Skip to content

[dbsp] Transactional bootstrapping. #4736

@ryzhyk

Description

@ryzhyk

Currently bootstrapping is implemented by replaying the contents of internal indexes chunk-by-chunk. This is (1) inefficient compared to huge steps, (2) can produce mutually canceling outputs. We should instead take advantage of huge steps for bootstrapping. The prerequisite for doing this is that we should support huge steps for all programs, including programs with recursion.

Metadata

Metadata

Assignees

Labels

DBSP coreRelated to the core DBSP library

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions