forked from OSchip/llvm-project
7397dcb403
This patch adds an initial ShuffleVectorInst::isInsertSubvectorMask helper to recognize 2-op shuffles where the lowest elements of one of the sources are being inserted into the "in-place" other operand, this includes "concat_vectors" patterns as can be seen in the Arm shuffle cost changes. This also helped fix a x86 issue with irregular/length-changing SK_InsertSubvector costs - I'm hoping this will help with D107188 This doesn't currently attempt to work with 1-op shuffles that could either be a "widening" shuffle or a self-insertion. The self-insertion case is tricky, but we currently always match this with the existing SK_PermuteSingleSrc logic. The widening case will be addressed in a follow up patch that treats the cost as 0. Masks with a high number of undef elts will still struggle to match optimal subvector widths - its currently bounded by minimum-width possible insertion, whilst some cases would benefit from wider (pow2?) subvectors. Differential Revision: https://reviews.llvm.org/D107228 |
||
|---|---|---|
| .. | ||
| add-cast-vect.ll | ||
| arith-overflow.ll | ||
| arith-ssat.ll | ||
| arith-usat.ll | ||
| arith.ll | ||
| cast.ll | ||
| cast_ldst.ll | ||
| cmps.ll | ||
| control-flow.ll | ||
| divrem.ll | ||
| fparith.ll | ||
| freeshift.ll | ||
| gep.ll | ||
| immediates.ll | ||
| insertelement.ll | ||
| intrinsic-cost-kinds.ll | ||
| lit.local.cfg | ||
| load_store.ll | ||
| logicalop.ll | ||
| memcpy.ll | ||
| mul-cast-vect.ll | ||
| mve-abs.ll | ||
| mve-active_lane_mask.ll | ||
| mve-cmp.ll | ||
| mve-gather-scatter-cost.ll | ||
| mve-minmax.ll | ||
| mve-vecreduce-add.ll | ||
| reduce-add.ll | ||
| reduce-and.ll | ||
| reduce-or.ll | ||
| reduce-smax.ll | ||
| reduce-smin.ll | ||
| reduce-umax.ll | ||
| reduce-umin.ll | ||
| select.ll | ||
| shl-cast-vect.ll | ||
| shuffle.ll | ||
| sub-cast-vect.ll | ||
| target-intrinsics.ll | ||