Plus assorted cleanups, passes tests/fp/fp-test.
Changes in v2:
* Add UInt256 type (david)
* Rewrite and inline shift256RightJamming. This keeps the whole
UInt256 in registers, avoiding long sequences of loads and stores.
* Add x86_64 assembly for double shifts. I don't know why the
compiler can't recognize this pattern, but swapping values in
and out of %cl (the only register in the base isa that can
hold a variable shift) is really ugly.
* Add ppc64 assembly.
r~
Richard Henderson (10):
softfloat: Use mulu64 for mul64To128
softfloat: Use int128.h for some operations
softfloat: Tidy a * b + inf return
softfloat: Add float_cmask and constants
softfloat: Inline pick_nan_muladd into its caller
softfloat: Implement float128_muladd
softfloat: Use x86_64 assembly for {add,sub}{192,256}
softfloat: Use x86_64 assembly for sh[rl]_double
softfloat: Use aarch64 assembly for {add,sub}{192,256}
softfloat: Use ppc64 assembly for {add,sub}{192,256}
include/fpu/softfloat-macros.h | 109 +++---
include/fpu/softfloat.h | 2 +
fpu/softfloat.c | 620 ++++++++++++++++++++++++++++++---
tests/fp/fp-test.c | 2 +-
tests/fp/wrap.c.inc | 12 +
5 files changed, 652 insertions(+), 93 deletions(-)
--
2.25.1
