Hi, we have gcc_jit_context_new_rvalue_from_int and gcc_jit_context_new_rvalue_from_long. What about gcc_jit_context_new_rvalue_from_long_long and other integral types? It could be emulated with initialized globals (and relying on the optimization pass), but only when the binary representation on the target is known, which it generally isn't unless the target is always the host. Thanks, Marc
Hi! I posted a patch addressing long long some time ago. https://gcc.gnu.org/pipermail/jit/2021q4/001375.html long long should cover all integral types on all platforms as you can pick the result type as you want, assuming unsigned long long never has more bits than long long :/ Regards, Petter Från: Jit <jit-bounces+tomner=kth.se@gcc.gnu.org> för Marc Nieper-Wißkirchen via Jit <jit@gcc.gnu.org> Skickat: den 13 december 2021 15:18 Till: jit@gcc.gnu.org Ämne: gcc_jit_context_new_rvalue_from_XXX Hi, we have gcc_jit_context_new_rvalue_from_int and gcc_jit_context_new_rvalue_from_long. What about gcc_jit_context_new_rvalue_from_long_long and other integral types? It could be emulated with initialized globals (and relying on the optimization pass), but only when the binary representation on the target is known, which it generally isn't unless the target is always the host. Thanks, Marc
Thanks for your quick and helpful response!
Getting this into the trunk would be great!
Am Mo., 13. Dez. 2021 um 15:54 Uhr schrieb Petter Tomner <tomner@kth.se>:
> Hi!
>
> I posted a patch addressing long long some time ago.
>
> https://gcc.gnu.org/pipermail/jit/2021q4/001375.html
>
> long long should cover all integral types on all platforms as you can
> pick the result type as you want, assuming unsigned long long never
> has more bits than long long :/
>
> Regards, Petter
>
>
>
> Från: Jit <jit-bounces+tomner=kth.se@gcc.gnu.org> för Marc
> Nieper-Wißkirchen via Jit <jit@gcc.gnu.org>
> Skickat: den 13 december 2021 15:18
> Till: jit@gcc.gnu.org
> Ämne: gcc_jit_context_new_rvalue_from_XXX
>
> Hi,
>
> we have
>
> gcc_jit_context_new_rvalue_from_int
>
> and
>
> gcc_jit_context_new_rvalue_from_long.
>
> What about
>
> gcc_jit_context_new_rvalue_from_long_long
>
> and other integral types?
>
> It could be emulated with initialized globals (and relying on the
> optimization pass), but only when the binary representation on the target
> is known, which it generally isn't unless the target is always the host.
>
> Thanks,
>
> Marc
>
This hasn't made it into the trunk yet, has it?
Is there a technical reason for the delay?
Thanks,
Marc
Am Mo., 13. Dez. 2021 um 16:12 Uhr schrieb Marc Nieper-Wißkirchen
<marc.nieper@gmail.com>:
>
> Thanks for your quick and helpful response!
>
> Getting this into the trunk would be great!
>
> Am Mo., 13. Dez. 2021 um 15:54 Uhr schrieb Petter Tomner <tomner@kth.se>:
>>
>> Hi!
>>
>> I posted a patch addressing long long some time ago.
>>
>> https://gcc.gnu.org/pipermail/jit/2021q4/001375.html
>>
>> long long should cover all integral types on all platforms as you can
>> pick the result type as you want, assuming unsigned long long never
>> has more bits than long long :/
>>
>> Regards, Petter
>>
>>
>>
>> Från: Jit <jit-bounces+tomner=kth.se@gcc.gnu.org> för Marc Nieper-Wißkirchen via Jit <jit@gcc.gnu.org>
>> Skickat: den 13 december 2021 15:18
>> Till: jit@gcc.gnu.org
>> Ämne: gcc_jit_context_new_rvalue_from_XXX
>>
>> Hi,
>>
>> we have
>>
>> gcc_jit_context_new_rvalue_from_int
>>
>> and
>>
>> gcc_jit_context_new_rvalue_from_long.
>>
>> What about
>>
>> gcc_jit_context_new_rvalue_from_long_long
>>
>> and other integral types?
>>
>> It could be emulated with initialized globals (and relying on the
>> optimization pass), but only when the binary representation on the target
>> is known, which it generally isn't unless the target is always the host.
>>
>> Thanks,
>>
>> Marc
>>