mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-23 04:52:54 +02:00
6d1fcfb76b
There are no vldN/vstN f16 variants, even with +fullfp16. We could use the i16 variants, but, in practice, even with +fullfp16, the f16 sequence leading to the i16 shuffle usually gets scalarized. We'd need to improve our support for f16 codegen before getting there. Teach the cost model to consider f16 interleaved operations as expensive. Otherwise, we are all but guaranteed to end up with a large block of scalarized vector code. llvm-svn: 294819 |
||
---|---|---|
.. | ||
arm-ieee-vectorize.ll | ||
arm-unroll.ll | ||
gather-cost.ll | ||
gcc-examples.ll | ||
interleaved_cost.ll | ||
lit.local.cfg | ||
mul-cast-vect.ll | ||
vector_cast.ll | ||
width-detect.ll |