diff options
| author | Simon Pilgrim <llvm-dev@redking.me.uk> | 2021-11-17 18:06:49 +0000 |
|---|---|---|
| committer | Simon Pilgrim <llvm-dev@redking.me.uk> | 2021-11-17 18:59:23 +0000 |
| commit | e76032c17301bf929b16a56f8d1dbb366c3c49d8 (patch) | |
| tree | 078839ce4246ead820796c04df34fed003605efc /libcxx/src/algorithm.cpp | |
| parent | 69f35f89691255eeed0dac26b2b642fea5c7db93 (diff) | |
| download | llvm-e76032c17301bf929b16a56f8d1dbb366c3c49d8.zip llvm-e76032c17301bf929b16a56f8d1dbb366c3c49d8.tar.gz llvm-e76032c17301bf929b16a56f8d1dbb366c3c49d8.tar.bz2 | |
[X86] LowerRotate - improve vXi8 rotate-by-scalar lowering with direct use of (extended) shift-by-scalar helpers.
If we're rotating vXi8 by a splatted amount, then unpack to vXi16, perform a SHL by the (extended) scalar, and then pack the results.
This is a vector equivalent to the "rotl(x,y) -> (((aext(x) << bw) | zext(x)) << (y & (bw-1))) >> bw" style expansion we do for scalars in LowerFunnelShift.
I think we can usefully use this for other vector types and vector funnel-shifts in the future, depending how we expand beyond D113192 for matching rotations/funnel-shifts for more type/ops.
Diffstat (limited to 'libcxx/src/algorithm.cpp')
0 files changed, 0 insertions, 0 deletions