Commits on Mar 31, 2022

1. feat: compute buffer size ahead of time …

   The current implementation loops over message fields and each field
   encoder creates a `Uint8Array` which is added to a `Uint8ArrayList`
   and stitched together at the end of the process.
   
   Other implementations compute the expected length of the serialized
   message, allocate a buffer that size then have encoders write their
   data into it at the correct offsets.
   
   This PR makes protons compute the final message buffer size first in
   the same way.
   
   Thing is, it doesn't seem to make a huge amount of difference to performance.
   
   Before:
   
   ```
   Running "Encode/Decode" suite...
   Progress: 100%
   
     pbjs:
       12 166 ops/s, ±3.92%   | 5.12% slower
   
     protons:
       9 755 ops/s, ±2.19%    | slowest, 23.93% slower
   
     protobufjs:
       12 823 ops/s, ±2.02%   | fastest
   
   Finished 3 cases!
     Fastest: protobufjs
     Slowest: protons
   ```
   
   After:
   
   ```
   Running "Encode/Decode" suite...
   Progress: 100%
   
     pbjs:
       11 866 ops/s, ±3.43%   | 2.05% slower
   
     protons:
       9 356 ops/s, ±2.45%    | slowest, 22.77% slower
   
     protobufjs:
       12 114 ops/s, ±2.16%   | fastest
   
   Finished 3 cases!
     Fastest: protobufjs
     Slowest: protons
   ```

   

   achingbrain committed Mar 31, 2022
   Configuration menu

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

   Copy the full SHA

   fecfedd View commit details

   Browse the repository at this point in the history