Re: [PATCH] target/ppc: Merge COMPUTE_CLASS and COMPUTE_FPRF

[Alex Bennée]

Richard Henderson <richard.henderson@linaro.org> writes:

> Instead of computing an artifical "class" bitmask then
> converting that to the fprf value, compute the final
> value from the start.
>
> Reorder the tests to check the most likely cases first.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

We see a slight performance boost although my baseline run was a bit
noisy:

➜  # before
🕙07:31:14 alex@zen:qemu.git/builds/all  (b44e6e6) (REBASING 1/2) [$?] took 10m 
➜  hyperfine -w 2 "./qemu-ppc64 
~/lsrc/tests/lame.git-svn/builds/ppc64/frontend/lame -h pts-trondheim-3.wav 
pts-trondheim-3.mp3"
Benchmark 1: ./qemu-ppc64 ~/lsrc/tests/lame.git-svn/builds/ppc64/frontend/lame 
-h pts-trondheim-3.wav pts-trondheim-3.mp3
  Time (mean ± σ):     649.858 s ± 58.351 s    [User: 649.272 s, System: 0.448 
s]
  Range (min … max):   616.148 s … 764.585 s    10 runs

# after

hyperfine "./qemu-ppc64 ~/lsrc/tests/lame.git-svn/builds/ppc64/frontend/lame -h 
pts-trondheim-3.wav pts-trondheim-3.mp3"
Benchmark 1: ./qemu-ppc64 ~/lsrc/tests/lame.git-svn/builds/ppc64/frontend/lame 
-h pts-trondheim-3.wav pts-trondheim-3.mp3
  Time (mean ± σ):     599.968 s ±  8.014 s    [User: 599.656 s, System: 0.181 
s]
  Range (min … max):   591.171 s … 615.912 s    10 runs

Anyway

Tested-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro