ENH: Extend the functionlty of C++ type np::Half
#23298
Conversation
e96f394 to
943b2ef
Compare
| /// If there are no hardware optimization available, rounding will always | ||
| /// be set to ties to even. |
There was a problem hiding this comment.
Doesn't this result in a double round-to-even?
Considering a reduced case of floats with 8, 4, and 2 bit of mantissa;
01011111->01(direct rounding from 8 to 2 bits)01011111->0110->10(rounding via the intermediate 4 bits)
There was a problem hiding this comment.
Agreed, I reverted the change and reinstated the previous behavior, which relies on explicit conversion instead of using single-precision as an argument when hardware-native conversion is available
943b2ef to
c872b25
Compare
c872b25 to
1ca3358
Compare
- optimize float/double conversions on x86, requires for now
raising up the baseline features to `f16c` at least
during the build.
- optimize float/double conversions on ppc64le, requires for now
raising up the baseline features to `VSX3` at least
during the build.
- Brings `np::Half` to npymath
1ca3358 to
ea15a57
Compare
|
close/reopen CI errors seems unrelated |
There was a problem hiding this comment.
Don't want to nitpick grammar here and it is mostly a reorganization (thanks for spotting that subtle rounding issue Eric! I am very sure we have tests to spot such an error luckily.).
Thanks Sayed, lets get this in and unblock the ufunc work!
EDIT: In case anyone wonders, the exported symbols stay identical (except that the private namespace is added), so that is fine.
|
|
||
| /// @name Comparison with no guarantee of NaN behavior | ||
| /// @{ | ||
| constexpr bool Less(Half r) const |
There was a problem hiding this comment.
Seems fair, I wonder whether Less and LessEqual is a typical nomenclature for no-nan versions.
There was a problem hiding this comment.
I think its fine since comparison operators are defined to handles NaNs. using methods Less or LessEqual instead of comparison operators and also instead of std::less or std::less_equal suppose to infuriates any C++ coder for the reason.
TBD
add support for Arm native half-precision operationsrevert this support due to massive fp exceptions triggered by testing unite
test_half.py,optimize float/double conversions on x86, which requires for now raising the baseline features to
f16cat least during the build.optimize float/double conversions on ppc64le, which requires for now raising the baseline features to
VSX3at least during the build.Enable
np::Halffor npymath