Date: Mon, 27 Apr 2026 20:59:35 -0700 From: Mark Millard <marklmi@yahoo.com> To: freebsd-arm <freebsd-arm@freebsd.org> Subject: FYI for main armv7 on aarch64: Sizable armv7 build time differences on a Windows Dev Kit 2023: 04:30:43 vs. 08:10:38 Message-ID: <efe16e21-7347-46ba-9629-10697fb37dc9@yahoo.com>
index | next in thread | raw e-mail
Note: I've not done an analogous comparison/contrast for just-aarch64
directly. I was not expecting nor looking-for the scale of
poudriere-devel bulk time differences observed: such was not why I was
doing the builds.
Unfortunately the context has my personal builds mixed in for the
environment the builds were done in:
) The aarch64 boot-kernel is my personal build, targeting Cortex-A76 (so
matching the RPi5B and compatible with Windows Dev Kit 2023.) and is a
non-debug kernel build. Both the poudriere bulk runs are from the same
boot of the environment, so the same kernel was in use.
) The boot world is an officially distributed pkgbase armv7 world build.
Again: same boot-world from the same boot of the environment was in use.
Note the boot world has:
# ls -lod /etc/malloc.conf
lrwxr-xr-x 1 root wheel - 10 Mar 29 10:38 /etc/malloc.conf -> junk:false
) One poudriere-devel jail world is also such an officially distributed
pkgbase armv7 build.
) The other poudriere-devel jail world is an install of my personal
armv7 world build, targeting Cortex-A7 explicitly, and, likely far more
importantly, is a non-debug world build, for example:
WITHOUT_ASSERT_DEBUG=yes
WITH_MALLOC_PRODUCTION=
WITHOUT_PTHREADS_ASSERTIONS=yes
WITH_LLVM_LINK_STATIC_LIBRARIES=
WITHOUT_LLVM_ASSERTIONS=yes
Both poudriere-devel bulk runs allowed up to 8 builders, each builder
was allowed up to 8 make jobs: there are 8 cores and ALLOW_MAKE_JOBS=yes
was in use without explicit constraints.
So what varies here is the poudriere-devel jail world's.
[Note: The rust build failure for armv7 is known in modern times and I
let it attempt the rust builds anyway.]
So what are the timings like? . . .
) My personal armv7 poudriere-devel jail set up to target Cortex-A7:
. . .
[00:03:06] [05] [00:02:22] Finished devel/libtextstyle |
libtextstyle-1.0: Success
[00:03:21] [01] [00:01:25] Finished textproc/libxml2 | libxml2-2.15.3:
Success
[00:06:07] [07] [00:05:24] Finished lang/perl5.42 | perl5-5.42.2: Success
. . .
[00:08:19] [04] [00:06:23] Finished security/nmap | nmap-7.99_1: Success
[00:08:49] [02] [00:06:53] Finished lang/python311 |
python311-3.11.15_2: Success
. . .
[00:43:42] [01] [00:12:11] Finished devel/git@default | git-2.54.0:
Success
[00:47:13] [06] [00:32:45] Finished devel/freebsd-gcc15@armv7 |
armv7-gcc15-15.2.0_1: Success
[04:29:49] [08] [04:15:35] Finished devel/llvm22@default |
llvm22-22.1.3: Success
[04:30:47] [05] [04:13:12] Finished lang/rust | rust-1.94.0: Failed: build
. . .
[04:31:19] [main-CA7-default] [2026-04-26_18h29m11s] [committing] Time:
04:30:43
Queued: 97 Inspected: 41 Ignored: 0 Built: 50 Failed: 1
Skipped: 5 Fetched: 0 Remaining: 0
) An official pkgbase armv7 poudriere-devel jail:
. . .
[00:04:36] [05] [00:04:12] Finished devel/libtextstyle |
libtextstyle-1.0: Success
[00:05:17] [03] [00:02:25] Finished textproc/libxml2 | libxml2-2.15.3:
Success
[00:10:57] [07] [00:10:33] Finished lang/perl5.42 | perl5-5.42.2: Success
. . .
[00:14:35] [01] [00:11:43] Finished security/nmap | nmap-7.99_1: Success
[00:15:03] [02] [00:12:11] Finished lang/python311 |
python311-3.11.15_2: Success
. . .
[01:08:38] [02] [00:14:51] Finished devel/git@default | git-2.54.0:
Success
[01:21:20] [06] [00:55:24] Finished devel/freebsd-gcc15@armv7 |
armv7-gcc15-15.2.0_1: Success
[08:04:08] [01] [07:31:58] Finished lang/rust | rust-1.94.0: Failed: build
. . .
[08:10:37] [08] [07:45:12] Finished devel/llvm22@default |
llvm22-22.1.3: Success
. . .
[08:10:53] [main-armv7-default] [2026-04-27_10h18m44s] [committing]
Time: 08:10:38
Queued: 97 Inspected: 41 Ignored: 0 Built: 50 Failed: 1
Skipped: 5 Fetched: 0 Remaining: 0
The context difference is the jail world, not the boot-kernel, nor the
boot-world.
[I showed a variety of matching example builder times for comparison and
contrast.]
--
===
Mark Millard
marklmi at yahoo.com
home |
help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?efe16e21-7347-46ba-9629-10697fb37dc9>