On 2/16/26 22:59, Mark Millard wrote:
> Prior numbers were based on just (nearly?) from scratch builds of the
> port-packages. This illustrates also with much smaller overall builds.
>
>
> Examples from official ampere4/5 builder records:
>
> Built+Failed for the below:
> ampere4 143arm64 1475+107 vs. 35759+111
> ampere5 150arm64 1476+130 vs. 35748+127
Typo. Both were ampere4, no ampere5 example is listed.
>
> ampere4 143arm64 qt6-webengine-6.10.1_2 20:10:22 vs. 27:06:38
> ampere5 150arm64 qt6-webengine-6.10.1_2 29:54:49 vs. 41:25:10
> approx. ratio (ampere5/ampere4) 1.48 vs. 1.53
Typo. Both were ampere4, no ampere5 example is listed. Ratios are
(150arm64/143arm64).
My actual focus is on the shorter times, were resource competition with
other builds that could lead to extra delays was likely significantly
smaller.
>
> So the shorter times are from much smaller incremental builds
> and the larger times are from (near?) from-scratch builds.
>
> ampere4 and ampere5 are reportedly nearly identical systems --so not
> expected to have large performance ratios if the same work is being done.
>
>
> Note: Various other long build time port packages do not have such large
> ratios.
>
>
My understanding is that beefy20 and beefy21 are very similar hardware.
I focus on the shortest times reported of the builds that I looked at
the qt6-webengine-6.10.1_2 records for (143amd64 and 150amd64 in 2026).
Built+Failed for the below:
beefy20 143amd64 qt6-webengine-6.10.1_2 1823+59
beefy21 150amd64 qt6-webengine-6.10.1_2 1474+92
beefy20 143amd64 qt6-webengine-6.10.1_2 09:05:49
beefy21 150amd64 qt6-webengine-6.10.1_2 11:26:20
approx. ratio (beefy21/beefy20) 1.26
So, while not as large of a ratio as for ampere{4,5}, still a notable one.
--
===
Mark Millard
marklmi at yahoo.com
