intel/cs_intrinsics: Handle 64-bit intrinsics
authorJason Ekstrand <jason.ekstrand@intel.com>
Thu, 15 Nov 2018 03:57:59 +0000 (21:57 -0600)
committerMarge Bot <eric+marge@anholt.net>
Wed, 12 Aug 2020 10:11:06 +0000 (10:11 +0000)
commit8e1de8e5ac90a9dd0a2fb9310cb36371a9d12dce
treedf9d9f0aa431ec97b920fcaeaad5cdff5bd652c7
parent8d62735b470c0851a567bd2cc0759c18af987923
intel/cs_intrinsics: Handle 64-bit intrinsics

It's safe to do the math in 32 bits because they're all local workgroup
calculations.  We just need to do a conversion at the end.  For a couple
of intrinsics, we just turn them into 32-bit intrinsics and add a u2u64.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>
src/intel/compiler/brw_nir_lower_cs_intrinsics.c