November 19

Hi,

This week I managed to fix the final bugs of the _d_arrayctor lowering.
In order to remove the call to __ArrayDtor, of which I spoke in last week's post, I resorted to declaring the array returned by _d_arrayctor inside a union. This is somewhat unorthodox, but Adam Ruppe assured me that this is the accepted way of handling array destruction myself.

As a result, this milestone I plan to run a benchmark for testing the performance increase when using the new lowering for _d_array{,set}ctor. I am also looking to finish replacing the lowerings to _d_arrayappendT and _d_arrayappendcTX with templates. I started working on this during the previous milestone and managed fixed a bug where CTFE did not correctly interpret the lowering to _d_arrayappendcTX.

After this, if time allows it, I'll move away for a while from lowerings involving array manipulation to hooks that manipulate somewhat simpler objects, such as _d_delstruct, _d_newitem{T,iT,U} or _d_newThrowable. I plan to make this shift because I've seen array-based generate some rather time-consuming bugs and I hope that by doing so, I'll be able to get more work finished and get stuck less.

Thanks,
Teodor

5 days ago

On Friday, 19 November 2021 at 21:45:46 UTC, Teodor Dutu wrote:

>

Hi,

This week I managed to fix the final bugs of the _d_arrayctor lowering.
In order to remove the call to __ArrayDtor, of which I spoke in last week's post, I resorted to declaring the array returned by _d_arrayctor inside a union. This is somewhat unorthodox, but Adam Ruppe assured me that this is the accepted way of handling array destruction myself.

[...]

👍