Enum UnifiedError 

pub enum UnifiedError {
    Allocation(String),
    Cuda(String),
    ZeroSize,
    TooLarge {
        requested: u64,
        limit: u64,
    },
    NotSupported {
        feature: &'static str,
        backing: &'static str,
    },
}

Errors raised by UnifiedBuffer operations.

Variants

Allocation(String)

The underlying allocator failed.

Cuda(String)

A CUDA API call failed.

ZeroSize

Zero-byte allocation requested (not supported).

TooLarge

Requested allocation exceeds the configured / hard-coded cap.

Distinct from UnifiedError::Allocation so callers can plumb structured requested / limit figures all the way through to tensor_wasm_core::error::TensorWasmError::MemoryExhausted without resorting to substring-matching on a message.

Fields

requested: u64

Bytes the caller asked for.

limit: u64

Hard cap enforced by the host (bytes).

NotSupported

A UnifiedBacking method is not available on the active backing.

Surfaced by trait methods (e.g. prefetch_to_device on the cudarc stub or the Box<[u8]> host fallback) when the underlying backing has no implementation. Carries the method name and the backing tag so operator tooling and tests can match on the exact gap without scraping driver error strings.

Fields

feature: &'static str

Stable identifier for the method that has no implementation on this backing (e.g. "prefetch_to_device").

backing: &'static str

Stable identifier for the backing that lacks the feature (e.g. "cudarc-stub", "host-box").

Trait Implementations

impl Debug for UnifiedError

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

Formats the value using the given formatter.

impl Display for UnifiedError

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

Formats the value using the given formatter.

impl Error for UnifiedError

fn source(&self) -> Option<&(dyn Error + 'static)>

Returns the lower-level source of this error, if any.

fn description(&self) -> &str

👎Deprecated since 1.42.0:

use the Display impl or to_string()

fn cause(&self) -> Option<&dyn Error>

👎Deprecated since 1.33.0:

replaced by Error::source, which can support downcasting

fn provide<'a>(&'a self, request: &mut Request<'a>)

🔬This is a nightly-only experimental API. (error_generic_member_access)

Provides type-based access to context intended for error reports.

impl From<UnifiedError> for TensorWasmError

Cross-crate conversion so tensor-wasm-mem errors flow into the workspace’s unified TensorWasmError type without manual mapping at every call site.

• ZeroSize → TensorWasmError::Serialization with a descriptive message (zero-byte allocations are caller bugs, not memory exhaustion).
• Allocation { .. } → TensorWasmError::Serialization carrying the detail string. Exhaustion is NOT routed through this variant — pool / buffer exhaustion is reported as the structured UnifiedError::TooLarge { requested, limit } instead, which maps directly to MemoryExhausted below. Any remaining Allocation payload reaching this conversion is a caller bug (e.g. minimum > maximum, bad alignment) and is surfaced as Serialization accordingly.
• TooLarge { requested, limit } → TensorWasmError::MemoryExhausted { requested, limit } (1:1, structured).
• Cuda { .. } → TensorWasmError::CudaError (1:1 mapping).

fn from(e: UnifiedError) -> Self

Converts to this type from the input type.