glsl/softfp64: Add fisfinite lowering.
authorEmma Anholt <emma@anholt.net>
Thu, 23 Mar 2023 19:14:05 +0000 (12:14 -0700)
committerMarge Bot <emma+marge@anholt.net>
Thu, 6 Apr 2023 02:32:01 +0000 (02:32 +0000)
commit261c1f9ee140e77007dfba2da49aefa6aa0fd3fd
tree172844fc8bf2b8ebf7ebd1883fa9fc39fb37da8f
parent7ff899dd6e830f7dd5e55261585c0a92c91403e0
glsl/softfp64: Add fisfinite lowering.

This is generated by nir_lower_frexp, and if we leave fisfinite in place
then the late algebraic pass lowering it to this pattern will cause an
un-lowered fabs64 to be emitted.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22083>
src/compiler/glsl/float64.glsl
src/compiler/nir/nir_lower_double_ops.c