diff options
author | Peter Klausler <pklausler@nvidia.com> | 2024-11-14 14:57:19 -0800 |
---|---|---|
committer | GitHub <noreply@github.com> | 2024-11-14 14:57:19 -0800 |
commit | ebc0163cea1cb1ad44f9c438064a52df7e5fc517 (patch) | |
tree | a1d2bba0b3c6894989c3f6c8b19d9ab55232bf7a /llvm/lib/CodeGen | |
parent | 2bc30f37ce7143fd30f21bd232e14aa787f6b08f (diff) | |
download | llvm-ebc0163cea1cb1ad44f9c438064a52df7e5fc517.zip llvm-ebc0163cea1cb1ad44f9c438064a52df7e5fc517.tar.gz llvm-ebc0163cea1cb1ad44f9c438064a52df7e5fc517.tar.bz2 |
[flang] INT2 & INT8 can't be specific intrinsic functions (#115360)
I recently added support for the extension intrinsic functions INT2 and
INT8, and took the shortcut of defining them as specific intrinsic
functions that map to the standard INT() with hard-wired KIND= values
for the result. This works fine for references to these functions, but
leads to a compiler crash for an attempt to use their names in contexts
other than calling them, since their argument types aren't restricted to
single types and no concrete interface can be characterized for them. So
move them out of the table of specific intrinsic functions and into the
general table of intrinsics, and then handle them afterwards as if they
had been INT().
Fixes https://github.com/llvm/llvm-project/issues/115324.
Diffstat (limited to 'llvm/lib/CodeGen')
0 files changed, 0 insertions, 0 deletions