[PATCH] [auto-libm-test-in] added pair of inputs for hypotf in binary32

[PATCH] [auto-libm-test-in] added pair of inputs for hypotf in binary32

[Paul Zimmermann]

This pair yields an error of 1 ulp in binary32, whereas the current
maximal known error for hypotf on x86_64 is zero:

Checking hypot with glibc-2.37
hypot 0 -1 -0x1.003222p-20,-0x1.6a2d58p-32 [0.501] 0.500001 0.500000001392678
libm gives 0x1.003224p-20
mpfr gives 0x1.003222p-20

See https://sourceware.org/pipermail/libc-alpha/2023-February/145432.html
and https://sourceware.org/pipermail/libc-alpha/2023-February/145442.html
---
 math/auto-libm-test-in | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in
index 393d9fcfaa..c400cafb4b 100644
--- a/math/auto-libm-test-in
+++ b/math/auto-libm-test-in
@@ -5773,6 +5773,8 @@ hypot 0x1p-16384 0x1p-16446
 hypot 0x1p-16383 0x1p-16494
 
 hypot -0x1.fa7deap+0 0x1.a761bab383ac8p+0
+# the following inputs yield an error of 1 ulp on x86_64 in binary32
+hypot -0x1.003222p-20 -0x1.6a2d58p-32
 
 j0 -1.0
 j0 0.0
-- 
2.39.1

Re: [PATCH] [auto-libm-test-in] added pair of inputs for hypotf in binary32

[Carlos O'Donell]

On 2/12/23 02:01, Paul Zimmermann via Libc-alpha wrote:
> This pair yields an error of 1 ulp in binary32, whereas the current
> maximal known error for hypotf on x86_64 is zero:
> 
> Checking hypot with glibc-2.37
> hypot 0 -1 -0x1.003222p-20,-0x1.6a2d58p-32 [0.501] 0.500001 0.500000001392678
> libm gives 0x1.003224p-20
> mpfr gives 0x1.003222p-20

LGTM. Passed CI for i686. Early in development is a good time to add more inputs.

Reviewed-by: Carlos O'Donell <carlos@redhat.com>


> See https://sourceware.org/pipermail/libc-alpha/2023-February/145432.html
> and https://sourceware.org/pipermail/libc-alpha/2023-February/145442.html
> ---
>  math/auto-libm-test-in | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in
> index 393d9fcfaa..c400cafb4b 100644
> --- a/math/auto-libm-test-in
> +++ b/math/auto-libm-test-in
> @@ -5773,6 +5773,8 @@ hypot 0x1p-16384 0x1p-16446
>  hypot 0x1p-16383 0x1p-16494
>  
>  hypot -0x1.fa7deap+0 0x1.a761bab383ac8p+0
> +# the following inputs yield an error of 1 ulp on x86_64 in binary32
> +hypot -0x1.003222p-20 -0x1.6a2d58p-32

OK. Double checked results.

Notes:
hypot(-0x1.003222p-20, -0x1.6a2d58p-32) == 0x1.003222fffffffp-20

>  
>  j0 -1.0
>  j0 0.0

-- 
Cheers,
Carlos.