On Tuesday, 1 August 2017 at 21:27:32 UTC, Stefan Koch wrote:
[ ... ]
Hi Guys,
I've fixed a few ABI bugs and as a result my alternative to
std.bitmanip.bitfields complies now.
I've also made an intrinsic for the concat operation.
Which causes ~= to be 6-10x faster when it's heavily used.
The huge space requirements for heavy ~= are not yet solved
though.
To pass the unittests regardless which does ~= 4096 times on a
moderately long string;
I've had to bump the ctfe heapSpace.
Therefore newCTFE will allocate 1GB at startup which causes it to
die on 32bit targets.
This issue will be worked when _most_ of the ABI is finalized
(i.e. when classes and unions are in)