(I've swtiched from sage-support to to sage-devel.)

I can test and review. I think I know what to do and I would have just
tried to implement this myself, except that it would take me a while to
figure out how to make the change in such a way that it fits into the Sage
build process.

I spent some time trying to understand the issue and what this PARI stack
is all about. What I think I understand is something like:

- PARI uses it's own internal stack for quick memory allocation
- At the "top level", if I'm using PARI as a C-library, or through the GP
interpreter, the stack is empty between function calls, so it can be
- While in use, though, the stack can't really be resized, because
references to memory allocated on the stack don't use the stack pointer,
and resizing might require moving the stack
- If the stack runs out of space, PARI throws some error, and once upon a
time Sage would notice this error, increase the stack size, and repeat
- Now instead Sage just sets the stack size to be really big, which
probably is generally fine because modern systems generally don't
allocation memory until it is actually touched

(Is that all correct?)

Maybe this change in Sage corresponds somewhat to a change in the way that
PARI manages its stack. But from reading the PARI source code I can't
actually see that PARI is managing the stack in any way that is actually
distinguishable from just calling malloc() to allocate the stack and then
using it until it is full. Probably the use of MAP_NORESERVE is intended to
have some effect, but on Linux MAP_NORESERVE appears to do nothing.
Meanwhile, the paristack_resize() function just does some arithmetic, and
doesn't actually touch the allocated stack space at all. Luckily, the
paristack_resize() function does exist, though, and probably gets properly
used even though it doesn't really do anything, so a call to mprotect can
be added there, and meanwhile the mmap call can use PROT_NONE. Maybe those
are the only spots where the PARI source code needs to be changed.

(I'm not completely sure I'm correct about all of that.)

I'm probably going to try to modify a clean copy of PARI to do this, or
just write some completely separate test code to check that an mmap call
with PROT_NONE will work like we think it will work.

On Tue, Sep 20, 2016 at 12:02 PM, Jeroen Demeyer <jdeme...@cage.ugent.be>

> On 2016-09-20 12:54, Jonathan Bober wrote:
>>  From reading what you've sent, I guess that what you have in mind is
>> calling mmap with PROT_NONE and then calling mprotect() to change that
>> to read/write whenever growing the stack? That seems like it might be a
>> reasonable thing to do (though I'm only basing that on spending a few
>> minutes reading what you sent, not from any actual knowledge that I had
>> before).
> Yes, that is my idea.
> I'm willing to devote some time (not today) to figuring out what the
>> right thing to do is (maybe the above is already the right thing) /
>> implementing this / reviewing this.
> I don't mind implementing it. What I *do* mind is that I implement it and
> that the patch rots away on Sage Trac in needs_review state (highly
> specialized patches like these have a higher chance of that). That's why I
> asked for a commitment to review.
> --
> You received this message because you are subscribed to the Google Groups
> "sage-support" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-support+unsubscr...@googlegroups.com.
> To post to this group, send email to sage-supp...@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-support.
> For more options, visit https://groups.google.com/d/optout.

You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Reply via email to