How can I write compile-time (pure) BigInt computations?

[Gor Gyolchanyan]

there's the __ctfe check, that could be used to write an alternative
implementation of BigInt (which would be arbitrarily slow, which won't
matter) without assembly.

On Tue, Oct 25, 2011 at 4:53 PM, Timon Gehr <timon.gehr at gmx.ch> wrote:
> On 10/25/2011 02:37 PM, Gor Gyolchanyan wrote:
>>
>> That's because N is not a compile-time value. Initializing global
>> variable is allowed only with compile-time values, but the variable
>> themselves are run-time values.
>> This could probably be fixed, since the initial values of global
>> variables are always known.
>>
>> Anyway, if the value of N (and probable X too) is not subject to
>> modification, you should make then either immutable or, better yet,
>> enum.
>> This way they WILL be compile-time values and you won't get that error.
>>
>> On Tue, Oct 25, 2011 at 4:31 PM, Chris Dew<cmsdew at gmail.com>  wrote:
>>>
>>> Stuck again, looks like BigInt doesn't work with CTFE for other
>>> reasons...
>>>
>>> Is there a way to create BigInts at compile time (I've also tried from
>>> a string)?  Or is it back to C code generation in Python :-(
>>>
>>> Thanks anyway,
>>>
>>> Chris.
>>>
>>>
>>> ./gcd.d(11): Error: static variable N cannot be read at compile time
>>> ./gcd.d(11): Error: cannot evaluate gcd((BigInt __ctmp1335;
>>>  , __ctmp1335).this(48),N) at compile time
>>> Failed: dmd  -v -o- './gcd.d' -I'.'>./gcd.d.deps
>>>
>>>
>>>
>>> import std.stdio;
>>> import std.bigint;
>>>
>>> T gcd(T)(T a, T b) {
>>>  if (b == 0) return a;
>>>  return gcd(b, a % b);
>>> }
>>>
>>>
>>> BigInt N = cast(BigInt)1000000000L;
>>> BigInt X = gcd(cast(BigInt)48, N);
>>>
>>> int main() {
>>>  writefln("%s", X);
>>>  return 0;
>>> }
>>>
>
> BigInt operates using hand-written assembly, if available for the target
> architecture. inline asm cannot be used in CTFE. CTFE for BigInt would
> specifically have to be supported by a patch to Phobos.
>