This is conceptually very similar to the strcspn(3) implementations
from D41557, but we can't do the fast paths the same way.
Also like for the strcspn(3) implementation, new unit tests are
provided for strspn(3), superseeding the rudimentary tests from
NetBSD.
Performance is comparable with glibc.
os: FreeBSD
arch: amd64
cpu: 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz
│ strspn.x86-64-v2.out │ strspn.pre.out │ strspn.scalar.out │
│ sec/op │ sec/op vs base │ sec/op vs base │
Short1 66.77µ ± 0% 141.36µ ± 0% +111.70% (p=0.000 n=20) 64.83µ ± 0% -2.91% (p=0.000 n=20)
Mid1 15.59µ ± 0% 79.25µ ± 0% +408.35% (p=0.000 n=20) 30.82µ ± 0% +97.67% (p=0.000 n=20)
Long1 6.030µ ± 0% 48.440µ ± 0% +703.35% (p=0.000 n=20) 21.386µ ± 0% +254.67% (p=0.000 n=20)
Short5 67.11µ ± 0% 182.21µ ± 0% +171.49% (p=0.000 n=20) 127.97µ ± 0% +90.68% (p=0.000 n=20)
Mid5 15.59µ ± 0% 87.71µ ± 0% +462.75% (p=0.000 n=20) 56.11µ ± 0% +260.01% (p=0.000 n=20)
Long5 6.030µ ± 0% 49.626µ ± 1% +723.05% (p=0.000 n=20) 35.214µ ± 0% +484.03% (p=0.000 n=20)
Short20 77.76µ ± 0% 352.34µ ± 1% +353.12% (p=0.000 n=20) 165.16µ ± 0% +112.41% (p=0.000 n=20)
Mid20 26.61µ ± 0% 133.80µ ± 0% +402.84% (p=0.000 n=20) 67.05µ ± 0% +151.97% (p=0.000 n=20)
Long20 14.04µ ± 0% 62.11µ ± 1% +342.24% (p=0.000 n=20) 35.22µ ± 0% +150.81% (p=0.000 n=20)
Short40 171.6µ ± 1% 588.1µ ± 1% +242.74% (p=0.000 n=20) 203.5µ ± 0% +18.61% (p=0.000 n=20)
Mid40 70.10µ ± 1% 197.90µ ± 0% +182.32% (p=0.000 n=20) 74.47µ ± 0% +6.24% (p=0.000 n=20)
Long40 35.18µ ± 0% 51.69µ ± 2% +46.92% (p=0.000 n=20) 35.45µ ± 0% +0.74% (p=0.000 n=20)
geomean 29.78µ 118.4µ +297.52% 60.20µ +102.13%
│ strspn.x86-64-v2.out │ strspn.pre.out │ strspn.scalar.out │
│ B/s │ B/s vs base │ B/s vs base │
Short1 1785.3Mi ± 0% 843.3Mi ± 0% -52.76% (p=0.000 n=20) 1838.7Mi ± 0% +2.99% (p=0.000 n=20)
Mid1 7.467Gi ± 0% 1.469Gi ± 0% -80.33% (p=0.000 n=20) 3.778Gi ± 0% -49.41% (p=0.000 n=20)
Long1 19.307Gi ± 0% 2.403Gi ± 0% -87.55% (p=0.000 n=20) 5.444Gi ± 0% -71.81% (p=0.000 n=20)
Short5 1776.2Mi ± 0% 654.2Mi ± 0% -63.17% (p=0.000 n=20) 931.5Mi ± 0% -47.56% (p=0.000 n=20)
Mid5 7.469Gi ± 0% 1.327Gi ± 0% -82.23% (p=0.000 n=20) 2.075Gi ± 0% -72.22% (p=0.000 n=20)
Long5 19.307Gi ± 0% 2.346Gi ± 1% -87.85% (p=0.000 n=20) 3.306Gi ± 0% -82.88% (p=0.000 n=20)
Short20 1533.1Mi ± 0% 338.3Mi ± 1% -77.93% (p=0.000 n=20) 721.8Mi ± 0% -52.92% (p=0.000 n=20)
Mid20 4480.0Mi ± 0% 890.9Mi ± 0% -80.11% (p=0.000 n=20) 1778.0Mi ± 0% -60.31% (p=0.000 n=20)
Long20 8.290Gi ± 0% 1.874Gi ± 1% -77.39% (p=0.000 n=20) 3.305Gi ± 0% -60.13% (p=0.000 n=20)
Short40 694.7Mi ± 1% 202.7Mi ± 1% -70.82% (p=0.000 n=20) 585.7Mi ± 0% -15.69% (p=0.000 n=20)
Mid40 1700.6Mi ± 1% 602.4Mi ± 0% -64.58% (p=0.000 n=20) 1600.7Mi ± 0% -5.88% (p=0.000 n=20)
Long40 3.309Gi ± 0% 2.252Gi ± 2% -31.93% (p=0.000 n=20) 3.284Gi ± 0% -0.74% (p=0.000 n=20)
geomean 3.909Gi 1006.9Mi -74.84% 1.934Gi -50.53%
os: Linux
arch: x86_64
cpu:
│ strspn.glibc.out │
│ sec/op │
Short1 71.64µ ± 0%
Mid1 18.82µ ± 0%
Long1 6.029µ ± 0%
Short5 71.63µ ± 1%
Mid5 18.82µ ± 0%
Long5 6.029µ ± 0%
Short20 217.1µ ± 0%
Mid20 81.78µ ± 0%
Long20 35.77µ ± 0%
Short40 266.8µ ± 0%
Mid40 95.84µ ± 0%
Long40 35.77µ ± 0%
geomean 42.85µ
│ strspn.glibc.out │
│ B/s │
Short1 1.625Gi ± 0%
Mid1 6.186Gi ± 0%
Long1 19.31Gi ± 0%
Short5 1.625Gi ± 1%
Mid5 6.187Gi ± 0%
Long5 19.31Gi ± 0%
Short20 549.2Mi ± 0%
Mid20 1.423Gi ± 0%
Long20 3.254Gi ± 0%
Short40 446.8Mi ± 0%
Mid40 1.215Gi ± 0%
Long40 3.254Gi ± 0%
geomean 2.717GiSponsored by: The FreeBSD Foundation