[Patch] New -fstack-check implementation (2/n)
Eric Botcazou
ebotcazou@adacore.com
Wed Nov 4 21:47:00 GMT 2009
> The middle-end parts are OK. This is OK if it is OK with the x86
> backend maintainers.
Thanks a lot. I installed yesterday the middle-end parts as well as the
couple of testcases. However, I overlooked that, on x86 and x86-64 Linux,
the Ada runtime does some pattern matching on the form of probes to make sure
that the stack pointer points to a valid page by the time execution resumes
after the stack overflow exception is propagated.
So I've just installed as obvious the small tweak to the x86 back-end that
changes the form of probes to IOR insns, they are shorter than the current
probes and have been used in the Windows probing code since the beginning.
I'll re-resubmit the rest of the patch to the x86 back-end for approval.
Tested on i586-suse-linux and x86_64-suse-linux,
2009-11-04 Eric Botcazou <ebotcazou@adacore.com>
PR target/10127
PR ada/20548
* config/i386/i386.md (probe_stack_range): New expander.
(logical operation peepholes): Do not split stack checking probes.
2009-11-04 Eric Botcazou <ebotcazou@adacore.com>
* ada/acats/norun.lst: Remove the stack checking tests.
* ada/acats/run_acats: Limit the stack to 8MB.
--
Eric Botcazou
-------------- next part --------------
A non-text attachment was scrubbed...
Name: p.diff
Type: text/x-diff
Size: 2686 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20091104/8df83f06/attachment.bin>
More information about the Gcc-patches
mailing list