Re: [systemd-devel] Is LTO worth it?

2021-01-20 Thread Cristian Rodríguez
On Mon, Jan 11, 2021 at 12:39 PM Lennart Poettering wrote: > https://fedoraproject.org/wiki/LTOByDefault > > (I think Suse is even further ahead on this) Yes. lto is already on by default, on worthiness of t..the jury is still out.. ___ systemd-devel

Re: [systemd-devel] Is LTO worth it?

2021-01-13 Thread Zbigniew Jędrzejewski-Szmek
On Mon, Jan 11, 2021 at 06:05:26PM +0100, Michael Biebl wrote: > Am Mo., 11. Jan. 2021 um 16:39 Uhr schrieb Lennart Poettering > : > > https://fedoraproject.org/wiki/LTOByDefault > > Interestingly, that wiki page says, that LTO should produce smaller > binaries, which clearly isn't the case here.

Re: [systemd-devel] Is LTO worth it?

2021-01-13 Thread Reindl Harald
Am 11.01.21 um 18:10 schrieb Michael Biebl: Am Mo., 11. Jan. 2021 um 18:07 Uhr schrieb Reindl Harald : it don't make sense using different flags in CI and production builds, especially LTO which often points out otherwise unvisible bugs Such as? I don't remember any bug report which was

Re: [systemd-devel] Is LTO worth it?

2021-01-13 Thread Reindl Harald
Am 11.01.21 um 16:05 schrieb Michael Biebl: Am Mo., 11. Jan. 2021 um 15:42 Uhr schrieb Reindl Harald : it shouldn't if properly used - means param with cpu-cores, just -flto alone is terrible slow -flto=%(nproc) %{__make} %{?_smp_mflags} The package uses meson's -Db_lto=true export

Re: [systemd-devel] Is LTO worth it?

2021-01-13 Thread Reindl Harald
Am 11.01.21 um 18:05 schrieb Michael Biebl: Am Mo., 11. Jan. 2021 um 16:39 Uhr schrieb Lennart Poettering : https://fedoraproject.org/wiki/LTOByDefault Interestingly, that wiki page says, that LTO should produce smaller binaries, which clearly isn't the case here. I wonder whether the wiki

Re: [systemd-devel] Is LTO worth it?

2021-01-13 Thread Reindl Harald
Am 11.01.21 um 15:34 schrieb Michael Biebl: we've been using LTO in the Debian build for as long as I can remember, but I begin to question whether that is a good idea. On Debian sid (gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.1), an LTO build almost takes

Re: [systemd-devel] Is LTO worth it?

2021-01-11 Thread Lennart Poettering
On Mo, 11.01.21 18:05, Michael Biebl (mbi...@gmail.com) wrote: > Concerning the build speed: I wonder whether at least disabling LTO on > our CI would make sense. We don't really care for fast/small > executables there. So, systemd's testsuite is certainly not supposed to be a testsuite for

Re: [systemd-devel] Is LTO worth it?

2021-01-11 Thread Michael Biebl
Am Mo., 11. Jan. 2021 um 18:26 Uhr schrieb Reindl Harald : > Am 11.01.21 um 18:10 schrieb Michael Biebl: > > Am Mo., 11. Jan. 2021 um 18:07 Uhr schrieb Reindl Harald > > : > >> it don't make sense using different flags in CI and production builds, > >> especially LTO which often points out

Re: [systemd-devel] Is LTO worth it?

2021-01-11 Thread Michael Biebl
Am Mo., 11. Jan. 2021 um 18:10 Uhr schrieb Michael Biebl : > > Am Mo., 11. Jan. 2021 um 18:07 Uhr schrieb Reindl Harald > : > > it don't make sense using different flags in CI and production builds, > > especially LTO which often points out otherwise unvisible bugs > > Such as? I don't remember

Re: [systemd-devel] Is LTO worth it?

2021-01-11 Thread Michael Biebl
Am Mo., 11. Jan. 2021 um 18:07 Uhr schrieb Reindl Harald : > it don't make sense using different flags in CI and production builds, > especially LTO which often points out otherwise unvisible bugs Such as? I don't remember any bug report which was uncovered by LTO being enabled.

Re: [systemd-devel] Is LTO worth it?

2021-01-11 Thread Michael Biebl
Am Mo., 11. Jan. 2021 um 16:39 Uhr schrieb Lennart Poettering : > https://fedoraproject.org/wiki/LTOByDefault Interestingly, that wiki page says, that LTO should produce smaller binaries, which clearly isn't the case here. I wonder whether the wiki is incorrect or whether this is a toolchain

Re: [systemd-devel] Is LTO worth it?

2021-01-11 Thread Michael Biebl
Am Mo., 11. Jan. 2021 um 16:12 Uhr schrieb Reindl Harald : > export LDFLAGS="-Wl,--as-needed -Wl,-z,now -Wl,-z,relro -pie %{optflags} > -flto=%(nproc)" > %meson Thanks for the hint. I tried it, but it didn't really help. ___ systemd-devel mailing list

Re: [systemd-devel] Is LTO worth it?

2021-01-11 Thread Lennart Poettering
On Mo, 11.01.21 15:34, Michael Biebl (mbi...@gmail.com) wrote: > Hi, > > we've been using LTO in the Debian build for as long as I can > remember, but I begin to question whether that is a good idea. > On Debian sid (gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU > Binutils for Debian)

Re: [systemd-devel] Is LTO worth it?

2021-01-11 Thread Michael Biebl
Am Mo., 11. Jan. 2021 um 15:42 Uhr schrieb Reindl Harald : > it shouldn't if properly used - means param with cpu-cores, just -flto > alone is terrible slow > > -flto=%(nproc) > %{__make} %{?_smp_mflags} The package uses meson's -Db_lto=true ___

[systemd-devel] Is LTO worth it?

2021-01-11 Thread Michael Biebl
Hi, we've been using LTO in the Debian build for as long as I can remember, but I begin to question whether that is a good idea. On Debian sid (gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.1), an LTO build almost takes twice as long as a no-LTO build. I also