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.
Bump autoconf requirement to 2.71 to allow regenerating configure on
more recent distributions. autoconf 2.71 has been in Fedora since F36
and is the current version in Debian stable (bookworm). It appears to
be current in Gentoo as well.
All sysdeps configure and preconfigure scripts have also been
regenerated; all changes are trivial transformations that do not affect
functionality.
Signed-off-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Return value from *scanf and *asprintf routines are now properly checked
in test-scanf-ldbl-compat-template.c and test-printf-ldbl-compat.c.
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
This allows to include bits/syslog-decl.h in include/sys/syslog.h and
therefore be able to create the libc_hidden_builtin_proto (__syslog_chk)
prototype.
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
If libc_hidden_builtin_{def,proto} isn't properly set for *_chk routines,
there are unwanted PLT entries in libc.so.
There is a special case with __asprintf_chk:
If ldbl_* macros are used for asprintf, ABI gets broken on s390x,
if it isn't, ppc64le isn't building due to multiple asm redirections.
This is due to the inclusion of bits/stdio-lbdl.h for ppc64le whereas it
isn't for s390x. This header creates redirections, which are not
compatible with the ones generated using libc_hidden_def.
Yet, we can't use libc_hidden_ldbl_proto on s390x since it will not
create a simple strong alias (e.g. as done on x86_64), but a versioned
alias, leading to ABI breakage.
This results in errors on s390x:
/usr/bin/ld: glibc/iconv/../libio/bits/stdio2.h:137: undefined reference
to `__asprintf_chk'
Original __asprintf_chk symbols:
00000000001395b0 T __asprintf_chk
0000000000177e90 T __nldbl___asprintf_chk
__asprintf_chk symbols with ldbl_* macros:
000000000012d590 t ___asprintf_chk
000000000012d590 t __asprintf_chk@@GLIBC_2.4
000000000012d590 t __GI___asprintf_chk
000000000012d590 t __GL____asprintf_chk___asprintf_chk
0000000000172240 T __nldbl___asprintf_chk
__asprintf_chk symbols with the patch:
000000000012d590 t ___asprintf_chk
000000000012d590 T __asprintf_chk
000000000012d590 t __GI___asprintf_chk
0000000000172240 T __nldbl___asprintf_chk
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
The *_chk routines naming doesn't match the name that would be generated
using libc_hidden_ldbl_proto. Since the macro is needed for some of
these *_chk functions for _FORTIFY_SOURCE to be enabled, that needed to
be fixed.
While at it, all the *_chk function get renamed appropriately for
consistency, even if not strictly necessary.
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
Reviewed-by: Paul E. Murphy <murphyp@linux.ibm.com>
Since the _FORTIFY_SOURCE feature uses some routines of Glibc, they need to
be excluded from the fortification.
On top of that:
- some tests explicitly verify that some level of fortification works
appropriately, we therefore shouldn't modify the level set for them.
- some objects need to be build with optimization disabled, which
prevents _FORTIFY_SOURCE to be used for them.
Assembler files that implement architecture specific versions of the
fortified routines were not excluded from _FORTIFY_SOURCE as there is no
C header included that would impact their behavior.
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
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>
This patch redirects the error functions to the appropriate
longdouble variants which enables the compiler to optimize
for the abi ieeelongdouble.
Signed-off-by: Sachin Monga <smonga@linux.ibm.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 8 (which is sizeof of uint32_t minus
MANTISSA_WIDTH plus the signal bit):
while (ex > ey)
{
mx << 8;
ex -= 8;
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 | 17.2549 | 12.0318
x86_64 (Ryzen 9) | normal | 85.4096 | 49.9641
x86_64 (Ryzen 9) | close-exponents | 19.1072 | 15.8224
aarch64 (N1) | subnormal | 10.2182 | 6.81778
aarch64 (N1) | normal | 60.0616 | 20.3667
aarch64 (N1) | close-exponents | 11.5256 | 8.39685
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, and multiplication):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
armhf (N1) | subnormal | 11.6662 | 10.8955
armhf (N1) | normal | 69.2759 | 34.1524
armhf (N1) | close-exponents | 13.6472 | 18.2131
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 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>
They are both used by __libc_freeres to free all library malloc
allocated resources to help tooling like mtrace or valgrind with
memory leak tracking.
The current scheme uses assembly markers and linker script entries
to consolidate the free routine function pointers in the RELRO segment
and to be freed buffers in BSS.
This patch changes it to use specific free functions for
libc_freeres_ptrs buffers and call the function pointer array directly
with call_function_static_weak.
It allows the removal of both the internal macros and the linker
script sections.
Checked on x86_64-linux-gnu, i686-linux-gnu, and aarch64-linux-gnu.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
C2x adds binary integer constants starting with 0b or 0B, and supports
those constants for the %i scanf format (in addition to the %b format,
which isn't yet implemented for scanf in glibc). Implement that scanf
support for glibc.
As with the strtol support, this is incompatible with previous C
standard versions, in that such an input string starting with 0b or 0B
was previously required to be parsed as 0 (with the rest of the input
potentially matching subsequent parts of the scanf format string).
Thus this patch adds 12 new __isoc23_* functions per long double
format (12, 24 or 36 depending on how many long double formats the
glibc configuration supports), with appropriate header redirection
support (generally very closely following that for the __isoc99_*
scanf functions - note that __GLIBC_USE (DEPRECATED_SCANF) takes
precedence over __GLIBC_USE (C2X_STRTOL), so the case of GNU
extensions to C89 continues to get old-style GNU %a and does not get
this new feature). The function names would remain as __isoc23_* even
if C2x ends up published in 2024 rather than 2023.
When scanf %b support is added, I think it will be appropriate for all
versions of scanf to follow C2x rules for inputs to the %b format
(given that there are no compatibility concerns for a new format).
Tested for x86_64 (full glibc testsuite). The first version was also
tested for powerpc (32-bit) and powerpc64le (stdio-common/ and wcsmbs/
tests), and with build-many-glibcs.py.
The patch suppress the same warnings from 87c266d758,
that shows issues for microblaze, mips soft-fp, nios2, and or1k.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Fix the following issues with built-in function use in
sysdeps/ieee754/ldbl-128 and sysdeps/ieee754/float128:
* fabsl used __builtin_fabsf128 unconditionally, breaking the build
with GCC 6 for several architectures; it should use __builtin_fabsl
with an appropriate redirection in float128_private.h. (I'm not
particularly concerned with building glibc with GCC 6; rather, I
want to be able to run the tgmath.h tests with GCC 6, which is a
significantly different case for tgmath.h compared to GCC 7 and
later because of the lack of _FloatN / _FloatNx support in the
compiler, and at present running the tests with a compiler means
building glibc with that compiler.)
* Some (conditional) uses of built-in functions had been added to
ldbl-128 without appropriate float128_private.h remapping (there was
remapping for the macros controlling whether the built-in functions
are used, just not for the functions themselves).
* s_llrintl.c called __builtin_round not __builtin_llrintl, which is
obviously wrong.
Tested with build-many-glibcs.py for aarch64-linux-gnu, GCC 6 (where
it fixes the glibc build) and GCC 12, and with the glibc testsuite for
x86_64.
vfprintf is entangled with vfwprintf (of course), __printf_fp,
__printf_fphex, __vstrfmon_l_internal, and the strfrom family of
functions. The latter use the internal snprintf functionality,
so vsnprintf is converted as well.
The simples conversion is __printf_fphex, followed by
__vstrfmon_l_internal and __printf_fp, and finally
__vfprintf_internal and __vfwprintf_internal. __vsnprintf_internal
and strfrom* are mostly consuming the new interfaces, so they
are comparatively simple.
__printf_fp is a public symbol, so the FILE *-based interface
had to preserved.
The __printf_fp rewrite does not change the actual binary-to-decimal
conversion algorithm, and digits are still not emitted directly to
the target buffer. However, the staging buffer now uses bytes
instead of wide characters, and one buffer copy is eliminated.
The changes are at least performance-neutral in my testing.
Floating point printing and snprintf improved measurably, so that
this Lua script
for i=1,5000000 do
print(i, i * math.pi)
end
runs about 5% faster for me. To preserve fprintf performance for
a simple "%d" format, this commit has some logic changes under
LABEL (unsigned_number) to avoid additional function calls. There
are certainly some very easy performance improvements here: binary,
octal and hexadecimal formatting can easily avoid the temporary work
buffer (the number of digits can be computed ahead-of-time using one
of the __builtin_clz* built-ins). Decimal formatting can use a
specialized version of _itoa_word for base 10.
The existing (inconsistent) width handling between strfmon and printf
is preserved here. __print_fp_buffer_1 would have to use
__translated_number_width to achieve ISO conformance for printf.
Test expectations in libio/tst-vtables-common.c are adjusted because
the internal staging buffer merges all virtual function calls into
one.
In general, stack buffer usage is greatly reduced, particularly for
unbuffered input streams. __printf_fp can still use a large buffer
in binary128 mode for %g, though.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
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>
GCC 13 has added more _FloatN and _FloatNx versions of existing
<math.h> and <complex.h> built-in functions, for use in libstdc++-v3.
This breaks the glibc build because of how those functions are defined
as aliases to functions with the same ABI but different types. Add
appropriate -fno-builtin-* options for compiling relevant files, as
already done for the case of long double functions aliasing double
ones and based on the list of files used there.
I fixed some mistakes in that list of double files that I noticed
while implementing this fix, but there may well be more such
(harmless) cases, in this list or the new one (files that don't
actually exist or don't define the named functions as aliases so don't
need the options). I did try to exclude cases where glibc doesn't
define certain functions for _FloatN or _FloatNx types at all from the
new uses of -fno-builtin-* options. As with the options for double
files (see the commit message for commit
49348beafe, "Fix build with GCC 10 when
long double = double."), it's deliberate that the options are used
even if GCC currently doesn't have a built-in version of a given
functions, so providing some level of future-proofing against more
such built-in functions being added in future.
Tested with build-many-glibcs.py for aarch64-linux-gnu
powerpc-linux-gnu powerpc64le-linux-gnu x86_64-linux-gnu (compilers
and glibcs builds) with GCC mainline.
Detecting an overflow edge case depended on signed overflow of a long
long. Replace the additions and the overflow checks by
__builtin_add_overflow().
Reviewed-by: Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
Avoid moving code across SET_RESTORE_ROUNDL in order to fix
[BZ #29463].
Tested-by: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
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>
GCC 13 adds support for _FloatN and _FloatNx types in C++, so breaking
the installed glibc headers that assume such support is not present.
GCC mostly works around this with fixincludes, but that doesn't help
for building glibc and its tests (glibc doesn't itself contain C++
code, but there's C++ code built for tests). Update glibc's
bits/floatn-common.h and bits/floatn.h headers to handle the GCC 13
support directly.
In general the changes match those made by fixincludes, though I think
the ones in sysdeps/powerpc/bits/floatn.h, where the header tests
__LDBL_MANT_DIG__ == 113 or uses #elif, wouldn't match the existing
fixincludes patterns.
Some places involving special C++ handling in relation to _FloatN
support are not changed. There's no need to change the
__HAVE_FLOATN_NOT_TYPEDEF definition (also in a form that wouldn't be
matched by the fixincludes fixes) because it's only used in relation
to macro definitions using features not supported for C++
(__builtin_types_compatible_p and _Generic). And there's no need to
change the inline function overloads for issignaling, iszero and
iscanonical in C++ because cases where types have the same format but
are no longer compatible types are handled automatically by the C++
overload resolution rules.
This patch also does not change the overload handling for iseqsig, and
there I think changes *are* needed, beyond those in this patch or made
by fixincludes. The way that overload is defined, via a template
parameter to a structure type, requires overloads whenever the types
are incompatible, even if they have the same format. So I think we
need to add overloads with GCC 13 for every supported _FloatN and
_FloatNx type, rather than just having one for _Float128 when it has a
different ABI to long double as at present (but for older GCC, such
overloads must not be defined for types that end up defined as
typedefs for another type).
Tested with build-many-glibcs.py: compilers build for
aarch64-linux-gnu ia64-linux-gnu mips64-linux-gnu powerpc-linux-gnu
powerpc64le-linux-gnu x86_64-linux-gnu; glibcs build for
aarch64-linux-gnu ia64-linux-gnu i686-linux-gnu mips-linux-gnu
mips64-linux-gnu-n32 powerpc-linux-gnu powerpc64le-linux-gnu
x86_64-linux-gnu.
math/test-float128-y1 fails on x86_64 and ppc64el with gcc 12 and -O3,
because code inside a block guarded by SET_RESTORE_ROUNDL is being moved
after the rounding mode has been restored. Use math_force_eval to
prevent this (and insert some math_opt_barrier calls to prevent code
from being moved before the rounding mode is set).
Fixes#29463
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.
Both float, double, and _Float128 are assumed to be supported
(float and double already only uses builtins). Only long double
is parametrized due GCC bug 29253 which prevents its usage on
powerpc.
It allows to remove i686, ia64, x86_64, powerpc, and sparc arch
specific implementation.
On ia64 it also fixes the sNAN handling:
math/test-float64x-fabs
math/test-ldouble-fabs
Checked on x86_64-linux-gnu, i686-linux-gnu, powerpc-linux-gnu,
powerpc64-linux-gnu, sparc64-linux-gnu, and ia64-linux-gnu.
Converting double precision constants to float is now affected by the
runtime dynamic rounding mode instead of being evaluated at compile
time with default rounding mode (except static object initializers).
This can change the computed result and cause performance regression.
The known correctness issues (increased ulp errors) are already fixed,
this patch fixes remaining cases of unnecessary runtime conversions.
Add float M_* macros to math.h as new GNU extension API. To avoid
conversions the new M_* macros are used and instead of casting double
literals to float, use float literals (only required if the conversion
is inexact).
The patch was tested on aarch64 where the following symbols had new
spurious conversion instructions that got fixed:
__clog10f
__gammaf_r_finite@GLIBC_2.17
__j0f_finite@GLIBC_2.17
__j1f_finite@GLIBC_2.17
__jnf_finite@GLIBC_2.17
__kernel_casinhf
__lgamma_negf
__log1pf
__y0f_finite@GLIBC_2.17
__y1f_finite@GLIBC_2.17
cacosf
cacoshf
casinhf
catanf
catanhf
clogf
gammaf_positive
Fixes bug 28713.
Reviewed-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
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
s_cosf.c and s_sinf.c have
if (abstop12 (y) < abstop12 (pio4))
where abstop12 takes a float argument, but pio4 is static const double.
pio4 is used only in calls to abstop12 and never in arithmetic. Apply
-static const double pio4 = 0x1.921FB54442D18p-1;
+static const float pio4 = 0x1.921FB6p-1f;
to fix:
FAIL: math/test-float-cos
FAIL: math/test-float-sin
FAIL: math/test-float-sincos
FAIL: math/test-float32-cos
FAIL: math/test-float32-sin
FAIL: math/test-float32-sincos
when compiling with GCC 12.
Reviewed-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
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.
This implementation is based on '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 subnormal results.
The main advantage of the new algorithm is its precision. With a
random 1e9 input pairs in the range of [LDBL_MIN, LDBL_MAX], glibc
current implementation shows around 0.05% results with an error of
1 ulp (453266 results) while the new implementation only shows
0.0001% of total (1280).
Checked on aarch64-linux-gnu and x86_64-linux-gnu.
[1] https://arxiv.org/pdf/1904.09481.pdf
This implementation is based on '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 subnormal results.
The main advantage of the new algorithm is its precision. With a
random 1e8 input pairs in the range of [LDBL_MIN, LDBL_MAX], glibc
current implementation shows around 0.02% results with an error of
1 ulp (23158 results) while the new implementation only shows
0.0001% of total (111).
[1] https://arxiv.org/pdf/1904.09481.pdf
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
