[RISC-V] Support for trapping math in RISC-V

[RISC-V] Support for trapping math in RISC-V

[Krishna Narayanan]

[-- Attachment #1: Type: text/plain, Size: 343 bytes --]

Hi all,
Is the RISC-V community planning to add support for trapping math in RISC-V
in the near future!?
This LLVM thread
https://discourse.llvm.org/t/trapping-math-for-risc-v/72168/7  suggests a
software emulation of traps, is the GNU team also on similar lines to this
implementation or has some different views.

Thanks,
Krishna Narayanan.

Re: [RISC-V] Support for trapping math in RISC-V

[Joseph Myers]

On Wed, 24 Apr 2024, Krishna Narayanan via Gcc wrote:

> Hi all,
> Is the RISC-V community planning to add support for trapping math in RISC-V
> in the near future!?
> This LLVM thread
> https://discourse.llvm.org/t/trapping-math-for-risc-v/72168/7  suggests a
> software emulation of traps, is the GNU team also on similar lines to this
> implementation or has some different views.

That strategy sounds both very slow and not even complete (it doesn't 
cover the case of exact underflow, which doesn't raise the exception flag, 
but should signal the exception when trapping is enabled).

Traps are not part of standard C.  I'm not aware of any implementations of 
the TS 18661-5 interfaces to alternate exception handling either.

-- 
Joseph S. Myers
josmyers@redhat.com