On Sunday, 20 November 2016 at 09:02:30 UTC, Stefan Koch wrote:
On Saturday, 19 November 2016 at 11:22:18 UTC, Stefan Koch
wrote:
On Saturday, 19 November 2016 at 09:42:33 UTC, Stefan Koch
wrote:
Another small update.
I got rid of the heapClearing overhead.
By using ddmds allocator.
Because of that I was able to shave off 20ms overhead.
Allocation of dynamic arrays works :)
The following code will work just fine with the new engine.
uint[] repeat(uint[] a, uint cnt)
{
uint[] result = [];
uint pos;
result.length = cast(uint) (a.length * cnt);
while(cnt--) foreach(c;a)
{
result[pos++] = c;
}
return result;
}
static assert(repeat([1,2,3,4], 4) ==
[1,2,3,4, 1,2,3,4, 1,2,3,4, 1,2,3,4]
);
There is an important change coming up.
I am going to rewrite the 64 bit support.
The after the change the interpreter will be slower when
working with 64bit values.
However it will be much easier for the jit.
Also I have a lead on howto fix the bugs in break and continue
handling.
A nice weekend to all of you!
I discovered a new bug in the handling of do-while statements.
Which I subsequently fixed.
I also fixed the C backend.
It would interpret a certain conditional different from all other
back-ends.
This makes it oblivious that better documentation of the code-gen
interface is needed.
Furthermore I need to extend my bc_tests. to make sure the
interpretation is the same.
Such that this bug cannot happen again.
As always a happy weekend to all of you.
- I really need a drink ... preferably root-beer.
