forked from OSchip/llvm-project
1669fddc9f
This patch updates LowerMatrixIntrinsics to preserve the alignment specified at the original load/stores and the align attribute for the pointer argument of the column.major.load/store intrinsics. We can always use the specified alignment for the load of the first column. For subsequent columns, the alignment may need to be reduced. For ConstantInt strides, compute the offset for the start of the column in bytes and use commonAlignment to get the largest valid alignment. For non-ConstantInt strides, we need to take the common alignment of the initial alignment and the element size in bytes. Reviewers: anemet, Gerolf, hfinkel, andrew.w.kaylor, LuoYuanke, rjmccall Reviewed By: rjmccall Differential Revision: https://reviews.llvm.org/D81960 |
||
|---|---|---|
| .. | ||
| bigger-expressions-double.ll | ||
| const-gep.ll | ||
| load-align-volatile.ll | ||
| multiply-add-sub-double-row-major.ll | ||
| multiply-double-contraction-fmf.ll | ||
| multiply-double-contraction.ll | ||
| multiply-double-row-major.ll | ||
| multiply-double.ll | ||
| multiply-float-contraction-fmf.ll | ||
| multiply-float-contraction.ll | ||
| multiply-float.ll | ||
| multiply-fused-multiple-blocks.ll | ||
| multiply-fused-volatile.ll | ||
| multiply-fused.ll | ||
| multiply-i32-row-major.ll | ||
| multiply-i32.ll | ||
| propagate-backward.ll | ||
| propagate-backwards-unsupported.ll | ||
| propagate-forward.ll | ||
| propagate-mixed-users.ll | ||
| propagate-multiple-iterations.ll | ||
| remarks-inlining.ll | ||
| remarks-shared-subtrees.ll | ||
| remarks.ll | ||
| store-align-volatile.ll | ||
| strided-load-double.ll | ||
| strided-load-float.ll | ||
| strided-load-i32.ll | ||
| strided-store-double.ll | ||
| strided-store-float.ll | ||
| strided-store-i32.ll | ||
| transpose-double-row-major.ll | ||
| transpose-double.ll | ||
| transpose-float-row-major.ll | ||
| transpose-float.ll | ||
| transpose-i32-row-major.ll | ||
| transpose-i32.ll | ||