qbit/go
1
0
Fork 0
mirror of https://github.com/golang/go synced 2024-11-21 23:24:41 -07:00
Code Releases Activity

test/bench/shootout: more speedups

Browse Source 
Most significant in mandelbrot, from avoiding MOVSD between registers,
but there are others.

R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/6258063
This commit is contained in:
Rob Pike 2012-05-30 12:31:28 -07:00
parent a768de8347
commit 6f3ffd4d79
1 changed files with 32 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
test/bench/shootout/timing.log
View File

@ -1023,6 +1023,7 @@ meteor 2098
gc meteor-contest 0.13u 0.00s 0.13r # -13%
gc_B meteor-contest 0.13u 0.00s 0.13r # -7%
# May 30, 2012.
# After http://codereview.appspot.com/6261051, restoring old code generated
# for floating-point constants. Mandelbrot is back to its previous numbers.
@ -1031,3 +1032,34 @@ mandelbrot 16000
gccgo -O2 mandelbrot.go 41.72u 0.01s 41.90r
gc mandelbrot 60.62u 0.00s 60.76r
gc_B mandelbrot 60.68u 0.00s 60.82r
# May 30, 2012.
# After http://codereview.appspot.com/6248068, better FP code
# by avoiding MOVSD between registers.
# Plus some other timing changes that have crept in from other speedups,
# from garbage collection to Printf.
fasta -n 25000000
gc fasta 1.76u 0.00s 1.76r # -12%
gc_B fasta 1.71u 0.00s 1.72r # -12%
nbody -n 50000000
gc nbody 17.56u 0.00s 17.60r # -8%
gc_B nbody 17.30u 0.00s 17.34r # -10%
fannkuch 12
gc fannkuch-parallel 155.92u 0.01s 44.05r # -15%
k-nucleotide 1000000
gc k-nucleotide 9.22u 0.01s 9.26r # -5%
gc k-nucleotide-parallel 9.23u 0.03s 3.26r # -9%
gc_B k-nucleotide 9.22u 0.03s 9.28r # -2%
mandelbrot 16000
gc mandelbrot 44.80u 0.00s 44.90r # -27%
gc_B mandelbrot 44.81u 0.00s 44.92r # -26%
pidigits 10000
gc pidigits 3.51u 0.00s 3.52r # -6%
gc_B pidigits 3.51u 0.00s 3.52r # -6%