diff options
author | Joseph Myers <joseph@codesourcery.com> | 2015-09-28 22:11:22 +0000 |
---|---|---|
committer | Joseph Myers <joseph@codesourcery.com> | 2015-09-28 22:11:22 +0000 |
commit | a5721ebc68cc8b761b0ffdf6dcae52723369e071 (patch) | |
tree | e49b70539a6b7390f60cac5dc2330dae7dd39a1d /sysdeps/ieee754/dbl-64 | |
parent | 60cf80f09d029257caedc0c8abe7e3e09c64e6c7 (diff) | |
download | glibc-a5721ebc68cc8b761b0ffdf6dcae52723369e071.tar.gz glibc-a5721ebc68cc8b761b0ffdf6dcae52723369e071.tar.xz glibc-a5721ebc68cc8b761b0ffdf6dcae52723369e071.zip |
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/ieee754/dbl-64')
-rw-r--r-- | sysdeps/ieee754/dbl-64/x2y2m1.c | 22 | ||||
-rw-r--r-- | sysdeps/ieee754/dbl-64/x2y2m1f.c | 4 |
2 files changed, 10 insertions, 16 deletions
diff --git a/sysdeps/ieee754/dbl-64/x2y2m1.c b/sysdeps/ieee754/dbl-64/x2y2m1.c index c96dae59e1..b040097d18 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 43a8acf8a8..835f6a0e45 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) |