huff0: Speed up compression of short blocks

[greatroar]

This is ugly, but it forces the compiler to merge loads and stores of nodeElts.

name                                      old speed      new speed      delta
Compress4XReuseNone/digits-8               433MB/s ± 1%   434MB/s ± 0%     ~     (p=0.393 n=10+10)
Compress4XReuseNone/gettysburg-8           245MB/s ± 1%   254MB/s ± 1%   +3.55%  (p=0.000 n=10+10)
Compress4XReuseNone/twain-8                363MB/s ± 0%   363MB/s ± 0%     ~     (p=0.360 n=8+10)
Compress4XReuseNone/low-ent.10k-8          465MB/s ± 0%   467MB/s ± 0%   +0.32%  (p=0.008 n=9+10)
Compress4XReuseNone/superlow-ent-10k-8     305MB/s ± 2%   304MB/s ± 1%     ~     (p=0.143 n=10+10)
Compress4XReuseNone/case1-8               13.1MB/s ± 1%  14.6MB/s ± 2%  +11.44%  (p=0.000 n=10+10)
Compress4XReuseNone/case2-8               10.9MB/s ± 1%  12.3MB/s ± 2%  +12.28%  (p=0.000 n=10+9)
Compress4XReuseNone/case3-8               11.5MB/s ± 1%  12.9MB/s ± 2%  +12.37%  (p=0.000 n=10+10)
Compress4XReuseNone/pngdata.001-8          300MB/s ± 0%   298MB/s ± 2%     ~     (p=0.762 n=8+10)
Compress4XReuseNone/normcount2-8          31.5MB/s ± 1%  33.7MB/s ± 2%   +6.87%  (p=0.000 n=9+10)
Compress4XReuseAllow/digits-8              435MB/s ± 1%   432MB/s ± 2%     ~     (p=0.043 n=10+10)
Compress4XReuseAllow/gettysburg-8          272MB/s ± 1%   281MB/s ± 1%   +3.19%  (p=0.000 n=10+10)
Compress4XReuseAllow/twain-8               363MB/s ± 0%   361MB/s ± 1%     ~     (p=0.150 n=9+10)
Compress4XReuseAllow/low-ent.10k-8         469MB/s ± 1%   469MB/s ± 0%     ~     (p=0.616 n=10+10)
Compress4XReuseAllow/superlow-ent-10k-8    305MB/s ± 0%   306MB/s ± 0%   +0.44%  (p=0.001 n=9+9)
Compress4XReuseAllow/case1-8              15.8MB/s ± 0%  18.0MB/s ± 2%  +13.88%  (p=0.000 n=9+10)
Compress4XReuseAllow/case2-8              13.3MB/s ± 0%  15.4MB/s ± 1%  +15.86%  (p=0.000 n=9+10)
Compress4XReuseAllow/case3-8              14.1MB/s ± 0%  15.8MB/s ± 1%  +12.15%  (p=0.000 n=10+8)
Compress4XReuseAllow/pngdata.001-8         302MB/s ± 1%   304MB/s ± 0%   +0.68%  (p=0.000 n=9+10)
Compress4XReuseAllow/normcount2-8         40.2MB/s ± 1%  43.8MB/s ± 2%   +8.86%  (p=0.000 n=10+10)
Compress4XReusePrefer/digits-8             436MB/s ± 0%   436MB/s ± 1%     ~     (p=0.560 n=9+9)
Compress4XReusePrefer/gettysburg-8         422MB/s ± 1%   421MB/s ± 2%     ~     (p=0.579 n=10+10)
Compress4XReusePrefer/twain-8              363MB/s ± 1%   365MB/s ± 0%     ~     (p=0.018 n=10+10)
Compress4XReusePrefer/low-ent.10k-8        473MB/s ± 0%   472MB/s ± 0%     ~     (p=0.021 n=7+8)
Compress4XReusePrefer/superlow-ent-10k-8   312MB/s ± 1%   312MB/s ± 0%     ~     (p=0.278 n=10+9)
Compress4XReusePrefer/case1-8              134MB/s ± 1%   134MB/s ± 1%     ~     (p=0.780 n=9+10)
Compress4XReusePrefer/case2-8              122MB/s ± 2%   122MB/s ± 0%     ~     (p=1.000 n=10+8)
Compress4XReusePrefer/case3-8              129MB/s ± 3%   129MB/s ± 0%     ~     (p=0.698 n=10+10)
Compress4XReusePrefer/pngdata.001-8        313MB/s ± 1%   313MB/s ± 0%     ~     (p=0.481 n=10+10)
Compress4XReusePrefer/normcount2-8         179MB/s ± 2%   182MB/s ± 0%   +1.96%  (p=0.000 n=10+9)
Compress4XSizes/digits-100-8              57.6MB/s ± 1%  61.7MB/s ± 1%   +7.11%  (p=0.000 n=9+10)
Compress4XSizes/digits-200-8               100MB/s ± 2%   108MB/s ± 2%   +7.90%  (p=0.000 n=10+10)
Compress4XSizes/digits-500-8               191MB/s ± 1%   201MB/s ± 1%   +5.55%  (p=0.000 n=10+9)
Compress4XSizes/digits-1000-8              273MB/s ± 1%   283MB/s ± 1%   +3.75%  (p=0.000 n=10+10)
Compress4XSizes/digits-5000-8              416MB/s ± 0%   418MB/s ± 1%     ~     (p=0.165 n=10+10)
Compress4XSizes/digits-10000-8             434MB/s ± 1%   437MB/s ± 1%   +0.65%  (p=0.002 n=10+10)
Compress4XSizes/digits-50000-8             434MB/s ± 0%   434MB/s ± 0%     ~     (p=0.604 n=10+9)
[Geo mean]                                 152MB/s        157MB/s        +3.33%

[klauspost]

Wonderful! Never took the time to optimize the table building. Great stuff!