The CORE-MATH implementation is correctly rounded (for any rounding mode).
This can be checked by exhaustive tests in a few minutes since there are
less than 2^32 values to check against for example GNU MPFR.
This patch also adds some bench values for tgammaf.
Tested on x86_64 and x86 (cfarm26).
With the initial GNU libc code it gave on an Intel(R) Core(TM) i7-8700:
"tgammaf": {
"": {
"duration": 3.50188e+09,
"iterations": 2e+07,
"max": 602.891,
"min": 65.1415,
"mean": 175.094
}
}
With the new code:
"tgammaf": {
"": {
"duration": 3.30825e+09,
"iterations": 5e+07,
"max": 211.592,
"min": 32.0325,
"mean": 66.1649
}
}
With the initial GNU libc code it gave on cfarm26 (i686):
"tgammaf": {
"": {
"duration": 3.70505e+09,
"iterations": 6e+06,
"max": 2420.23,
"min": 243.154,
"mean": 617.509
}
}
With the new code:
"tgammaf": {
"": {
"duration": 3.24497e+09,
"iterations": 1.8e+07,
"max": 1238.15,
"min": 101.155,
"mean": 180.276
}
}
Signed-off-by: Alexei Sibidanov <sibid@uvic.ca>
Signed-off-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Changes in v2:
- include <math.h> (fix the linknamespace failures)
- restored original benchtests/strcoll-inputs/filelist#en_US.UTF-8 file
- restored original wrapper code (math/w_tgammaf_compat.c),
except for the dealing with the sign
- removed the tgammaf/float entries in all libm-test-ulps files
- address other comments from Joseph Myers
(https://sourceware.org/pipermail/libc-alpha/2024-July/158736.html)
Changes in v3:
- pass NULL argument for signgam from w_tgammaf_compat.c
- use of math_narrow_eval
- added more comments
Changes in v4:
- initialize local_signgam to 0 in math/w_tgamma_template.c
- replace sysdeps/ieee754/dbl-64/gamma_productf.c by dummy file
Changes in v5:
- do not mention local_signgam any more in math/w_tgammaf_compat.c
- initialize local_signgam to 1 instead of 0 in w_tgamma_template.c
and added comment
Changes in v6:
- pass NULL as 2nd argument of __ieee754_gammaf_r in
w_tgammaf_compat.c, and check for NULL in e_gammaf_r.c
Changes in v7:
- added Signed-off-by line for Alexei Sibidanov (author of the code)
Changes in v8:
- added Signed-off-by line for Paul Zimmermann (submitted of the patch)
Changes in v9:
- address comments from review by Adhemerval Zanella
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Left shift of ki is undefined when ki<0, copy the logic from exp,
which uses unsigned arithmetics, to fix it.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Fix BZ #31759 by not defining nearbyint aliases when used in IFUNC.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
Remove the error handling wrapper from exp10. This is very similar to
the changes done to exp and exp2, except that we also need to handle
pow10 and pow10l.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
New implementation is based on the existing exp/exp2, with different
reduction constants and polynomial. Worst-case error in round-to-
nearest is 0.513 ULP.
The exp/exp2 shared table is reused for exp10 - .rodata size of
e_exp_data increases by 64 bytes.
As for exp/exp2, targets with single-instruction rounding/conversion
intrinsics can use them by toggling TOINT_INTRINSICS=1 and adding the
necessary code to their math_private.h.
Improvements on Neoverse V1 compared to current GLIBC master:
exp10 thruput: 3.3x in [-0x1.439b746e36b52p+8 0x1.34413509f79ffp+8]
exp10 latency: 1.8x in [-0x1.439b746e36b52p+8 0x1.34413509f79ffp+8]
Tested on:
aarch64-linux-gnu (TOINT_INTRINSICS, fma contraction) and
x86_64-linux-gnu (!TOINT_INTRINSICS, no fma contraction)
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
With GCC 14 on 32-bit x86 the compiler emits a maybe-uninitialized
warning:
../sysdeps/ieee754/dbl-64/k_rem_pio2.c: In function '__kernel_rem_pio2':
../sysdeps/ieee754/dbl-64/k_rem_pio2.c:364:20: error: 'fq' may be used uninitialized [-Werror=maybe-uninitialized]
364 | y[0] = fq[0]; y[1] = fq[1]; y[2] = fw;
| ~~^~~
This is similar to the warning that is suppressed in the other branch of
the switch. Help the compiler knowing that the variable is always
initialized, which also makes the suppression obsolete.
On Skylake, it changes log1p bench performance by:
Before After Improvement
max 63.349 58.347 8%
min 4.448 5.651 -30%
mean 12.0674 10.336 14%
The minimum code path is
if (hx < 0x3FDA827A) /* x < 0.41422 */
{
if (__glibc_unlikely (ax >= 0x3ff00000)) /* x <= -1.0 */
{
...
}
if (__glibc_unlikely (ax < 0x3e200000)) /* |x| < 2**-29 */
{
math_force_eval (two54 + x); /* raise inexact */
if (ax < 0x3c900000) /* |x| < 2**-54 */
{
...
}
else
return x - x * x * 0.5;
FMA and non-FMA code sequences look similar. Non-FMA version is slightly
faster. Since log1p is called by asinh and atanh, it improves asinh
performance by:
Before After Improvement
max 75.645 63.135 16%
min 10.074 10.071 0%
mean 15.9483 14.9089 6%
and improves atanh performance by:
Before After Improvement
max 91.768 75.081 18%
min 15.548 13.883 10%
mean 18.3713 16.8011 8%
On Skylake, it improves expm1 bench performance by:
Before After Improvement
max 70.204 68.054 3%
min 20.709 16.2 22%
mean 22.1221 16.7367 24%
NB: Add
extern long double __expm1l (long double);
extern long double __expm1f128 (long double);
for __typeof (__expm1l) and __typeof (__expm1f128) when __expm1 is
defined since __expm1 may be expanded in their declarations which
causes the build failure.
Replace the loop-over-scalar placeholder routines with optimised
implementations from Arm Optimized Routines (AOR).
Also add some headers containing utilities for aarch64 libmvec
routines, and update libm-test-ulps.
Data tables for new routines are used via a pointer with a
barrier on it, in order to prevent overly aggressive constant
inlining in GCC. This allows a single adrp, combined with offset
loads, to be used for every constant in the table.
Special-case handlers are marked NOINLINE in order to confine the
save/restore overhead of switching from vector to normal calling
standard. This way we only incur the extra memory access in the
exceptional cases. NOINLINE definitions have been moved to
math_private.h in order to reduce duplication.
AOR exposes a config option, WANT_SIMD_EXCEPT, to enable
selective masking (and later fixing up) of invalid lanes, in
order to trigger fp exceptions correctly (AdvSIMD only). This is
tested and maintained in AOR, however it is configured off at
source level here for performance reasons. We keep the
WANT_SIMD_EXCEPT blocks in routine sources to greatly simplify
the upstreaming process from AOR to glibc.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
Optimize the fast paths (x < y) and (x/y < 2^12). Delay handling of special
cases to reduce the number of instructions executed before the fast paths.
Performance improvements for fmod:
Skylake Zen2 Neoverse V1
subnormals 11.8% 4.2% 11.5%
normal 3.9% 0.01% -0.5%
close-exponents 6.3% 5.6% 19.4%
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The error handling is moved to sysdeps/ieee754 version with no SVID
support. The compatibility symbol versions still use the wrapper
with SVID error handling around the new code. There is no new symbol
version nor compatibility code on !LIBM_SVID_COMPAT targets
(e.g. riscv).
The ia64 is unchanged, since it still uses the arch specific
__libm_error_region on its implementation. For both i686 and m68k,
which provive arch specific implementation, wrappers are added so
no new symbol are added (which would require to change the
implementations).
It shows an small improvement, the results for fmod:
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
x86_64 (Ryzen 9) | subnormals | 12.5049 | 9.40992
x86_64 (Ryzen 9) | normal | 296.939 | 296.738
x86_64 (Ryzen 9) | close-exponents | 16.0244 | 13.119
aarch64 (N1) | subnormal | 6.81778 | 4.33313
aarch64 (N1) | normal | 155.620 | 152.915
aarch64 (N1) | close-exponents | 8.21306 | 5.76138
armhf (N1) | subnormal | 15.1083 | 14.5746
armhf (N1) | normal | 244.833 | 241.738
armhf (N1) | close-exponents | 21.8182 | 22.457
Checked on x86_64-linux-gnu, i686-linux-gnu, and aarch64-linux-gnu.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
This uses a new algorithm similar to already proposed earlier [1].
With x = mx * 2^ex and y = my * 2^ey (mx, my, ex, ey being integers),
the simplest implementation is:
mx * 2^ex == 2 * mx * 2^(ex - 1)
while (ex > ey)
{
mx *= 2;
--ex;
mx %= my;
}
With mx/my being mantissa of double floating pointer, on each step the
argument reduction can be improved 11 (which is sizeo of uint64_t minus
MANTISSA_WIDTH plus the signal bit):
while (ex > ey)
{
mx << 11;
ex -= 11;
mx %= my;
} */
The implementation uses builtin clz and ctz, along with shifts to
convert hx/hy back to doubles. Different than the original patch,
this path assume modulo/divide operation is slow, so use multiplication
with invert values.
I see the following performance improvements using fmod benchtests
(result only show the 'mean' result):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
x86_64 (Ryzen 9) | subnormals | 19.1584 | 12.5049
x86_64 (Ryzen 9) | normal | 1016.51 | 296.939
x86_64 (Ryzen 9) | close-exponents | 18.4428 | 16.0244
aarch64 (N1) | subnormal | 11.153 | 6.81778
aarch64 (N1) | normal | 528.649 | 155.62
aarch64 (N1) | close-exponents | 11.4517 | 8.21306
I also see similar improvements on arm-linux-gnueabihf when running on
the N1 aarch64 chips, where it a lot of soft-fp implementation (for
modulo, clz, ctz, and multiplication):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
armhf (N1) | subnormal | 15.908 | 15.1083
armhf (N1) | normal | 837.525 | 244.833
armhf (N1) | close-exponents | 16.2111 | 21.8182
Instead of using the math_private.h definitions, I used the
math_config.h instead which is used on newer math implementations.
Co-authored-by: kirill <kirill.okhotnikov@gmail.com>
[1] https://sourceware.org/pipermail/libc-alpha/2020-November/119794.html
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
This patch is using the corresponding GCC builtin for logbf, logb,
logbl and logbf128 if the USE_FUNCTION_BUILTIN macros are defined to one
in math-use-builtins-function.h.
Co-Authored-By: Xi Ruoyao <xry111@xry111.site>
This patch is using the corresponding GCC builtin for llrintf, llrint,
llrintl and llrintf128 if the USE_FUNCTION_BUILTIN macros are defined to one
in math-use-builtins-function.h.
Co-Authored-By: Xi Ruoyao <xry111@xry111.site>
This patch is using the corresponding GCC builtin for lrintf, lrint,
lrintl and lrintf128 if the USE_FUNCTION_BUILTIN macros are defined to one
in math-use-builtins-function.h.
Co-Authored-By: Xi Ruoyao <xry111@xry111.site>
This works around a gcc issue where it const folded inf/inf into nan,
preventing the invalid exception to be signalled.
(x-x)/(x-x) is more robust against optimizations and works for all
out of bounds values including x==nan.
The gcc issue https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95115
should be fixed on release branches starting from gcc-10, but it is
better to change the code in case glibc is built with older gcc.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
The compiler may substitute calls to sin or cos with calls to sincos, thus
we should have the same optimized implementations for sincos. The
optimized implementations may produce results that differ, that also makes
sure that the sincos call aggrees with the sin and cos calls.
I used these shell commands:
../glibc/scripts/update-copyrights $PWD/../gnulib/build-aux/update-copyright
(cd ../glibc && git commit -am"[this commit message]")
and then ignored the output, which consisted lines saying "FOO: warning:
copyright statement not found" for each of 7061 files FOO.
I then removed trailing white space from math/tgmath.h,
support/tst-support-open-dev-null-range.c, and
sysdeps/x86_64/multiarch/strlen-vec.S, to work around the following
obscure pre-commit check failure diagnostics from Savannah. I don't
know why I run into these diagnostics whereas others evidently do not.
remote: *** 912-#endif
remote: *** 913:
remote: *** 914-
remote: *** error: lines with trailing whitespace found
...
remote: *** error: sysdeps/unix/sysv/linux/statx_cp.c: trailing lines
The macro TAYLOR_SIN adds the term `-0.5*da*a^2 + da` in hopes
of regaining some precision as a function of da. However the
comment says we add the term `-0.5*da*a^2 + 0.5*da` which is
different. This fix updates the comment to reflect the
code and also simplifies the calculation by replacing `a` with `x`
because they always have the same value.
Signed-off-by: Akila Welihinda <akilawelihinda@ucla.edu>
Reviewed-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
The error handling is moved to sysdeps/ieee754 version with no SVID
support. The compatibility symbol versions still use the wrapper with
SVID error handling around the new code. There is no new symbol version
nor compatibility code on !LIBM_SVID_COMPAT targets (e.g. riscv).
Only ia64 is unchanged, since it still uses the arch specific
__libm_error_region on its implementation.
Checked on x86_64-linux-gnu, i686-linux-gnu, and aarch64-linux-gnu.
Improve hypot performance significantly by using fma when available. The
fma version has twice the throughput of the previous version and 70% of
the latency. The non-fma version has 30% higher throughput and 10%
higher latency.
Max ULP error is 0.949 with fma and 0.792 without fma.
Passes GLIBC testsuite.
This implementation is based on the 'An Improved Algorithm for
hypot(a,b)' by Carlos F. Borges [1] using the MyHypot3 with the
following changes:
- Handle qNaN and sNaN.
- Tune the 'widely varying operands' to avoid spurious underflow
due the multiplication and fix the return value for upwards
rounding mode.
- Handle required underflow exception for denormal results.
The main advantage of the new algorithm is its precision: with a
random 1e9 input pairs in the range of [DBL_MIN, DBL_MAX], glibc
current implementation shows around 0.34% results with an error of
1 ulp (3424869 results) while the new implementation only shows
0.002% of total (18851).
The performance result are also only slight worse than current
implementation. On x86_64 (Ryzen 5900X) with gcc 12:
Before:
"hypot": {
"workload-random": {
"duration": 3.73319e+09,
"iterations": 1.12e+08,
"reciprocal-throughput": 22.8737,
"latency": 43.7904,
"max-throughput": 4.37184e+07,
"min-throughput": 2.28361e+07
}
}
After:
"hypot": {
"workload-random": {
"duration": 3.7597e+09,
"iterations": 9.8e+07,
"reciprocal-throughput": 23.7547,
"latency": 52.9739,
"max-throughput": 4.2097e+07,
"min-throughput": 1.88772e+07
}
}
Co-Authored-By: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Checked on x86_64-linux-gnu and aarch64-linux-gnu.
[1] https://arxiv.org/pdf/1904.09481.pdf
This patch adds the narrowing fused multiply-add functions from TS
18661-1 / TS 18661-3 / C2X to glibc's libm: ffma, ffmal, dfmal,
f32fmaf64, f32fmaf32x, f32xfmaf64 for all configurations; f32fmaf64x,
f32fmaf128, f64fmaf64x, f64fmaf128, f32xfmaf64x, f32xfmaf128,
f64xfmaf128 for configurations with _Float64x and _Float128;
__f32fmaieee128 and __f64fmaieee128 aliases in the powerpc64le case
(for calls to ffmal and dfmal when long double is IEEE binary128).
Corresponding tgmath.h macro support is also added.
The changes are mostly similar to those for the other narrowing
functions previously added, especially that for sqrt, so the
description of those generally applies to this patch as well. As with
sqrt, I reused the same test inputs in auto-libm-test-in as for
non-narrowing fma rather than adding extra or separate inputs for
narrowing fma. The tests in libm-test-narrow-fma.inc also follow
those for non-narrowing fma.
The non-narrowing fma has a known bug (bug 6801) that it does not set
errno on errors (overflow, underflow, Inf * 0, Inf - Inf). Rather
than fixing this or having narrowing fma check for errors when
non-narrowing does not (complicating the cases when narrowing fma can
otherwise be an alias for a non-narrowing function), this patch does
not attempt to check for errors from narrowing fma and set errno; the
CHECK_NARROW_FMA macro is still present, but as a placeholder that
does nothing, and this missing errno setting is considered to be
covered by the existing bug rather than needing a separate open bug.
missing-errno annotations are duly added to many of the
auto-libm-test-in test inputs for fma.
This completes adding all the new functions from TS 18661-1 to glibc,
so will be followed by corresponding stdc-predef.h changes to define
__STDC_IEC_60559_BFP__ and __STDC_IEC_60559_COMPLEX__, as the support
for TS 18661-1 will be at a similar level to that for C standard
floating-point facilities up to C11 (pragmas not implemented, but
library functions done). (There are still further changes to be done
to implement changes to the types of fromfp functions from N2548.)
Tested as followed: natively with the full glibc testsuite for x86_64
(GCC 11, 7, 6) and x86 (GCC 11); with build-many-glibcs.py with GCC
11, 7 and 6; cross testing of math/ tests for powerpc64le, powerpc32
hard float, mips64 (all three ABIs, both hard and soft float). The
different GCC versions are to cover the different cases in tgmath.h
and tgmath.h tests properly (GCC 6 has _Float* only as typedefs in
glibc headers, GCC 7 has proper _Float* support, GCC 8 adds
__builtin_tgmath).
As described in bug 28358, the round-to-odd computations used in the
libm functions that round their results to a narrower format can yield
spurious underflow exceptions in the following circumstances: the
narrowing only narrows the precision of the type and not the exponent
range (i.e., it's narrowing _Float128 to _Float64x on x86_64, x86 or
ia64), the architecture does after-rounding tininess detection (which
applies to all those architectures), the result is inexact, tiny
before rounding but not tiny after rounding (with the chosen rounding
mode) for _Float64x (which is possible for narrowing mul, div and fma,
not for narrowing add, sub or sqrt), so the underflow exception
resulting from the toward-zero computation in _Float128 is spurious
for _Float64x.
Fixed by making ROUND_TO_ODD call feclearexcept (FE_UNDERFLOW) in the
problem cases (as indicated by an extra argument to the macro); there
is never any need to preserve underflow exceptions from this part of
the computation, because the conversion of the round-to-odd value to
the narrower type will underflow in exactly the cases in which the
function should raise that exception, but it may be more efficient to
avoid the extra manipulation of the floating-point environment when
not needed.
Tested for x86_64 and x86, and with build-many-glibcs.py.
include/math.h has a mechanism to redirect internal calls to various
libm functions, that can often be inlined by the compiler, to call
non-exported __* names for those functions in the case when the calls
aren't inlined, with the redirection being disabled when
NO_MATH_REDIRECT. Add fma to the functions to which this mechanism is
applied.
At present, libm-internal fma calls (generally to __builtin_fma*
functions) are only done when it's known the call will be inlined,
with alternative code not relying on an fma operation being used in
the caller otherwise. This patch is in preparation for adding the TS
18661 / C2X narrowing fma functions to glibc; it will be natural for
the narrowing function implementations to call the underlying fma
functions unconditionally, with this either being inlined or resulting
in an __fma* call. (Using two levels of round-to-odd computation like
that, in the case where there isn't an fma hardware instruction, isn't
optimal but is certainly a lot simpler for the initial implementation
than writing different narrowing fma implementations for all the
various pairs of formats.)
Tested with build-many-glibcs.py that installed stripped shared
libraries are unchanged by the patch (using
<https://sourceware.org/pipermail/libc-alpha/2021-September/130991.html>
to fix installed library stripping in build-many-glibcs.py). Also
tested for x86_64.
This patch adds the narrowing square root functions from TS 18661-1 /
TS 18661-3 / C2X to glibc's libm: fsqrt, fsqrtl, dsqrtl, f32sqrtf64,
f32sqrtf32x, f32xsqrtf64 for all configurations; f32sqrtf64x,
f32sqrtf128, f64sqrtf64x, f64sqrtf128, f32xsqrtf64x, f32xsqrtf128,
f64xsqrtf128 for configurations with _Float64x and _Float128;
__f32sqrtieee128 and __f64sqrtieee128 aliases in the powerpc64le case
(for calls to fsqrtl and dsqrtl when long double is IEEE binary128).
Corresponding tgmath.h macro support is also added.
The changes are mostly similar to those for the other narrowing
functions previously added, so the description of those generally
applies to this patch as well. However, the not-actually-narrowing
cases (where the two types involved in the function have the same
floating-point format) are aliased to sqrt, sqrtl or sqrtf128 rather
than needing a separately built not-actually-narrowing function such
as was needed for add / sub / mul / div. Thus, there is no
__nldbl_dsqrtl name for ldbl-opt because no such name was needed
(whereas the other functions needed such a name since the only other
name for that entry point was e.g. f32xaddf64, not reserved by TS
18661-1); the headers are made to arrange for sqrt to be called in
that case instead.
The DIAG_* calls in sysdeps/ieee754/soft-fp/s_dsqrtl.c are because
they were observed to be needed in GCC 7 testing of
riscv32-linux-gnu-rv32imac-ilp32. The other sysdeps/ieee754/soft-fp/
files added didn't need such DIAG_* in any configuration I tested with
build-many-glibcs.py, but if they do turn out to be needed in more
files with some other configuration / GCC version, they can always be
added there.
I reused the same test inputs in auto-libm-test-in as for
non-narrowing sqrt rather than adding extra or separate inputs for
narrowing sqrt. The tests in libm-test-narrow-sqrt.inc also follow
those for non-narrowing sqrt.
Tested as followed: natively with the full glibc testsuite for x86_64
(GCC 11, 7, 6) and x86 (GCC 11); with build-many-glibcs.py with GCC
11, 7 and 6; cross testing of math/ tests for powerpc64le, powerpc32
hard float, mips64 (all three ABIs, both hard and soft float). The
different GCC versions are to cover the different cases in tgmath.h
and tgmath.h tests properly (GCC 6 has _Float* only as typedefs in
glibc headers, GCC 7 has proper _Float* support, GCC 8 adds
__builtin_tgmath).
We stopped adding "Contributed by" or similar lines in sources in 2012
in favour of git logs and keeping the Contributors section of the
glibc manual up to date. Removing these lines makes the license
header a bit more consistent across files and also removes the
possibility of error in attribution when license blocks or files are
copied across since the contributed-by lines don't actually reflect
reality in those cases.
Move all "Contributed by" and similar lines (Written by, Test by,
etc.) into a new file CONTRIBUTED-BY to retain record of these
contributions. These contributors are also mentioned in
manual/contrib.texi, so we just maintain this additional record as a
courtesy to the earlier developers.
The following scripts were used to filter a list of files to edit in
place and to clean up the CONTRIBUTED-BY file respectively. These
were not added to the glibc sources because they're not expected to be
of any use in future given that this is a one time task:
https://gist.github.com/siddhesh/b5ecac94eabfd72ed2916d6d8157e7dchttps://gist.github.com/siddhesh/15ea1f5e435ace9774f485030695ee02
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
This patch is using the corresponding GCC builtin for roundevenf,
roundeven and roundevenl if the USE_FUNCTION_BUILTIN macros are defined
to one in math-use-builtins.h.
These builtin functions is supported since GCC 10.
The code of the generic implementation is not changed.
Signed-off-by: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
This patch redirect roundeven function for futhermore changes.
Signed-off-by: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
With this patch, the maximal known error for tgamma is now reduced to 9 ulps
for dbl-64, for all rounding modes. Since exhaustive testing is not possible
for dbl-64, it might be that there are still cases with an error larger than
9 ulps, but all known cases are fixed (intensive tests were done to find cases
with large errors).
Tested on x86_64 and powerpc (and by Adhemerval Zanella on aarch64, arm,
s390x, sparc, and i686).
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Previous commit was missing deleted files in sysdeps/ieee754/dbl-64.
Finally remove all mpa related files, headers, declarations, probes, unused
tables and update makefiles.
Reviewed-By: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Finally remove all mpa related files, headers, declarations, probes, unused
tables and update makefiles.
Reviewed-By: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Remove slow paths in tan. Add ULP annotations. Merge 'number' into 'mynumber'.
Remove unused entries from tan constants.
Reviewed-By: Paul Zimmermann <Paul.Zimmermann@inria.fr>
This patch series removes all remaining slow paths and related code.
First asin/acos, tan, atan, atan2 implementations are updated, and the final
patch removes the unused mpa files, headers and probes. Passes buildmanyglibc.
Remove slow paths from asin/acos. Add ULP annotations based on previous slow
path checks (which are approximate). Update AArch64 and x86_64 libm-test-ulps.
Reviewed-By: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Remove the wordsize-64 implementations by merging them into the main dbl-64
directory. The second patch just moves all wordsize-64 files and removes a
few wordsize-64 uses in comments and Implies files.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Remove the wordsize-64 implementations by merging them into the main dbl-64
directory. The first patch adds special cases needed for 32-bit targets
(FIX_INT_FP_CONVERT_ZERO and FIX_DBL_LONG_CONVERT_OVERFLOW) to the
wordsize-64 versions. This has no effect on 64-bit targets since they don't
define these macros.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
I used these shell commands:
../glibc/scripts/update-copyrights $PWD/../gnulib/build-aux/update-copyright
(cd ../glibc && git commit -am"[this commit message]")
and then ignored the output, which consisted lines saying "FOO: warning:
copyright statement not found" for each of 6694 files FOO.
I then removed trailing white space from benchtests/bench-pthread-locks.c
and iconvdata/tst-iconv-big5-hkscs-to-2ucs4.c, to work around this
diagnostic from Savannah:
remote: *** pre-commit check failed ...
remote: *** error: lines with trailing whitespace found
remote: error: hook declined to update refs/heads/master
asin and acos have slow paths for rounding the last bit that cause some
calls to be 500-1500x slower than average calls.
These slow paths are rare, a test of a trillion (1.000.000.000.000)
random inputs between -1 and 1 showed 32870 slow calls for acos and 4473
for asin, with most occurrences between -1.0 .. -0.9 and 0.9 .. 1.0.
The slow paths claim correct rounding and use __sin32() and __cos32()
(which compare two result candidates and return the closest one) as the
final step, with the second result candidate (res1) having a small offset
applied from res. This suggests that res and res1 are intended to be 1
ULP apart (which makes sense for rounding), barring bugs, allowing us to
pick either one and still remain within 1 ULP of the exact result.
Remove the slow paths as the accuracy is better than 1 ULP even without
them, which is enough for glibc.
Also remove code comments claiming correctly rounded results.
After slow path removal, checking the accuracy of 14.400.000.000 random
asin() and acos() inputs showed only three incorrectly rounded
(error > 0.5 ULP) results:
- asin(-0x1.ee2b43286db75p-1) (0.500002 ULP, same as before)
- asin(-0x1.f692ba202abcp-4) (0.500003 ULP, same as before)
- asin(-0x1.9915e876fc062p-1) (0.50000000001 ULP, previously exact)
The first two had the same error even before this commit, and they did
not use the slow path at all.
Checking 4934 known randomly found previously-slow-path asin inputs
shows 25 calls with incorrectly rounded results, with a maximum error of
0.500000002 ULP (for 0x1.fcd5742999ab8p-1). The previous slow-path code
rounded all these inputs correctly (error < 0.5 ULP).
The observed average speed increase was 130x.
Checking 36240 known randomly found previously-slow-path acos inputs
shows 42 calls with incorrectly rounded results, with a maximum error of
0.500000008 ULP (for 0x1.f63845056f35ep-1). The previous "exact"
slow-path code showed 34 calls with incorrectly rounded results, with the
same maximum error of 0.500000008 ULP (for 0x1.f63845056f35ep-1).
The observed average speed increase was 130x.
The functions could likely be trimmed more while keeping acceptable
accuracy, but this at least gets rid of the egregiously slow cases.
Tested on x86_64.
In TS 18661-1, getpayload had an unspecified return value for a
non-NaN argument, while C2x requires the return value -1 in that case.
This patch implements the return value of -1. I don't think this is
worth having a new symbol version that's an alias of the old one,
although occasionally we do that in such cases where the new function
semantics are a refinement of the old ones (to avoid programs relying
on the new semantics running on older glibc versions but not behaving
as intended).
Tested for x86_64 and x86; also ran math/ tests for aarch64 and
powerpc.
This came to light when adding hard-flaot support to ARC glibc port
without hardware sqrt support causing glibc build to fail:
| ../sysdeps/ieee754/dbl-64/e_sqrt.c: In function '__ieee754_sqrt':
| ../sysdeps/ieee754/dbl-64/e_sqrt.c:58:54: error: unused variable 'ty' [-Werror=unused-variable]
| double y, t, del, res, res1, hy, z, zz, p, hx, tx, ty, s;
The reason being EMULV() macro uses the hardware provided
__builtin_fma() variant, leaving temporary variables 'p, hx, tx, hy, ty'
unused hence compiler warning and ensuing error.
The intent of the patch was to fix that error, but EMULV is pervasive
and used fair bit indirectly via othe rmacros, hence this patch.
Functionally it should not result in code gen changes and if at all
those would be better since the scope of those temporaries is greatly
reduced now
Built tested with aarch64-linux-gnu arm-linux-gnueabi arm-linux-gnueabihf hppa-linux-gnu x86_64-linux-gnu arm-linux-gnueabihf riscv64-linux-gnu-rv64imac-lp64 riscv64-linux-gnu-rv64imafdc-lp64 powerpc-linux-gnu microblaze-linux-gnu nios2-linux-gnu hppa-linux-gnu
Also as suggested by Joseph [1] used --strip and compared the libs with
and w/o patch and they are byte-for-byte unchanged (with gcc 9).
| for i in `find . -name libm-2.31.9000.so`;
| do
| echo $i; diff $i /SCRATCH/vgupta/gnu2/install/glibcs/$i ; echo $?;
| done
| ./aarch64-linux-gnu/lib64/libm-2.31.9000.so
| 0
| ./arm-linux-gnueabi/lib/libm-2.31.9000.so
| 0
| ./x86_64-linux-gnu/lib64/libm-2.31.9000.so
| 0
| ./arm-linux-gnueabihf/lib/libm-2.31.9000.so
| 0
| ./riscv64-linux-gnu-rv64imac-lp64/lib64/lp64/libm-2.31.9000.so
| 0
| ./riscv64-linux-gnu-rv64imafdc-lp64/lib64/lp64/libm-2.31.9000.so
| 0
| ./powerpc-linux-gnu/lib/libm-2.31.9000.so
| 0
| ./microblaze-linux-gnu/lib/libm-2.31.9000.so
| 0
| ./nios2-linux-gnu/lib/libm-2.31.9000.so
| 0
| ./hppa-linux-gnu/lib/libm-2.31.9000.so
| 0
| ./s390x-linux-gnu/lib64/libm-2.31.9000.so
[1] https://sourceware.org/pipermail/libc-alpha/2019-November/108267.html
