EmilyMatt commented on PR #10833:
URL: https://github.com/apache/arrow-rs/pull/10833#issuecomment-5424950357

   > ```
   > arrow_ipc_stream_writer/StreamEncoder/encode_10/dict          1.19     
66.4±0.47µs        ? ?/sec    1.00     55.8±0.37µs        ? ?/sec
   > ```
   > 
   > 
   > ```
   > arrow_ipc_stream_writer/StreamEncoder/encode_10/dict          1.22     
67.1±0.77µs        ? ?/sec    1.00     55.0±0.42µs        ? ?/sec
   > ```
   > 
   > the benchmark its self looks fine so i dont think this is noise -
   > 
   > 
https://github.com/apache/arrow-rs/blob/2a82e5920421693d5a1c9419f3651cd95d80e37f/arrow-ipc/benches/ipc_writer.rs#L119
   
   I can't seem to reproduce anything conclusive locally
   
     ### StreamWriter/write_10/zstd
   
     | variant | mean | median | min | max | sd | paired Δ vs main (95% CI) | 
wins |
     |---|---|---|---|---|---|---|---|
     | main | 4.8460 | 4.8189 | 4.7925 | 4.9493 | 0.0626 | — | — |
     | `reserve` | 4.7904 | 4.7872 | 4.7435 | 4.8635 | 0.0416 | −1.13% [−2.63, 
+0.36] | 5/6 |
     | `reserve_exact` | 4.7692 | 4.7660 | 4.7539 | 4.7946 | 0.0165 | −1.57% 
[−2.98, −0.16] | 6/6 |
   
     ### StreamEncoder/encode_10/zstd
   
     | variant | mean | median | min | max | sd | paired Δ vs main (95% CI) | 
wins |
     |---|---|---|---|---|---|---|---|
     | main | 4.7536 | 4.7498 | 4.7248 | 4.7888 | 0.0238 | — | — |
     | `reserve` | 4.8166 | 4.7561 | 4.6901 | 5.1938 | 0.1886 | +1.33% [−2.99, 
+5.65] | 3/6 |
     | `reserve_exact` | 4.7793 | 4.7699 | 4.6992 | 4.9262 | 0.0799 | +0.54% 
[−1.21, +2.29] | 3/6 |
     
   (I thought maybe the issue is that the benchmark oes a single encode which 
would be better with reserve_exact as that might save a reallocation in some 
specific cases, but even if that proves better for those cases, it's irrelevant 
for real life use, since after the encode() call there will be others, meaning 
the vec would be reallocated anyway after the reserve_exact, making it worse, 
rather than better)


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to