Something like attached.
I will clean it up next week and submit it then.
It should also fix some arm64be related issues too.

Thanks,
Andrew Pinski

On Mon, Feb 3, 2020 at 6:17 PM Ian Lance Taylor <iant@golang.org> wrote:
>
> On Sun, Feb 2, 2020 at 2:27 AM Andreas Schwab <schwab@linux-m68k.org> wrote:
> >
> > I'm getting these errors on aarch64 with -mabi=ilp32:
> >
> > ../../../../libgo/go/runtime/mpagealloc.go:226:38: error: shift count overflow
> >   226 |  chunks [1 << pallocChunksL1Bits]*[1 << pallocChunksL2Bits]pallocData
> >       |                                      ^
> > ../../../../libgo/go/runtime/mgcscavenge.go:487:15: error: shift count overflow
> >   487 |    l2 := (*[1 << pallocChunksL2Bits]pallocData)(atomic.Loadp(unsafe.Pointer(&s.chunks[i.l1()])))
> >       |               ^
> > ../../../../libgo/go/runtime/mpagealloc.go:138:22: error: shift count overflow
> >   138 |   return uint(i) & (1<<pallocChunksL2Bits - 1)
> >       |                      ^
> > ../../../../libgo/go/runtime/mpagealloc.go:129:21: error: integer constant overflow
> >   129 |   return uint(i) >> pallocChunksL1Shift
> >       |                     ^
> > ../../../../libgo/go/runtime/mpagealloc_64bit.go:34:2: error: integer constant overflow
> >    34 |  summaryL0Bits,
> >       |  ^
>
> I'm not sure that gccgo ever fully worked with aarch64 -mabi=ilp32.
> In Go I think that will have to be represented with a new GOARCH
> value, arm64p32.
>
> Ian