• Ferhat =?UTF-8?B?S3VydHVsbXXFnw==?= (2/2) May 10 2023 Do we have any implementation of ref-counted class instances? İf
• Richard (Rikki) Andrew Cattermole (7/7) May 10 2023 We do not, no.
• Ferhat =?UTF-8?B?S3VydHVsbXXFnw==?= (4/12) May 10 2023 Thank you Rikki for the clarification. I think d need this
• Richard (Rikki) Andrew Cattermole (5/7) May 10 2023 Yeah we do.

Ferhat =?UTF-8?B?S3VydHVsbXXFnw==?= <aferust gmail.com> writes:

Do we have any implementation of ref-counted class instances? İf 
not why?

"Richard (Rikki) Andrew Cattermole" <richard cattermole.co.nz> writes:

We do not, no.

Reference counting is not currently in the language, although you can 
fake it with structs.

Its a huge shame, but it should be added at some point because the 
compiler would tell the backend that it is possible to elide pair calls.

However it'll have to wait for the DIP queue to reopen (slight process 
adjustment to make things easier is incoming).

Ferhat =?UTF-8?B?S3VydHVsbXXFnw==?= <aferust gmail.com> writes:

On Wednesday, 10 May 2023 at 18:15:36 UTC, Richard (Rikki) Andrew 
Cattermole wrote:
 We do not, no.

 Reference counting is not currently in the language, although 
 you can fake it with structs.

 Its a huge shame, but it should be added at some point because 
 the compiler would tell the backend that it is possible to 
 elide pair calls.

 However it'll have to wait for the DIP queue to reopen (slight 
 process adjustment to make things easier is incoming).

Thank you Rikki for the clarification. I think d need this 
without ugly workarounds.

"Richard (Rikki) Andrew Cattermole" <richard cattermole.co.nz> writes:

On 11/05/2023 6:23 AM, Ferhat Kurtulmuş wrote:
 Thank you Rikki for the clarification. I think d need this without ugly 
 workarounds.

Yeah we do.

Reference counting is expensive, and modern backends have the capability 
to elide pairs of calls. So this would be a massive improvement just 
from the perspective of performance alone.