Displaying 3 results from an estimated 3 matches for "splat_lo".
Did you mean:
splat_h
2011 Mar 18
0
[LLVMdev] Long-Term ISel Design
On Mar 17, 2011, at 9:32 AM, David A. Greene wrote:
> Chris Lattner <clattner at apple.com> writes:
>>> 1. We have special target-specific operators for certain shuffles in X86,
>>> such as X86unpckl.
>
>> It also eliminates a lot of fragility. Before doing this, X86
>> legalize would have to be very careful to specifically form shuffles
>> that
2011 Mar 27
2
[LLVMdev] Long-Term ISel Design
...easier. I'm trying to avoid having to write manual code to discover
patterns that isel is already perfectly capable of discovering on its
own.
Take the broadcast case. The .td is something like this (theoretical,
because I did end up creating an X86broadcast node) :
(set VR128:$dst, (v8f32 (splat_lo (v4f32 scalar_to_vector (f32 load addr:$src)),
(undef))))
It's fairly straightforward.
The way things are now, in legalize I have to look for broadcast-like
operations and make sure they come from memory, something like:
if (...we have avx, this is a splat f...
2011 Mar 17
2
[LLVMdev] Long-Term ISel Design
Chris Lattner <clattner at apple.com> writes:
>> 1. We have special target-specific operators for certain shuffles in X86,
>> such as X86unpckl.
> It also eliminates a lot of fragility. Before doing this, X86
> legalize would have to be very careful to specifically form shuffles
> that it knew isel would turn into (e.g.) unpck operations. Now
> instead of