projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 942c994) | patch
[SystemZ] Fix 128-bit strict FMA expansion pre-z14
authorUlrich Weigand <ulrich.weigand@de.ibm.com>
Wed, 11 Dec 2019 15:28:14 +0000 (16:28 +0100)
committerUlrich Weigand <ulrich.weigand@de.ibm.com>
Wed, 11 Dec 2019 15:32:08 +0000 (16:32 +0100)
commitac473394ff04ac3e33c15b8358b68a78834b8424
treefc97a4a9f0f5bd6d210c933839568fee6cb49cfbtree | snapshot
parent942c9946cc2fba6acd3805cfeaca90007f532f1fcommit | diff
[SystemZ] Fix 128-bit strict FMA expansion pre-z14

Before z14, we did not have any FMA instruction for 128-bit
floating-point, so the @llvm.fma.f128 intrinsic needs to be
expanded to a libcall on those platforms.

This worked correctly for regular FMA, but was implemented
incorrectly for the strict version.  This was not noticed
because we did not have test coverage for this case.

This patch fixes that incorrect expansion and adds the
missing test cases.
llvm/lib/Target/SystemZ/SystemZISelLowering.cpp diff | blob | history
llvm/test/CodeGen/SystemZ/fp-mul-13.ll [new file with mode: 0644] blob
llvm/test/CodeGen/SystemZ/fp-strict-mul-12.ll [new file with mode: 0644] blob
llvm/test/CodeGen/SystemZ/fp-strict-mul-13.ll [new file with mode: 0644] blob
Domain: System / Toolchain;