Re: regression in Newlib 4.2.0 (lgammaf)
Andoni Arregi <[email protected]>
| Newsgroups | gmane.comp.lib.newlib |
|---|---|
| Organization | GTD GmbH |
| Message-ID | <[email protected]> |
Dear Paul, here again, although we can provide a fix, we cannot understand well enough why this didn't seem to be present in Newlib 4.1.0. The source code regarding the incorrect limit at |x|<2**-70 has been there since decades. Best regards On Sat, 2022-02-05 at 07:38 +0100, Paul Zimmermann wrote: > Hi, > > a third regression in Newlib 4.2.0 for single precision: > > Checking lgamma with newlib-4.2.0.20211231 > MPFR library: 4.1.0 > MPFR header: 4.1.0 (based on 4.1.0) > Checking function mylgammaf with MPFR_RNDN > libm wrong by up to 8.93e+43 ulp(s) for x=-0x1p-70 > mylgamma gives inf > mpfr_mylgamma gives 0x1.842994p+5 > > The largest error for lgammaf in Newlib 4.1.0 was 7.50e6 ulps > (for x=-0x1.3a7fcap+1). > > Paul -- Andoni Arregi Geschäftsführer GTD GmbH Ravensburger Str. 32a, 88677 Markdorf T: +49 7544 96440 22 | M: +49 151 65620499 | F: +49 7544 96440 29 http://www.gtd-gmbh.de [email protected]