Re: GNAT.Sockets Streaming inefficiency

[gautier_niouzes@hotmail.com]

Hello,

It could be related to a performance bottleneck in 'Read and 'Write, perhaps more than to the nature of the stream (socket vs. file vs. ...).

For instance, copying a file with different methods leads to the following timings (here using a byte buffer of sizes 1KB, then 1MB):

xxx'Write / xxx'Read (Stream attributes)......... 0.247916104 seconds
Workarounds with Stream_Element_Array buffer:
  copy (byte buffer from/to SE buffer)........... 0.005628329 seconds
  overlay (read), unchecked_conversion (write)... 0.005184935 seconds
  overlay (read and write)....................... 0.005185743 seconds
Factor (Copy)    44.047905515
Factor (UC)      47.814698544
Factor (Overlay) 47.807248450
Buffer size in bytes..... 1024

xxx'Write / xxx'Read (Stream attributes)......... 0.326904305 seconds
Workarounds with Stream_Element_Array buffer:
  copy (byte buffer from/to SE buffer)........... 0.002427558 seconds
  overlay (read), unchecked_conversion (write)... 0.001585229 seconds
  overlay (read and write)....................... 0.001425978 seconds
Factor (Copy)    134.663849432
Factor (UC)      206.218978456
Factor (Overlay) 229.249192484
Buffer size in bytes..... 1048576

This is with GNAT GPL 2016 / Windows with full "fast mode" switches. There were similar factors when the issue was discoverd in ~2008.
The test procedure is test/test_stream_performance.adb in the Zip-Ada project.
HTH
_________________________ 
Gautier's Ada programming 
http://gautiersblog.blogspot.com/search/label/Ada 
NB: follow the above link for a valid e-mail address