FLAC v1.4.3 (Release)
https://github.com/xiph/flac/releases/tag/1.4.3
Full range of compiles now available on Rarewares. :)
FLAC v1.4.3 (GCC v13.1) & FLAC v1.4.3 (CLANG v16.0.5) Win Binaries
Thanks to all involved in the making of FLAC v1.4.3.
But, I have one question. I tried the new builds, posted above, they all still say "reference libFLAC 1.4.2 20221022"
Is that intended?
Hmmm, the Rarewares compiles all show "reference libFLAC 1.4.3 20230623".
@TuNk77 Tested encoding with all of John's Binaries & DLL's and they all show "reference libFLAC 1.4.3 20230623".
Same thing goes for the binaries i posted above.
I'm sorry, I made a mistake, I forgot to change the commandline encoder settings in foobar2000.
I have set a customized parameter in the the commandline encoder settings in foobar2000.
Previously I just changed "additional commandline encoder paths" in foobar2000, no wonder that the newly encoded file still said "reference libFLAC 1.4.2 20221022"
Thanks, again guys :)
:) ;) O:)
These compiles are using the CPU optimization x86-64-v3 similar to haswell including AVX2.
Inside are builds with Clang 16.0.5, GCC 13.1.0 and a "disable-asm-optimizations" version with faster 16bit performance but slower 24bit performance for the use in CD format apps like CUETools or EAC.
Oddly enough, when using the RareWares AVX2 compiles, I get intermittent access violation errors (-1073741819), but (so far) only when running the test command (-t) and only when testing multiple files in parallel. I can't reproduce with the vanilla RareWare compiles. ¯\_(ツ)_/¯
Windows 11, Ryzen 3900X (which ostensibly supports AVX2)
Oddly enough, when using the RareWares AVX2 compiles, I get intermittent access violation errors (-1073741819), but (so far) only when running the test command (-t) and only when testing multiple files in parallel. I can't reproduce with the vanilla RareWare compiles. ¯\_(ツ)_/¯
Windows 11, Ryzen 3900X (which ostensibly supports AVX2)
Hmmm, thanks for the info. It's a GCC 13.1.0 compile. I'll try a different version of the 13.1.0 compiler and make it available for testing.
Recompile for testing: https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip (https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip)
Recompile for testing: https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip (https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip)
I'm unable to reproduce the error after multiple runs with this revised compile. 8)
Recompile for testing: https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip (https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip)
I'm unable to reproduce the error after multiple runs with this revised compile. 8)
OK, thanks for that. I'll replace the one on Rarewares.
Edit: Done.
32-bit libflac from rarewares is not working with SoX 14.4.2 on windows 7. libflac from official github release is working.
Ah, it's the good old naming problem - libfac.dll versus libflac_dynamic.dll. Thanks for the notification, I'll sort it in the morning.
Are AVX2 compiles actually faster?
Any chances for AVX512 compile for testing? Since I have Ryzen 7800X3D. I could benchmark if I could know how to do it.
Ah, it's the good old naming problem - libfac.dll versus libflac_dynamic.dll. Thanks for the notification, I'll sort it in the morning.
Uploaded a set of dlls without 'dynamic' in the name, so two sets are available one with and the other without 'dynamic' in the name. Functionally they are identical.
Ah, it's the good old naming problem - libfac.dll versus libflac_dynamic.dll. Thanks for the notification, I'll sort it in the morning.
Unfortunately, it ssems that problem is not filename. Actually, i do rename them all to "libflac-8.dll" (this is the name of libflac dll in vanilla release of SoX 14.4.2 from sourceforge).
Apologies for that, I found an Intel compile option that was causing the problem. Recompiled and freshly uploaded. I've removed the 'dynamic' named options as I don't believe anything uses that name.
@itisljar Please use the 'FLAC v1.4.x Performance Tests' thread to post speed tests etc.
https://hydrogenaud.io/index.php/topic,123025.0.html
I've removed the 'dynamic' named options as I don't believe anything uses that name.
On Windows, CUETools and CDex both reference "libFLAC_dynamic.dll".
I've removed the 'dynamic' named options as I don't believe anything uses that name.
On Windows, CUETools and CDex both reference "libFLAC_dynamic.dll".
Oh, marvelous!! I'll reinstate new versions. I'd love to know how we got into the dual naming position, it's irritating, to say the least!
Edit: 'Dynamic' named dlls available again at Rarewares.
One or more of the messages of this topic have been moved to the FLAC v1.4.x Performance Tests (https://hydrogenaud.io/index.php/topic,123025.0.html) thread - https://hydrogenaud.io/index.php/topic,123025.msg1029226.html#msg1029226
Apologies for that, I found an Intel compile option that was causing the problem. Recompiled and freshly uploaded.
Thanks. It works now.
Apologies for that, I found an Intel compile option that was causing the problem. Recompiled and freshly uploaded.
Thanks. It works now.
Excellent! Thanks for letting me know.
FLAC 1.4.3 20230623
Built on June 23, 2023, Visual Studio 2022 (v143) 17.6.4 10.0.22621.0
+ libFLAC_dynamic.dll
https://github.com/xiph/flac/commits/master
Just a very quick and nooby question: is it safe to replace older versions of these files with thee latest FLAC 1.4.3? For example, when I search flac.exe I can see EAC, Flac Frontend, Trader's Little Helper uses flac.exe. Can I replace those outdated versions?
FartMasterExtreme, you should be OK to do so.
FartMasterExtreme, you should be OK to do so.
Not necessarily? Which builds are static?
For those who don't know the impact of the programming lingo: Some of the builds available require a libflac.dll (or with a different name, but a .dll file) and some have a "stand-alone" flac.exe that can be copied around like this.
Is it so that every build posted in this thread is OK, but the official xiph release is not?
By the way,
@Case - you did provide some compile that in some situations performed well. Care to post an exe of 1.4.3 too?
Is it so that every build posted in this thread is OK, but the official xiph release is not?
The official build is OK too, as long as you also copy libFLAC.dll with flac.exe, as explained in the readme bundled with it.
Recompile for testing: https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip (https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip)
This build runs really fast on my computer, much faster than all the other builds available here.
Is it not safe?
Recompile for testing: https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip (https://www.rarewares.org/files/lossless/flac-1.4.3-x64-AVX2-T.zip)
This build runs really fast on my computer, much faster than all the other builds available here.
Is it not safe?
It's fine. It was simply a recompile using a different version of GCC 13.1.0 as the previous version of the compiler caused a problem.
This build runs really fast on my computer, much faster than all the other builds available here.
Performance test thread: https://hydrogenaud.io/index.php/topic,123025.0.html
Yes indeed the speed differences might be substantial. See the Ozz build at the last line of my table here, it is not friends with my CPU: https://hydrogenaud.io/index.php/topic,123025.msg1029768.html#msg1029768 (that one is in "speed" terms, not in "time" terms).
Concerning the build you link to, which uses AVX2 instructions, I made a mistake in #338 and #347 (https://hydrogenaud.io/index.php/topic,123025.msg1030749.html#msg1030749): I intended to use that build in those tests, but by mistake picked the slower of
@john33 's builds.
Note, it is nothing "wrong" about making a build which does not require a more recent CPU - especially the official Xiph build should give priority to compatibility. But since I intended to test the faster builds, it was a "mistake" of mine to leave out the one I intended to test.
(Even newer CPUs can try Wombat's build with AVX-512: the "v4" of the two posted at https://hydrogenaud.io/index.php/topic,123025.msg1030680.html#msg1030680 .
I have given that a few tests at https://hydrogenaud.io/index.php/topic,123025.msg1030848.html#msg1030848 and https://hydrogenaud.io/index.php/topic,124437.msg1030916.html#msg1030916 ; on this particular CPU, it seems it does slight improvements at -8 and slower.
But nothing like the big differences quoted above.)
If the AVX-512 "v4" helps and interest is there i can do a reference 1.43 compile the same way.
FLAC v1.4.3 (GCC v14.1.0) & FLAC v1.4.3 (CLANG v18.1.4) Win Binaries
Built May 10, 2024