News for hackers. Stuff that matters.

A comment on The Cost of Exception Handling.

Re:The Cost of Exception Handling by mrjin on Saturday November 12, 2022 @04:24AM (#33571528)
Took a glance, it seemed the author took exception handling as simple as try {} catch{} and thus never discuss OS' role in exception handling.

But one of the most expensive part in user mode exception handling is context switches between user mode and kernel mode.

[ Parent | on HN ]
Re:The Cost of Exception Handling by pjc50 on Saturday November 12, 2022 @04:53AM (#33571672)
Is that incurred for user-initiated exceptions (not traps or Windows "SEH")? In which implementations?
[ Parent | on HN ]
Re:The Cost of Exception Handling by ishitatsuyuki on Saturday November 12, 2022 @04:54AM (#33571676)
It is… not? It simply invokes the unwinder / exception handler, like _Unwind_RaiseException. For normal exceptions, no signals are involved. You can get an idea of what is executed at http://stffrdhrn.github.io/software/toolchain/openrisc/2020/....
[ Parent | on HN ]
Re:The Cost of Exception Handling by tsimionescu on Saturday November 12, 2022 @04:56AM (#33571680)
Are you referring to interrupts by any chance? The OS has nothing whatsoever to do with C++ exceptions in user-mode code.
[ Parent | on HN ]
Re:The Cost of Exception Handling by bregma on Saturday November 12, 2022 @07:07AM (#33572319)
Usermode exception handling performs no system calls. Perhaps you're thinking of user handling of system exceptions on Windows, which are pretty much unrelated to C++ exceptions?
[ Parent | on HN ]