The highest tagged major version is
Documentation
¶
Overview ¶
Package thresholdvm re-exports the canonical Threshold (FHE / MPC) VM from github.com/luxfi/chains/thresholdvm so existing callers that imported github.com/luxfi/node/vms/thresholdvm pre-extraction keep working without source changes.
New code should import the canonical path:
"github.com/luxfi/chains/thresholdvm"
This file is a thin alias wrapper kept for backward compatibility.
Index ¶
Constants ¶
const ( GPUBackendNone = thresholdvm.GPUBackendNone GPUBackendCUDA = thresholdvm.GPUBackendCUDA GPUBackendHIP = thresholdvm.GPUBackendHIP GPUBackendMetal = thresholdvm.GPUBackendMetal GPUBackendVulkan = thresholdvm.GPUBackendVulkan GPUBackendWebGPU = thresholdvm.GPUBackendWebGPU )
Backend constants re-exported. Matches the chains/thresholdvm dlopen probe order: cuda → hip → metal → vulkan → webgpu.
Variables ¶
var ( ErrInvalidOperation = thresholdvm.ErrInvalidOperation NewClient = thresholdvm.NewClient )
var ErrGPUNotAvailable = thresholdvm.ErrGPUNotAvailable
ErrGPUNotAvailable is the canonical "no plugin loaded" error. Callers `errors.Is(err, thresholdvm.ErrGPUNotAvailable)` to distinguish a fallback-to-CPU condition from a hard launcher failure.
Functions ¶
This section is empty.
Types ¶
type Block ¶
type Block = thresholdvm.Block
type BlockError ¶
type BlockError = thresholdvm.BlockError
type Client ¶
type Client = thresholdvm.Client
type GPUBackend ¶ added in v1.29.0
type GPUBackend = thresholdvm.GPUBackend
func GPUBackendInstance ¶ added in v1.29.0
func GPUBackendInstance() *GPUBackend
GPUBackendInstance returns the dlopen'd GPU plugin handle resolved at chains/thresholdvm package init. nil means no plugin was loaded (CPU-only mode). The handle is shared across the whole process — both the node and chains paths see the same instance.
The name is GPUBackendInstance (not GPUBackend / Backend) because Go disallows a function named the same as a type alias in the same package, and `Backend` is already a domain term in the threshold state machine. Callers write:
if g := thresholdvm.GPUBackendInstance(); g != nil && g.IsAvailable() {
_, err := g.CeremonyApply(desc, ops, ceremonies)
...
}
This mirrors the cevm pattern `cevm.AvailableBackends()` / `cevm.LibraryABIVersion()` — discovery via package-scope function, not via a global variable.
func SelectGPUBackend ¶ added in v1.29.0
func SelectGPUBackend() (*GPUBackend, string)
SelectGPUBackend returns the resolved GPU plugin (or nil) and a single human-readable diagnostic string. luxd startup logs use this to surface "thresholdvm-gpu backend=<name>" lines alongside the cevm backend selection, matching the cevm.go pattern at ~/work/lux/chains/evm/backend_cgo.go.
Calling this multiple times is cheap — the underlying probe runs once at package init via sync.Once in chains/thresholdvm/backend.go.
type GPUBackendKind ¶ added in v1.29.0
type GPUBackendKind = thresholdvm.GPUBackendKind
type GPUCeremony ¶ added in v1.29.0
type GPUCeremony = thresholdvm.GPUCeremony
type GPUCeremonyOp ¶ added in v1.29.0
type GPUCeremonyOp = thresholdvm.GPUCeremonyOp
type GPUContribution ¶ added in v1.29.0
type GPUContribution = thresholdvm.GPUContribution
type GPUContributionOp ¶ added in v1.29.0
type GPUContributionOp = thresholdvm.GPUContributionOp
type GPUKeyShare ¶ added in v1.29.0
type GPUKeyShare = thresholdvm.GPUKeyShare
type GPUMPCVMRoundDescriptor ¶ added in v1.29.0
type GPUMPCVMRoundDescriptor = thresholdvm.GPUMPCVMRoundDescriptor
type GPUMPCVMState ¶ added in v1.29.0
type GPUMPCVMState = thresholdvm.GPUMPCVMState
type GPUMPCVMTransitionResult ¶ added in v1.29.0
type GPUMPCVMTransitionResult = thresholdvm.GPUMPCVMTransitionResult
type Operation ¶
type Operation = thresholdvm.Operation
Source Files