Skip to content

runtime/race: eliminate cgo call overheads #4249

@dvyukov

Description

@dvyukov
Currently the callchain for memory accesses is:

runtime.raceread() -> runtime/race.Read() -> cgo -> runtime.syscallenter()
-> indirect call -> __tsan_read()

We need to eliminate the cgo overheads and maybe some forwarding as well.

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions