Page: Previous  1, 2, 3, 4  Next

Thanks!
alienbob
Status: New User - Welcome
Joined: 09 Apr 2013
Posts: 1
Location: United Kingdom
Reply Quote
Thank you for this, needed to convert several hundred .flacs to .ogg on my CentOS (release 5.3) running Subsonic, this did it flawlessly!!

Cheers

Ab
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 3986
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
glad it was of use, I use this one all the time myself as well, and it 'just works', which is what I want.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 3986
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
I forgot about this thread, etique's suggestions for cleaning directories was a good one, not sure why that never got done, maybe it was because I thought 3.0 would be coming soon, heh. That's now implemented with the new --clean option. Use with --test option first to make sure nothing unexpected happens!!

Note that acxi 3.0 uses File::Find for all file and directory syncing, which allows for much better and more granular control.

Some items get adjusted, ` is removed, though I'm not sure it needs to be anymore with the new syntax. I should have checked that for the new acxi.

Anyway, 3.0 is very much cleaner than 2.9 was, it's similar internally to inxi, which was also rewritten in 2018 to Perl 5, so I wanted their styles to be similar internally.

Note that acxi now uses Perl builtins for cp/copy, find, and mkdir. And it would also use native Perl features for rm if that was added. The goal is that the only subshells used are the direct lame, oggenc, and metaflac that acxi uses to encode the compressed files. That way acxi will be cross-platform without any further work.

I still use acxi all the time, heh.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 3986
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
I've iimproved the --clean option to now double verify all deletions, I have stuff in my compressed directory that is not in my source directory, so I realized it can't just blindly delete, now it will confirm first the directories, then the files, as 2 groups, you can exit that, and probably should, if you see something you were not expecting to see happen. Automatic deletions are very dangerous, so i decided that a double confirm coupled with a listing of files/directories to be deleted would be an absolute requirement. This feature can also be used with the --test option, though you will see what is going to happen before you confirm it no matter what.

The forking item is interesting but I'm not sure I'll do that for this 3.0.0 release, if there is real interest, maybe I will do it for the 3.1.0 release, if that ever happens.
Back to top
prupert
Status: New User - Welcome
Joined: 12 Jul 2019
Posts: 4
Location: United Kingdom
Reply Quote
:: etique wrote ::
Ah ah, glad you like !

Well, unfortunately I had to gave up on multithreading...

Just to use forking instead :)

And you need to add a line "use Parallel::ForkManager;" at the beginning (and add the proper Fork module... wasn't there by default on Ubuntu 11.10)

Hope you like!


FYI and sorry for the thread necro, but I updated this forking approach to the most recent version of acxi on github and it seems to work well. Happy to share if interested.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 3986
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
Can I see the code?

I'd be curious about forking logic and code.
Back to top
prupert
Status: New User - Welcome
Joined: 12 Jul 2019
Posts: 4
Location: United Kingdom
Reply Quote
Sure thing, here's a link to the modified file:

[link]

(I am assuming its best not to paste all the code here on the forum).

Its still making use of that forking Perl module and I've hard coded the number of cores (7) in the line:
my $pm = new Parallel::ForkManager(7);

But as the original poster said this can be added as an option in the config file.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 3986
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
that looks pretty doable. Having the module load be dynamically tested is not a problem, it would simply exit on failure to load if forked was requested but not present on the system.

I don't see any real reason not to add this logic to acxi, with extra tests etc to make sure module is available, thanks for posting this.

What happens if you request too many forks, more than the system can handle? or is that not how it works? Is the fork max directly connected to the thread max, say, for a multithreaded cpu like recent intel or amd ryzen, the thread count is 2xcore count.

I guess I can find out by testing it.

This will make a nice 3.2.0, though I also had something else in mind, but I'm too lazy to make, heh.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 3986
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
tested this, very impressive.

without forking, converting a test flac directory with 15 files on this cpu:

:: Code ::
CPU:
  Topology: 6-Core model: AMD Ryzen 5 2600 bits: 64 type: MT MCP
  L2 cache: 3072 KiB
  Speed: 1746 MHz min/max: 1550/3400 MHz


:: Code ::
acxi
....
real    0m31.877s
user    0m38.232s
sys     0m0.244s


:: Code ::
acxi --fork 12
....
real    0m6.256s
user    0m44.100s
sys     0m0.239s


:: Code ::
acxi --fork 4
....
real    0m11.250s
user    0m41.608s
sys     0m0.256s


and just to see how it works with > number of cpu threads:

:: Code ::
acxi --fork 24
....
real    0m5.255s
user    0m48.730s
sys     0m0.327s


so pretty impressive.

I'll have this on github later today, thanks.

This uses the new config value $FORK, defaults to 1, so no fork, tests user data, tests module presence if $FORK > 1, exits.

Haven't updated the man page, will do that later.

I didn't realize the speedup would be so big, that's very impressive, glad you posted, thanks.

I've pushed 3.1.4 with forking stuff just so it can be checked etc, I'll do a bigger test on it on my main music system, but that one is only dual core so it won't be a huge speedup.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 3986
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
Note that there's an oddity which is somewhat expected in terms of the output/return values of each fork, for debugging purposes those would be nice to get resolved, but not a huge deal.
Back to top
Display posts from previous:   
Page: Previous  1, 2, 3, 4  Next
All times are GMT - 8 Hours