[bitc-dev] Bitc and Simd
Jonathan S. Shapiro
shap at eros-os.org
Fri Aug 13 12:02:50 PDT 2010
On Fri, Aug 13, 2010 at 11:08 AM, orthochronous <orthochronous at gmail.com>wrote:
> As I said at the outset, I don't expect anything to be done about
> actually implementing [SIMD] now, but it'd be reassuring to check
> that the foundations being laid aren't precluding this being
> implemented efficiently in the future.
You and Ben are hereby elected.
My problem with SIMD, right now, is in three parts:
- I don't know enough about it to deal with it.
- I'm already at my overflow threshold; I don't have the conceptual
capacity to take this on right now.
- As you pointed out, there isn't any general consensus about how SIMD
compute models work across the candidate processors.
Given this, I think your basic model - that we should handle them for now
with intrinsics - is correct.
I'm also fairly confident that we are in no *worse* a position than anybody
else. Ben's points about alignment and offset specification are well taken;
alignment iswhat the (undocumented) declarations part of the structure
definition mechanism is for, and I know about the need to deal with offsets.
Other than that, what do we *know* we are missing?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the bitc-dev