Fix erf underflow handling near 0 (bug 16516).

Bug 16516 reports spurious underflows from erf (for all floating-point
types), when the result is close to underflowing but does not actually
underflow.

erf (x) is about (2/sqrt(pi))*x for x close to 0, so there are
subnormal arguments for which it does not underflow.  The various
implementations do (x + efx*x) (for efx = 2/sqrt(pi) - 1), for greater
accuracy than if just using a single multiplication by an
approximation to 2/sqrt(pi) (effectively, this way there are a few
more bits in the approximation to 2/sqrt(pi)).  This can introduce
underflows when efx*x underflows even though the final result does
not, so a scaled calculation with 8*efx is done in these cases - but 8
is not a big enough scale factor to avoid all such underflows.  16 is
(any underflows with a scale factor of 16 would only occur when the
final result underflows), so this patch changes the code to use that
factor.  Rather than recomputing all the values of the efx8 variable,
it is removed, leaving it to the compiler's constant folding to
compute 16*efx.  As such scaling can also lose underflows when the
final scaling down happens to be exact, appropriate checks are added
to ensure underflow exceptions occur when required in such cases.

Tested x86_64 and x86; no ulps updates needed.  Also spot-checked for
powerpc32 and mips64 to verify the changes to the ldbl-128ibm and
ldbl-128 implementations.

	[BZ #16516]
	* sysdeps/ieee754/dbl-64/s_erf.c (efx8): Remove variable.
	(__erf): Scale by 16 instead of 8 in potentially underflowing
	case.  Ensure exception if result actually underflows.
	* sysdeps/ieee754/flt-32/s_erff.c (efx8): Remove variable.
	(__erff): Scale by 16 instead of 8 in potentially underflowing
	case.  Ensure exception if result actually underflows.
	* sysdeps/ieee754/ldbl-128/s_erfl.c: Include <float.h>.
	(efx8): Remove variable.
	(__erfl): Scale by 16 instead of 8 in potentially underflowing
	case.  Ensure exception if result actually underflows.
	* sysdeps/ieee754/ldbl-128ibm/s_erfl.c: Include <float.h>.
	(efx8): Remove variable.
	(__erfl): Scale by 16 instead of 8 in potentially underflowing
	case.  Ensure exception if result actually underflows.
	* sysdeps/ieee754/ldbl-96/s_erfl.c: Include <float.h>.
	(efx8): Remove variable.
	(__erfl): Scale by 16 instead of 8 in potentially underflowing
	case.  Ensure exception if result actually underflows.
	* math/auto-libm-test-in: Add more tests of erf.
	* math/auto-libm-test-out: Regenerated.

diff --git a/ChangeLog b/ChangeLog
index 26720a95f0576ecb538fcfb63a9a6ccc496a25af..62192f3b9350fd3f03701fe9905d6f6fa8104e83 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,27 @@
+2014-05-14  Joseph Myers  <joseph@codesourcery.com>
+
+       [BZ #16516]
+       * sysdeps/ieee754/dbl-64/s_erf.c (efx8): Remove variable.
+       (__erf): Scale by 16 instead of 8 in potentially underflowing
+       case.  Ensure exception if result actually underflows.
+       * sysdeps/ieee754/flt-32/s_erff.c (efx8): Remove variable.
+       (__erff): Scale by 16 instead of 8 in potentially underflowing
+       case.  Ensure exception if result actually underflows.
+       * sysdeps/ieee754/ldbl-128/s_erfl.c: Include <float.h>.
+       (efx8): Remove variable.
+       (__erfl): Scale by 16 instead of 8 in potentially underflowing
+       case.  Ensure exception if result actually underflows.
+       * sysdeps/ieee754/ldbl-128ibm/s_erfl.c: Include <float.h>.
+       (efx8): Remove variable.
+       (__erfl): Scale by 16 instead of 8 in potentially underflowing
+       case.  Ensure exception if result actually underflows.
+       * sysdeps/ieee754/ldbl-96/s_erfl.c: Include <float.h>.
+       (efx8): Remove variable.
+       (__erfl): Scale by 16 instead of 8 in potentially underflowing
+       case.  Ensure exception if result actually underflows.
+       * math/auto-libm-test-in: Add more tests of erf.
+       * math/auto-libm-test-out: Regenerated.
+
 2014-05-14  Andreas Schwab  <schwab@suse.de>
 
        * elf/rtld.c (_dl_start_final, _dl_start, init_tls, dl_main):

diff --git a/NEWS b/NEWS
index f06e33524cb30d58785e089f3197f9c9d89a96bb..9293f802f8fda0af6cb459a16ed15fd6adff23ff 100644 (file)
--- a/NEWS
+++ b/NEWS
@@ -11,12 +11,13 @@ Version 2.20
 
   6804, 9894, 12994, 13347, 13651, 14308, 14770, 15119, 15347, 15514, 15804,
   15894, 16002, 16064, 16198, 16284, 16348, 16349, 16357, 16362, 16447,
-  16532, 16545, 16574, 16599, 16600, 16609, 16610, 16611, 16613, 16619,
-  16623, 16629, 16632, 16634, 16639, 16642, 16648, 16649, 16670, 16674,
-  16677, 16680, 16683, 16689, 16695, 16701, 16706, 16707, 16712, 16713,
-  16714, 16731, 16739, 16740, 16743, 16754, 16758, 16759, 16760, 16770,
-  16786, 16789, 16791, 16799, 16800, 16815, 16823, 16824, 16831, 16838,
-  16854, 16876, 16877, 16885, 16888, 16890, 16912, 16916, 16922, 16932.
+  16516, 16532, 16545, 16574, 16599, 16600, 16609, 16610, 16611, 16613,
+  16619, 16623, 16629, 16632, 16634, 16639, 16642, 16648, 16649, 16670,
+  16674, 16677, 16680, 16683, 16689, 16695, 16701, 16706, 16707, 16712,
+  16713, 16714, 16731, 16739, 16740, 16743, 16754, 16758, 16759, 16760,
+  16770, 16786, 16789, 16791, 16799, 16800, 16815, 16823, 16824, 16831,
+  16838, 16854, 16876, 16877, 16885, 16888, 16890, 16912, 16916, 16922,
+  16932.
 
 * The minimum Linux kernel version that this version of the GNU C Library
   can be used with is 2.6.32.

diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in
index 420f657d81df8c32e8eda6d44aa39a6d866a4af0..da93a302cc211a69fa12adea868643affd5758ed 100644 (file)
--- a/math/auto-libm-test-in
+++ b/math/auto-libm-test-in
@@ -796,6 +796,10 @@ erf 4.125
 erf 27.0
 erf -27.0
 erf -0x1.fffffffffffff8p-2
+erf 0x1.c5bf94p-127
+erf 0x3.8b7fa8p-128
+erf -0x3.8b7f12369ded8p-1024
+erf 0x3.8b7f12369ded5518p-16384
 
 erfc 0.0
 erfc -0

diff --git a/math/auto-libm-test-out b/math/auto-libm-test-out
index 4a56804e3dfb645daef5e155d64269769280803f..a0d919a8bdc7a03bd84b27dcd2066eb6555de970 100644 (file)
--- a/math/auto-libm-test-out
+++ b/math/auto-libm-test-out
@@ -87801,6 +87801,206 @@ erf -0x1.fffffffffffff8p-2
 = erf tonearest ldbl-128ibm -0x7.ffffffffffffep-4L : -0x8.53f7ae0c76e8f9c90d4d08ca18p-4L : inexact-ok
 = erf towardzero ldbl-128ibm -0x7.ffffffffffffep-4L : -0x8.53f7ae0c76e8f9c90d4d08ca14p-4L : inexact-ok
 = erf upward ldbl-128ibm -0x7.ffffffffffffep-4L : -0x8.53f7ae0c76e8f9c90d4d08ca14p-4L : inexact-ok
+erf 0x1.c5bf94p-127
+= erf downward flt-32 0x3.8b7f28p-128f : 0x4.000018p-128f : inexact-ok
+= erf tonearest flt-32 0x3.8b7f28p-128f : 0x4.000018p-128f : inexact-ok
+= erf towardzero flt-32 0x3.8b7f28p-128f : 0x4.000018p-128f : inexact-ok
+= erf upward flt-32 0x3.8b7f28p-128f : 0x4.00002p-128f : inexact-ok
+= erf downward dbl-64 0x3.8b7f28p-128 : 0x4.000018956724p-128 : inexact-ok
+= erf tonearest dbl-64 0x3.8b7f28p-128 : 0x4.000018956724p-128 : inexact-ok
+= erf towardzero dbl-64 0x3.8b7f28p-128 : 0x4.000018956724p-128 : inexact-ok
+= erf upward dbl-64 0x3.8b7f28p-128 : 0x4.0000189567244p-128 : inexact-ok
+= erf downward ldbl-96-intel 0x3.8b7f28p-128L : 0x4.0000189567240f1p-128L : inexact-ok
+= erf tonearest ldbl-96-intel 0x3.8b7f28p-128L : 0x4.0000189567240f1p-128L : inexact-ok
+= erf towardzero ldbl-96-intel 0x3.8b7f28p-128L : 0x4.0000189567240f1p-128L : inexact-ok
+= erf upward ldbl-96-intel 0x3.8b7f28p-128L : 0x4.0000189567240f18p-128L : inexact-ok
+= erf downward ldbl-96-m68k 0x3.8b7f28p-128L : 0x4.0000189567240f1p-128L : inexact-ok
+= erf tonearest ldbl-96-m68k 0x3.8b7f28p-128L : 0x4.0000189567240f1p-128L : inexact-ok
+= erf towardzero ldbl-96-m68k 0x3.8b7f28p-128L : 0x4.0000189567240f1p-128L : inexact-ok
+= erf upward ldbl-96-m68k 0x3.8b7f28p-128L : 0x4.0000189567240f18p-128L : inexact-ok
+= erf downward ldbl-128 0x3.8b7f28p-128L : 0x4.0000189567240f10919b31d6bcf4p-128L : inexact-ok
+= erf tonearest ldbl-128 0x3.8b7f28p-128L : 0x4.0000189567240f10919b31d6bcf4p-128L : inexact-ok
+= erf towardzero ldbl-128 0x3.8b7f28p-128L : 0x4.0000189567240f10919b31d6bcf4p-128L : inexact-ok
+= erf upward ldbl-128 0x3.8b7f28p-128L : 0x4.0000189567240f10919b31d6bcf8p-128L : inexact-ok
+= erf downward ldbl-128ibm 0x3.8b7f28p-128L : 0x4.0000189567240f10919b31d6bcp-128L : inexact-ok
+= erf tonearest ldbl-128ibm 0x3.8b7f28p-128L : 0x4.0000189567240f10919b31d6bcp-128L : inexact-ok
+= erf towardzero ldbl-128ibm 0x3.8b7f28p-128L : 0x4.0000189567240f10919b31d6bcp-128L : inexact-ok
+= erf upward ldbl-128ibm 0x3.8b7f28p-128L : 0x4.0000189567240f10919b31d6bep-128L : inexact-ok
+erf 0x3.8b7fa8p-128
+= erf downward flt-32 0x3.8b7fa8p-128f : 0x4.0000a8p-128f : inexact-ok
+= erf tonearest flt-32 0x3.8b7fa8p-128f : 0x4.0000a8p-128f : inexact-ok
+= erf towardzero flt-32 0x3.8b7fa8p-128f : 0x4.0000a8p-128f : inexact-ok
+= erf upward flt-32 0x3.8b7fa8p-128f : 0x4.0000bp-128f : inexact-ok
+= erf downward dbl-64 0x3.8b7fa8p-128 : 0x4.0000a90421a6p-128 : inexact-ok
+= erf tonearest dbl-64 0x3.8b7fa8p-128 : 0x4.0000a90421a64p-128 : inexact-ok
+= erf towardzero dbl-64 0x3.8b7fa8p-128 : 0x4.0000a90421a6p-128 : inexact-ok
+= erf upward dbl-64 0x3.8b7fa8p-128 : 0x4.0000a90421a64p-128 : inexact-ok
+= erf downward ldbl-96-intel 0x3.8b7fa8p-128L : 0x4.0000a90421a623e8p-128L : inexact-ok
+= erf tonearest ldbl-96-intel 0x3.8b7fa8p-128L : 0x4.0000a90421a623e8p-128L : inexact-ok
+= erf towardzero ldbl-96-intel 0x3.8b7fa8p-128L : 0x4.0000a90421a623e8p-128L : inexact-ok
+= erf upward ldbl-96-intel 0x3.8b7fa8p-128L : 0x4.0000a90421a623fp-128L : inexact-ok
+= erf downward ldbl-96-m68k 0x3.8b7fa8p-128L : 0x4.0000a90421a623e8p-128L : inexact-ok
+= erf tonearest ldbl-96-m68k 0x3.8b7fa8p-128L : 0x4.0000a90421a623e8p-128L : inexact-ok
+= erf towardzero ldbl-96-m68k 0x3.8b7fa8p-128L : 0x4.0000a90421a623e8p-128L : inexact-ok
+= erf upward ldbl-96-m68k 0x3.8b7fa8p-128L : 0x4.0000a90421a623fp-128L : inexact-ok
+= erf downward ldbl-128 0x3.8b7fa8p-128L : 0x4.0000a90421a623ebfa28a3ab477p-128L : inexact-ok
+= erf tonearest ldbl-128 0x3.8b7fa8p-128L : 0x4.0000a90421a623ebfa28a3ab4774p-128L : inexact-ok
+= erf towardzero ldbl-128 0x3.8b7fa8p-128L : 0x4.0000a90421a623ebfa28a3ab477p-128L : inexact-ok
+= erf upward ldbl-128 0x3.8b7fa8p-128L : 0x4.0000a90421a623ebfa28a3ab4774p-128L : inexact-ok
+= erf downward ldbl-128ibm 0x3.8b7fa8p-128L : 0x4.0000a90421a623ebfa28a3ab46p-128L : inexact-ok
+= erf tonearest ldbl-128ibm 0x3.8b7fa8p-128L : 0x4.0000a90421a623ebfa28a3ab48p-128L : inexact-ok
+= erf towardzero ldbl-128ibm 0x3.8b7fa8p-128L : 0x4.0000a90421a623ebfa28a3ab46p-128L : inexact-ok
+= erf upward ldbl-128ibm 0x3.8b7fa8p-128L : 0x4.0000a90421a623ebfa28a3ab48p-128L : inexact-ok
+erf -0x3.8b7f12369ded8p-1024
+= erf downward flt-32 -0x0p+0f : -0x0p+0f : inexact-ok
+= erf tonearest flt-32 -0x0p+0f : -0x0p+0f : inexact-ok
+= erf towardzero flt-32 -0x0p+0f : -0x0p+0f : inexact-ok
+= erf upward flt-32 -0x0p+0f : -0x0p+0f : inexact-ok
+= erf downward dbl-64 -0x0p+0 : -0x0p+0 : inexact-ok
+= erf tonearest dbl-64 -0x0p+0 : -0x0p+0 : inexact-ok
+= erf towardzero dbl-64 -0x0p+0 : -0x0p+0 : inexact-ok
+= erf upward dbl-64 -0x0p+0 : -0x0p+0 : inexact-ok
+= erf downward ldbl-96-intel -0x0p+0L : -0x0p+0L : inexact-ok
+= erf tonearest ldbl-96-intel -0x0p+0L : -0x0p+0L : inexact-ok
+= erf towardzero ldbl-96-intel -0x0p+0L : -0x0p+0L : inexact-ok
+= erf upward ldbl-96-intel -0x0p+0L : -0x0p+0L : inexact-ok
+= erf downward ldbl-96-m68k -0x0p+0L : -0x0p+0L : inexact-ok
+= erf tonearest ldbl-96-m68k -0x0p+0L : -0x0p+0L : inexact-ok
+= erf towardzero ldbl-96-m68k -0x0p+0L : -0x0p+0L : inexact-ok
+= erf upward ldbl-96-m68k -0x0p+0L : -0x0p+0L : inexact-ok
+= erf downward ldbl-128 -0x0p+0L : -0x0p+0L : inexact-ok
+= erf tonearest ldbl-128 -0x0p+0L : -0x0p+0L : inexact-ok
+= erf towardzero ldbl-128 -0x0p+0L : -0x0p+0L : inexact-ok
+= erf upward ldbl-128 -0x0p+0L : -0x0p+0L : inexact-ok
+= erf downward ldbl-128ibm -0x0p+0L : -0x0p+0L : inexact-ok
+= erf tonearest ldbl-128ibm -0x0p+0L : -0x0p+0L : inexact-ok
+= erf towardzero ldbl-128ibm -0x0p+0L : -0x0p+0L : inexact-ok
+= erf upward ldbl-128ibm -0x0p+0L : -0x0p+0L : inexact-ok
+= erf downward flt-32 -0x8p-152f : -0x1p-148f : inexact-ok underflow errno-erange-ok
+= erf tonearest flt-32 -0x8p-152f : -0x8p-152f : inexact-ok underflow errno-erange-ok
+= erf towardzero flt-32 -0x8p-152f : -0x8p-152f : inexact-ok underflow errno-erange-ok
+= erf upward flt-32 -0x8p-152f : -0x8p-152f : inexact-ok underflow errno-erange-ok
+= erf downward dbl-64 -0x8p-152 : -0x9.06eba8214db7p-152 : inexact-ok
+= erf tonearest dbl-64 -0x8p-152 : -0x9.06eba8214db68p-152 : inexact-ok
+= erf towardzero dbl-64 -0x8p-152 : -0x9.06eba8214db68p-152 : inexact-ok
+= erf upward dbl-64 -0x8p-152 : -0x9.06eba8214db68p-152 : inexact-ok
+= erf downward ldbl-96-intel -0x8p-152L : -0x9.06eba8214db688ep-152L : inexact-ok
+= erf tonearest ldbl-96-intel -0x8p-152L : -0x9.06eba8214db688dp-152L : inexact-ok
+= erf towardzero ldbl-96-intel -0x8p-152L : -0x9.06eba8214db688dp-152L : inexact-ok
+= erf upward ldbl-96-intel -0x8p-152L : -0x9.06eba8214db688dp-152L : inexact-ok
+= erf downward ldbl-96-m68k -0x8p-152L : -0x9.06eba8214db688ep-152L : inexact-ok
+= erf tonearest ldbl-96-m68k -0x8p-152L : -0x9.06eba8214db688dp-152L : inexact-ok
+= erf towardzero ldbl-96-m68k -0x8p-152L : -0x9.06eba8214db688dp-152L : inexact-ok
+= erf upward ldbl-96-m68k -0x8p-152L : -0x9.06eba8214db688dp-152L : inexact-ok
+= erf downward ldbl-128 -0x8p-152L : -0x9.06eba8214db688d71d48a7f6bffp-152L : inexact-ok
+= erf tonearest ldbl-128 -0x8p-152L : -0x9.06eba8214db688d71d48a7f6bffp-152L : inexact-ok
+= erf towardzero ldbl-128 -0x8p-152L : -0x9.06eba8214db688d71d48a7f6bfe8p-152L : inexact-ok
+= erf upward ldbl-128 -0x8p-152L : -0x9.06eba8214db688d71d48a7f6bfe8p-152L : inexact-ok
+= erf downward ldbl-128ibm -0x8p-152L : -0x9.06eba8214db688d71d48a7f6cp-152L : inexact-ok
+= erf tonearest ldbl-128ibm -0x8p-152L : -0x9.06eba8214db688d71d48a7f6cp-152L : inexact-ok
+= erf towardzero ldbl-128ibm -0x8p-152L : -0x9.06eba8214db688d71d48a7f6bcp-152L : inexact-ok
+= erf upward ldbl-128ibm -0x8p-152L : -0x9.06eba8214db688d71d48a7f6bcp-152L : inexact-ok
+= erf downward dbl-64 -0x3.8b7f12369ded8p-1024 : -0x4.0000000000004p-1024 : inexact-ok
+= erf tonearest dbl-64 -0x3.8b7f12369ded8p-1024 : -0x4.0000000000004p-1024 : inexact-ok
+= erf towardzero dbl-64 -0x3.8b7f12369ded8p-1024 : -0x4p-1024 : inexact-ok
+= erf upward dbl-64 -0x3.8b7f12369ded8p-1024 : -0x4p-1024 : inexact-ok
+= erf downward ldbl-96-intel -0x3.8b7f12369ded8p-1024L : -0x4.0000000000003098p-1024L : inexact-ok
+= erf tonearest ldbl-96-intel -0x3.8b7f12369ded8p-1024L : -0x4.000000000000309p-1024L : inexact-ok
+= erf towardzero ldbl-96-intel -0x3.8b7f12369ded8p-1024L : -0x4.000000000000309p-1024L : inexact-ok
+= erf upward ldbl-96-intel -0x3.8b7f12369ded8p-1024L : -0x4.000000000000309p-1024L : inexact-ok
+= erf downward ldbl-96-m68k -0x3.8b7f12369ded8p-1024L : -0x4.0000000000003098p-1024L : inexact-ok
+= erf tonearest ldbl-96-m68k -0x3.8b7f12369ded8p-1024L : -0x4.000000000000309p-1024L : inexact-ok
+= erf towardzero ldbl-96-m68k -0x3.8b7f12369ded8p-1024L : -0x4.000000000000309p-1024L : inexact-ok
+= erf upward ldbl-96-m68k -0x3.8b7f12369ded8p-1024L : -0x4.000000000000309p-1024L : inexact-ok
+= erf downward ldbl-128 -0x3.8b7f12369ded8p-1024L : -0x4.00000000000030934524cf4ab6e4p-1024L : inexact-ok
+= erf tonearest ldbl-128 -0x3.8b7f12369ded8p-1024L : -0x4.00000000000030934524cf4ab6ep-1024L : inexact-ok
+= erf towardzero ldbl-128 -0x3.8b7f12369ded8p-1024L : -0x4.00000000000030934524cf4ab6ep-1024L : inexact-ok
+= erf upward ldbl-128 -0x3.8b7f12369ded8p-1024L : -0x4.00000000000030934524cf4ab6ep-1024L : inexact-ok
+= erf downward ldbl-128ibm -0x3.8b7f12369ded8p-1024L : -0x4.0000000000004p-1024L : inexact-ok underflow errno-erange-ok
+= erf tonearest ldbl-128ibm -0x3.8b7f12369ded8p-1024L : -0x4.0000000000004p-1024L : inexact-ok underflow errno-erange-ok
+= erf towardzero ldbl-128ibm -0x3.8b7f12369ded8p-1024L : -0x4p-1024L : inexact-ok underflow errno-erange-ok
+= erf upward ldbl-128ibm -0x3.8b7f12369ded8p-1024L : -0x4p-1024L : inexact-ok underflow errno-erange-ok
+erf 0x3.8b7f12369ded5518p-16384
+= erf downward flt-32 0x8p-152f : 0x8p-152f : inexact-ok underflow errno-erange-ok
+= erf tonearest flt-32 0x8p-152f : 0x8p-152f : inexact-ok underflow errno-erange-ok
+= erf towardzero flt-32 0x8p-152f : 0x8p-152f : inexact-ok underflow errno-erange-ok
+= erf upward flt-32 0x8p-152f : 0x1p-148f : inexact-ok underflow errno-erange-ok
+= erf downward dbl-64 0x8p-152 : 0x9.06eba8214db68p-152 : inexact-ok
+= erf tonearest dbl-64 0x8p-152 : 0x9.06eba8214db68p-152 : inexact-ok
+= erf towardzero dbl-64 0x8p-152 : 0x9.06eba8214db68p-152 : inexact-ok
+= erf upward dbl-64 0x8p-152 : 0x9.06eba8214db7p-152 : inexact-ok
+= erf downward ldbl-96-intel 0x8p-152L : 0x9.06eba8214db688dp-152L : inexact-ok
+= erf tonearest ldbl-96-intel 0x8p-152L : 0x9.06eba8214db688dp-152L : inexact-ok
+= erf towardzero ldbl-96-intel 0x8p-152L : 0x9.06eba8214db688dp-152L : inexact-ok
+= erf upward ldbl-96-intel 0x8p-152L : 0x9.06eba8214db688ep-152L : inexact-ok
+= erf downward ldbl-96-m68k 0x8p-152L : 0x9.06eba8214db688dp-152L : inexact-ok
+= erf tonearest ldbl-96-m68k 0x8p-152L : 0x9.06eba8214db688dp-152L : inexact-ok
+= erf towardzero ldbl-96-m68k 0x8p-152L : 0x9.06eba8214db688dp-152L : inexact-ok
+= erf upward ldbl-96-m68k 0x8p-152L : 0x9.06eba8214db688ep-152L : inexact-ok
+= erf downward ldbl-128 0x8p-152L : 0x9.06eba8214db688d71d48a7f6bfe8p-152L : inexact-ok
+= erf tonearest ldbl-128 0x8p-152L : 0x9.06eba8214db688d71d48a7f6bffp-152L : inexact-ok
+= erf towardzero ldbl-128 0x8p-152L : 0x9.06eba8214db688d71d48a7f6bfe8p-152L : inexact-ok
+= erf upward ldbl-128 0x8p-152L : 0x9.06eba8214db688d71d48a7f6bffp-152L : inexact-ok
+= erf downward ldbl-128ibm 0x8p-152L : 0x9.06eba8214db688d71d48a7f6bcp-152L : inexact-ok
+= erf tonearest ldbl-128ibm 0x8p-152L : 0x9.06eba8214db688d71d48a7f6cp-152L : inexact-ok
+= erf towardzero ldbl-128ibm 0x8p-152L : 0x9.06eba8214db688d71d48a7f6bcp-152L : inexact-ok
+= erf upward ldbl-128ibm 0x8p-152L : 0x9.06eba8214db688d71d48a7f6cp-152L : inexact-ok
+= erf downward flt-32 0x0p+0f : 0x0p+0f : inexact-ok
+= erf tonearest flt-32 0x0p+0f : 0x0p+0f : inexact-ok
+= erf towardzero flt-32 0x0p+0f : 0x0p+0f : inexact-ok
+= erf upward flt-32 0x0p+0f : 0x0p+0f : inexact-ok
+= erf downward dbl-64 0x0p+0 : 0x0p+0 : inexact-ok
+= erf tonearest dbl-64 0x0p+0 : 0x0p+0 : inexact-ok
+= erf towardzero dbl-64 0x0p+0 : 0x0p+0 : inexact-ok
+= erf upward dbl-64 0x0p+0 : 0x0p+0 : inexact-ok
+= erf downward ldbl-96-intel 0x0p+0L : 0x0p+0L : inexact-ok
+= erf tonearest ldbl-96-intel 0x0p+0L : 0x0p+0L : inexact-ok
+= erf towardzero ldbl-96-intel 0x0p+0L : 0x0p+0L : inexact-ok
+= erf upward ldbl-96-intel 0x0p+0L : 0x0p+0L : inexact-ok
+= erf downward ldbl-96-m68k 0x0p+0L : 0x0p+0L : inexact-ok
+= erf tonearest ldbl-96-m68k 0x0p+0L : 0x0p+0L : inexact-ok
+= erf towardzero ldbl-96-m68k 0x0p+0L : 0x0p+0L : inexact-ok
+= erf upward ldbl-96-m68k 0x0p+0L : 0x0p+0L : inexact-ok
+= erf downward ldbl-128 0x0p+0L : 0x0p+0L : inexact-ok
+= erf tonearest ldbl-128 0x0p+0L : 0x0p+0L : inexact-ok
+= erf towardzero ldbl-128 0x0p+0L : 0x0p+0L : inexact-ok
+= erf upward ldbl-128 0x0p+0L : 0x0p+0L : inexact-ok
+= erf downward ldbl-128ibm 0x0p+0L : 0x0p+0L : inexact-ok
+= erf tonearest ldbl-128ibm 0x0p+0L : 0x0p+0L : inexact-ok
+= erf towardzero ldbl-128ibm 0x0p+0L : 0x0p+0L : inexact-ok
+= erf upward ldbl-128ibm 0x0p+0L : 0x0p+0L : inexact-ok
+= erf downward dbl-64 0x4p-1076 : 0x4p-1076 : inexact-ok underflow errno-erange-ok
+= erf tonearest dbl-64 0x4p-1076 : 0x4p-1076 : inexact-ok underflow errno-erange-ok
+= erf towardzero dbl-64 0x4p-1076 : 0x4p-1076 : inexact-ok underflow errno-erange-ok
+= erf upward dbl-64 0x4p-1076 : 0x8p-1076 : inexact-ok underflow errno-erange-ok
+= erf downward ldbl-96-intel 0x4p-1076L : 0x4.8375d410a6db4468p-1076L : inexact-ok
+= erf tonearest ldbl-96-intel 0x4p-1076L : 0x4.8375d410a6db4468p-1076L : inexact-ok
+= erf towardzero ldbl-96-intel 0x4p-1076L : 0x4.8375d410a6db4468p-1076L : inexact-ok
+= erf upward ldbl-96-intel 0x4p-1076L : 0x4.8375d410a6db447p-1076L : inexact-ok
+= erf downward ldbl-96-m68k 0x4p-1076L : 0x4.8375d410a6db4468p-1076L : inexact-ok
+= erf tonearest ldbl-96-m68k 0x4p-1076L : 0x4.8375d410a6db4468p-1076L : inexact-ok
+= erf towardzero ldbl-96-m68k 0x4p-1076L : 0x4.8375d410a6db4468p-1076L : inexact-ok
+= erf upward ldbl-96-m68k 0x4p-1076L : 0x4.8375d410a6db447p-1076L : inexact-ok
+= erf downward ldbl-128 0x4p-1076L : 0x4.8375d410a6db446b8ea453fb5ff4p-1076L : inexact-ok
+= erf tonearest ldbl-128 0x4p-1076L : 0x4.8375d410a6db446b8ea453fb5ff8p-1076L : inexact-ok
+= erf towardzero ldbl-128 0x4p-1076L : 0x4.8375d410a6db446b8ea453fb5ff4p-1076L : inexact-ok
+= erf upward ldbl-128 0x4p-1076L : 0x4.8375d410a6db446b8ea453fb5ff8p-1076L : inexact-ok
+= erf downward ldbl-128ibm 0x4p-1076L : 0x4p-1076L : inexact-ok underflow errno-erange-ok
+= erf tonearest ldbl-128ibm 0x4p-1076L : 0x4p-1076L : inexact-ok underflow errno-erange-ok
+= erf towardzero ldbl-128ibm 0x4p-1076L : 0x4p-1076L : inexact-ok underflow errno-erange-ok
+= erf upward ldbl-128ibm 0x4p-1076L : 0x8p-1076L : inexact-ok underflow errno-erange-ok
+= erf downward ldbl-96-intel 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000028p-16384L : inexact-ok
+= erf tonearest ldbl-96-intel 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000028p-16384L : inexact-ok
+= erf towardzero ldbl-96-intel 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000028p-16384L : inexact-ok
+= erf upward ldbl-96-intel 0x3.8b7f12369ded5518p-16384L : 0x4.000000000000003p-16384L : inexact-ok
+= erf downward ldbl-96-m68k 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000028p-16384L : inexact-ok
+= erf tonearest ldbl-96-m68k 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000028p-16384L : inexact-ok
+= erf towardzero ldbl-96-m68k 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000028p-16384L : inexact-ok
+= erf upward ldbl-96-m68k 0x3.8b7f12369ded5518p-16384L : 0x4.000000000000003p-16384L : inexact-ok
+= erf downward ldbl-128 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000029274014aceae4p-16384L : inexact-ok
+= erf tonearest ldbl-128 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000029274014aceae4p-16384L : inexact-ok
+= erf towardzero ldbl-128 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000029274014aceae4p-16384L : inexact-ok
+= erf upward ldbl-128 0x3.8b7f12369ded5518p-16384L : 0x4.0000000000000029274014aceae8p-16384L : inexact-ok
 erfc 0.0
 = erfc downward flt-32 0x0p+0f : 0x1p+0f : inexact-ok
 = erfc tonearest flt-32 0x0p+0f : 0x1p+0f : inexact-ok

diff --git a/sysdeps/ieee754/dbl-64/s_erf.c b/sysdeps/ieee754/dbl-64/s_erf.c
index 3f37397224fc78e2df2c8e3a05d3e2666ce48ea6..ea0a73424ecf325b13595b99741859ae25e7c48f 100644 (file)
--- a/sysdeps/ieee754/dbl-64/s_erf.c
+++ b/sysdeps/ieee754/dbl-64/s_erf.c
@@ -128,7 +128,6 @@ static const double
  * Coefficients for approximation to  erf on [0,0.84375]
  */
   efx = 1.28379167095512586316e-01, /* 0x3FC06EBA, 0x8214DB69 */
-  efx8 = 1.02703333676410069053e+00, /* 0x3FF06EBA, 0x8214DB69 */
   pp[] = { 1.28379167095512558561e-01, /* 0x3FC06EBA, 0x8214DB68 */
           -3.25042107247001499370e-01, /* 0xBFD4CD7D, 0x691CB913 */
           -2.84817495755985104766e-02, /* 0xBF9D2A51, 0xDBD7194F */
@@ -211,7 +210,16 @@ __erf (double x)
       if (ix < 0x3e300000)              /* |x|<2**-28 */
        {
          if (ix < 0x00800000)
-           return 0.125 * (8.0 * x + efx8 * x);  /*avoid underflow */
+           {
+             /* Avoid spurious underflow.  */
+             double ret = 0.0625 * (16.0 * x + (16.0 * efx) * x);
+             if (fabs (ret) < DBL_MIN)
+               {
+                 double force_underflow = ret * ret;
+                 math_force_eval (force_underflow);
+               }
+             return ret;
+           }
          return x + efx * x;
        }
       z = x * x;

diff --git a/sysdeps/ieee754/flt-32/s_erff.c b/sysdeps/ieee754/flt-32/s_erff.c
index 7c0958964850cad36161851480d97c774cd71a0f..d23cff9293894a5ebe51a7925a5eb34b579041f7 100644 (file)
--- a/sysdeps/ieee754/flt-32/s_erff.c
+++ b/sysdeps/ieee754/flt-32/s_erff.c
@@ -33,7 +33,6 @@ erx =  8.4506291151e-01, /* 0x3f58560b */
  * Coefficients for approximation to  erf on [0,0.84375]
  */
 efx =  1.2837916613e-01, /* 0x3e0375d4 */
-efx8=  1.0270333290e+00, /* 0x3f8375d4 */
 pp0  =  1.2837916613e-01, /* 0x3e0375d4 */
 pp1  = -3.2504209876e-01, /* 0xbea66beb */
 pp2  = -2.8481749818e-02, /* 0xbce9528f */
@@ -111,8 +110,16 @@ float __erff(float x)
        if(ix < 0x3f580000) {           /* |x|<0.84375 */
            if(ix < 0x31800000) {       /* |x|<2**-28 */
                if (ix < 0x04000000)
-                   /*avoid underflow */
-                   return (float)0.125*((float)8.0*x+efx8*x);
+                 {
+                   /* Avoid spurious underflow.  */
+                   float ret = 0.0625f * (16.0f * x + (16.0f * efx) * x);
+                   if (fabsf (ret) < FLT_MIN)
+                     {
+                       float force_underflow = ret * ret;
+                       math_force_eval (force_underflow);
+                     }
+                   return ret;
+                 }
                return x + efx*x;
            }
            z = x*x;

diff --git a/sysdeps/ieee754/ldbl-128/s_erfl.c b/sysdeps/ieee754/ldbl-128/s_erfl.c
index ef65ed89229f97d33a9ea7fe49d696d493f69558..35ca8c1360a9d07c2d75d3acdae2781ced7dc4ef 100644 (file)
--- a/sysdeps/ieee754/ldbl-128/s_erfl.c
+++ b/sysdeps/ieee754/ldbl-128/s_erfl.c
@@ -97,6 +97,7 @@
  */
 
 #include <errno.h>
+#include <float.h>
 #include <math.h>
 #include <math_private.h>
 
@@ -143,9 +144,7 @@ tiny = 1e-4931L,
   one = 1.0L,
   two = 2.0L,
   /* 2/sqrt(pi) - 1 */
-  efx = 1.2837916709551257389615890312154517168810E-1L,
-  /* 8 * (2/sqrt(pi) - 1) */
-  efx8 = 1.0270333367641005911692712249723613735048E0L;
+  efx = 1.2837916709551257389615890312154517168810E-1L;
 
 
 /* erf(x)  = x  + x R(x^2)
@@ -782,7 +781,16 @@ __erfl (long double x)
       if (ix < 0x3fc60000) /* |x|<2**-57 */
        {
          if (ix < 0x00080000)
-           return 0.125 * (8.0 * x + efx8 * x);        /*avoid underflow */
+           {
+             /* Avoid spurious underflow.  */
+             long double ret =  0.0625 * (16.0 * x + (16.0 * efx) * x);
+             if (fabsl (ret) < LDBL_MIN)
+               {
+                 long double force_underflow = ret * ret;
+                 math_force_eval (force_underflow);
+               }
+             return ret;
+           }
          return x + efx * x;
        }
       y = a + a * neval (z, TN1, NTN1) / deval (z, TD1, NTD1);

diff --git a/sysdeps/ieee754/ldbl-128ibm/s_erfl.c b/sysdeps/ieee754/ldbl-128ibm/s_erfl.c
index 95dc4158451cdc558a1dff24300bc1c448977950..f55e8b7879d0439e61348a11a6965294a8b496b0 100644 (file)
--- a/sysdeps/ieee754/ldbl-128ibm/s_erfl.c
+++ b/sysdeps/ieee754/ldbl-128ibm/s_erfl.c
@@ -102,6 +102,7 @@
  */
 
 #include <errno.h>
+#include <float.h>
 #include <math.h>
 #include <math_private.h>
 #include <math_ldbl_opt.h>
@@ -149,9 +150,7 @@ tiny = 1e-300L,
   one = 1.0L,
   two = 2.0L,
   /* 2/sqrt(pi) - 1 */
-  efx = 1.2837916709551257389615890312154517168810E-1L,
-  /* 8 * (2/sqrt(pi) - 1) */
-  efx8 = 1.0270333367641005911692712249723613735048E0L;
+  efx = 1.2837916709551257389615890312154517168810E-1L;
 
 
 /* erf(x)  = x  + x R(x^2)
@@ -801,10 +800,17 @@ __erfl (long double x)
          if (ix < 0x00800000)
            {
              /* erf (-0) = -0.  Unfortunately, for IBM extended double
-                0.125 * (8.0 * x + efx8 * x) for x = -0 evaluates to 0.  */
+                0.0625 * (16.0 * x + (16.0 * efx) * x) for x = -0
+                evaluates to 0.  */
              if (x == 0)
                return x;
-             return 0.125 * (8.0 * x + efx8 * x);      /*avoid underflow */
+             long double ret = 0.0625 * (16.0 * x + (16.0 * efx) * x);
+             if (fabsl (ret) < LDBL_MIN)
+               {
+                 long double force_underflow = ret * ret;
+                 math_force_eval (force_underflow);
+               }
+             return ret;
            }
          return x + efx * x;
        }

diff --git a/sysdeps/ieee754/ldbl-96/s_erfl.c b/sysdeps/ieee754/ldbl-96/s_erfl.c
index 47e4b9e9099ddb70717836e760d23db6682e18ce..c27de812ca23a7f26aeb32cc5cc2b1116b7bf34c 100644 (file)
--- a/sysdeps/ieee754/ldbl-96/s_erfl.c
+++ b/sysdeps/ieee754/ldbl-96/s_erfl.c
@@ -105,6 +105,7 @@
 
 
 #include <errno.h>
+#include <float.h>
 #include <math.h>
 #include <math_private.h>
 
@@ -120,8 +121,6 @@ tiny = 1e-4931L,
  */
   /* 2/sqrt(pi) - 1 */
   efx = 1.2837916709551257389615890312154517168810E-1L,
-  /* 8 * (2/sqrt(pi) - 1) */
-  efx8 = 1.0270333367641005911692712249723613735048E0L,
 
   pp[6] = {
     1.122751350964552113068262337278335028553E6L,
@@ -272,7 +271,16 @@ __erfl (long double x)
       if (ix < 0x3fde8000) /* |x|<2**-33 */
        {
          if (ix < 0x00080000)
-           return 0.125 * (8.0 * x + efx8 * x);        /*avoid underflow */
+           {
+             /* Avoid spurious underflow.  */
+             long double ret = 0.0625 * (16.0 * x + (16.0 * efx) * x);
+             if (fabsl (ret) < LDBL_MIN)
+               {
+                 long double force_underflow = ret * ret;
+                 math_force_eval (force_underflow);
+               }
+             return ret;
+           }
          return x + efx * x;
        }
       z = x * x;