diff options
author | Kirill Radkin <116365474+kr-sc@users.noreply.github.com> | 2025-05-08 13:29:27 +0300 |
---|---|---|
committer | GitHub <noreply@github.com> | 2025-05-08 18:29:27 +0800 |
commit | b3ef15aa00c94aa937cb40cd7f9483140c62514d (patch) | |
tree | 5bd348cfc164a1f0b87d01b88c018bd18bd3a9e1 /llvm/lib/CodeGen/MachineRegisterInfo.cpp | |
parent | 20169cb4ac328be419033eda26555d2d79898b84 (diff) | |
download | llvm-b3ef15aa00c94aa937cb40cd7f9483140c62514d.zip llvm-b3ef15aa00c94aa937cb40cd7f9483140c62514d.tar.gz llvm-b3ef15aa00c94aa937cb40cd7f9483140c62514d.tar.bz2 |
[RISCV] Fix generation of DWARF info for vector segmented types (#137941)
In DWARF info RISC-V Vector types are presented as DW_TAG_array_type
with tags DW_AT_type (what elements does this array consist of) and
DW_TAG_subrange_type. DW_TAG_subrange_type have DW_AT_upper_bound tag
which contain upper bound value for this array.
For now, it's generate same DWARF info about length of segmented types
and their corresponding non-tuple types.
For example, vint32m4x2_t and vint32m4_t have DW_TAG_array_type with
same DW_AT_type and DW_TAG_subrange_type, it means that this types have
same length, which is not correct
(vint32m4x2_t length is twice as big as vint32m4_t)
Diffstat (limited to 'llvm/lib/CodeGen/MachineRegisterInfo.cpp')
0 files changed, 0 insertions, 0 deletions