FloatZone: Accelerating Memory Error Detection Using the Floating Point Unit [pdf]

Re:FloatZone: Accelerating Memory Error Detection Using the ... by moonchild on Tuesday July 11, 2023 @07:21AM (#36679400)

Very cute! A related trick—which I believe has been used by at least one gc's read barrier—is to turn a branch into a load from an address which will be unmapped if the branch should be taken.

[ Parent | on HN ]

Re:FloatZone: Accelerating Memory Error Detection Using the ... by barbito on Wednesday July 12, 2023 @08:29AM (#36693576)

I didn't understand the idea. Can you please provide more details?

[ Parent | on HN ]

Re:FloatZone: Accelerating Memory Error Detection Using the ... by pkhuong on Wednesday July 12, 2023 @09:23AM (#36694161)

You encode the test condition in address `ea` and execute some quasi no-op like `test eax, [ea]`. The address space is setup such that `ea` is unmapped / unreadable if the test condition says the code should trap.

[ Parent | on HN ]