[colug-432] RISC V effort, and distribution building
R P Herrold
herrold at owlriver.com
Fri Mar 23 12:18:04 EDT 2018
Nice COLUG meeting
I mentioned a current RISC V synthetic machine family (with a
tangible one coming). RISC V has a large number of options --
register count, presence of hardware maths, bus widths, more.
The useful target has 'all' of the processor features, and the
64 bit bus, which is the 'big' $1k unit mentioned below, and
which represents real competition to ARM's campaign to get
into the DC
https://riscv.org/
Outlinks to the processor design details and philosophy
https://riscv.org/risc-v-foundation/
https://www.sifive.com/
There is a low power preliminary item available on short
delivery turn. I have contributed a couple times to this and
I see that I am listed on their 'backers' page -- sort of a
surprise. I found that these 'arduino profile' boards are not
powerful enough, nor 'sufficiently' ram'd to handle much of a
distribution, beyond what one gets on a re-purposed wireless
router and DD-WRT or such. The sifive folks are
slip-streaming in fixes / improvements to the chip over time
https://www.crowdsupply.com/sifive/hifive1
https://www.sifive.com/documentation/chips/freedom-e310-g000-manual/
https://www.sifive.com/documentation/chips/freedom-e310-g000-datasheet/
https://info.sifive.com/the-sifive-download-part-vi
(newsletter)
https://www.sifive.com/blog/
working up toward a release the big kahuna, already in the
pipeline
https://www.crowdsupply.com/sifive/hifive-unleashed
The eight SMA RF connectors to use impedance controlled
transmission lines are intriguing -- one assumes multi-gigibit
per second serial commo channels to sub-systems. I've seen
approaches like that in IBM mainframe kit on the Z and Power
families
https://github.com/riscv/riscv-pk
some random links
https://github.com/rwmjones/fedora-riscv-bootstrap/
https://fedorapeople.org/groups/risc-v/RPMS/
his builder described, and the stage 1 packages
https://github.com/rwmjones/fedora-riscv
Much of the package queueing and re-queueing, when improperly
solved dependencies and problems with prior builds are spotted
are manual. They are working toward get a Koji (Fedora's
Swiss army knife tool for managing lots of packages), and Mock
(a builder, which can chroot, and depsolve) going
As that that pair of build automation should then be able to
walk through the build rounds the needed 4 times (against the
software simulation, as they await generally available
hardware -- they have some access to one)
It does not QA it of course, but it gets rid of the monkey
work of resubmission
I run something similar with a pair called:
drunkard-build.sh
and its friend:
rph-rpmbuild.sh
and NFS, and a thoughtfully designed directory structure, and
mirroring of SRPMs. tool:
srcfind
says I presently have"
847611 /home/herrold/.tmp/srcfind.sh.cache.txt
7862 /home/herrold/.tmp/srcfind.sh.cache-ORC.txt
rather a lot of packages ;)
Some of the tools used for the cAos (post Red Hat
RHL -> RHEL cut-over) buildsystem are out at:
ftp://ftp.owlriver.com/pub/mirror/ORC/buildfarm/
and a support note to get people going burning our first
bootable ISO (cinch) which had custom initscripts
http://www.owlriver.com/projects/caos/cd-burning/
back to RISC-V
http://delorie.com/riscv/rpm-install-latest.html
QEMU provides a cross compile environment
https://github.com/riscv/riscv-qemu/commits/qemu-upstream-v5
The distribution (fedora / Rawhide) building is on
irc.freenode.net in channel: #fedora-riscv
I have back logs as follows
./2017/freenode/#fedora-riscv.12.log
./2018/freenode/#fedora-riscv.02.log
./2018/freenode/#fedora-riscv.01.log
./2018/freenode/#fedora-riscv.03.log
If interested, lemme know and I can post. I don't currently
have a preferred 'mark-up' tool for raw IRC logs -- I have
written such in the last for CentOS, but was never happy with
it. It really needed extension to do multiple passes and find
the more frequent posters for colorization, to better suppress
management messages like /part and /join and some other stuff
I am forgetting off the top of my head
-- Russ herrold
More information about the colug-432
mailing list