LilyPond 2.25.14
We are happy to announce the release of LilyPond 2.25.14. This is termed a development release, but these are usually reliable for testing new features and recent bug fixes. However, if you require stability, we recommend using version 2.24.3, the current stable release. Please refer to the Installing section in the Learning Manual for instructions how to set up the provided binaries: https://lilypond.org/doc/v2.25/Documentation/learning/installing signature.asc Description: This is a digitally signed message part
Re: LilyPond 2.25.14
On Sat, 2024-03-23 at 14:25 +0100, Jonas Hahnfeld wrote: > We are happy to announce the release of LilyPond 2.25.14. This is > termed a development release, but these are usually reliable for > testing new features and recent bug fixes. Following this release, I made two special builds with the changes proposed in https://gitlab.com/lilypond/lilypond/-/merge_requests/2283 that would be great to get more widespread testing: Native Apple Silicon build: https://cloud.hahnjo.de/s/x9D62eASSn6Ng7D This is a native arm64 version for macOS 12+; if you have a Mac, please test and (hopefully) confirm the performance advantage over running the x86_64 build via Rosetta 2 🙂 Windows build with Guile JIT: https://cloud.hahnjo.de/s/Ek5x9rybpiPNtoj This turns on just-in-time compilation that was added in Guile 3.0, but we had to keep disabled on Windows until now. Please test, especially on larger scores where this should provide a performance advantage. Cheers Jonas signature.asc Description: This is a digitally signed message part
Re: LilyPond 2.25.14
WOW!!! Apple Silicon support!!! On Sat, Mar 23, 2024 at 8:03 AM Jonas Hahnfeld via LilyPond user discussion wrote: > > On Sat, 2024-03-23 at 14:25 +0100, Jonas Hahnfeld wrote: > > We are happy to announce the release of LilyPond 2.25.14. This is > > termed a development release, but these are usually reliable for > > testing new features and recent bug fixes. > > Following this release, I made two special builds with the changes > proposed in https://gitlab.com/lilypond/lilypond/-/merge_requests/2283 > that would be great to get more widespread testing: > > Native Apple Silicon build: https://cloud.hahnjo.de/s/x9D62eASSn6Ng7D > This is a native arm64 version for macOS 12+; if you have a Mac, please > test and (hopefully) confirm the performance advantage over running the > x86_64 build via Rosetta 2 🙂 > > Windows build with Guile JIT: https://cloud.hahnjo.de/s/Ek5x9rybpiPNtoj > This turns on just-in-time compilation that was added in Guile 3.0, but > we had to keep disabled on Windows until now. Please test, especially > on larger scores where this should provide a performance advantage. > > Cheers > Jonas
Re: LilyPond 2.25.14
On 3/23/2024 10:02 AM, Jonas Hahnfeld via LilyPond user discussion wrote: Windows build with Guile JIT: https://cloud.hahnjo.de/s/Ek5x9rybpiPNtoj This turns on just-in-time compilation that was added in Guile 3.0, but we had to keep disabled on Windows until now. Please test, especially on larger scores where this should provide a performance advantage. Carver MSDM.ly, subject of many past performance-test posts. Inte Core i5-3450 24 GB RAM Windows 10 22H2 real 0m32.109s user 0m0.000s sys 0m0.015s -- Karlin High Missouri, USA
Re: LilyPond 2.25.14
Le samedi 23 mars 2024 à 16:02 +0100, Jonas Hahnfeld via Discussions on LilyPond development a écrit : > Native Apple Silicon build: https://cloud.hahnjo.de/s/x9D62eASSn6Ng7D > This is a native arm64 version for macOS 12+ Great! OOI, what machine did you use for that? signature.asc Description: This is a digitally signed message part
Re: LilyPond 2.25.14
Hi Jonas; I've tried engraving a couple of randomly chosen previous engravings (using LP 2.24.x) using the new LP 2.25.14 version on my Mac (Apple Silicon). Since none of my engravings are large, most of them are one to two pages in length, I thought there would be no noticable decrease in engraving time, but I was wrong! It IS noticably faster for even small files :-) One thing I noticed is what one of my engravings had significant change in apperance in the second and third page, as if I had inserted a forced page break in two or three places. There were no conversion errors or warnings in any of my engravings. I don't know what was the cause of the speedup, but it is appreciated! I have been using the MacPorts version of Lilypond 2.24.3 compared to this recent test release. Thanks to all who have made this an awesome product! Ken Wolcott On Sat, Mar 23, 2024 at 10:20 AM Kenneth Wolcott wrote: > > WOW!!! Apple Silicon support!!! > > On Sat, Mar 23, 2024 at 8:03 AM Jonas Hahnfeld via LilyPond user > discussion wrote: > > > > On Sat, 2024-03-23 at 14:25 +0100, Jonas Hahnfeld wrote: > > > We are happy to announce the release of LilyPond 2.25.14. This is > > > termed a development release, but these are usually reliable for > > > testing new features and recent bug fixes. > > > > Following this release, I made two special builds with the changes > > proposed in https://gitlab.com/lilypond/lilypond/-/merge_requests/2283 > > that would be great to get more widespread testing: > > > > Native Apple Silicon build: https://cloud.hahnjo.de/s/x9D62eASSn6Ng7D > > This is a native arm64 version for macOS 12+; if you have a Mac, please > > test and (hopefully) confirm the performance advantage over running the > > x86_64 build via Rosetta 2 🙂 > > > > Windows build with Guile JIT: https://cloud.hahnjo.de/s/Ek5x9rybpiPNtoj > > This turns on just-in-time compilation that was added in Guile 3.0, but > > we had to keep disabled on Windows until now. Please test, especially > > on larger scores where this should provide a performance advantage. > > > > Cheers > > Jonas >
Re: LilyPond 2.25.14
On Sat, 2024-03-23 at 11:41 -0500, Karlin High wrote: > On 3/23/2024 10:02 AM, Jonas Hahnfeld via LilyPond user discussion wrote: > > Windows build with Guile JIT: https://cloud.hahnjo.de/s/Ek5x9rybpiPNtoj > > This turns on just-in-time compilation that was added in Guile 3.0, but > > we had to keep disabled on Windows until now. Please test, especially on > > larger scores where this should provide a performance advantage. > > Carver MSDM.ly, subject of many past performance-test posts. > > Inte Core i5-3450 > 24 GB RAM > Windows 10 22H2 > > real 0m32.109s > user 0m0.000s > sys 0m0.015s Would be interesting to see how this compares to the "vanilla" package here: https://gitlab.com/lilypond/lilypond/-/packages/23923980 ... But I guess I can also do this test on my end since the score is public. signature.asc Description: This is a digitally signed message part
Re: LilyPond 2.25.14
On Sat, 2024-03-23 at 17:46 +0100, Jean Abou Samra wrote: > Le samedi 23 mars 2024 à 16:02 +0100, Jonas Hahnfeld via Discussions on > LilyPond development a écrit : > > Native Apple Silicon build: https://cloud.hahnjo.de/s/x9D62eASSn6Ng7D > > This is a native arm64 version for macOS 12+ > > Great! OOI, what machine did you use for that? Ah yes, I forgot to mention I used the one macOS node (Mac mini 2020, Apple M1) from https://portal.cfarm.net/ This could be very nice going forward because anybody can just request an account. signature.asc Description: This is a digitally signed message part
Re: LilyPond 2.25.14
On Sat, 2024-03-23 at 11:39 -0700, Kenneth Wolcott wrote: > Hi Jonas; > > I've tried engraving a couple of randomly chosen previous engravings > (using LP 2.24.x) using the new LP 2.25.14 version on my Mac (Apple > Silicon). > > Since none of my engravings are large, most of them are one to two > pages in length, I thought there would be no noticable decrease in > engraving time, but I was wrong! It IS noticably faster for even > small files :-) > > One thing I noticed is what one of my engravings had significant > change in apperance in the second and third page, as if I had inserted > a forced page break in two or three places. There were no conversion > errors or warnings in any of my engravings. > > I don't know what was the cause of the speedup, but it is > appreciated! I have been using the MacPorts version of Lilypond 2.24.3 > compared to this recent test release. Thanks for testing! The performance advantage probably comes from Guile 3.0 in our binaries of 2.25.14 versus Guile 2.2 with the MacPorts version, having a faster startup time which matters especially for smaller files. If you have some time, I would be really interested in seeing a comparison with LilyPond 2.25.14 from https://gitlab.com/lilypond/lilypond/-/packages/23923980 when the x86_64 executable is run through Rosetta on Apple Silicon... Jonas signature.asc Description: This is a digitally signed message part
Re: LilyPond 2.25.14
On 3/23/2024 2:00 PM, Jonas Hahnfeld wrote: Would be interesting to see how this compares to the "vanilla" package I tried it, 3 runs. First one maybe unfair because font cache initialization or something. real0m48.643s user0m0.000s sys 0m0.031s real0m31.457s user0m0.015s sys 0m0.000s real0m31.591s user0m0.000s sys 0m0.031s The another with the JIT version: real0m32.319s user0m0.000s sys 0m0.030s But I expect I should get the computer otherwise idle as far as possible for this. I had other software running, but nothing with heavy loads. -- Karlin High Missouri, USA
Re: LilyPond 2.25.14
Hi Jonas; I hope that the following information is useful: Lilypond performance stats for the engraving of one very small and one medium sized pieces native = Apple Silicon build foreign = x86_64 other = MacPorts LP 2.24.3 ~/Downloads/native_lilypond-2.25.14/bin/lilypond ./Danny_Boy_updated.ly 0.53s user 0.06s system 87% cpu 0.683 total ~/Downloads/foreign_lilypond-2.25.14/bin/lilypond ./Danny_Boy_updated.ly 0.93s user 0.09s system 101% cpu 1.010 total /opt/local/bin/lilypond ./Danny_Boy.ly 0.48s user 0.06s system 104% cpu 0.515 total ~/Downloads/native_lilypond-2.25.14/bin/lilypond ./Waltz_Nr2_updated.ly 1.07s user 0.10s system 102% cpu 1.145 total ~/Downloads/foreign_lilypond-2.25.14/bin/lilypond ./Waltz_Nr2_updated.ly 1.76s user 0.15s system 100% cpu 1.907 total /opt/local/bin/lilypond ./Waltz_Nr2.ly 1.06s user 0.12s system 103% cpu 1.140 total for i in ./Danny_Boy.ly ./Danny_Boy_updated.ly ./Waltz_Nr2.ly ./Waltz_Nr2_updated.ly for> do for> wc $i for> done 232 8144641 ./Danny_Boy.ly 232 8144642 ./Danny_Boy_updated.ly 9393892 19981 ./Waltz_Nr2.ly 9393892 19982 ./Waltz_Nr2_updated.ly Ken Wolcott On Sat, Mar 23, 2024 at 12:09 PM Jonas Hahnfeld wrote: > > On Sat, 2024-03-23 at 11:39 -0700, Kenneth Wolcott wrote: > > Hi Jonas; > > > > I've tried engraving a couple of randomly chosen previous engravings > > (using LP 2.24.x) using the new LP 2.25.14 version on my Mac (Apple > > Silicon). > > > > Since none of my engravings are large, most of them are one to two > > pages in length, I thought there would be no noticable decrease in > > engraving time, but I was wrong! It IS noticably faster for even > > small files :-) > > > > One thing I noticed is what one of my engravings had significant > > change in apperance in the second and third page, as if I had inserted > > a forced page break in two or three places. There were no conversion > > errors or warnings in any of my engravings. > > > > I don't know what was the cause of the speedup, but it is > > appreciated! I have been using the MacPorts version of Lilypond 2.24.3 > > compared to this recent test release. > > Thanks for testing! The performance advantage probably comes from Guile > 3.0 in our binaries of 2.25.14 versus Guile 2.2 with the MacPorts > version, having a faster startup time which matters especially for > smaller files. If you have some time, I would be really interested in > seeing a comparison with LilyPond 2.25.14 from > https://gitlab.com/lilypond/lilypond/-/packages/23923980 when the > x86_64 executable is run through Rosetta on Apple Silicon... > > Jonas
Re: LilyPond 2.25.14
Hi Jonas, thanks for working on this! Windows build with Guile JIT: https://cloud.hahnjo.de/s/Ek5x9rybpiPNtoj This turns on just-in-time compilation that was added in Guile 3.0, but we had to keep disabled on Windows until now. Please test, especially on larger scores where this should provide a performance advantage. I gave it a try with a medium-sized score (appr. 60 pages A4), 10 times in a row: with JIT: real   4m42.513s user   0m0.090s sys 0m0.185s without JIT: real   4m45.295s user   0m0.046s sys 0m0.277s Just out of interest, do we tweak the JIT_THRESHOLD somewhere or do we use the default value? Michael
Re: LilyPond 2.25.14
On 2024-03-23 6:25 am, Jonas Hahnfeld via Discussions on LilyPond development wrote: We are happy to announce the release of LilyPond 2.25.14. This is termed a development release, but these are usually reliable for testing new features and recent bug fixes. However, if you require stability, we recommend using version 2.24.3, the current stable release. Please refer to the Installing section in the Learning Manual for instructions how to set up the provided binaries: https://lilypond.org/doc/v2.25/Documentation/learning/installing /opt/lilypond/2.25.14/bin/lilypond: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.28' not found (required by /opt/lilypond/2.25.14/bin/lilypond) Something changed from .13 to .14? I'm presuming my system is just out-of-date, and this might be expected. -- Aaron Hill
