On Tuesday, 9 June 2015 at 20:01:24 UTC, Walter Bright wrote:
On 6/9/2015 4:25 AM, "Ola Fosheim =?UTF-8?B?R3LDuHN0YWQi?=
<[email protected]>" wrote:
When constexpr is part of the prototype you know that it isn't
supposed to do
substantial work.
That doesn't follow.
What doesn't follow? All my constexpr functions are simple. I
absolutely don't want to solve N-Queens problems at CT.
Without it, you risk the compiler spending a lot of time
computing
constants/tables that should be done at runtime to save
compile time and object
code size. And since C++ programmers expect that level of
control, it would have
ended up as a pragma if it was not part of the language?
This is just made up problems.
Nah, sometimes I want my big lookup tables to be evaluated at
runtime so I don't get bloated binaries. But neither C++ or D
cuts it. Because sometimes I want the code that uses constant
indexing to look it up in the LUT, but actually generate the LUT
at runtime.
Can D check that a parameter is a CT value so that one can
overload on it? That would work.
