diff options
| author | Adhemerval Zanella <adhemerval.zanella@linaro.org> | 2024-12-20 13:18:32 -0300 |
|---|---|---|
| committer | Adhemerval Zanella <adhemerval.zanella@linaro.org> | 2025-02-12 16:31:57 -0300 |
| commit | 95a01ea9551f09c5e6accd07105bac360f2d38e6 (patch) | |
| tree | e7fdad0a78404c49b4c345699d3d51a61db851e7 /math | |
| parent | 1cd9ccd8c0b23a3106f08f6e29bcbb1f57d3334c (diff) | |
| download | glibc-95a01ea9551f09c5e6accd07105bac360f2d38e6.tar.xz glibc-95a01ea9551f09c5e6accd07105bac360f2d38e6.zip | |
math: Use atanpif from CORE-MATH
The CORE-MATH implementation is correctly rounded (for any rounding mode)
and shows better performance to the generic atanpif.
The code was adapted to glibc style and to use the definition of
math_config.h (to handle errno, overflow, and underflow).
Benchtest on x64_64 (Ryzen 9 5900X, gcc 14.2.1), aarch64 (Neoverse-N1,
gcc 13.3.1), and powerpc (POWER10, gcc 13.2.1):
latency master patched improvement
x86_64 66.3296 52.7558 20.46%
x86_64v2 66.0429 51.4007 22.17%
x86_64v3 60.6294 48.7876 19.53%
aarch64 (Neoverse) 24.3163 20.9110 14.00%
power8 16.5766 13.3620 19.39%
power10 16.5115 13.4072 18.80%
reciprocal-throughput master patched improvement
x86_64 30.8599 16.0866 47.87%
x86_64v2 29.2286 15.4688 47.08%
x86_64v3 23.0960 12.8510 44.36%
aarch64 (Neoverse) 15.4619 10.6752 30.96%
power8 7.9200 5.2483 33.73%
power10 6.8539 4.6262 32.50%
Reviewed-by: DJ Delorie <dj@redhat.com>
Diffstat (limited to 'math')
| -rw-r--r-- | math/auto-libm-test-in | 1 | ||||
| -rw-r--r-- | math/auto-libm-test-out-atanpi | 25 |
2 files changed, 26 insertions, 0 deletions
diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in index 4f194da19d..7dd9cde9c9 100644 --- a/math/auto-libm-test-in +++ b/math/auto-libm-test-in @@ -936,6 +936,7 @@ atanpi -0x3.eb8e18p+0 atanpi 0x3.53c188p+0 atanpi -0x1.58c83p+0 atanpi 0x1.626772p-1 +atanpi 0x1p-126 atanpi min atanpi -min atanpi min_subnorm diff --git a/math/auto-libm-test-out-atanpi b/math/auto-libm-test-out-atanpi index 9f9ded81d1..4d3454b4ae 100644 --- a/math/auto-libm-test-out-atanpi +++ b/math/auto-libm-test-out-atanpi @@ -1061,6 +1061,31 @@ atanpi 0x1.626772p-1 = atanpi tonearest ibm128 0xb.133b9p-4 : 0x3.156898f544dca52f89f818e127p-4 : inexact-ok = atanpi towardzero ibm128 0xb.133b9p-4 : 0x3.156898f544dca52f89f818e127p-4 : inexact-ok = atanpi upward ibm128 0xb.133b9p-4 : 0x3.156898f544dca52f89f818e128p-4 : inexact-ok +atanpi 0x1p-126 += atanpi downward binary32 0x4p-128 : 0x1.45f3p-128 : inexact-ok underflow errno-erange-ok += atanpi tonearest binary32 0x4p-128 : 0x1.45f308p-128 : inexact-ok underflow errno-erange-ok += atanpi towardzero binary32 0x4p-128 : 0x1.45f3p-128 : inexact-ok underflow errno-erange-ok += atanpi upward binary32 0x4p-128 : 0x1.45f308p-128 : inexact-ok underflow errno-erange-ok += atanpi downward binary64 0x4p-128 : 0x1.45f306dc9c882p-128 : inexact-ok += atanpi tonearest binary64 0x4p-128 : 0x1.45f306dc9c883p-128 : inexact-ok += atanpi towardzero binary64 0x4p-128 : 0x1.45f306dc9c882p-128 : inexact-ok += atanpi upward binary64 0x4p-128 : 0x1.45f306dc9c883p-128 : inexact-ok += atanpi downward intel96 0x4p-128 : 0x1.45f306dc9c882a52p-128 : inexact-ok += atanpi tonearest intel96 0x4p-128 : 0x1.45f306dc9c882a54p-128 : inexact-ok += atanpi towardzero intel96 0x4p-128 : 0x1.45f306dc9c882a52p-128 : inexact-ok += atanpi upward intel96 0x4p-128 : 0x1.45f306dc9c882a54p-128 : inexact-ok += atanpi downward m68k96 0x4p-128 : 0x1.45f306dc9c882a52p-128 : inexact-ok += atanpi tonearest m68k96 0x4p-128 : 0x1.45f306dc9c882a54p-128 : inexact-ok += atanpi towardzero m68k96 0x4p-128 : 0x1.45f306dc9c882a52p-128 : inexact-ok += atanpi upward m68k96 0x4p-128 : 0x1.45f306dc9c882a54p-128 : inexact-ok += atanpi downward binary128 0x4p-128 : 0x1.45f306dc9c882a53f84eafa3ea69p-128 : inexact-ok += atanpi tonearest binary128 0x4p-128 : 0x1.45f306dc9c882a53f84eafa3ea6ap-128 : inexact-ok += atanpi towardzero binary128 0x4p-128 : 0x1.45f306dc9c882a53f84eafa3ea69p-128 : inexact-ok += atanpi upward binary128 0x4p-128 : 0x1.45f306dc9c882a53f84eafa3ea6ap-128 : inexact-ok += atanpi downward ibm128 0x4p-128 : 0x1.45f306dc9c882a53f84eafa3eap-128 : inexact-ok += atanpi tonearest ibm128 0x4p-128 : 0x1.45f306dc9c882a53f84eafa3ea8p-128 : inexact-ok += atanpi towardzero ibm128 0x4p-128 : 0x1.45f306dc9c882a53f84eafa3eap-128 : inexact-ok += atanpi upward ibm128 0x4p-128 : 0x1.45f306dc9c882a53f84eafa3ea8p-128 : inexact-ok atanpi min = atanpi downward binary32 0x4p-128 : 0x1.45f3p-128 : inexact-ok underflow errno-erange-ok = atanpi tonearest binary32 0x4p-128 : 0x1.45f308p-128 : inexact-ok underflow errno-erange-ok |