Where does warp collide?

In general relativity, to warp spacetime is to curve its geometry so that distances and the paths of light and matter bend. Alcubierre's 1994 The warp drive: Hyper-fast travel within general relativity proposed a metric that contracts space ahead of a craft and expands it behind, carrying a flat bubble through at effectively faster-than-light speed without the craft ever locally outrunning light. The referent is a deliberate deformation of the spacetime metric, described by curvature, and it shares only the word with threads on a loom or a bundle of processor threads.[1]

Unit

The degree of curvature imposed on the spacetime metric

Recovery means

Not applicable

Measured by

Described by the metric tensor and its curvature, and by the exotic energy the configuration would require

In NVIDIA's GPU architecture a warp is a group of parallel threads (32 on Tesla-class hardware) scheduled and executed together in lockstep, all running the same instruction on different data. Lindholm and colleagues' 2008 NVIDIA Tesla: A unified graphics and computing architecture defined the warp as the unit the hardware scheduler issues at once, so that divergence within a warp costs performance. The name is borrowed from weaving (a row of threads advancing together), but the referent is a scheduling unit inside a processor, not cloth and not spacetime.[2]

Unit

The number of threads executed together per issued instruction

Recovery means

Not applicable

Measured by

By warp size, occupancy, and branch (divergence) efficiency