On 8/4/2026 10:51 AM, Jeffrey Law wrote:
On 7/21/2026 9:28 AM, [email protected] wrote:
From: Kyrylo Tkachov <[email protected]>
The stage 2 and stage 3 object comparison runs one command per object
in a
serial shell loop. Everything after it waits while an otherwise idle
machine
compares thousands of objects.
Generate one make target per object and run the comparisons through
$(MAKE),
so they share the original jobserver. Keep the generated makefile
and result
shards invocation-specific, clean them on normal and signal exits,
and sort
the shards before creating the traditional .bad_compare diagnostic.
Files in
the exclusion list still only warn. Comparator statuses other than
zero or
one now report an operational error instead of being treated as equal.
The tail-based fallback for cmp implementations without byte skipping
used
fixed temporary names. Give those files process-specific names,
clean them
with a trap, and preserve the same status contract for tail failures and
signals.
On highly parallel AArch64 and x86_64 systems this reduces comparison
wall
time by about 95%.
ChangeLog:
* Makefile.tpl ([+compare-target+]): Run object comparisons
through an
invocation-specific parallel sub-make. Collect sorted failure
shards
and clean temporary state on exit.
* Makefile.in: Regenerate.
* configure: Regenerate.
config/ChangeLog:
* acx.m4 (ACX_PROG_CMP_IGNORE_INITIAL): Use process-specific
temporary
files and clean them on exit. Diagnose tail failures as errors.
Signed-off-by: Kyrylo Tkachov <[email protected]>
Given the utility of the bootstrap comparison test, I hope you you
tested with a forced failure to ensure it'll trigger failures properly.
Assuming you've done a force failure test, OK.
BTW, I think this cut about 2hrs off the bootstrap time for
hppa-linux-gnu on my 40c systems. Similarly for alpha. I suspect the
others will show similarly sized gains (m68k, loongarch, sh4, etc).
Essentially everything where I'm using QEMU user mode emulation to
provide an emulated native environment suitable for bootstrapping and
regression testing I would expect to see comparable gains. It's a full
10%+ improvement in the total cycle time.
Jeff
