Home
last modified time | relevance | path

Searched refs:ulp (Results 1 – 18 of 18) sorted by relevance

/glibc-2.36/math/
Dlibm-test-support.c190 compare_ulp_data (const void *key, const void *ulp) in compare_ulp_data() argument
193 const struct ulp_data *ulpdat = ulp; in compare_ulp_data()
310 print_function_ulps (const char *function_name, FLOAT ulp) in print_function_ulps() argument
315 FTOSTR (ustrn, FSTR_MAX, "%.0f", FUNC (ceil) (ulp)); in print_function_ulps()
576 (FUNC(fabs) ((given) - (expected)) / ulp (expected))
580 ulp (FLOAT value) in ulp() function
582 FLOAT ulp; in ulp() local
597 ulp = FUNC(ldexp) (1.0, MIN_EXP - MANT_DIG); in ulp()
601 ulp = FUNC(ldexp) (1.0, FUNC(ilogb) (value) - MANT_DIG + 1); in ulp()
609 return ulp; in ulp()
[all …]
Dgen-libm-test.py612 ulp = p_ulps.normal[func][flt]
613 ulp_str = str(ulp) if ulp else '-'
DREADME.libm-test72 ./testrun.sh math/test-double -u --ignore-max-ulp=yes
Dgen-fromfp-tests-inputs203 /* 1, 1.5, 2, 2.5, 3, 3.5 +/- 1ulp. */
Dlibm-test-roundeven.inc148 /* 0.5, 1.5, 2.5, 3.5 +/- 1ulp. */
Dauto-libm-test-in6042 # Values +/- 10ulp from overflow threshold. (Values very close to
6044 # where a result inaccurate by a few ulp could differ from the ideal
6045 # result in whether it overflows; +/- 10ulp is sufficient for overflow
6313 # Integers +/- 1ulp for binary128 (gen-auto-libm-tests will round these
6314 # to produce integers +/- 1ulp for other formats).
6437 # +/- 1ulp.
Dlibm-test-ufromfpx.inc3570 /* 1, 1.5, 2, 2.5, 3, 3.5 +/- 1ulp. */
Dlibm-test-ufromfp.inc3570 /* 1, 1.5, 2, 2.5, 3, 3.5 +/- 1ulp. */
Dlibm-test-fromfp.inc5200 /* 1, 1.5, 2, 2.5, 3, 3.5 +/- 1ulp. */
Dlibm-test-fromfpx.inc5200 /* 1, 1.5, 2, 2.5, 3, 3.5 +/- 1ulp. */
/glibc-2.36/sunrpc/
Dxdr_intXX_t.c115 xdr_uint32_t (XDR *xdrs, uint32_t *ulp) in libc_hidden_nolink_sunrpc()
120 return XDR_PUTINT32 (xdrs, (int32_t *) ulp); in libc_hidden_nolink_sunrpc()
122 return XDR_GETINT32 (xdrs, (int32_t *) ulp); in libc_hidden_nolink_sunrpc()
Dxdr.c209 xdr_u_long (XDR *xdrs, u_long *ulp) in libc_hidden_def()
220 *ulp = (uint32_t) tmp; in libc_hidden_def()
226 && (uint32_t) *ulp != *ulp) in libc_hidden_def()
229 return XDR_PUTLONG (xdrs, (long *) ulp); in libc_hidden_def()
/glibc-2.36/manual/
Dmath.texi1315 an infinite-precision result that is within a few ulp (in both real
1331 within 0.5ulp of the value actually passed, where ``ulp'' is
1340 the error bound being the same as that for division (3ulp).
1354 If the mathematical result is more than a few ulp above the overflow
/glibc-2.36/
DNEWS3353 [21094] math: cosf(1.57079697) has 3 ulp error on targets where the
3356 [21112] math: powf has large ulp errors with base close to 1 and exponent
/glibc-2.36/ChangeLog.old/
DChangeLog.104543 (parse_opt): Parse --ignore-max-ulp.
4603 * math/gen-libm-test.pl (parse_ulps): Ignore 0 has ulp.
DChangeLog.132180 * sysdeps/i386/fpu/libm-test-ulps: Adjust ulp for yn for float.
12268 * math/libm-test.inc (print_ulps): Print ulp values rounded to
DChangeLog.189400 (ulp): Likewise.
10240 (ulp): Likewise.
27604 errors up to 0.5 ulp.
39481 * sysdeps/x86_64/fpu/libm-test-ulp: Likewise.
66100 (ulp): Use MIN_EXP - MANT_DIG.
66101 (check_ulp): Verify subnormal ulps. Only allow a range of +/- 1 ulp.
66293 (ulp): New function.
66296 (check_ulp): Test ulp() implementation.
77866 (MPNSIZE): Increase to represent 10^n where 2^-n is 1/4 ulp of the
77869 lie within an exact representation of 1/2 ulp of the result.
[all …]
DChangeLog.162265 * math/libm-test.inc (check_float_internal): Allow ulp <= 0.5.