SSD on the Netwinder
daniel at gimpelevich.san-francisco.ca.us
Sun Apr 21 08:54:23 PDT 2019
On Sat, 2019-04-20 at 20:33 -0400, Ralph Siemsen wrote:
> On Sat, Apr 20, 2019 at 11:45:56AM -0700, Daniel Gimpelevich wrote:
> >Rereading the https://wiki.debian.org/ArmEabiPort page, it's unclear
> >whether gcc8 tweaks will allow the armel port to support StrongARM as
> >v4t. You think it might be possible?
> I did not see any mention of gcc8 on that page, did I miss something?
> Note that the page is actually fairly old. There have been some recent
> cosmetic changes, but no substantive changes since at least 2013. Note
> that at that time, gcc-4.1 was considered "current".
> As for supporting StrongARM, it seems to me that armel will not support
> it. Debian wanted EABI, which makes sense for newer hardware, as it has
> much better floating point. But EABI also mandates Thumb interworking,
> which means using "bx" instruction. That's not possible on StrongARM.
The discussions on that page propose, among other things, mandating
Thumb interworking for only armv5 and later (Scenario #3). Most of those
proposals long predated the armhf port, so the possibility I was
referring to would be of banishing Thumb interworking (Scenario #4) from
armel built with gcc8 and --march=armv4t in favor of having only armhf
built with Thumb interworking, especially since there are competing
armhf ports: Debian (armv7), Raspbian (armv6), Ubuntu (armv7). I'm
hoping that after these years, the libgcc problem mentioned in Scenario
#7 no longer applies.
More information about the Netwinder