This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 0/3] aarch64: Update ld.so for vector abi
- From: Szabolcs Nagy <szabolcs dot nagy at arm dot com>
- To: Richard Henderson <rth at twiddle dot net>, libc-alpha at sourceware dot org
- Cc: nd at arm dot com, marcus dot shawcroft at linaro dot org, Richard Henderson <richard dot henderson at linaro dot org>
- Date: Thu, 2 Aug 2018 10:26:08 +0100
- Subject: Re: [PATCH 0/3] aarch64: Update ld.so for vector abi
- References: <20180801222347.18903-1-rth@twiddle.net> <076e9d14-708a-dec8-2111-0ff6f7369c61@twiddle.net>
On 01/08/18 23:32, Richard Henderson wrote:
On 08/01/2018 06:23 PM, rth@twiddle.net wrote:
There is a new calling convention defined for vectorized functions [1].
I should have added that it appears that armclang is already
using this ABI, and QEMU has received a bug report about "not
working properly" with dynamic linking.
Whee!
this abi is only usable without lazy binding otherwise
it's not backward compatible with existing dynamic linkers.
the pcs document does not talk about this since lazy
binding is for elf targets only, i think it's possible
to make the code generation backward compatible: e.g.
make the vector pcs attribute imply the noplt attribute,
this fixes pic/pie code, for executables somehow bindnow
should be turned on (or move the corresponding relocations
out from DT_JMPREL part of the relocation table so the
dynamic linker processes them at load time)