Enum ShaderAction 

pub enum ShaderAction {
    ListKernels,
    Doctor,
    Roofline {
        kernel: String,
        n: u64,
        json: bool,
    },
    ProfileHardware {
        export: Option<PathBuf>,
        json: bool,
    },
    Generate {
        kernel: String,
        target: String,
        schedule: ScheduleArgs,
        out: Option<PathBuf>,
        json: bool,
    },
    Validate {
        input: Option<PathBuf>,
        kernel: String,
        target: String,
        schedule: ScheduleArgs,
        manifest: Option<PathBuf>,
        json: bool,
    },
    Certify {
        kernel: String,
        schedule: ScheduleArgs,
        length: usize,
        warmups: usize,
        samples: usize,
        manifest: Option<PathBuf>,
        cache_dir: Option<PathBuf>,
        dry_run: bool,
    },
    Tune {
        kernel: String,
        length: usize,
        warmups: usize,
        initial_samples: usize,
        finalist_samples: usize,
        finalists: usize,
        max_candidates: usize,
        manifest: Option<PathBuf>,
        cache_dir: Option<PathBuf>,
        dry_run: bool,
    },
    AutoTuneAll {
        length: usize,
        warmups: usize,
        max_candidates: usize,
        cache_dir: Option<PathBuf>,
        update_local_manifest: bool,
        dry_run: bool,
        budget_ms: Option<u64>,
        thermal_limit: Option<f32>,
    },
}

Variants

ListKernels

List deterministic kernels currently known to WGSL Forge.

Doctor

Check that the native backend toolchains (wgpu adapter, DXC, CUDA) are present and report how the Forge will degrade if any are missing.

Roofline

Print the roofline estimate (FLOP/byte, memory- vs compute-bound) for a kernel.

Fields

kernel: String

n: u64

Representative problem size (output elements / records).

json: bool

ProfileHardware

Probe the local adapter and print a rich hardware/topology profile.

Fields

export: Option<PathBuf>

Write the profile JSON to this path (also prints the topology hash).

json: bool

Generate

Generate one deterministic WGSL module.

Fields

kernel: String

target: String

Target backend (wgsl, msl, hlsl, ptx, cuda, spirv)

schedule: ScheduleArgs

out: Option<PathBuf>

Write WGSL to this path instead of stdout.

json: bool

Emit the complete generated record as JSON.

Validate

Run Naga parsing and semantic validation.

Fields

input: Option<PathBuf>

Validate an existing WGSL file; otherwise generate the selected kernel.

kernel: String

target: String

Target backend (wgsl, msl, hlsl, ptx, cuda, spirv)

schedule: ScheduleArgs

manifest: Option<PathBuf>

Write a Naga-level certification manifest.

json: bool

Certify

Create a real pipeline, compare GPU output with the CPU oracle, and profile it.

Fields

kernel: String

schedule: ScheduleArgs

length: usize

warmups: usize

samples: usize

manifest: Option<PathBuf>

cache_dir: Option<PathBuf>

Also store the adapter-keyed certification in this cache directory.

dry_run: bool

Prune/emit/validate only; do not dispatch on the GPU.

Tune

Search the bounded schedule space and certify the fastest correct variant.

Fields

kernel: String

length: usize

warmups: usize

initial_samples: usize

finalist_samples: usize

finalists: usize

max_candidates: usize

manifest: Option<PathBuf>

cache_dir: Option<PathBuf>

Also store the adapter-keyed tuning record in this cache directory.

dry_run: bool

Report adapter-pruned candidate counts only; do not dispatch on the GPU.

AutoTuneAll

Tune every GPU-certifiable kernel, reusing the topology-keyed cache.

Fields

length: usize

warmups: usize

max_candidates: usize

cache_dir: Option<PathBuf>

Cache directory to read existing manifests from and (with –update-local-manifest) write new ones to.

update_local_manifest: bool

Persist freshly tuned manifests to the cache, keyed by topology.

dry_run: bool

List what would be tuned vs. served from cache; do not dispatch.

budget_ms: Option<u64>

Wall-clock tuning budget in milliseconds. Tuning stops before starting a kernel once this elapses (kernel-granular; a true mid-dispatch interrupt is not possible with the synchronous wgpu poll). Remaining kernels are reported as skipped-due-to-budget.

thermal_limit: Option<f32>

GPU thermal ceiling in degrees Celsius. Between kernels, if the GPU temperature (read via nvidia-smi) is at/above this, tuning waits and re-polls before continuing. Without an nvidia-smi sensor this flag has no effect and a single warning is printed.

Trait Implementations

impl Debug for ShaderAction

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl FromArgMatches for ShaderAction

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<'b>( &mut self, __clap_arg_matches: &mut ArgMatches, ) -> Result<(), Error>

Assign values from ArgMatches to self.

impl Subcommand for ShaderAction

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

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

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

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

fn has_subcommand(__clap_name: &str) -> bool

Test whether Self can parse a specific subcommand