Well, it's finally here. 0.10 has a lot of improvements. See the news page for the full list but here's a summary: * both encoder and decoder are significantly faster * flac now has gzip-like command-line usage (should be more intuitive) * the -# options have been tweaked for efficiency; -5 is the new default * flac now autodetects WAVE input * flac now ignores (with warning) unsupported WAVE subchunks * new -o and --delete-input-file options * XMMS plugin tweaked so that visualizations should not be so slow * new SEEKTABLE metadata block for storing seek points (speeds up seeking) http://flac.sourceforge.net/news.html This is very likely the last beta as I am happy enough with the performance. If there are no serious bugs this should become 1.0. Josh __________________________________________________ Do You Yahoo!? Get personalized email addresses from Yahoo! Mail - only $35 a year! http://personal.mail.yahoo.com/
Josh Coalson <xflac@yahoo.com> wrote:> This is very likely the last beta as I am happy enough with the > performance.Indeed, performance is very much improved. I've noticed that the assemblerized i386 version creates a different output file from the C version (alpha, sparc). Is that a known effect? -- Christian "naddy" Weisgerber naddy@mips.inka.de
> I've noticed that the assemblerized i386 version creates a different > output file from the C version (alpha, sparc). Is that a known > effect? >yes. the base type C floating point type used in libFLAC is now 32-bit float. so in C-land intermediate results tend to be affected by that precision whereas in the assembly code, we are free to stay in the 80-bit FPU stack as long as possible to keep the accuracy up. in some cases it will help the compression. but the resulting files in either case are valid. Josh __________________________________________________ Do You Yahoo!? Get personalized email addresses from Yahoo! Mail - only $35 a year! http://personal.mail.yahoo.com/