[PATCH 0/5] New double precision exp, exp2, log, log2 and pow functions

[Szabolcs Nagy]

The implementations are based on code from
https://github.com/ARM-software/optimized-routines
and provided here under 3 clause BSD license.

Tested on aarch64 and arm targets.

(I'll reply with git format-patch attached because git
send email does not work for me and thunderbird mangles
things if i inline).

Szabolcs Nagy (5):
Use uint32_t sign argument to math error functions
New exp and exp2 implementations
New log implementation
New log2 implementation
New pow implementation

newlib/libc/include/machine/ieeefp.h | 23 ++
newlib/libm/common/Makefile.am | 4 +-
newlib/libm/common/Makefile.in | 75 ++++-
newlib/libm/common/exp.c | 165 ++++++++++
newlib/libm/common/exp2.c | 152 +++++++++
newlib/libm/common/exp_data.c | 611 +++++++++++++++++++++++++++++++++++
newlib/libm/common/fdlibm.h | 3 +
newlib/libm/common/log.c | 175 ++++++++++
newlib/libm/common/log2.c | 157 +++++++++
newlib/libm/common/log2_data.c | 208 ++++++++++++
newlib/libm/common/log_data.c | 510 +++++++++++++++++++++++++++++
newlib/libm/common/math_config.h | 125 ++++++-
newlib/libm/common/math_err.c | 101 ++++++
newlib/libm/common/math_errf.c | 12 +-
newlib/libm/common/pow.c | 377 +++++++++++++++++++++
newlib/libm/common/pow_log_data.c | 187 +++++++++++
newlib/libm/common/s_log2.c | 2 +
newlib/libm/common/sf_pow.c | 6 +-
newlib/libm/math/e_exp.c | 2 +
newlib/libm/math/e_log.c | 2 +
newlib/libm/math/e_pow.c | 2 +
newlib/libm/math/w_exp.c | 2 +
newlib/libm/math/w_exp2.c | 2 +
newlib/libm/math/w_log.c | 2 +
newlib/libm/math/w_pow.c | 3 +-
25 files changed, 2889 insertions(+), 19 deletions(-)
create mode 100644 newlib/libm/common/exp.c
create mode 100644 newlib/libm/common/exp2.c
create mode 100644 newlib/libm/common/exp_data.c
create mode 100644 newlib/libm/common/log.c
create mode 100644 newlib/libm/common/log2.c
create mode 100644 newlib/libm/common/log2_data.c
create mode 100644 newlib/libm/common/log_data.c
create mode 100644 newlib/libm/common/math_err.c
create mode 100644 newlib/libm/common/pow.c
create mode 100644 newlib/libm/common/pow_log_data.c

--.
2.14.1

[Szabolcs Nagy]

[Szabolcs Nagy]

[Sebastian Huber]

Hello,

+ THIS SOFTWARE IS PROVIDED BY Arm LTD ``AS IS AND ANY EXPRESS OR IMPLIED

is there a typo in this line of the license?
--
Sebastian Huber, embedded brains GmbH

Address : Dornierstr. 4, D-82178 Puchheim, Germany
Phone : [phone]
Fax : [phone]
E-Mail : ***@embedded-brains.de
PGP : Public key available on request.

Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.

[Szabolcs Nagy]

Post by Sebastian Huber
Hello,

+   THIS SOFTWARE IS PROVIDED BY Arm LTD ``AS IS AND ANY EXPRESS OR IMPLIED

is there a typo in this line of the license?

it is not a typo.

i think either using all uppercase name or the official
name of the company are both fine, the BSD license template
has "THE COPYRIGHT HOLDERS AND CONTRIBUTORS" here but you
are supposed to put an actual name there to be clear, it
seems others tend to use all uppercase name, but i doubt
that makes a legal difference and using the official name
is generally better in legal text (this is just a quirk of
the bsd license that it has too many configurable bits and
everybody does it slightly differently).

[Szabolcs Nagy]

[Szabolcs Nagy]

[Szabolcs Nagy]

[Corinna Vinschen]

Post by Szabolcs Nagy
The implementations are based on code from
https://github.com/ARM-software/optimized-routines
and provided here under 3 clause BSD license.
Tested on aarch64 and arm targets.
(I'll reply with git format-patch attached because git
send email does not work for me and thunderbird mangles
things if i inline).
Use uint32_t sign argument to math error functions
New exp and exp2 implementations
New log implementation
New log2 implementation
New pow implementation

Series pushed.


Thanks,
Corinna