Bug#921198: mipsel speed: UTM8 vs Loongson 3A (was: Bug#921198: hard to track down)

[YunQiang Su]

Andreas Metzler <ametzler at bebt.de> 于2019年2月3日周日 下午6:51写道：
>
> On 2019-02-03 Willi Mann <willi at debian.org> wrote:
> > just for the record, I tried to reproduce the build failure on eller
> > (mipsel porterbox), but it builds fine there. I'm wondering whether the
> > timeout set on line 232 of tests/srp.c (40 seconds) is too short for
> > slower build hosts.
>
> Hello,
>
> The two builds on mipsel-aql-01.debian and mipsel-aql-02.debian.org
> (both Processor: LS3A-RS780-1w (Quad Core Loongson 3A))failed, the third
> try on mipsel-sil-01.debian.org (Processor: Rhino Labs UTM8) succeeded.
>
> Eller also has the same hardware (Processor: Rhino Labs UTM8) as the
> succeeding build-machine and gnutls builds reliably there. The failing
> test (srp) takes 22 seconds there.

Cavium's CPU works well much better than Loongson's

>
> Is the Loongson that much slower?

In generally, it is not due to Loongson is slower, but about the
Loongson's CPU is buggy.
https://patchwork.linux-mips.org/patch/21333/

We are working on add more Cavium's machines.
And Loongson claims that their future CPU won't be so buggy.

@Aurelien, gnutls28 should be another packages that should be pinned
on Cavium's CPU.

>
> cu Andreas
> (x-debbugs-cc-ing both Willi and porter mailing list)
> --
> `What a good friend you are to him, Dr. Maturin. His other friends are
> so grateful to you.'
> `I sew his ears on from time to time, sure'
>


-- 
YunQiang Su