intel/peephole_ffma: Fix swizzle propagation
authorJason Ekstrand <jason.ekstrand@intel.com>
Wed, 17 Oct 2018 16:34:32 +0000 (11:34 -0500)
committerJason Ekstrand <jason.ekstrand@intel.com>
Fri, 11 Jan 2019 16:44:08 +0000 (10:44 -0600)
commit1ede463b6e66eb0a6df5250261810b6985c35eb9
tree5ddf4e5e2cfc134d69816c195d3da2ce0695c376
parentae683ed3bce6ae7d22fcc8e5ef844b1e0f26b93b
intel/peephole_ffma: Fix swizzle propagation

The num_components value passed into get_mul_for_src is used to only
compose the parts of the swizzle that we know will be used so we don't
compose invalid swizzle components.  However, we had a bug where we
passed the number of components of the add all the way through.  For the
given source, we need the number of components read from that source.
In the case where we have a narrow add, say 2 components, that is
sourced from a chain of wider instructions, we may not compose all the
swizzles.  All we really need to do is pass through the right number of
components at each level.

Fixes: 2231cf0ba3a "nir: Fix output swizzle in get_mul_for_src"
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
src/intel/compiler/brw_nir_opt_peephole_ffma.c