https://bugs.freedesktop.org/show_bug.cgi?id=100613
--- Comment #14 from Roland Scheidegger <srol...@vmware.com> ---
(In reply to Ray Strode [halfline] from comment #13)
> Created attachment 131000 [details] [review]
> patch that didn't help at all
>
> Hi,
> (In reply to Roland Scheidegger from comment #10)
> > I'm not sure maybe loading as 3x16 plus vector padding, analogous to what is
> > done for 3x32 would work? The code for doing so should work, it's purely not
> > done this way because codegen on x86 produces a (albeit correct) mess - but
> > it might fare better on ppc...
> So to be clear, I don't really grok the code. I did have a go at what I
> thought you meant, using the quick-and-dirty proof-of-concept patch above
> but it failed in the same way as doing the scalar fetch.
This looks good to me. I actually thought the code would be easily switchable,
but I suppose I nuked that when I noticed there's no point on doing 3x16bit
fetches on x86... (The fetch_dst_type.length adjustment seems unnecessary but
harmless.)
So, if that doesn't help, this at least suggests loading 3x16 plus pad is the
same as loading 1x48 plus zext. This is actually good news as it would be
really confusing otherwise... So it just needs a vector_justify somewhere.
--
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
