This series adds new codegen tests for converting any of our seven
floating point modes to any of those seven. It also fixes a bunch of
bugs so that these testcases pass.
Tested on powerpc64-linux {-m32,-m64} (a power7); on powerpc64le-linux,
both on power8 and on power9; and on AIX.
Also tested the new testcases with options
{-mlong-double-64,-mlong-double-128}
{-mabi=ibmlongdouble,-mabi=ieeelongdouble}
{-mcpu=power4,-mcpu=970,-mcpu=power6,-mcpu=power7,-mcpu=power9}
{-mabi=elfv1/-mbig,-mabi=elfv2/-mlittle}
{-m32,-m64}
(but not -mabi=ieeelongdouble before power7 because that is not
supported).
Committing to trunk.
Segher
Segher Boessenkool (6):
rs6000: Use more correct names for some trunc/extend libcalls
rs6000: Use correct names for some trunc/extend libcalls
rs6000: Improve truncifsf2
rs6000: Fix testsuite bug in check_ppc_float128_hw_available
rs6000: New testsuite selectors
rs6000: New testcase fp-convert.c
gcc/config/rs6000/rs6000.c | 24 +++---
gcc/config/rs6000/rs6000.md | 23 ++----
gcc/testsuite/gcc.target/powerpc/convert-fp-128.c | 99 +++++++++++++++++++++++
gcc/testsuite/gcc.target/powerpc/convert-fp-64.c | 61 ++++++++++++++
gcc/testsuite/lib/target-supports.exp | 41 +++++++++-
5 files changed, 217 insertions(+), 31 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/powerpc/convert-fp-128.c
create mode 100644 gcc/testsuite/gcc.target/powerpc/convert-fp-64.c
--
1.8.3.1
