Chris Wright wrote:
On Sun, 13 Mar 2016 16:34:44 -0700, Adam Wilson wrote:
Is this a debate about precise vs. non-precise GC or are we just
bikeshedding about terminology and technical details?
You made a large number of assertions about garbage collection and they
were almost all wrong.
It's not about minutiae -- D can't reasonably have a precise collector,
and you were saying that nothing else useful can happen before you make
the GC precise. A non-precise collector can and often should be
concurrent, and you were saying this was useless.
These are big things. They affect prioritization of major projects.
Getting them wrong could mislead a GSoC student to the point of not
accomplishing much when they could have plucked lower-hanging fruit
successfully.
Erm. That seems a bit extreme. You are asserting that D can't have a
precise garbage collector, yet only have presented one case where the
compiler cannot provide definitive information about the classification
of pointer. Unions. Ok, so Unions, a relatively low-usage scenario with
a valid work-around (assume it's a pointer). We can teach the compiler
and GC how to pin C-style pointers so that's not it. Lastly, Rainer
seemed to think a precise GC could be done, and he then went and did it
... so "can't reasonably have a precise collector" is a factually
incorrect assertion.
My GC lingo may be rusty, and I will admit to using superlatives
incorrectly myself, but based on the existing evidence my assertions are
hardly "almost all wrong". I'll clean up my lingo and retool my language
if you'll tone down the superlatives.
Of course GSoC students need guidance, I've been a GSoC Mentor for DLang
before. If I didn't think that a precise GC was not only extremely
important to D, but also within the ability of the student, I would not
have pushed for it.
"low-hanging fruit" is the reason D's quality is so spotty. We now have
a chance with GSoC to get some quality time dedicated to a pain point,
we should utilize that time to work on foundational issues. Then we (the
community) can flame each other to death about the best algorithms to
achieve the performance we desire. That is assuming we haven't scared
him off.
--
// Adam Wilson
// quiet.dlang.dev
