] on error.

fn try_parse() -> Result<Self, Error>

Parse from std::env::args_os(), return Err on error.

fn parse_from<I, T>(itr: I) -> Self

where I: IntoIterator<Item = T>, T: Into<OsString> + Clone,

Parse from iterator, [exit][Error::exit] on error.

fn try_parse_from<I, T>(itr: I) -> Result<Self, Error>

where I: IntoIterator<Item = T>, T: Into<OsString> + Clone,

Parse from iterator, return Err on error.

fn update_from<I, T>(&mut self, itr: I)

where I: IntoIterator<Item = T>, T: Into<OsString> + Clone,

Update from iterator, [exit][Error::exit] on error.

fn try_update_from<I, T>(&mut self, itr: I) -> Result<(), Error>

where I: IntoIterator<Item = T>, T: Into<OsString> + Clone,

Update from iterator, return Err on error.