>
> Put it this way - Im willing to drop 32K of memory for the right results
> and test the patch if you do it. To start with I can get that 32K back
> simply by putting Andy's select/poll patches into my kernel and non inlining
> the copy_to/from_user which btw my benchmarks say speeds up the machine
>
> Alan
>
>
Hello Alan,
You tweaked my curiosity, so I de-inlined copy_to/from_user and did
some comparisons. Bottom line is that on my machine it doesn't seem
to care either way.. performance stayed the same. I did Bonnie and
some network throughput tests in four configs. I'm only posting the
Bonnie runs in the interest of brevity.
-Mike
16 meg size because I wanted to test buffercache not cheap drives.
kernel is 2.1.77 (minute differences are test jitter imho)
no fpu_memcpy no inlines
-------Sequential Output-------- ---Sequential Input--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block---
MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU
16 6246 99.9 17211 98.7 7490 100.1 5274 99.2 27632 99.5
no fpu_memcpy with inlines
-------Sequential Output-------- ---Sequential Input--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block---
MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU
16 6246 99.9 17469 100.2 7730 100.0 5246 98.9 27644 99.6
fpu_memcpy no inlines
-------Sequential Output-------- ---Sequential Input--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block---
MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU
16 7433 99.8 31556 100.2 10745 100.3 5405 99.0 41660 99.2
fpu_memcpy with inlines
-------Sequential Output-------- ---Sequential Input--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block---
MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU
16 7435 99.8 31514 100.0 11102 99.6 5555 99.4 42526 98.6