Re: [fluid-dev] parallel loading of soundfont samples

[Marcus Weseloh]

Hi again,

just to highlight the much faster speed of SF2 loading, here is the
previous test run again, but this time with the same Soundfont in SF2
format (converted with Polyphone). The resulting SF2 is 204 MB (SF3
was 40MB):

$ time ./test.sh 1
real 0m0,366s
user 0m0,093s
sys 0m0,163s

$ time ./test.sh 2
real 0m0,457s
user 0m0,196s
sys 0m0,454s

$ time ./test.sh 3
real 0m0,521s
user 0m0,378s
sys 0m0,754s

$ time ./test.sh 4
real 0m0,595s
user 0m0,476s
sys 0m0,964s

So loading the SF2 version of a converted SF3 soundfont is around 20
times faster on a single core. So if you can find a good way to
convert SF3 to SF2 and have the disk space, using SF2 soundfonts is by
far the best way to speed up loading times.

Cheers
Marcus

Am Mi., 28. Okt. 2020 um 14:50 Uhr schrieb Marcus Weseloh <marcus@weseloh.cc>:
>
> Hi,
>
> Am Mi., 28. Okt. 2020 um 13:10 Uhr schrieb Ceresa Jean-Jacques
> <jean-jacques.ceresa@orange.fr>:
> > Now we should try the suggested test to run 2 fluidsynth console 
> > applications (a1, a2) at the same time (not sequencially) and measure the 
> > total time.
>
> That's a really good idea to quickly test if parallel loading has
> merit. A quick test on my machine (i5-3210M CPU @ 2.50GHz × 4) and the
> MuseScore_General.sf3 (ca. 40MB) Soundfont on SSD with the following
> test script:
>
> --------------
> #!/bin/bash
> for i in `seq $1`; do
>     ./fluidsynth -a file ./MuseScore_General.sf3 &
> done
> wait
> --------------
>
> $ time ./test.sh 1
> real 0m6,110s
> user 0m5,829s
> sys 0m0,172s
>
> $ time ./test.sh 2
> real 0m6,658s
> user 0m12,615s
> sys 0m0,344s
>
> $ time ./test.sh 3
> real 0m8,884s
> user 0m23,626s
> sys 0m0,544s
>
> $ time ./test.sh 4
> real 0m10,444s
> user 0m38,182s
> sys 0m0,851s
>
> With each test the number of used cores in the test was maxed out at
> 100% CPU load.
>
> So it looks like parallel loading would be beneficial.
>
> Cheers
> Marcus