Recent mainline progress on libstdc++/9876

[Paolo Carlini]

Hi everyone,

I'd like to announce that in mainline we are making
progress on that benchmark.

To wit, on my P4-2400 linux platform
(-O2, iterations = 100000000):

3.3
---
time ./stdio
4.500u 0.250s 0:04.79 99.1%     0+0k 0+0io 218pf+0w

time ./stdio_unlocked
2.040u 0.250s 0:02.31 99.1%     0+0k 0+0io 217pf+0w

time ./iostreams
3.750u 0.330s 0:04.12 99.0%     0+0k 0+0io 219pf+0w


3.4
---
time ./stdio
4.500u 0.330s 0:04.88 98.9%     0+0k 0+0io 217pf+0w

time ./stdio_unlocked
2.020u 0.270s 0:02.31 99.1%     0+0k 0+0io 216pf+0w

time ./iostreams
1.970u 0.420s 0:02.39 100.0%    0+0k 0+0io 215pf+0w

I can't tell exactly, what was worth this measurable
improvement: perhaps some simplifications suggested by
Nathan and Pétur, perhaps something else.

Anyway, I'm quite happy with that but before really celebrating
I encourage feedback for platforms other than mine.

Paolo.