[Date Prev][Date Next][Thread Prev][Thread Next]
- Subject: Re: Avoiding FFI- allocations + using SSE-vectors
- From: Mike Pall <mikelu-1202@...>
- Date: Mon, 6 Feb 2012 14:19:42 +0100
Adam Strzelecki wrote:
> Can I ask you if there would be a limit for vector size?
There is no theoretical limit (other than memory). And I don't
think I've added an implementation limit, anywhere. However, you
can't do much with these vector types right now.
Once full vector support is there, there'll be some practical
limits. That is, if you still want good performance. Really big
value types cause some implementation problems, due to their
specific semantics (one might run out of spill slots etc.).
And vector operation splitting needs to be made smarter. Otherwise
it could stay dumb, as it's only used as a fallback for CPUs that
don't support a particular vector length (the same code runs
everywhere, but it would be slower on some CPUs).
I guess there's a cross-over point where an array of vectors would
be a better choice. The behavior is different, of course. E.g.
you'd have to do the operation splitting yourself. But you can
hide this with a metatype (wrap it in a struct).