This is a grab-bag collection of cleanups, at least one of which have been on the to-do list since day we abandoned compatibility with upstream softfloat.
The compare changes have near-zero effect for float32/float64, since that's the way they are implemented now. For float128, it simply removes unneeded complexity. For floatx80, this fixes bugs vs pseudo-denormals which were only fixed in floatx80_compare_internal. We could just remove them, as these interfaces are not used by x87. But I thought it'd be best to keep the interface compatibility for now. r~ Richard Henderson (10): softfloat: Use post test for floatN_mul softfloat: Replace flag with bool softfloat: Change tininess_before_rounding to bool softfloat: Name rounding mode enum softfloat: Name compare relation enum softfloat: Inline float32 compare specializations softfloat: Inline float64 compare specializations softfloat: Inline float128 compare specializations softfloat: Inline floatx80 compare specializations softfloat: Return bool from all classification predicates include/fpu/softfloat-helpers.h | 27 +- include/fpu/softfloat-macros.h | 24 +- include/fpu/softfloat-types.h | 28 +- include/fpu/softfloat.h | 336 +++++--- target/i386/ops_sse.h | 8 +- fpu/softfloat-specialize.inc.c | 32 +- fpu/softfloat.c | 1305 ++++--------------------------- target/arm/sve_helper.c | 8 +- target/arm/vfp_helper.c | 14 +- target/hppa/op_helper.c | 7 +- target/i386/fpu_helper.c | 8 +- target/m68k/fpu_helper.c | 6 +- target/m68k/softfloat.c | 70 +- target/mips/msa_helper.c | 10 +- target/openrisc/fpu_helper.c | 4 +- target/ppc/int_helper.c | 13 +- target/s390x/fpu_helper.c | 22 +- target/s390x/vec_fpu_helper.c | 2 +- target/sparc/fop_helper.c | 4 +- target/unicore32/ucf64_helper.c | 6 +- target/xtensa/fpu_helper.c | 6 +- tests/fp/fp-test.c | 2 +- 22 files changed, 547 insertions(+), 1395 deletions(-) -- 2.20.1
