[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gcl-devel] [PATCH 00/11] Add support for AArch64
From: |
Camm Maguire |
Subject: |
Re: [Gcl-devel] [PATCH 00/11] Add support for AArch64 |
Date: |
Mon, 16 Jun 2014 13:34:20 -0400 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/23.4 (gnu/linux) |
Greetings!
Will Newton <address@hidden> writes:
>> More on your patches later, but dlopen is really a last resort. The
>> limitation on open file descriptors alone makes it almost useless for
>> acl2. We have many custreloc targets making use of some sort of
>> trampoline, and I'd be happy to implement the same here, if machine
>> access can be made available. The quickest though non-optimal custreloc
>> port is usually via some -mlongcalls switch to gcc. Is this available
>> here?
>
> gcc aarch64 does not have -mlong-calls, and the closest equivalent,
> -mcmodel=large is incompatible with PIC. Could you point me to the
> code where other architectures implement trampolines? I had expected
> to find something like that but it wasn't obvious to me where the work
> was done.
>
Look at the mips, ppc, alpha reloc headers under h/. Mips is the most
complex, so I suggest starting with alpha.
Also, while I greatly appreciate your work on master, which will be
incorporated, this is the bleeding edge branch. To get a quick
port into the soon to be released 2.6.11, please look at the cygwin
branch in git. (So named, as cywgin, ppc64, and solaris will be newly
supported targets.) The release cycle (thus far) has basically been --
bold ideas into master and branches thereof, then every year or so move
the best ideas into the stable series 2.6.x. These are released as
Debian packages to test the supported apps on the autobuilders. So
2.6.10-1 in Debian is the official 2.6.10 release, and 2.6.10-14 (now)
is the test of the forthcoming 2.6.11.
To get a trampoline working should not demand a lot of speed. I've used
aranym to maintain m68k even :-). But you might be in the best position
to suggest a patch here. Even if gcc does not support longcalls, if you
know how to load a 64bit word into register and jump there, and how to
set the return, and have maybe an objdump snippet, we can work from that.
Take care,
--
Camm Maguire address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah
- [Gcl-devel] [PATCH 02/11] o/main.c: Ensure cs_org is always pointer aligned, (continued)
- [Gcl-devel] [PATCH 02/11] o/main.c: Ensure cs_org is always pointer aligned, Will Newton, 2014/06/11
- [Gcl-devel] [PATCH 06/11] Check for preceding NULL in find_init_name1, Will Newton, 2014/06/11
- [Gcl-devel] [PATCH 07/11] Fix handling of varargs in sSuniversal_error_handler, Will Newton, 2014/06/11
- [Gcl-devel] [PATCH 05/11] Always pass a va_list as the last arg of call_proc_new, Will Newton, 2014/06/11
- [Gcl-devel] [PATCH 10/11] Reduce threshold for increasing maxpage, Will Newton, 2014/06/11
- [Gcl-devel] [PATCH 09/11] Avoid looking up dynamic symbols in the executable, Will Newton, 2014/06/11
- [Gcl-devel] [PATCH 08/11] Handle dladdr failure correctly, Will Newton, 2014/06/11
- [Gcl-devel] [PATCH 11/11] Hack to disable x86 floating point code, Will Newton, 2014/06/11
- Re: [Gcl-devel] [PATCH 00/11] Add support for AArch64, Camm Maguire, 2014/06/11