Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: SRLA: A lossless audio codec focused on decode speed and compression rate (Read 15373 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #25 –
Quote from: genuine on
Also, the executable files I downloaded don't do compression, they just seem to rewrite the file.
genuine, thanks for using the codec. High-entropy audio (such as white noise) is generally hard to compress. If not, please provide me with some information so that I can help with this problem.

1. What is the command line you entered to execute the codec?
2. What is your running OS?
3. What is the input audio format?
 - Currently, SRLA supports 8-24bit WAV files only. Also, I noticed that SRLA does not support WAVEFORMATEXTENSIBLE WAV files.

It would be a great help if you could provide an input audio file.

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #26 –
I also used 3 versions of SRLA (latest, latest_avx2 and latest_sse41). The results are same(Windows10 X64). The processor I tried Intel Core i5-2300.

srla.exe -e test.wav out.srla
finished: 9580592 -> 9576864 ( 99.96 %)

flac.exe -o out.flac -5 test.wav
5427759

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #27 –
I think codec behaves like this in the old CPUs. But at least CPU had AVX support. Anyway, now i tried a CPU with AXV2 supported.

20 tracks, 44.1, 16 bit, stereo (840,911,660 bytes)

SRLA 0.0.10 (avx2) -m0 -> 580,672,062 bytes
Encoding 61.921 s, Decoding 37.472 s

SRLA 0.0.10 (avx2) -m4 -> 553,542,096 bytes
Encoding 92.226 s, Decoding 38.284 s

SRLA 0.0.7 (avx2) -> 552,888,600 bytes
Encoding 95.892 s, Decoding 20.501 s

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #28 –
Quote from: genuine on
I think codec behaves like this in the old CPUs. But at least CPU had AVX support.
Thank you for reporting it. I found that the codec tool unintentionally invokes AVX instructions.
I try to handle this problem by building a Win32 application. Can you try these codecs instead?
archive URL: https://github.com/aikiriao/SRLA/actions/runs/13852910671/artifacts/2751364580

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #29 –
I tested SLRA v0.0.10 on my usual testset of 16 files (stereo, 44.1Khz)

Code: [Select]
-m 4         29.305.243 [00:00:00]
-m 6         29.121.839 [00:00:01]
-m 6 -V 2    29.008.537 [00:00:02]
-m 6 -V 4    28.925.896 [00:00:16]
-m 6 -V 8    28.897.557 [01:02:52]
sac v0.7.17  26.691.259 [00:00:22]

 

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #30 –
Quote from: aikiriao on
Thank you for reporting it. I found that the codec tool unintentionally invokes AVX instructions.
I try to handle this problem by building a Win32 application. Can you try these codecs instead?
archive URL: https://github.com/aikiriao/SRLA/actions/runs/13852910671/artifacts/2751364580
This is the results with CPU without AVX2 support.
finished: 5292048 -> 5295586 (100.07 %)
finished: 3377108 -> 3379370 (100.07 %)
finished: 4410076 -> 4413031 (100.07 %)
finished: 4061588 -> 4064301 (100.07 %)
finished: 2757092 -> 2757972 (100.03 %)
finished: 4320784 -> 4323673 (100.07 %)
finished: 3588140 -> 3590534 (100.07 %)

However, I did a new test with the same processor with AVX2 supported. This time the decode time is much better.

MIX 10 tracks, 44.1, 16 bit, stereo (904,185,144 bytes)

SRLA 0.0.10 (avx2) -m 4 -> 545,936,257 bytes
Encoding 115.741 s, Decoding 7.716 s

SRLA 0.0.10 (avx2) -m 0 -> 604,178,724 bytes
Encoding 92.803 s, Decoding 6.493 s

FLAC 1.5 -5 --no-md5 -> 556,590,777
Encoding 8.402 s, Decoding 4.711 s

FLAC 1.5 -0 --no-md5 -> 607,321,979
Encoding 4.336 s, Decoding 5.227 s

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #31 –
Quote from: genuine on
This is the results with CPU without AVX2 support.
Thank you for reporting it. It is difficult to identify the cause of this problem, but I want to clarify it to guarantee backward compatibility.
I build the codec for the x86 CPU host. Can you run codecs?
https://github.com/aikiriao/SRLA/actions/runs/13900369490/artifacts/2764678034

And can you run a debug build codec attached to this message? This is because I want to clarify whether this problem is a compiler matter.
I'm sorry for bothering your environment.

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #32 –
This is really an interesting situation. If you have the chance to try this on a CPU that does not support AVX2, you can see the result. I don't think the issue is related to x86 or x64. Again, it works flawlessly on AVX2 supported CPUs.

srla.exe -e test.wav out.srla
finished: 9580592 -> 9576864 ( 99.96 %)

For a 50 mb file, it uses 4 times the ram (200 mb) and works very slowly. It may be because it is "Debug". And for all the other files I tried, I get the following error during the process.
Assertion failed: ((k2 + 1) + 1) <= 32, file C:\Users\aikiriao-ryzen\Desktop\works\SRLA\libs\srla_coder\src\srla_coder.c, line 584

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #33 –
Quote from: genuine on
Assertion failed: ((k2 + 1) + 1) <= 32, file C:\Users\aikiriao-ryzen\Desktop\works\SRLA\libs\srla_coder\src\srla_coder.c, line 584
Thanks again. It is beneficial to me.
I found that I use LZCNT operation, which some Intel CPUs do not support.
Instead, I built another codec that uses a more portable operation. Can you try this?
https://github.com/aikiriao/SRLA/actions/runs/13911348204/artifacts/2768614862

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #34 –
I think there was a problem with the forum and what i wrote was deleted. Okay, it works fine as it is.
12 tracks, 44.1, 16 bit, stereo (585,144,122 bytes)

Quote
SRLA 0.0.10 (sse41) -m 6 -> 393,959,679 bytes
Encoding 189.021 s, Decoding 7.496 s

SRLA 0.0.10 (sse41) -m 4 -> 394,469,100 bytes
Encoding 117.127 s, Decoding 7.056 s

SRLA 0.0.10 (sse41) -m 0 -> 421,118,005 bytes
Encoding 94.398 s, Decoding 5.071 s
Quote
FLAC 1.5 -8 --no-md5 -> 401,192,090
Encoding 23.172 s, Decoding 5.210 s

FLAC 1.5 -5 --no-md5 -> 403,602,181
Encoding 8.053 s, Decoding 4.849 s

FLAC 1.5 -0 --no-md5 -> 432,450,718
Encoding 4.350 s, Decoding 5.107 s
Quote
HALAC 0.3.8 -normal -> 404,597,742
Encoding 3.150 s, Decoding 4.239 s

HALAC 0.3.8 -fast -> 411,483,100
Encoding 2.586 s, Decoding 3.819 s

HALAC 0.3.8 -ufast -> 425,071,585
Encoding 2.313 s, Decoding 2.748 s

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #35 –
Thank you genuine for your testing and patience.

I just released v0.0.12:
https://github.com/aikiriao/SRLA/releases/tag/v0.0.12
This version newly included long-term (pitch) prediction, a well-known method for reducing redundancy, especially for tonal signals (such as vocal).
And also I redo my benchmarks, including HALAC V.0.3.8
https://github.com/aikiriao/SRLA/blob/main/evaluation/decodespeed_vs_compressionrate_4096_total.png
As you can see in this graph, HALAC is also truly fast in my environment.

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #36 –
Who uses these new codecs?

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #37 –
Quote from: mycroft on
Who uses these new codecs?
According to this idea, there is no need for anything new. There is no need for science, technology or research. What is done is done, what is over is over. Some things should remain the same forever. I don't have it, you shouldn't have it. Nobody should disturb their comfort.
It's been a long time since we've seen new audio codecs. But the reactions to them are quite bizarre.

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #38 –
• Listen to the music, not the media it's on
• The older, the 'lossier'

Re: SRLA: A lossless audio codec focused on decode speed and compression rate

Reply #39 –
Quote from: mycroft on
Who uses these new codecs?
It is too late for world domination - if that was the purpose with yours?

Luckily, lossless codecs can just be replaced by decoding, and so those who once chose Shorten or Real Audio Lossless (or, *coughs*, sonic) still got their music. Maybe WMAL too ... maybe.