New stack implementation

[chfast]

Comparing results/master to results/new
Benchmark                            Time             CPU      Time Old      Time New       CPU Old       CPU New
-----------------------------------------------------------------------------------------------------------------
sha1_shifts/empty                 -0.1163         -0.1163            53            47            53            47
sha1_shifts/1351                  -0.1393         -0.1393          1031           888          1031           888
sha1_shifts/2737                  -0.1386         -0.1386          2006          1728          2006          1728
sha1_shifts/5311                  -0.1390         -0.1390          3908          3365          3908          3365
sha1_shifts/65536                 -0.1399         -0.1399         47639         40976         47639         40976
stop                             +14.1512        +14.1515             0             1             0             1
blake2b_huff/empty                -0.0215         -0.0215            65            64            65            64
blake2b_huff/abc                  -0.0198         -0.0198            65            64            65            64
blake2b_huff/2805nulls            -0.1305         -0.1305           489           425           489           425
blake2b_huff/2805aa               -0.1301         -0.1301           489           425           489           425
blake2b_huff/5610nulls            -0.1393         -0.1393           913           786           913           786
blake2b_huff/8415nulls            -0.1400         -0.1400          1315          1131          1315          1131
blake2b_huff/65536nulls           -0.1475         -0.1475          9921          8457          9921          8458
sha1_divs/empty                   -0.0604         -0.0604            92            87            92            87
sha1_divs/1351                    -0.0777         -0.0777          1846          1703          1846          1703
sha1_divs/2737                    -0.0790         -0.0790          3604          3319          3603          3319
sha1_divs/5311                    -0.0791         -0.0791          7026          6470          7026          6470
sha1_divs/65536                   -0.0837         -0.0837         85873         78683         85871         78684
blake2b_shifts/empty              +0.0000         +0.0000             0             0             0             0
blake2b_shifts/2805nulls          -0.1172         -0.1172          6553          5785          6553          5785
blake2b_shifts/5610nulls          -0.1096         -0.1096         13080         11647         13080         11647
blake2b_shifts/8415nulls          -0.1154         -0.1154         19629         17363         19629         17363
blake2b_shifts/65536nulls         -0.1612         -0.1612        153859        129053        153852        129053

[axic]

Wow, 17% improvement in the best case?

[chfast]

Wow, 17% improvement in the best case?

Where do you see it? It's up to 14% faster in some cases and 14x slower for "stop" which does nothing.

[codecov-io]

Codecov Report

Merging #79 into master will decrease coverage by 0.1%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master      #79      +/-   ##
==========================================
- Coverage   98.07%   97.96%   -0.11%     
==========================================
  Files          16       16              
  Lines        1712     1669      -43     
  Branches      161      161              
==========================================
- Hits         1679     1635      -44     
- Misses         24       25       +1     
  Partials        9        9

[chfast]

Wow, 17% improvement in the best case?

Where do you see it? It's up to 14% faster in some cases and 14x slower for "stop" which does nothing.

Oh yes, there is -0.1612 but the blake2b_shifts/65536nulls is hardly relevant as it consumes too much memory and we get gas usage 1.5G/s.