aboutsummaryrefslogtreecommitdiff
path: root/sysdeps
diff options
context:
space:
mode:
authorJoseph Myers <joseph@codesourcery.com>2015-09-28 22:11:22 +0000
committerJoseph Myers <joseph@codesourcery.com>2015-09-28 22:11:22 +0000
commita5721ebc68cc8b761b0ffdf6dcae52723369e071 (patch)
treee49b70539a6b7390f60cac5dc2330dae7dd39a1d /sysdeps
parent60cf80f09d029257caedc0c8abe7e3e09c64e6c7 (diff)
downloadglibc-a5721ebc68cc8b761b0ffdf6dcae52723369e071.zip
glibc-a5721ebc68cc8b761b0ffdf6dcae52723369e071.tar.gz
glibc-a5721ebc68cc8b761b0ffdf6dcae52723369e071.tar.bz2
Fix clog, clog10 inaccuracy (bug 19016).
For arguments with X^2 + Y^2 close to 1, clog and clog10 avoid large errors from log(hypot) by computing X^2 + Y^2 - 1 in a way that avoids cancellation error and then using log1p. However, the thresholds for using that approach still result in log being used on argument as large as sqrt(13/16) > 0.9, leading to significant errors, in some cases above the 9ulp maximum allowed in glibc libm. This patch arranges for the approach using log1p to be used in any cases where |X|, |Y| < 1 and X^2 + Y^2 >= 0.5 (with the existing allowance for cases where one of X and Y is very small), adjusting the __x2y2m1 functions to work with the wider range of inputs. This way, log only gets used on arguments below sqrt(1/2) (or substantially above 1), where the error involved is much less. Tested for x86_64, x86, mips64 and powerpc. For the ulps regeneration I removed the existing clog and clog10 ulps before regenerating to allow any reduced ulps to appear. Tests added include those found by random test generation to produce large ulps either before or after the patch, and some found by trying inputs close to the (0.75, 0.5) threshold where the potential errors from using log are largest. [BZ #19016] * sysdeps/generic/math_private.h (__x2y2m1f): Update comment to allow more cases with X^2 + Y^2 >= 0.5. * sysdeps/ieee754/dbl-64/x2y2m1.c (__x2y2m1): Likewise. Add -1 as normal element in sum instead of special-casing based on values of arguments. * sysdeps/ieee754/dbl-64/x2y2m1f.c (__x2y2m1f): Update comment. * sysdeps/ieee754/ldbl-128/x2y2m1l.c (__x2y2m1l): Likewise. Add -1 as normal element in sum instead of special-casing based on values of arguments. * sysdeps/ieee754/ldbl-128ibm/x2y2m1l.c (__x2y2m1l): Likewise. * sysdeps/ieee754/ldbl-96/x2y2m1.c [FLT_EVAL_METHOD != 0] (__x2y2m1): Update comment. * sysdeps/ieee754/ldbl-96/x2y2m1l.c (__x2y2m1l): Likewise. Add -1 as normal element in sum instead of special-casing based on values of arguments. * math/s_clog.c (__clog): Handle more cases using log1p without hypot. * math/s_clog10.c (__clog10): Likewise. * math/s_clog10f.c (__clog10f): Likewise. * math/s_clog10l.c (__clog10l): Likewise. * math/s_clogf.c (__clogf): Likewise. * math/s_clogl.c (__clogl): Likewise. * math/auto-libm-test-in: Add more tests of clog and clog10. * math/auto-libm-test-out: Regenerated. * sysdeps/i386/fpu/libm-test-ulps: Update. * sysdeps/x86_64/fpu/libm-test-ulps: Likewise.
Diffstat (limited to 'sysdeps')
-rw-r--r--sysdeps/generic/math_private.h4
-rw-r--r--sysdeps/i386/fpu/libm-test-ulps80
-rw-r--r--sysdeps/ieee754/dbl-64/x2y2m1.c22
-rw-r--r--sysdeps/ieee754/dbl-64/x2y2m1f.c4
-rw-r--r--sysdeps/ieee754/ldbl-128/x2y2m1l.c22
-rw-r--r--sysdeps/ieee754/ldbl-128ibm/x2y2m1l.c24
-rw-r--r--sysdeps/ieee754/ldbl-96/x2y2m1.c4
-rw-r--r--sysdeps/ieee754/ldbl-96/x2y2m1l.c22
-rw-r--r--sysdeps/x86_64/fpu/libm-test-ulps80
9 files changed, 119 insertions, 143 deletions
diff --git a/sysdeps/generic/math_private.h b/sysdeps/generic/math_private.h
index a8f1a8e..cf1865d 100644
--- a/sysdeps/generic/math_private.h
+++ b/sysdeps/generic/math_private.h
@@ -365,8 +365,8 @@ extern double __slowpow (double __x, double __y, double __z);
extern void __docos (double __x, double __dx, double __v[]);
/* Return X^2 + Y^2 - 1, computed without large cancellation error.
- It is given that 1 > X >= Y >= epsilon / 2, and that either X >=
- 0.75 or Y >= 0.5. */
+ It is given that 1 > X >= Y >= epsilon / 2, and that X^2 + Y^2 >=
+ 0.5. */
extern float __x2y2m1f (float x, float y);
extern double __x2y2m1 (double x, double y);
extern long double __x2y2m1l (long double x, long double y);
diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps
index 32f24d0..438a390 100644
--- a/sysdeps/i386/fpu/libm-test-ulps
+++ b/sysdeps/i386/fpu/libm-test-ulps
@@ -836,12 +836,12 @@ ildouble: 3
ldouble: 3
Function: Real part of "clog":
-double: 3
-float: 2
-idouble: 3
-ifloat: 2
-ildouble: 4
-ldouble: 4
+double: 2
+float: 1
+idouble: 2
+ifloat: 1
+ildouble: 3
+ldouble: 3
Function: Imaginary part of "clog":
double: 1
@@ -864,10 +864,10 @@ ildouble: 2
ldouble: 2
Function: Real part of "clog10_downward":
-double: 5
-float: 4
-idouble: 5
-ifloat: 4
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
ildouble: 8
ldouble: 8
@@ -876,14 +876,14 @@ double: 1
float: 1
idouble: 1
ifloat: 1
-ildouble: 2
-ldouble: 2
+ildouble: 3
+ldouble: 3
Function: Real part of "clog10_towardzero":
-double: 5
-float: 4
-idouble: 5
-ifloat: 4
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
ildouble: 8
ldouble: 8
@@ -896,12 +896,12 @@ ildouble: 3
ldouble: 3
Function: Real part of "clog10_upward":
-double: 5
-float: 5
-idouble: 5
-ifloat: 5
-ildouble: 6
-ldouble: 6
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
+ildouble: 7
+ldouble: 7
Function: Imaginary part of "clog10_upward":
double: 1
@@ -912,12 +912,12 @@ ildouble: 3
ldouble: 3
Function: Real part of "clog_downward":
-double: 5
-float: 5
-idouble: 5
-ifloat: 5
-ildouble: 7
-ldouble: 7
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
+ildouble: 5
+ldouble: 5
Function: Imaginary part of "clog_downward":
double: 1
@@ -928,12 +928,12 @@ ildouble: 1
ldouble: 1
Function: Real part of "clog_towardzero":
-double: 5
-float: 5
-idouble: 5
-ifloat: 5
-ildouble: 8
-ldouble: 8
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
+ildouble: 5
+ldouble: 5
Function: Imaginary part of "clog_towardzero":
double: 1
@@ -944,12 +944,12 @@ ildouble: 1
ldouble: 1
Function: Real part of "clog_upward":
-double: 5
-float: 5
-idouble: 5
-ifloat: 5
-ildouble: 6
-ldouble: 6
+double: 2
+float: 3
+idouble: 2
+ifloat: 3
+ildouble: 4
+ldouble: 4
Function: Imaginary part of "clog_upward":
double: 1
diff --git a/sysdeps/ieee754/dbl-64/x2y2m1.c b/sysdeps/ieee754/dbl-64/x2y2m1.c
index c96dae5..b040097 100644
--- a/sysdeps/ieee754/dbl-64/x2y2m1.c
+++ b/sysdeps/ieee754/dbl-64/x2y2m1.c
@@ -80,32 +80,26 @@ compare (const void *p, const void *q)
}
/* Return X^2 + Y^2 - 1, computed without large cancellation error.
- It is given that 1 > X >= Y >= epsilon / 2, and that either X >=
- 0.75 or Y >= 0.5. */
+ It is given that 1 > X >= Y >= epsilon / 2, and that X^2 + Y^2 >=
+ 0.5. */
double
__x2y2m1 (double x, double y)
{
- double vals[4];
+ double vals[5];
SET_RESTORE_ROUND (FE_TONEAREST);
mul_split (&vals[1], &vals[0], x, x);
mul_split (&vals[3], &vals[2], y, y);
- if (x >= 0.75)
- vals[1] -= 1.0;
- else
- {
- vals[1] -= 0.5;
- vals[3] -= 0.5;
- }
- qsort (vals, 4, sizeof (double), compare);
+ vals[4] = -1.0;
+ qsort (vals, 5, sizeof (double), compare);
/* Add up the values so that each element of VALS has absolute value
at most equal to the last set bit of the next nonzero
element. */
- for (size_t i = 0; i <= 2; i++)
+ for (size_t i = 0; i <= 3; i++)
{
add_split (&vals[i + 1], &vals[i], vals[i + 1], vals[i]);
- qsort (vals + i + 1, 3 - i, sizeof (double), compare);
+ qsort (vals + i + 1, 4 - i, sizeof (double), compare);
}
/* Now any error from this addition will be small. */
- return vals[3] + vals[2] + vals[1] + vals[0];
+ return vals[4] + vals[3] + vals[2] + vals[1] + vals[0];
}
diff --git a/sysdeps/ieee754/dbl-64/x2y2m1f.c b/sysdeps/ieee754/dbl-64/x2y2m1f.c
index 43a8acf..835f6a0 100644
--- a/sysdeps/ieee754/dbl-64/x2y2m1f.c
+++ b/sysdeps/ieee754/dbl-64/x2y2m1f.c
@@ -21,8 +21,8 @@
#include <float.h>
/* Return X^2 + Y^2 - 1, computed without large cancellation error.
- It is given that 1 > X >= Y >= epsilon / 2, and that either X >=
- 0.75 or Y >= 0.5. */
+ It is given that 1 > X >= Y >= epsilon / 2, and that X^2 + Y^2 >=
+ 0.5. */
float
__x2y2m1f (float x, float y)
diff --git a/sysdeps/ieee754/ldbl-128/x2y2m1l.c b/sysdeps/ieee754/ldbl-128/x2y2m1l.c
index 11757c6..a0498c3 100644
--- a/sysdeps/ieee754/ldbl-128/x2y2m1l.c
+++ b/sysdeps/ieee754/ldbl-128/x2y2m1l.c
@@ -80,32 +80,26 @@ compare (const void *p, const void *q)
}
/* Return X^2 + Y^2 - 1, computed without large cancellation error.
- It is given that 1 > X >= Y >= epsilon / 2, and that either X >=
- 0.75 or Y >= 0.5. */
+ It is given that 1 > X >= Y >= epsilon / 2, and that X^2 + Y^2 >=
+ 0.5. */
long double
__x2y2m1l (long double x, long double y)
{
- long double vals[4];
+ long double vals[5];
SET_RESTORE_ROUNDL (FE_TONEAREST);
mul_split (&vals[1], &vals[0], x, x);
mul_split (&vals[3], &vals[2], y, y);
- if (x >= 0.75L)
- vals[1] -= 1.0L;
- else
- {
- vals[1] -= 0.5L;
- vals[3] -= 0.5L;
- }
- qsort (vals, 4, sizeof (long double), compare);
+ vals[4] = -1.0L;
+ qsort (vals, 5, sizeof (long double), compare);
/* Add up the values so that each element of VALS has absolute value
at most equal to the last set bit of the next nonzero
element. */
- for (size_t i = 0; i <= 2; i++)
+ for (size_t i = 0; i <= 3; i++)
{
add_split (&vals[i + 1], &vals[i], vals[i + 1], vals[i]);
- qsort (vals + i + 1, 3 - i, sizeof (long double), compare);
+ qsort (vals + i + 1, 4 - i, sizeof (long double), compare);
}
/* Now any error from this addition will be small. */
- return vals[3] + vals[2] + vals[1] + vals[0];
+ return vals[4] + vals[3] + vals[2] + vals[1] + vals[0];
}
diff --git a/sysdeps/ieee754/ldbl-128ibm/x2y2m1l.c b/sysdeps/ieee754/ldbl-128ibm/x2y2m1l.c
index a001b58..081fb98 100644
--- a/sysdeps/ieee754/ldbl-128ibm/x2y2m1l.c
+++ b/sysdeps/ieee754/ldbl-128ibm/x2y2m1l.c
@@ -80,13 +80,13 @@ compare (const void *p, const void *q)
}
/* Return X^2 + Y^2 - 1, computed without large cancellation error.
- It is given that 1 > X >= Y >= epsilon / 2, and that either X >=
- 0.75 or Y >= 0.5. */
+ It is given that 1 > X >= Y >= epsilon / 2, and that X^2 + Y^2 >=
+ 0.5. */
long double
__x2y2m1l (long double x, long double y)
{
- double vals[12];
+ double vals[13];
SET_RESTORE_ROUND (FE_TONEAREST);
union ibm_extended_long_double xu, yu;
xu.ld = x;
@@ -105,25 +105,19 @@ __x2y2m1l (long double x, long double y)
vals[8] *= 2.0;
vals[9] *= 2.0;
mul_split (&vals[11], &vals[10], yu.d[1].d, yu.d[1].d);
- if (xu.d[0].d >= 0.75)
- vals[1] -= 1.0;
- else
- {
- vals[1] -= 0.5;
- vals[7] -= 0.5;
- }
- qsort (vals, 12, sizeof (double), compare);
+ vals[12] = -1.0;
+ qsort (vals, 13, sizeof (double), compare);
/* Add up the values so that each element of VALS has absolute value
at most equal to the last set bit of the next nonzero
element. */
- for (size_t i = 0; i <= 10; i++)
+ for (size_t i = 0; i <= 11; i++)
{
add_split (&vals[i + 1], &vals[i], vals[i + 1], vals[i]);
- qsort (vals + i + 1, 11 - i, sizeof (double), compare);
+ qsort (vals + i + 1, 12 - i, sizeof (double), compare);
}
/* Now any error from this addition will be small. */
- long double retval = (long double) vals[11];
- for (size_t i = 10; i != (size_t) -1; i--)
+ long double retval = (long double) vals[12];
+ for (size_t i = 11; i != (size_t) -1; i--)
retval += (long double) vals[i];
return retval;
}
diff --git a/sysdeps/ieee754/ldbl-96/x2y2m1.c b/sysdeps/ieee754/ldbl-96/x2y2m1.c
index a6cc82c..2f6b0be 100644
--- a/sysdeps/ieee754/ldbl-96/x2y2m1.c
+++ b/sysdeps/ieee754/ldbl-96/x2y2m1.c
@@ -27,8 +27,8 @@
#else
/* Return X^2 + Y^2 - 1, computed without large cancellation error.
- It is given that 1 > X >= Y >= epsilon / 2, and that either X >=
- 0.75 or Y >= 0.5. */
+ It is given that 1 > X >= Y >= epsilon / 2, and that X^2 + Y^2 >=
+ 0.5. */
double
__x2y2m1 (double x, double y)
diff --git a/sysdeps/ieee754/ldbl-96/x2y2m1l.c b/sysdeps/ieee754/ldbl-96/x2y2m1l.c
index 11757c6..a0498c3 100644
--- a/sysdeps/ieee754/ldbl-96/x2y2m1l.c
+++ b/sysdeps/ieee754/ldbl-96/x2y2m1l.c
@@ -80,32 +80,26 @@ compare (const void *p, const void *q)
}
/* Return X^2 + Y^2 - 1, computed without large cancellation error.
- It is given that 1 > X >= Y >= epsilon / 2, and that either X >=
- 0.75 or Y >= 0.5. */
+ It is given that 1 > X >= Y >= epsilon / 2, and that X^2 + Y^2 >=
+ 0.5. */
long double
__x2y2m1l (long double x, long double y)
{
- long double vals[4];
+ long double vals[5];
SET_RESTORE_ROUNDL (FE_TONEAREST);
mul_split (&vals[1], &vals[0], x, x);
mul_split (&vals[3], &vals[2], y, y);
- if (x >= 0.75L)
- vals[1] -= 1.0L;
- else
- {
- vals[1] -= 0.5L;
- vals[3] -= 0.5L;
- }
- qsort (vals, 4, sizeof (long double), compare);
+ vals[4] = -1.0L;
+ qsort (vals, 5, sizeof (long double), compare);
/* Add up the values so that each element of VALS has absolute value
at most equal to the last set bit of the next nonzero
element. */
- for (size_t i = 0; i <= 2; i++)
+ for (size_t i = 0; i <= 3; i++)
{
add_split (&vals[i + 1], &vals[i], vals[i + 1], vals[i]);
- qsort (vals + i + 1, 3 - i, sizeof (long double), compare);
+ qsort (vals + i + 1, 4 - i, sizeof (long double), compare);
}
/* Now any error from this addition will be small. */
- return vals[3] + vals[2] + vals[1] + vals[0];
+ return vals[4] + vals[3] + vals[2] + vals[1] + vals[0];
}
diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps
index 12c3dd1..ef3ab70 100644
--- a/sysdeps/x86_64/fpu/libm-test-ulps
+++ b/sysdeps/x86_64/fpu/libm-test-ulps
@@ -869,11 +869,11 @@ ldouble: 3
Function: Real part of "clog":
double: 3
-float: 2
+float: 3
idouble: 3
-ifloat: 2
-ildouble: 4
-ldouble: 4
+ifloat: 3
+ildouble: 3
+ldouble: 3
Function: Imaginary part of "clog":
float: 1
@@ -883,9 +883,9 @@ ldouble: 1
Function: Real part of "clog10":
double: 3
-float: 3
+float: 4
idouble: 3
-ifloat: 3
+ifloat: 4
ildouble: 4
ldouble: 4
@@ -898,10 +898,10 @@ ildouble: 2
ldouble: 2
Function: Real part of "clog10_downward":
-double: 6
-float: 6
-idouble: 6
-ifloat: 6
+double: 5
+float: 4
+idouble: 5
+ifloat: 4
ildouble: 8
ldouble: 8
@@ -910,14 +910,14 @@ double: 2
float: 4
idouble: 2
ifloat: 4
-ildouble: 2
-ldouble: 2
+ildouble: 3
+ldouble: 3
Function: Real part of "clog10_towardzero":
double: 5
-float: 4
+float: 5
idouble: 5
-ifloat: 4
+ifloat: 5
ildouble: 8
ldouble: 8
@@ -930,28 +930,28 @@ ildouble: 3
ldouble: 3
Function: Real part of "clog10_upward":
-double: 8
+double: 6
float: 5
-idouble: 8
+idouble: 6
ifloat: 5
-ildouble: 6
-ldouble: 6
+ildouble: 7
+ldouble: 7
Function: Imaginary part of "clog10_upward":
double: 2
-float: 3
+float: 4
idouble: 2
-ifloat: 3
+ifloat: 4
ildouble: 3
ldouble: 3
Function: Real part of "clog_downward":
-double: 7
-float: 5
-idouble: 7
-ifloat: 5
-ildouble: 7
-ldouble: 7
+double: 4
+float: 3
+idouble: 4
+ifloat: 3
+ildouble: 5
+ldouble: 5
Function: Imaginary part of "clog_downward":
double: 1
@@ -962,28 +962,28 @@ ildouble: 1
ldouble: 1
Function: Real part of "clog_towardzero":
-double: 7
-float: 5
-idouble: 7
-ifloat: 5
-ildouble: 8
-ldouble: 8
+double: 4
+float: 4
+idouble: 4
+ifloat: 4
+ildouble: 5
+ldouble: 5
Function: Imaginary part of "clog_towardzero":
double: 1
-float: 2
+float: 3
idouble: 1
-ifloat: 2
+ifloat: 3
ildouble: 1
ldouble: 1
Function: Real part of "clog_upward":
-double: 8
-float: 5
-idouble: 8
-ifloat: 5
-ildouble: 6
-ldouble: 6
+double: 4
+float: 3
+idouble: 4
+ifloat: 3
+ildouble: 4
+ldouble: 4
Function: Imaginary part of "clog_upward":
double: 1