Re: [PATCH] comp.c: Use the newly added bitcast API for type coercion, when available. (feature/jit-improved-type-punning)

[Vibhav Pant]

On Wed, 2022-09-28 at 12:37 +0000, Andrea Corallo wrote:
> thanks for the patch, please find some comments below.
Hi Andrea,

Thanks for the feedback.

> 
> Could you clarify why we need this double cast in both cases here?
> 

gcc_jit_context_new_bitcast requires that the types being converted
between have the same layout and as such, doesn't allow converting
between an arbitrarily sized integer/boolean and a pointer. Casting it
to a uintptr/void* is still necessary, to ensure that it can be bitcast
into a void*/uintptr respectively.

> > 
> 
> Are this and the following new line added voluntarily?
> 

Oops, fixed.

> 
> 
> Which kind of tests did this patch went through?  I assumed you tried
> a
> bootstrap could you please confirm?  Also have comp tests been tried?
> 

I ran the regular test suite, the code bootstraps succesfully, and I
have been using it as a daily driver for more than a week. I've
attached the results of comp tests with the expensive selector enabled
below.

Thanks,
Vibhav


> Thanks
> 
>   Andrea

comp-tests.log
Description: Text Data

signature.asc
Description: This is a digitally signed message part