Commits on Jun 17, 2023

1. k12: improve API and support multithreaded computation …

   Use options style API, so that the common case is very simple:
   
       h := k12.NewDraft10()
   
   but we can provide options elegantly:
   
       h := k12.NewDraft10(
           WithContext([]byte("some context")),
           WithWorkers(runtime.NumCPU()),
       )
   
   Allows multithreaded computation with the WithWorkers() option.
   On M2 Pro scales well with a few workers, but isn't able to
   utilize all cores effectively.
   
       BenchmarkK12_100B-12          	 5223334	       228.8 ns/op	 437.07 MB/s
       BenchmarkK12_10K-12           	  105744	     11183 ns/op	 894.21 MB/s
       BenchmarkK12_100K-12          	   27141	     44364 ns/op	2254.06 MB/s
       BenchmarkK12_3M-12            	    1172	   1010401 ns/op	3243.07 MB/s
       BenchmarkK12_32M-12           	     100	  10033730 ns/op	3265.78 MB/s
       BenchmarkK12_327M-12          	      12	  98888208 ns/op	3313.64 MB/s
       BenchmarkK12_3276M-12         	       2	 991111938 ns/op	3306.19 MB/s
       BenchmarkK12x2_32M-12         	     183	   6510423 ns/op	5033.16 MB/s
       BenchmarkK12x2_327M-12        	      18	  63622058 ns/op	5150.41 MB/s
       BenchmarkK12x2_3276M-12       	       2	 632823584 ns/op	5178.06 MB/s
       BenchmarkK12x4_32M-12         	     364	   3300120 ns/op	9929.34 MB/s
       BenchmarkK12x4_327M-12        	      39	  29477854 ns/op	11116.14 MB/s
       BenchmarkK12x4_3276M-12       	       2	 581819167 ns/op	11263.98 MB/s
       BenchmarkK12x8_32M-12         	     520	   2301923 ns/op	14235.05 MB/s
       BenchmarkK12x8_327M-12        	      76	  15590312 ns/op	21018.18 MB/s
       BenchmarkK12x8_3276M-12       	       4	 296590469 ns/op	22096.46 MB/s
       BenchmarkK12xCPUs_32M-12      	     472	   2526827 ns/op	12968.04 MB/s
       BenchmarkK12xCPUs_327M-12     	      78	  15139957 ns/op	21643.39 MB/s
       BenchmarkK12xCPUs_3276M-12    	       4	 280958114 ns/op	23325.90 MB/s
   
   We only reach 23GB/s (at 12x) instead of the lower bound of 33GB/s
   expected with 10 performance cores.
   
   Adds {Max,Next}WriteSize to suggest the caller how big to choose
   their Write() calls.
   
   Adds a test to the CI, to check the k12 package with the data race
   detector on.

   

   bwesterb committed Jun 17, 2023
   Configuration menu

   • View commit details
   • Copy full SHA for 78965b7
   • Browse repository at this point

   Copy the full SHA

   78965b7 View commit details

   Browse the repository at this point in the history